Programmed −2/−1 Ribosomal Frameshifting in Simarteriviruses: an Evolutionarily Conserved Mechanism

Simarteriviruses are a group of arteriviruses infecting nonhuman primates, and a number of new species have been established in recent years. Although these arteriviruses are widely distributed among African nonhuman primates of different species, and some of them cause lethal hemorrhagic fever disease, this group of viruses has been undercharacterized. Since wild nonhuman primates are historically important sources or reservoirs of human pathogens, there is concern that simarteriviruses may be preemergent zoonotic pathogens. Thus, molecular characterization of simarteriviruses is becoming a priority in arterivirology. In this study, we demonstrated that an evolutionarily conserved ribosomal frameshifting mechanism is used by simarteriviruses and other distantly related arteriviruses for the expression of additional viral proteins. This mechanism is unprecedented in eukaryotic systems. Given the crucial role of ribosome function in all living systems, the potential impact of the in-depth characterization of this novel mechanism reaches beyond the field of virology.

Most RNA viruses have polycistronic genomes and have evolved strategies to overcome a limitation of the eukaryotic translation apparatus, namely that in general, only the 5=-most open reading frame (ORF) on an mRNA is translated. These include noncanonical translation mechanisms, such as programmed ribosomal frameshifting (PRF) and alternative initiation, and in addition, the expression of polyproteins that are subsequently cleaved by viral or host proteases. Viruses may also produce subgenomic mRNAs that are functionally monocistronic. Arteriviruses use several of these strategies to coordinate their complex replication cycles (15,16).
Arterivirus genomes vary in length between 12.5 and 15.5 kb and contain 10 to 15 known ORFs. All but two are located toward the 3= end of the genome and encode viral structural proteins that are translated from a nested set of subgenomic mRNAs (17). ORF1a and ORF1b, at the 5= end of the genome, comprise some three-quarters of the coding capacity and encode replicase-associated proteins. Translation of the genomic RNA yields the ORF1a polyprotein and, in addition, an ORF1ab fusion polyprotein following Ϫ1 PRF at the ORF1a/ORF1b junction (18,19). The two polyproteins are processed into individual functional nonstructural proteins (nsps) by ORF1a-encoded protease domains. In PRRSV, these comprise two papain-like proteases, PLP1␣ and PLP1␤, located in nsp1␣ and nsp1␤, respectively, a papain-like protease (PLP2) domain situated in the N terminus of nsp2, and a serine protease domain residing in nsp4. The rapid release of nsp1␣, nsp1␤, and nsp2 from the N terminus of the polyprotein is mediated by autocatalytic cleavage with PLP1␣ (between nsp1␣ and nsp1␤ [nsp1␣/ 1␤]), PLP1␤ (nsp1␤/2), and PLP2 (nsp2/3) (20). In the SHFV nsp1 region, three papainlike proteases (PLP␣, PLP␤, and PLP␥) are present within nsp1 subunits nsp1␣, nsp1␤, and nsp1␥. Similar to the case for PRRSV, these nsp1 subunits are released from the N terminus of SHFV polyproteins by autocleavages (21). Nsp2, the largest replicase subunit of arteriviruses, is a multifunctional protein that plays important roles in viral replication and virus-host interaction (20,(22)(23)(24)(25)(26)(27)(28)(29)(30). In addition to cleavage of the nsp2/3 site, the PLP2 domain functions as a cofactor for the serine protease during proteolytic processing of the C-terminal region of pp1a and pp1ab (20,31). The C terminus of nsp2 contains a highly conserved Cys-rich domain of unknown function and a multispanning transmembrane domain that plays a critical role in the formation of membranous structures (32).
PRRSV uses an unusual Ϫ2 PRF signal to direct efficient expression of an additional protein from the ϩ1 reading frame overlapping the nsp2-encoding region. The Ϫ2 PRF generates a transframe (TF) fusion protein, nsp2TF. It consists of the N-terminal two-thirds of nsp2, followed by a unique C-terminal domain that is encoded by a novel overlapping TF ORF (33). At the same frameshifting site, an immediate stop codon is generated by Ϫ1 PRF, which leads to the expression of a truncated nsp2, designated nsp2N (34). Remarkably, both cellular poly(rC) binding proteins (PCBPs) and viral nsp1␤ are required for efficient Ϫ2/Ϫ1 PRF (35). Sequence analysis shows that the signals for Ϫ2/Ϫ1 PRF, including a slippery sequence and downstream C-rich RNA motif, are highly conserved in all arterivirus genomes except that of EAV (36). However, variations in slippery sequences have been identified in several newly identified simarteriviruses. In this study, we demonstrate that Ϫ2/Ϫ1 PRF identified in PRRSV is evolutionarily conserved in non-EAV/-WPDV arteriviruses, with particular emphasis on simarteriviruses. This study provides additional insights into the biological characteristics of arteriviruses and advances our knowledge of noncanonical translation mechanisms in both virus infection and cellular systems. (39)(40)(41). The highly conserved motif consistently forms an ␣-helix. The longer distance between slippery sequence and C-rich sequence in WPDV and its different nsp1␤ protein conformation compared to those of other arteriviruses suggest that WPDV likely does not utilize frameshifting at this site. Taken together, these data suggest that Ϫ2/Ϫ1 PRF might be an evolutionarily conserved mechanism for the expression of additional viral proteins in non-EAV/-WPDV arteriviruses.
SHFV nsp1␤ transactivates ؊2/؊1 ribosomal frameshifting. In PRRSV, Ϫ2/Ϫ1 PRF results in the translation of two novel proteins, nsp2TF and nsp2N (34). To test our hypothesis that Ϫ2/Ϫ1 PRF is utilized to translate nsp2TF and nsp2N proteins in simarteriviruses, we initially characterized nsp2-related proteins in SHFV-infected MARC-145 cells. For the detection of nsp2-related proteins, we generated two monoclonal antibodies (mAbs) (mAb 133-243 and mAb 134-260) against the PLP2 domain, which is shared by SHFV nsp2 and the predicted nsp2TF and nsp2N. A rabbit polyclonal antibody (pAb) against the unique epitope (RLDSTVVFEETTPL) at the C terminus of nsp2TF was also generated for specific identification of nsp2TF. The N terminus of SHFV nsp2 (Gly484/Gly485/Gly486; the residue positions are based on the SHFV pp1a sequence) was previously determined by mass spectrometry sequence analysis (21). We further predicted the C terminus of SHFV nsp2 based on the putative cleavage site between nsp2 and nsp3. As shown in Fig. 2A, the cleavage site of SHFV nsp2/3 was predicted to be between Gly1236 and Gly1237 (the residue positions based on SHFV pp1a sequence) by analysis of sequence alignment with the nsp2/3 cleavage site in PRRSVs (20). Thus, the SHFV nsp2 gene is predicted to be nucleotides 1649 to 3901 of the SHFV genome, and the predicted molecular mass of the protein is 81.2 kDa. If Ϫ2/Ϫ1 PRF occurs, nsp2TF is translated from SHFV genomic nucleotides 1649 to 2860 fused to nucleotides 2859 to 3536, which results in a product with a predicted molecular mass of 68.7 kDa, and nsp2N is translated from SHFV genomic nucleotides 1649 to 2860 fused to nucleotides 2860 to 3093, resulting in a product with a predicted molecular mass of 52 kDa (Fig. 2B).
Western blot analysis was performed to identify nsp2-related proteins in SHFVinfected MARC-145 cells. As shown by the results in Fig. 2, four major virus-specific bands were detected by mAb 134-260 against SHFV PLP2 (Fig. 2C, left). The top three largest proteins appear to be nsp2, nsp2TF, and nsp2N, although their masses are larger than those that were predicted-a phenomenon that was previously reported for nsp2-related proteins of PRRSV (33). The rabbit polyclonal antibody specifically designed to detect SHFV nsp2TF recognized two protein bands in the Western blot analysis. The top band appeared to be a cellular protein, since this band was also detected in mock-infected cells, while the second large protein band was specific to nsp2TF (Fig. 2C, right), indicating that SHFV indeed expresses nsp2TF via Ϫ2 PRF. The fourth band, with a size of less than 50 kDa, is a C-terminally truncated isoform of nsp2, which may be generated by proteolytic cleavage by PLP2 or some other proteinase. The identity of this product needs to be further studied in the future. To confirm the identity of these nsp2-related proteins, immunoprecipitation (IP) was performed with anti-PLP2 mAb 133-243 using cell lysate of SHFV-infected MARC-145 cells, whereas cell lysate of mock-infected MARC-145 cells was used as a negative control. As expected, four major bands were detected in the IP product using mAb 134-260 (Fig. 2D, left), and the second-largest protein band was also recognized by the nsp2TF-specific pAb (Fig. 2D, right). An ectopic expression system was further employed to investigate whether the expression of SHFV nsp2TF and nsp2N requires nsp1␤ as a PRF transactivator. In HEK-293T cells transfected with a plasmid expressing SHFV nsp2 alone, the PRF products, nsp2TF and nsp2N, were not detected (Fig. 3A, lane marked "EV"). When HEK-293T cells were cotransfected with plasmids expressing nsp2 and nsp1␤, both PRF products, nsp2TF and nsp2N, were detected (Fig. 3A, lane marked "nsp1␤"). These results indicate that SHFV nsp1␤ is critical for the expression of nsp2TF and nsp2N. To further identify the key residues in SHFV nsp1␤, amino acids Gly109, Lys110, Tyr111, Arg114, and Arg115 in the conserved motif (Fig. 1A) were targeted for mutagenesis . SHFV nsp2 is predicted to be released from pp1a through proteolytic cleavage by the PLP2 domain. The cleavage site between SHFV nsp2 and nsp3 was predicted to be at Gly1236/Gly1237. For each sequence, the pp1a coordinate of the last amino acid in the alignment is specified. (B) Schematic diagram of putative SHFV nsp2, nsp2TF, and nsp2N. SHFV nsp2 is encoded by the SHFV genomic region, comprising nucleotides (nt) 1649 to 3901, whereas nsp2TF (nt 1649 to 2860 ϩ 2859 to 3536) and nsp2N (nt 1649 to 2860 ϩ 2860 to 3093) are translated through Ϫ2 PRF and Ϫ1 PRF, respectively. The epitopes recognized by antibodies are indicated by black arrows. (C) SHFV nsp2-related proteins detected by Western blot (WB) analysis. MARC-145 cells were infected with SHFV at an MOI of 0.01, and cell lysates were harvested at 36 hpi. The nsp2-related products, including nsp2, nsp2TF, nsp2N, and an unknown nsp2-related protein, marked with an asterisk (*), were detected by anti-PLP2 mAb 134-260. SHFV nsp2TF was also recognized by rabbit pAb against the nsp2TF C-terminal peptide (nsp2TFC). (D) SHFV nsp2-related proteins detected by immunoprecipitation and WB analysis. MARC-145 cells were infected with SHFV at an MOI of 0.01, and cell lysates were harvested at 36 hpi. The nsp2-related products were immunoprecipitated (IP) using anti-PLP2 mAb 133-243 and probed by WB with anti-PLP2 mAb 134-260 and rabbit anti-nsp2TFC pAb. An unknown nsp2-related protein is marked with an asterisk (*).
with alanine replacement. Consistent with our previous findings on PRRSV (34,38), mutation of R to A at position 114 (R114A) completely abolished the expression of nsp2TF and nsp2N, whereas the G109A and K110A mutations had no obvious effects on the expression of nsp2TF and nsp2N. The R115A mutant stimulated lower expression levels of nsp2TF and nsp2N, which may be due to the lower expression level of this nsp1␤ mutant. The Y111A mutant was also unable to stimulate the expression of nsp2TF and nsp2N, indicating that Tyr111 is another key residue for nsp1␤'s function in transactivation of Ϫ2/Ϫ1 PRF (Fig. 3A). This result was further confirmed with PRRSV: nsp2TF and nsp2N were not detectable in HEK-293T cells expressing PRRSV nsp1␤ mutants that contained the corresponding alanine substitution at Y125 (PRRSV-2) or Y131 (PRRSV-1) (Fig. 3B). Similar to the R115A mutant, a lower expression level was observed for the SHFV Y111A mutant (Fig. 3A). However, this lower expression should not be the direct reason leading to the loss of expression of nsp2TF and nsp2N, since the Y111A mutant, with an expression level similar to that of wild-type (WT) nsp1␤ in vY111A-infected cells, was unable to stimulate the expression of nsp2TF and nsp2N (see Fig. 5).

FIG 3
Identification of key residues involved in the PRF transactivation function of nsp1␤ from SHFV and PRRSV. (A) Immunoprecipitation and Western blot (WB) analysis of amino acid residues critical for the transactivation function of SHFV nsp1␤. HEK-293T cells were cotransfected with a plasmid expressing HA-tagged SHFV nsp2 and a plasmid expressing FLAG-tagged SHFV nsp1␤ or mutants thereof. An empty vector (EV) was used as the control. Whole-cell lysates (WCLs) were harvested at 36 h posttransfection (hpt). The expression of nsp2, nsp2TF, and nsp2N was detected by WB using anti-HA mAb, whereas FLAG-tagged nsp1␤ was detected using anti-FLAG mAb M2. The expression of HA-tagged nsp2TF was further confirmed by IP using anti-HA mAb and WB detection using anti-nsp2TF C-terminal peptide pAb (anti-TFC). (B) Immunoprecipitation and WB analysis of amino acid residues critical for the transactivation function of nsp1␤ from PRRSV-1 and PRRSV-2. HEK-293T cells were cotransfected with a plasmid expressing PRRSV nsp2 and a plasmid expressing FLAG-tagged PRRSV nsp1␤ or mutants thereof. An empty vector (EV) was used as a control. PRRSV nsp2-related proteins were immunoprecipitated with anti-PRRSV PLP2 mAb. The expression of nsp2, nsp2TF, and nsp2N was evaluated by WB using anti-PRRSV PLP2 domain mAb, and FLAG-tagged nsp1␤ was detected with anti-FLAG mAb M2. GAPDH was monitored as a loading control.

Ϫ2/Ϫ1 PRF in Simarteriviruses
Journal of Virology A dual luciferase reporter system was employed to confirm the activity of SHFV nsp1␤ in stimulating Ϫ2/Ϫ1 PRF at the predicted PRF signal in the SHFV genome. A dual luciferase reporter plasmid, pDluc-SHFV/WT, was generated using the approach that we described previously (34), in which 79 nt containing the SHFV PRF signal was inserted in the plasmid between its Renilla luciferase (Rluc) and firefly luciferase (Fluc) ORFs. In this construct, Fluc was encoded in the Ϫ2 frame relative to the upstream Rluc ORF. Besides an in-frame Rluc product (stop), two frameshifting products (Ϫ2FS and Ϫ1FS) could be expressed through Ϫ2/Ϫ1 PRF (Fig. 4A). The predicted molecular masses of the stop, Ϫ1FS, and Ϫ2FS products are 40.4 kDa, 43.4 kDa, and 100.1 kDa, respectively. In HEK-293T cells cotransfected with the reporter plasmid and a plasmid expressing SHFV nsp1␤, all three nsp2-related proteins were detected at the predicted sizes by a mAb against their shared N-terminal Rluc (Fig. 4B, lane marked "nsp1␤"). Only the in-frame translation product (stop) was detected in cells without the expression of SHFV nsp1␤ (Fig. 4B, lane marked "EV" in the pDluc-SHFV/WT-transfected cells). The ability of nsp1␤ mutants to activate Ϫ2/Ϫ1 PRF was also tested using this reporter system. In cells expressing the SHFV nsp1␤ Y111A or R114A mutants, Ϫ2FS and Ϫ1FS expression products were undetectable, indicating that Tyr111 and Arg114 are essential to nsp1␤'s ability to stimulate frameshifting (Fig. 4B). In cells expressing the SHFV nsp1␤ R115A mutant, reduced expression of the Ϫ2FS product was observed, whereas the Ϫ1FS product was undetectable. Compared to the reduced expression levels of nsp2TF and nsp2N stimulated by the R115A mutant in Fig. 3A, SHFV nsp1␤ R115A mutant in the luciferase reporter system may stimulate a much lower level of Ϫ1FS, which is under the detection limit of Western blot analysis.
These findings were confirmed in an in vitro translation assay using rabbit reticulocyte lysate (RRL) programmed with an in vitro-transcribed reporter mRNA and purified nsp1␤ protein (Fig. 4C). In this system, the predicted masses of stop, Ϫ1FS, and Ϫ2FS products are 40.4 kDa, 43.4 kDa, and 70.8 kDa. When in vitro translation was performed with only reporter mRNA from pDluc-SHFV/WT, the Ϫ2FS product was not detected and only a trace amount of Ϫ1FS was observed (Fig. 4C, lane marked "DB" [dilution buffer]). In contrast, with the addition of purified nsp1␤, Ϫ1FS and Ϫ2FS were detected at the predicted molecular masses. Furthermore, within the range of 0ϳ1 M of nsp1␤, the expression levels of the Ϫ1 and Ϫ2 PRF products were observed to be dose dependent ( Fig. 4C and D). When the concentration of nsp1␤ was higher than 1 M, both frameshifts reached their maximum efficiencies, which were 11.7% (Ϫ2 PRF) and 3.5% (Ϫ1 PRF) (Fig. 4D). Furthermore, the purified nsp1␤ R114A mutant was impaired in its ability to stimulate Ϫ2/Ϫ1 PRF, as demonstrated by the disappearance of the Ϫ2FS product, and reduction of Ϫ1FS product to the background level (Fig. 4E).
To further confirm these results in the context of SHFV infection, two recombinant viruses, vY111A and vR114A, were rescued using an SHFV reverse genetics system (constructed from variant strain NIH LVR42-0/M6941). The ability of vY111A and vR114A to express nsp2TF and nsp2N was evaluated by Western blot analysis using cell lysates of SHFV-infected MARC-145 cells. In MARC-145 cells infected with either of the two nsp1␤ mutants, nsp2TF was not detected and only a low expression level of nsp2N was detected (Fig. 5), which is consistent with the results generated in the in vitro expression systems.
Both the slippery sequence and C-rich motif are required for efficient ؊2/؊1 PRF in simian arteriviruses. As described above (Fig. 1A), the Ϫ2/Ϫ1 PRF signals (slippery sequence and C-rich RNA motif) were found in all simarteriviruses. However, four types of slippery sequence were observed in simarteriviruses of different species, namely G_GUU_UUU (SHFV), U_GUU_UUU (DeMAV, KRTGV, and PBJV), G_GUC_UCU (KKCBV, KRCV-1, KRCV-2, and MYBV-1), and U_UUC_UCU (FSVV, SHEV, and ZMbV-1). To test whether some of these variant sequences could support Ϫ2/Ϫ1 PRF in the context of the SHFV 3= stimulatory sequence and SHFV nsp1␤, plasmids pDluc-SHFV/SS1 and pDluc-SHFV/SS2 were created by introducing mutations at the SHFV slippery sequence in the dual luciferase reporter plasmid pDluc-SHFV/WT to mimic the PRF signals of distinct simarteriviruses (Fig. 6A). In the pDluc-SHFV/SS1 construct, the slippery se-quence (G_GUU_UUU) was changed to G_GUC_UCU, whereas in the pDluc-SHFV/SS2 construct, the slippery sequence was changed to U_UUC_UCU (Fig. 6A). In HEK-293T cells expressing SHFV nsp1␤, the SHFV/WT slippery sequence permits Ϫ2/Ϫ1 PRF, as evidenced by the detection of Ϫ1FS and Ϫ2FS expression products with mAb against Renilla luciferase in Western blots (Fig. 6B). In contrast, only the Ϫ2FS product was detected when the SHFV/SS1 or SHFV/SS2 construct was used. We further confirmed this result using an in vitro translation assay in RRL. The reporter mRNAs transcribed from pDluc-SHFV/SS1 or pDluc-SHFV/SS2 constructs and translated in RRL only generated stop and Ϫ2FS products in the presence of SHFV nsp1␤ (Fig. 6C). We further confirmed the results using simarterivirus KRCV-1. As indicated above, the genome of this virus contains the slippery sequence (G_GUC_UCU). The predicted coding regions of KRCV-1 nsp1␤ and nsp2 were cloned into eukaryotic expression vectors. The predicted molecular masses of KRCV-1 nsp2 and nsp2TF are 77.9 kDa and 64.7 kDa. In HEK-293T cells cotransfected with plasmids containing nsp1␤ and nsp2, the expression of nsp2TF was detected, but no nsp2N was detected. As we expected, no frameshifting products were detected in HEK-293T cells that were not transfected with the plasmid expressing nsp1␤ (Fig. 6D). These data indicate that the slippery sequence of KRCV-1 lacks the ability to support Ϫ1 PRF, which is consistent with the results generated with the luciferase reporter system. Subsequently, we investigated the function of the C-rich RNA motif in SHFV Ϫ2/Ϫ1 PRF. The plasmid pDluc-SHFV/CC1 was constructed by introducing synonymous mutations to disrupt two C-rich patches within the PRF signal (Fig. 6A). HEK-293T cells were cotransfected with pDluc-SHFV/CC1 and the plasmid expressing SHFV nsp1␤. No frameshifting products were detected (Fig. 6B). This result was further confirmed by in vitro translation assay using RRL. Again, no frameshifting products were detected in in vitro translation reactions when using the reporter mRNAs transcribed from the pDluc-SHFV/CC1 construct (Fig. 6C), thus indicating that the substitutions of C residues in the C-rich RNA motif knocked out both Ϫ2 and Ϫ1 PRF.
Next, we confirmed the data from in vitro analysis using recombinant viruses carrying the designed mutations in pDluc-SHFV/SS1, pDluc-SHFV/SS2, and pDluc-SHFV/ CC1. Using the SHFV reverse genetics system, three recombinant viruses were generated, which were designated vSS1, vSS2, and vCC1. Western blot analysis was performed to assess the expression of nsp2TF and nsp2N in WT-and mutant virus-infected MARC-145 cells. The results were consistent with the data generated in the in vitro expression system, in which the Ϫ1 PRF product (nsp2N) was not detected in vSS1 and vSS2, whereas expression of both Ϫ2 PRF and Ϫ1 PRF products was not detected in vCC1 (Fig. 6E). These data indicate that the C-rich RNA motif is required for efficient Ϫ2/Ϫ1 PRF in simarteriviruses and that X_XUC_UCU variants of the slippery sequence lack the ability to facilitate Ϫ1 PRF. SHFV nsp2TF was also recognized by rabbit pAb against the nsp2TF C-terminal peptide (nsp2TFC). The expression of SHFV nsp1␤ and mutants thereof was evaluated with mAb 76-69. GAPDH was monitored as a loading control.
Poly(rC) binding proteins are important for efficient ؊2/؊1 PRF in simarteriviruses. In our previous study, poly(rC) binding proteins (PCBPs) were demonstrated to be critical for Ϫ2/Ϫ1 PRF in PRRSV (35). To test whether PCBPs are also involved in stimulating Ϫ2/Ϫ1 PRF in simarteriviruses, in vitro translations were performed with SHFV reporter mRNA and the addition of nsp1␤ and/or PCBP2 in wheat germ extract (WGE). In vitro translation using RRL was included as the control. As expected, in RRL, both Ϫ2FS and Ϫ1FS were stimulated by the presence of nsp1␤ (Fig. 7A). In the WGE system, although there was some expression of Ϫ2FS and Ϫ1FS products in the absence of any exogenous protein, their levels were greatly stimulated only upon the addition of both nsp1␤ and PCBP2. The presence of nsp1␤ or PCBP2 alone showed no stimulatory effect on the translation of Ϫ2FS and Ϫ1FS products. (Note that, in contrast to RRL, WGE likely contains endogenous PCBPs that are divergent from those of mammalian cells and not active in the stimulation of PRF.) Interestingly, the frameshifting efficiencies for Ϫ1 PRF in WGE were much higher than those observed using RRL (Fig. 7B). With the addition of nsp1␤ and PCBP2 in WGE, the efficiency of Ϫ2 PRF increased from 0.9% to 9.8%, whereas the efficiency of Ϫ1 PRF increased from 8% to 29%. These data suggest that PCBPs are required for efficient Ϫ2/Ϫ1 PRF in simarteriviruses.
Previously, the interaction between PCBPs and PRRSV-1 nsp1␤ was determined to be required for nsp1␤'s ability to bind the PRRSV-1 Ϫ2/Ϫ1 PRF RNA signal (35). In this study, we further analyzed the interactions between PCBP1/2 and SHFV nsp1␤. In HEK-293T cells transfected with plasmids expressing FLAG-tagged SHFV nsp1␤ and PCBP2, the protein complex of nsp1␤ and PCBP2 was immunoprecipitated by anti-FLAG mAb and subsequently detected by Western blot analysis using anti-PCBP2 mAb (Fig. 8A). Consistently, in SHFV-infected MARC-145 cells, nsp1␤ was pulled down by the mAb against PCBP2 (Fig. 8B). Similarly, the interaction between SHFV nsp1␤ and PCBP1 was also demonstrated by immunoprecipitation and Western blot analysis (Fig. 8C).
The frameshift products play a role in SHFV replication in vitro. As described above, a panel of recombinant viruses containing mutations in nsp1␤ or the Ϫ2/Ϫ1 PRF signal regions was rescued using SHFV reverse genetics. Five recombinant viruses with deficiencies in the expression of nsp2N and/or nsp2TF were passaged five times (P5 viruses) in MARC-145 cells, and the introduced mutations were verified for P5 viruses (data not shown). The P3 recombinant viruses were used to evaluate viral growth kinetics in vitro, and WT SHFV recovered by the reverse genetics was included as the control. Growth kinetics analysis showed that vY111A, vR114A, and vCC1 attenuated viral growth in MARC-145 cells, whereas vSS1 and vSS2 displayed growth kinetics similar to that of WT virus (Fig. 9A). Before 48 h postinfection (hpi), the virus titers of vCC1 were reduced by about 1 log compared to the titer of the WT virus. In contrast, the growth of the vY111A and vR114A mutants was more significantly reduced than vCC1, with virus titers decreasing 1.5 to 3 log throughout the time course of the study (Fig. 9A). The plaque assay results consistently showed that the vCC1, vY111A, and vR114A mutants developed smaller plaques than those caused by WT virus (Fig. 9B). All three mutants displaying attenuated growth have lost the ability to express nsp2TF and nsp2N ( Fig. 5 and 6E). On the other hand, vSS1 and vSS2 express nsp2TF but not nsp2N and had growth kinetics more similar to that of WT virus. This  finding suggests that nsp2TF plays a role in SHFV replication, whereas nsp2N appears not to be important for viral growth in MARC-145 cells. Nonetheless, the relatively mild attenuation of vCC1, which expresses neither nsp2TF nor nps2N, confirms that neither protein is essential for viral growth in cell culture, but they may be important for maintaining maximal virus fitness.
Heterotypic arterivirus nsp1␤s stimulate ribosomal frameshifting on the SHFV ؊2/؊1 PRF signal. The genomes of most arteriviruses (with the exception of EAV and WPDV) share a highly conserved RNA-binding motif in nsp1␤, which is critical for nsp1␤'s function in Ϫ2/Ϫ1 PRF transactivation (Fig. 1B). To further determine whether this binding motif is evolutionarily conserved in the Arteriviridae family, we evaluated the ability of PRRSV-1 nsp1␤ to transactivate frameshifting upon the Ϫ2/Ϫ1 PRF signal from SHFV and vice versa. In vitro translation was performed in the RRL system with reporter mRNA of SHFV and nsp1␤ protein of PRRSV-1. As expected, Ϫ2 and Ϫ1 frameshift products were detected. Increasing concentrations of PRRSV-1 nsp1␤ led to dose-dependent expression of the two PRF products (Fig. 10A, left, and Fig. 10B). . Each data point shown represents the mean value from two independent experiments, and error bars show standard errors of the means (SEM). (C) Analysis of Ϫ2/Ϫ1 frameshifting at the SHFV PRF signal when stimulated by heterotypic arterivirus nsp1␤s. HEK-293T cells were cotransfected with the plasmid pDluc-SHFV/WT and a plasmid expressing heterotypic nsp1␤. Empty vector (EV) was used as a control. Nonframeshift, Ϫ1 PRF, and Ϫ2 PRF products (indicated as stop, Ϫ1FS, and Ϫ2FS, respectively) were detected by Western blotting using anti-Renilla luciferase mAb. FLAG-tagged nsp1␤ was detected with mAb M2, and EAV nsp1 was probed with mAb 12A4. GAPDH was detected as a loading control.
To confirm that the PRF transactivation mechanism of nsp1␤ is also conserved among other arteriviruses, we included additional heterotypic nsp1␤s in the assay. The Ϫ2FS product was detected in HEK-293T cells transfected with pDluc-SHFV/WT and a plasmid expressing nsp1␤ from arteriviruses of other species, including KRCV-1, PRRSV-1, PRRSV-2, and LDV. Again, the Ϫ2FS product was detected; however, the Ϫ1FS product was not observed, which may be due to the low efficiency of Ϫ1 PRF (Fig. 10C). No frameshifting product was detected in cells expressing EAV nsp1. Since the canonical Ϫ2/Ϫ1 PRF signal was identified in viruses of all known arteriviral species except EAV and WPDV, these data demonstrate that the transactivation function of nsp1␤ on Ϫ2/Ϫ1 PRF is evolutionarily conserved in non-EAV/-WPDV arteriviruses.

DISCUSSION
Arteriviruses are a group of mammalian positive-sense RNA viruses. Although most of them have not been associated with overt disease, some arteriviruses cause acute respiratory syndrome, abortion, lethal hemorrhagic fever, or neurological impairment (42)(43)(44). EAV, PRRSV-1, and PRRSV-2 are veterinary pathogens with significant economic impact (7). PBJV, SHEV, and SHFV are known etiologic agents of almost uniformly lethal viral hemorrhagic fever in macaques (8). Most related simarteriviruses have not been identified as pathogens, but their infectivity for and transmission ability among nonhuman primates cause concern regarding zoonotic transmission (14). Improved understanding of the biological characteristics of arteriviruses would facilitate the development of disease control strategies and may also advance our knowledge of the factors that drive zoonotic transmission of RNA viruses.
The arterivirus Ϫ2/Ϫ1 PRF and the involvement of a transactivating viral protein and host factors in ribosomal frameshifting are unprecedented in eukaryotic systems. Our recent studies revealed that the PRF products of PRRSV, nsp2TF, and nsp2N are important for viral replication. On the other hand, these proteins function as innate immune antagonists, suggesting that recombinant PRRSV with impaired nsp2TF/nsp2N expression could be developed as candidate vaccines (45). In a recent comparative genomic study, conserved ϩ1 and Ϫ2 PRF signals were identified in additional sex combs-like (ASXL) genes 1 and 2, respectively, and hypothesized to be utilized for the expression of a conserved overlapping ORF via PRF (46). ASXL genes encode important epigenetic and transcriptional regulatory proteins, and truncation or frameshift mutants of ASXL are linked to myeloid malignancies and genetic diseases. This study highlights the significance of the Ϫ2 PRF mechanism, suggesting that the mechanism could be more widely employed in regulating viral/host gene expression.
Comparative genomic analysis of 19 arteriviruses revealed that the key elements for Ϫ2/Ϫ1 PRF, including the slippery sequence and downstream C-rich RNA motif, are highly conserved in all known arteriviruses except EAV. Of note, the distance between the slippery sequence and downstream C-rich RNA motif is consistently 9 or 10 nt in all non-EAV arteriviruses, with the exception of that of WPDV, which has a stretch that is 9 nucleotides longer. WPDV is phylogenetically distant from other arteriviruses (47). It also lacks a long overlapping TF ORF (36); thus, like EAV, WPDV most likely does not utilize frameshifting in this region of the nsp2 gene. Our experimental data indicate that a Ϫ2/Ϫ1 PRF mechanism similar to that used by PRRSV is employed by simarteriviruses to express nsp2TF and/or nsp2N analogs. In addition, we also experimentally demonstrated that nsp1␤ proteins of other arteriviruses (KRCV-1, PRRSV-1, PRRSV-2, and LDV, but not EAV) are able to stimulate ribosomal frameshifting on the SHFV Ϫ2/Ϫ1 PRF RNA signal. These results indicate that Ϫ2/Ϫ1 PRF is an evolutionarily conserved mechanism used by most arteriviruses for the expression of additional viral proteins. Furthermore, in line with our previous study on PRRSV, at least nsp2TF plays an important role in SHFV replication. Previous studies suggest that the N-terminal PLP2 domain shared by nsp2-related proteins of arteriviruses (PRRSV-1, PRRSV-2, LDV, and SHFV) suppresses the host type I IFN response through its deubiquitination activity (30). In comparison with nsp2, PRRSV-2 nsp2TF and nsp2N have a stronger inhibitory effect on host innate immune responses (45). Therefore, we suspect that SHFV nsp2TF and nsp2N may also play important roles in SHFV infection and virus-host interaction.
Three variants of the G_GUU_UUU slippery sequence were identified in simarteriviruses, namely, U_GUU_UUU (DeMAV, KRTGV-1, and PBJV), G_GUC_UCU (KKCBV, KRCV-1, KRCV-2, and MYBV-1), and U_UUC_UCU (FSVV, SHEV, and ZMbV-1). In the case of PRRSV, we proposed that tandem slippage of ribosome-bound tRNAs on G_GUU_UUU allows complete A site re-pairing in both Ϫ1 and Ϫ2 frames (tRNA anticodon-mRNA codon pairing in 0 frame is 3=-AAG-5=:5=-UUU-3=; single tRNA Phe isoacceptor AAG). In contrast, however, tandem slippage of ribosome-bound tRNAs on G_GUC_UCU and U_UUC_UCU does not allow A site re-pairing in the Ϫ1 frame. Consistently, our results showed that the SHFV SS1 and SS2 mutants that mimic these slippery sequence variants only permit Ϫ2 PRF. Thus, those simarteriviruses carrying the slippery sequence of G_GUC_UCU or U_UUC_UCU are unlikely to be able to express nsp2N. As indicated above, the SHFV nsp2N is predicted to be an innate immune antagonist. It will be interesting to compare the pathogenicity of this group of simarteriviruses (lack of nsp2N expression) with that of SHFV (expression of nsp2N).
PRRSV nsp1␤ transactivates ribosomal frameshifting through a highly conserved ␣-helix motif, in which a universally conserved arginine (Arg128 in PRRSV-2 and Arg134 in PRRSV-1) is a key residue for nsp1␤ activity (34,38). Comparative sequence analysis of nsp1␤ of 19 arteriviruses shows that the highly conserved ␣-helix motif is present in all of them except EAV and WPDV. In our study, we observed that SHFV nsp1␤ stimulates Ϫ2/Ϫ1 PRF to express nsp2TF and nsp2N, and an alanine substitution at Arg114 completely impairs this process. Consistently, nsp1␤ of KRCV-1, a newly isolated simarterivirus, is also capable of stimulating Ϫ2 PRF. Based on sequence alignment, Arg114 in SHFV nsp1␤ is the residue corresponding to Arg128 in PRRSV-2 and Arg134 in PRRSV-1, and this arginine is universally conserved among all arteriviruses except EAV and WPDV. Besides this arginine, Tyr111 in SHFV is also highly conserved among all arteriviruses except EAV and WPDV. Of note, alanine substitutions introduced at this residue in SHFV and at analogous residues in PRRSV-1 (Tyr131) and PRRSV-2 (Tyr125) also impair nsp1␤'s ability to stimulate Ϫ2/Ϫ1 PRF of both PRRSV-1 and PRRSV-2. Remarkably, nsp1␤s from PRRSV-1, PRRSV-2, KRCV-1, and LDV are capable of stimulating ribosomal frameshifting on the SHFV Ϫ2/Ϫ1 PRF signal. The amino acid sequence identities of nsp1␤s among these arteriviruses range from 21.4% to 57.7%. However, in silico structure prediction shows that, with the exception of EAV nsp1 and WPDV nsp1␤, all arteriviral nsp1␤s share a similar 3-D structure, and the conserved ␣-helix motif is also found in these predicted structures. These data suggest that the nsp1␤ protein structure, especially the ␣-helix region, is essential for its PRF transactivation function.
Cellular PCBPs were initially identified as interacting partners of PRRSV nsp1␤, nsp9, and the genomic 5= untranslated region (5=-UTR) (48). Our recent study explored the function of PCBPs in enhancing Ϫ2/Ϫ1 PRF, together with viral protein nsp1␤ (35). We also observed that the enhancement of Ϫ2/Ϫ1 PRF in SHFV was dependent on the addition of both viral nsp1␤ and PCBP2 in WGE (Fig. 7A). Interestingly, the frameshifting efficiencies for Ϫ1 PRF in WGE were much higher than those observed using rabbit reticulocyte lysate (RRL). In our previous study on the PRRSV frameshift signal (35), we found efficient Ϫ2 PRF in RRL upon the addition of PRRSV nsp1␤ only. Additional supplementation with PCBP2 maintained Ϫ2 PRF but also led to an increase in Ϫ1 PRF. We reasoned that RRL may already contain endogenous PCBPs, with a balance of paralogs that favored Ϫ2 PRF over Ϫ1 PRF. Turning to WGE translations, we found that supplementation with PCBP2 led preferentially to Ϫ1 PRF, whereas supplementation with PCPB1 led preferentially to Ϫ2 PRF. This finding suggests that PCBP1 is the abundant form in RRL. Knockdown of either PCBP1 or PCBP2 in mammalian cells using small interfering RNAs (siRNAs) produced consistent results. In the current SHFV study, supplementation with PCBP2 could particularly promote Ϫ1 PRF in WGE due to the absence of competing endogenous mammalian PCBP1 (compare Fig. 7B with Fig. 4D). As we expected, the interaction between PCBP1/2 and nsp1␤ was detected by coim-munoprecipitation in HEK-293T cells ( Fig. 8A and C). In SHFV-infected MARC-145 cells, SHFV nsp1␤ was also coimmunoprecipitated with endogenous PCBP2 (Fig. 8B). These data suggest that PCBPs are required for efficient Ϫ2/Ϫ1 PRF in arteriviruses, and the interaction between nsp1␤ and PCBPs may be also required for the mechanism.
In conclusion, our study demonstrates that Ϫ2/Ϫ1 PRF is an evolutionarily conserved mechanism used by distantly related arteriviruses for the expression of additional viral proteins, and the PRF products are important for viral replication. This mechanism is unprecedented in eukaryotic systems, not only with the efficient shift to the Ϫ2 frame, but also with the involvement of a transactivating viral protein factor (nsp1␤) and host cellular protein (PCBPs). Given the crucial role of ribosome function in all living systems, the potential impact of the in-depth characterization of this novel mechanism reaches beyond the field of virology. Plasmids. Based on a previous study (21), SHFV nsp1␤ is predicted to contain amino acids (aa) 166 to 350 of ORF1a, while nsp2 contains aa 486 to 1236 of ORF1a. In this study, the coding sequences of nsp1␤ and nsp2 were inserted into the pcDNA 3.1 (ϩ) vector (Thermo Fisher Scientific, Waltham, MA) under the control of the cytomegalovirus (CMV) promoter and were also N-terminally tagged with the FLAG (DYKDDDDK) epitope or hemagglutinin (HA) epitope, designated pFLAG-SHFV-nsp1␤ and pHA-SHFV-nsp2, respectively. Mutations in nsp1␤ were generated using the QuikChange site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA) according to the manufacturer's instructions. Wild-type (WT) nsp1␤ and mutants thereof were cloned into the pGEX-6p-2 vector (GE Healthcare, Chicago, IL) for expression as glutathione S-transferase (GST)-tagged proteins. The coding regions of nsp1␤ and the PLP2 domain were cloned into the pET28a(ϩ) vector (MilliporeSigma, Burlington, MA). Plasmids expressing nsp1␤ and nsp2 proteins of PRRSV were described elsewhere (34). The predicted nsp1␤ gene sequence of KRCV-1 (GenBank accession no. HQ845737.1) was cloned into the p3xFLAG-Myc-CMV-24 expression vector (Sigma-Aldrich, St. Louis, MO), whereas the predicted nsp2 gene sequence of KRCV-1 (GenBank accession no. HQ845737.1) tagged with an HA epitope at the N terminus was cloned into the pcDNA 3.1 (ϩ) vector (Thermo Fisher Scientific, Waltham, MA). The LDV Plagemann strain (LDV-P) nsp1␤ gene was codon optimized for expression in human cells and cloned into the p3xFLAG-Myc-CMV-24 expression vector (Sigma-Aldrich, St. Louis, MO). The plasmid was designated pFLAG-LDV-nsp1␤. The plasmid expressing nsp1 of the EAV Bucyrus strain was described previously (50). A dual luciferase reporter plasmid, pDluc (51), was used for evaluation of in vitro programmed ribosomal frameshifting (PRF) efficiencies. The pDluc-SHFV/WT plasmid and mutants thereof were generated by inserting the SHFV Ϫ2/Ϫ1 PRF signal into pDluc between the Renilla and firefly luciferase reporter genes using the method we described previously (34). The pDluc-PRRSV/WT plasmid containing Ϫ2/Ϫ1 PRF signals of PRRSV-1 was described in our previous study (34). Human PCBP2 (NM_005016.5) cloned in pcDNA 3.1 (ϩ) was kindly provided by Asit K. Pattnaik, University of Nebraska-Lincoln, Lincoln, NE. For expression of the hexahistidine (His6)-tagged recombinant protein, the PCBP2 gene was cloned into the pET28a(ϩ) vector (MilliporeSigma, Burlington, MA) with a His6 tag on the N terminus.
In vitro characterization of recombinant viruses. Recombinant viruses were characterized by determining the viral growth kinetics. For multistep growth curves, MARC-145 cells were seeded in 24-well plates. When the cells reached 100% confluence, they were infected with parental virus or mutants thereof at an MOI of 0.01. Cell culture supernatants were collected at 12, 24, 36, 48, 60, and 72 hpi. Virus titer was measured by titration on MARC-145 cells and calculated as 50% tissue culture infective dose (TCID 50 )/ml according to the Reed and Muench method (55). To determine the plaque morphology of the parental virus and corresponding mutants, a plaque assay was performed in MARC-145 cells using the method described previously (33).