Semliki Forest Virus Chimeras with Functional Replicase Modules from Related Alphaviruses Survive by Adaptive Mutations in Functionally Important Hot Spots

Alphaviruses cause debilitating symptoms and have caused massive outbreaks. There are currently no approved antivirals or vaccines for treating these infections. ABSTRACT Alphaviruses (family Togaviridae) include both human pathogens such as chikungunya virus (CHIKV) and Sindbis virus (SINV) and model viruses such as Semliki Forest virus (SFV). The alphavirus positive-strand RNA genome is translated into nonstructural (ns) polyprotein(s) that are precursors for four nonstructural proteins (nsPs). The three-dimensional structures of nsP2 and the N-terminal 2/3 of nsP3 reveal that these proteins consist of several domains. Cleavage of the ns-polyprotein is performed by the strictly regulated protease activity of the nsP2 region. Processing results in the formation of a replicase complex that can be considered a network of functional modules. These modules work cooperatively and should perform the same task for each alphavirus. To investigate functional interactions between replicase components, we generated chimeras using the SFV genome as a backbone. The functional modules corresponding to different parts of nsP2 and nsP3 were swapped with their counterparts from CHIKV and SINV. Although some chimeras were nonfunctional, viruses harboring the CHIKV N-terminal domain of nsP2 or any domain of nsP3 were viable. Viruses harboring the protease part of nsP2, the full-length nsP2 of CHIKV, or the nsP3 macrodomain of SINV required adaptive mutations for functionality. Seven mutations that considerably improved the infectivity of the corresponding chimeric genomes affected functionally important hot spots recurrently highlighted in previous alphavirus studies. These data indicate that alphaviruses utilize a rather limited set of strategies to survive and adapt. Furthermore, functional analysis revealed that the disturbance of processing was the main defect resulting from chimeric alterations within the ns-polyprotein. IMPORTANCE Alphaviruses cause debilitating symptoms and have caused massive outbreaks. There are currently no approved antivirals or vaccines for treating these infections. Understanding the functions of alphavirus replicase proteins (nsPs) provides valuable information for both antiviral drug and vaccine development. The nsPs of all alphaviruses consist of similar functional modules; however, to what extent these are independent in functionality and thus interchangeable among homologous viruses is largely unknown. Homologous domain swapping was used to study the functioning of modules from nsP2 and nsP3 of other alphaviruses in the context of Semliki Forest virus. Most of the introduced substitutions resulted in defects in the processing of replicase precursors that were typically compensated by adaptive mutations that mapped to determinants of polyprotein processing. Understanding the principles of virus survival strategies and identifying hot spot mutations that permit virus adaptation highlight a route to the rapid development of attenuated viruses as potential live vaccine candidates.

A lphaviruses (family Togaviridae) are positive-strand RNA viruses. Many mosquitotransmitted alphaviruses are human pathogens that cause debilitating symptoms. Chikungunya virus (CHIKV) has recently emerged worldwide, resulting in epidemics that require increasing medical attention (1). The need for effective antiviral compounds and vaccines is not currently met and is steadily increasing together with the spread of the virus (2).
The genome of alphaviruses contains two main open reading frames (ORFs) that encode nonstructural (ns) and structural polyproteins (3). The ns-polyprotein P1234 is autocatalytically cleaved into four individual proteins (nsP1 to 4). Processing occurs in a temporally and sequentially controlled manner and is synchronized with other essential viral functions (4,5). nsPs have specific enzymatic activities and, together with host-derived factors, form a membrane-bound replicase complex (RC) that ensures viral genome replication and synthesis of subgenomic (SG) RNA that is translated into structural proteins (6).
Viral RC comprises several functional modules that correspond structurally to one or several protein domains. To date, the three-dimensional structures of three alphavirus nsPs have been published. nsP1 serves as a membrane anchor of RC. Twelve copies of nsP1 form a crown-shaped ring with a central 7-to 7.5-nm pore (7,8). nsP2 consists of two functional regions, the N-terminal RNA helicase region (nsP2h) and the C-terminal protease region (nsP2p), which function in a coordinated manner (9,10). nsP2h (for CHIKV, residues 1 to 464) consists of a uniquely folded N-terminal domain (NTD) followed by a superfamily 1 RNA helicase fold including two Rec-A-like domains, designated RecA1 and RecA2, and an extended a-helical connector domain (11). nsP2p (for CHIKV, residues 471 to 798) consists of papain-like protease and methyltransferase-like (MTL) domains (12,13). The domains are connected via a flexible linker that has an important role in viral RNA synthesis (14). The N-terminal macrodomain of nsP3 has ADP-ribosyl-binding and hydrolase activities that are essential for virus infectivity and virulence (15)(16)(17)(18). The central portion of nsP3 is referred to as the alphavirus unique domain (AUD); it binds zinc ions and RNA (19) and has multiple roles in RNA replication (20). The C-terminal hypervariable domain (HVD) of nsP3 is intrinsically disordered (21). The HVD is phosphorylated (22,23), is involved in reprogramming of host cell metabolic activity (24), and serves as a binding platform for host proteins that are important for alphavirus replication (25)(26)(27)(28)(29)(30).
In contrast to relatively well-characterized specific activities, the functional interactions of different domains of nsP2 and nsP3 with one another and with other viral proteins have not been sufficiently elucidated. Such interactions have traditionally been studied using point mutations, deletions, or insertions (4,31,32). This approach, however, is time-and labor-intensive because it relies on the meticulous introduction of specific mutations that have effects that are large enough to allow investigators to draw meaningful conclusions. Scanning mutagenesis has been used to map functionally important regions of nsPs (33). As a more general genetic shortcut, the homologous domain swapping approach, which can be viewed as the simultaneous introduction of multiple substitution mutations, has been used to map some of the determinants that are responsible for the specific phenotypes of alphavirus strains (34,35). Additionally, the exchange of regions that encode homologous proteins or their domains has been used to study alphavirus RNA replication and virus-host interactions (36,37; see companion article [38]), as well as in the analysis of determinants of the host range restrictions of an insect-specific alphavirus (39).
In this study, we used swapping of homologous domains to examine the functional interactions of different parts of nsP2 and nsP3. It was found that Semliki Forest virus (SFV) genomes harboring nsP2 or nsP2p of Sindbis virus (SINV) were not infectious, whereas those harboring nsP2 NTD or the nsP3 macro region of CHIKV had wild-type (WT)-like infectivity. Swapped genomes harboring nsP2 or nsP2p derived from the CHIKV or nsP3 macrodomain from SINV had low infectivity, but the corresponding viruses replicated to high titers after accumulating adaptive changes. In total, seven different adaptive substitutions that considerably improved the infectivity of the swapped genomes were identified. Interestingly, only three of these mutations also resulted in clearly increased RNA synthesis in the trans-replicase assay, and one acted by increasing the activity of the SG promoter. Our data highlight the importance of regions involved in the formation of alphavirus RC for virus infection, provide valuable information for understanding the interactions of nsPs, and describe an attractive approach that can be used in virus attenuation.
mutations that have the largest impact on the viability of rescued viruses, SFV CP2 , SFV CP , SFV SN , and SFV SM were passaged five times. Four plaque-purified isolates were obtained for each of these in cellulo-evolved viruses, their ns-regions were sequenced, and the nonsynonymous substitutions present in these regions were identified ( Fig.  2A; Table 1). The numbers indicate the amino acid residues flanking the swapped domains; sequences from CHIKV are shown in gray, and sequences from SINV are shown in black. The number of amino acid substitutions introduced by each of the swaps is shown, and the percent amino acid identity of the proteins encoded by the swapped regions is shown in parentheses. ICA titers are presented as PFU per 1 mg RNA and represent the averages of the data obtained in four independent experiments; ,2 indicates values below the detection limit. Plaque sizes are indicated as follows: large, .3 mm; small, 1 to 2 mm; very small, ,1 mm in diameter. P 0 titers are presented as PFU per ml and represent the averages of the titers measured in three independent experiments. Please note that P 0 stocks were harvested upon detection of CPE, i.e., at different times for different viruses. (B) Multistep growth curves of WT SFV, SFV CN , SFV CM , and SFV CA . BHK-21 cells were infected with P 0 stocks of the indicated viruses at an MOI of 0.1. The culture medium was sampled immediately (0 h) and at 3 h, 6 h, 9 h, 12 h, 24 h, and 48 h p.i., and the amount of infectious virus present was measured using a focusforming assay. Each data point represents the mean 6 standard deviation (SD) from three independent experiments. The dotted line indicates the limit of detection. Detected only in some replicas; in some, it was not possible to quantify, as P123 is not detectable.
c No increase in the nsP2/P123 ratio over time; possibly no cleavage at all.
All of the isolates of SFV SM contained only a V515E substitution in nsP2. Residue 515 corresponds to the S4 subsite residue of nsP2p, and variation at this site is linked to the efficiency of processing of the cleavage site between nsP1 and nsP2 (1/2 site) (4, 45). All four isolates of SFV SN contained a V73A substitution in the NTD region, and three of these isolates also contained a K415M substitution in the HVD of nsP3 ( Fig. 2A). Two isolates of SFV CP contained V2F and G460S substitutions in nsP2 and an R611K substitution in nsP4; these mutations also appeared in various combinations in the remaining SFV CP isolates ( Fig. 2A). Interestingly, the guanine-to-adenine substitution (nucleotide [nt] G7371A) responsible for R611K occurs at position 21 of the SFV SG promoter and was also present in all isolates of SFV CP2 . In addition, one isolate of SFV CP2 also contained the C483Y substitution in nsP4, a substitution that was previously shown to increase the fidelity of CHIKV replication (46,47). Two isolates of SFV CP2 contained identical sets of three mutations (E228K and A796V in nsP2 and R611K in nsP4) and differed in the presence or absence of the C483Y substitution in nsP4 ( Fig. 2A).
The mutations found in isolates of SFV SN decreased the infectivity of the recombinant genomes but did not prevent rescue and replication of infectious virus ( Fig. 2B; Table 1). In contrast, E228K, A796V, C483Y, and R611K all markedly increased the infectivity of SFV CP2 (Fig. 2B; Table 1). A similar observation was made for all three substitutions identified in isolates of SFV CP and for the V515E substitution found in isolates of SFV SM (Fig. 2B; Table 1). Because mutations introduced into the ns-region can have different effects in vertebrate and mosquito cells (20,48), Aedes albopictus C6/36 cells were infected with the obtained P 0 stocks. Most of the rescued viruses replicated well in mosquito cells. The exceptions were SFV CP12V2F and SFV CP12G460S , which had titers .50-fold lower in mosquito cells than in BHK-21 cells. Because this feature was not observed for SFV CP14R611K (Fig. 2B; Table 1), the defective replication of these viruses in mosquito cells (and/of at lower 28°C temperature) must originate from the V2F and G460S substitutions or from a combination of these substitutions with swapping of the nsP2p region.
In cellulo-evolved viruses and viruses harboring adaptive mutations are attenuated. The multistep growth curves of SFV SM12V515E and SFV SM , obtained as a result of in cellulo evolution, were similar. Only a slight delay between 6 and 12 h p.i. was observed for SFV SM ; thereafter, the titers of the two viruses were identical (Fig. 3A). This finding is consistent with the idea that V515E is the primary adaptive mutation found in the SFV SM stock ( Fig. 2A). The growth dynamics of these viruses were similar to that of WT SFV; their titers were, however, consistently ;10-fold lower, with the largest difference observed at 12 h p.i. (Fig. 3A). For SFV CP2 and SFV CP , the evolved stocks consist mostly of viruses that harbored combinations of the adaptive mutations ( Fig. 2A). Analysis of the growth curve of evolved stock of SFV CP2 revealed that it grew more slowly than WT SFV; the difference was more prominent at early time points. The growth of SFV CP212A796V was slower than that of SFV CP2 (Fig. 3A), indicating that the adaptation that restored the infectivity of the RNA genome ( Fig. 2B) did not fully restore replication of the virus. Compared to WT SFV, SFV CP grew more slowly and reached lower final titers (Fig. 3B). The growth rates of SFV CP12V2F and of SFV CP12G460S were almost identical to that of SFV CP (Fig. 3B). SFV CP14R611K , on the other hand, grew faster than SFV CP and reached titers that were only ;60-fold lower than those of WT SFV. Taken together, these findings indicate that the growth of all three derivatives of SFV CP that harbored single adaptive mutations was rather similar to that of SFV CP and suggest that the combinations of adaptive mutations found in the evolved stock ( Fig. 2A) offered limited growth advantages.
The 2/3 sites of SFV CP212E228K , SFV CP214C483Y , and SFV CP214R611K accumulate additional mutations. Evolved SFV CP2 grew faster and had higher titers than SFV CP212A796V (Fig. 3A), indicating an additive effect of combinations of different mutations ( Fig. 2A). It was therefore hypothesized that additional mutations with similar effects may also accumulate during rescue and passage of SFV CP212E228K , SFV CP212A796V , SFV CP214C483Y , and SFV CP214R611K . Sequence analysis of the region encoding the C-terminal fragment of nsP1, nsP2, and the N-terminal part of nsP3 revealed no additional mutations in the SFV CP212A796V stock. In contrast, genomes containing additional substitutions in the region corresponding to the 2/3 site of ns-polyprotein were found in the SFV CP212E228K , SFV CP214C483Y , and SFV CP214R611K stocks ( Table 2), indicating that there was selection pressure for a decreased rate of processing at the 2/3 site. Unexpectedly, however, these changes did not provide any growth advantage; SFV CP212E228K12C798F and SFV CP214C483Y13A1P were both strongly attenuated; no plaques were detected in ICA, and only mild cytopathic effects (CPE) were observed (Table 1). This finding may indicate that the viruses in which the C798F and A1P mutations were found also contained additional mutation(s) in regions that were not analyzed in this experiment.
Swapping of nsP2, its regions, and the macrodomain of nsP3 results in defects in P1234 processing. The known functions of swapped domains/proteins and the localization of the majority of adaptive mutations ( Fig. 2A; Tables 1 and 2) strongly suggest that they may impact P1234 processing. Due to the lethal phenotype of SFV SP2 and SFV SP , the adaptations of SFV CP2 , SFV CP , SFV SN , and SFV SM processing were studied using in vitro-translated polyproteins.
In WT SFV, the P123 and P12/P23/P34 polyproteins, as well as mature nsPs, were observed at the 09 time point. During the 30-min chase, the amounts of P12/P23/P34 drastically decreased, and P123 became undetectable (Fig. 4). Processing of the 1/2 site of SFV CP2 was faster than that of WT SFV, most likely being the reason for the low infectivity of this construct and the need for adaptive mutations (4). Interestingly, the polyproteins with the mobilities of P12/P23 and P34 were more stable than their counterparts in WT SFV4 (Fig. 4A). As expected, A796V substitution reduced the speed of P123 processing. The E228K substitution, located in a RecA1 domain that is not known to affect the protease activity of nsP2, resulted in a similar, albeit less pronounced, effect. As expected, substitutions in nsP4 had no effect on P1234 processing (Fig. 4A). Adding secondary mutations affecting the 2/3 site to the ns-polyproteins of SFV CP212E228K or SFV CP214C483Y severely suppressed ns-polyprotein processing; P1234 became detectable, P123 was stabilized, and mature nsP2 could not be detected (Fig.  4A). Lack of release of mature nsP2, a major factor of cytotoxicity of Old World alphaviruses (49,50), is therefore the most likely reason for the lack of CPE in cells infected with the corresponding viruses. The C798F substitution also resulted in the accumulation of unidentified processing/degradation products, while the A1P substitution resulted in the accumulation of a smaller polyprotein, most likely P23 (Fig. 4A).
SFV CM was found to have slightly reduced processing of P123 and somewhat increased stability of the processing intermediate (presumably P23). These effects are consistent with the previously described role of the nsP3 macrodomain in 2/3 site processing (31). We were unable to detect P123 in SFV CN (Fig. 4B). The acceleration of 1/2 site processing, which was probably not prominent enough to inhibit virus rescue ( Fig. 1A; Table 1), may explain the attenuated growth of SFV CN (Fig. 1B). In contrast to SFV CP2 , the ns-polyprotein of SFV CP was processed very inefficiently; P123 was stabilized, other processing intermediates were present in reduced amounts, mature nsPs were found only in trace amounts, and no significant increase in the nsP2 to P123 ratio was observed during the chase (Fig. 4B). Thus, chimeric nsP2 ( Fig. 1A) cannot efficiently cleave the 1/2 site of SFV, and this in turn also inhibits processing of the 2/3 site (5). The V2F and G460S substitutions did not result in detectable changes (Fig. 4B), indicating that they increased SFV CP infectivity (Fig. 2B) by other mechanism(s) and/or that the effect of these mutations on P1234 processing is more prominent in infected cells.
The processing of P1234 of SFV SN was similar to that of WT SFV; the observed small reduction in processing efficiency is not likely to be the cause of the reduced infectivity of SFV SN genomes. Neither of the mutations found in the evolved SFV SN stock had a clear effect on processing (Fig. 4C). The processing of the ns-polyprotein of SFV SM was unique; the nsP2/P123 ratio was similar to that found in WT SFV, but the amount of P12/P23 was drastically reduced (Fig. 4C). As P23 is essential for SFV RC formation (51), the defect in this mutant probably originates from accelerated cleavage of the 2/3 site, an event in which the nsP3 macrodomain has an important role (31). Consistently, adding the V515E substitution to SFV SM not only resulted in a slowdown of P123 processing but also stabilized P12/P23 processing intermediates (Fig. 4C). Finally, swaps resulting in nonviable SFV SP2 and SFV SP (Fig. 1B) were found to inhibit the formation of individual nsPs. In contrast, P1234 was prominent in these mutants, indicating that nsP2 of SINV and a chimeric nsP2 harboring nsP2p of SINV were unable to process the 3/4 site of SFV ns-polyprotein. This property may also be the cause of the lethal phenotype, as cleavage of the 3/4 site is absolutely required for SFV infectivity (40).
Mutations introduced into the 1/2 cleavage site affect the rescue of SFV CP and SFV CP2 by changing ns-polyprotein processing. To confirm the link between ns-polyprotein processing and rescue of chimeric viruses, two types of mutations were introduced into the P-side of the 1/2 site in SFV CP and SFV CP2 (Fig. 5A and B; Table 1). In SFV CP11G536V and SFV CP211G536V , the P2 Gly residue was substituted by Ala; this mutation abolishes cleavage of the corresponding site (52). In SFV CP11Y533D11H534R and SFV CP211Y533D11H534R , the P5 Tyr and P4 His residues of the SFV 1/2 site were substituted by Asp and Arg, which are found in the 1/2 site of CHIKV. These two changes reintroduce a match between the sequence of the P-side of the 1/2 site and the nsP2p domain (Fig. 5B); they were also expected to accelerate the processing of the 1/2 site due to the introduction of a favorable P4 Arg residue (4,53).
ICA revealed no plaque formation by SFV CP11G536V or SFV CP211G536V . Furthermore, in contrast to SFV CP and SFV CP2 , no infectious progeny was obtained ( Fig. 5B; Table 1). An in vitro translation and processing assay revealed the presence of minor quantities of a protein with an electrophoretic mobility corresponding to that of nsP2. However, the  Amino acid no.  793  794  795  796  797  798  1  2  3  4  Cleavage site residue  P6  P5  P4  P3  P2  P1  P19  P29  P39 P49 a Mutated residues found in novel SFV CP2 variants are shown in boldface and italics.

FIG 4
Swapping of the nsP2, nsP2 NTD, nsP2p, or nsP3 macrodomains often results in altered nspolyprotein processing. In vitro translation was conducted using the TNT SP6 rabbit reticulocyte system, and plasmids containing icDNAs of the indicated viruses were used as templates. After 45 min, the reaction was stopped with cycloheximide, and samples were collected immediately (09) or after incubation for an additional 30 min (309). The produced ns-polyproteins and their processing products were separated using SDS-PAGE, and the densities of bands corresponding to nsP2 and P123 were quantified using ImageQuant software. (A) Translation of pSFV CP2 and its variants containing adaptive mutations. (B) Translation of other icDNAs harboring CHIKV-derived sequences. (C) Translation of icDNAs harboring SINV-derived sequences. The positions of P1234, P123, and P12/P23/P34 and those of these mature nsPs are shown. The ratio shows the amount of label that was incorporated into mature nsP2 relative to the amount that was incorporated into unprocessed P123; the nsP2/P123 ratio of WT SFV at 09 was taken as 100. Each number represents the mean 6 SD of the values obtained in three independent experiments. * indicates complete processing of P123 (corresponding density at background level). An additional band visible in samples of SFV CP214C483Y most likely represents Cterminally truncated P34. It was detected using only one batch of the TNT SP6 rabbit reticulocyte system and probably represents an artifact generated by premature termination of transcription in the region corresponding to the C483Y substitution. amount of this product did not increase during the chase, and the increase in the nsP2/P123 ratio was due to a decrease in the amount of P123 present, presumably due to slow processing of the 2/3 site (Fig. 5C). These data indicate that processing at the 1/2 site is essential for replication of the virus and that the viability of SFV CP2 , which is characterized by accelerated processing of P123 ( Fig. 4A; 5C), cannot be restored by blocking the cleavage of the 1/2 site. Substitution of the P5 and P4 residues increased the rescue of the corresponding viruses. For SFV CP11Y533D11H534R , the number of plaques in ICA was greatly increased, and the rescued virus grew to final titers ;3-fold higher than those of SFV CP (Fig. 5B;  Fig. 4. Each number represents the mean 6 SD from three independent experiments. * indicates complete processing of P123 (corresponding density at background level); the numbers in parentheses indicate a lack of increase in the amount of product with the same mobility as that of nsP2 during the chase. Table 1). An in vitro translation and processing assay revealed that, in contrast to selected adaptive mutations, the substitution of P5 and P4 residues accelerated ns-polyprotein processing, which became similar to that in WT SFV (Fig. 5C). These data show that the defect caused by swapping of the nsP2p region can, like other defects resulting from swapping, be rescued by changes in ns-polyprotein processing, confirming processing as one of the key factors in the viability of viruses with swapped genomes. Interestingly, mutations at the P5 and P4 positions resulted in detectable rescue of SFV CP2 , although that mutant suffers from accelerated processing of P123 (Fig. 4A and  5C). The effect on the rescue was, however, very minor, as SFV CP211Y533D11H534R formed extremely small plaques that were observed only in some of the independent experiments. Furthermore, the rescued virus grew to titers that were ;50-fold lower than those of SFV CP2 (Fig. 5B), indicating that further acceleration of 1/2 site processing was unfavorable for the virus. It is therefore plausible that the observed rescue in ICA was not due to the change in 1/2 site processing speed per se; it is more likely that the match between the P-side of the 1/2 site and the nsP2p region allowed better coordination of events leading to the formation of functional replicases that compensated to some degree for the unfavorable acceleration of ns-polyprotein processing.
Swaps and adaptive mutations have an impact on viral RNA synthesis. Mutations in the ns-region of the alphavirus genome can alter virus infection by affecting the activity of viral RNA replicase. Hence, a highly efficient SFV trans-replication assay (53) was used to evaluate the effects of swaps and adaptive mutations. In this system, nspolyprotein is expressed from CMV-P1234-SFV, and template RNA is expressed from HSPolI-FG-SFV plasmids (Fig. 6A). Increases in firefly luciferase (Fluc) and Gaussia luciferase (Gluc) activities in the presence of active replicase serve as proxies for the synthesis of full-length "genomic" RNAs (here referred to as "replication") and SG RNAs (here referred to as "transcription"), respectively (54).
The replicases of SFV CM and SFV CA had activities similar to those of WT SFV and SFV CH containing HVD of CHIKV nsP3 (37). Thus, the replacement of any domain of SFV nsP3 with its CHIKV counterpart had no significant negative effect (P . 0.1 in all cases) on viral RNA synthesis ( Fig. 6B and C; Table 1). The replicase of SFV CN had significantly reduced replication activity (P , 0.01) and a less pronounced but still significant (P , 0.05) reduction in SG RNA synthesis. Consistent with the lethal phenotype (Fig. 1B) and lack of/strong reduction in 3/4 site processing (Fig. 4C), the activities of SFV SP2 and SFV SP replicases were close to the background level ( Fig. 6B and C; Table 1). Taken together, the results of the trans-replication assay accurately reflected the properties of the corresponding viruses and their genomes.
The replication and transcription activities of the SFV SM replicase were notably low, but introduction of the V515M substitution into nsP2 significantly increased both of these activities (P , 0.001 and P , 0.05, respectively) ( Fig. 6B and C; Table 1). Thus, adaptation of SFV SM does occur through an increase in the ability of its replicase to synthesize viral RNAs. The ability of SFV SN replicase to replicate template RNAs was similar to that of SFV CN replicase. At the same time, its capacity to synthesize SG RNAs was ∼6-fold lower, and the difference was statistically significant (P , 0.01) (Fig. 6C). A73V substitution in nsP2 or K415M substitution in nsP3 failed to increase the activities of SFV SN replicase significantly, confirming that these changes are not associated with virus adaptation.
The ability of SFV CP replicase to replicate template RNA was similar to that of WT SFV replicase, and none of the adaptive mutations increased it significantly any further (Fig. 6B). In contrast, the ability of this replicase to produce SG RNAs was significantly reduced (P , 0.05). It was slightly increased by the V2F and G460S substitutions ( Fig.  6C; Table 1); however, the differences associated with these mutations were not statistically significant. The most likely reason for these effects is highly inefficient P123 processing (Fig. 4B). Both the replication and the transcription activities of the SFV CP2 replicase were significantly lower (P , 0.01 and P , 0.05, respectively) than those of the WT SFV replicase (Fig. 6B and C). A796V substitution increased both of these activities ∼10-fold. E228K substitution had a smaller activating effect on both replication and SG RNA synthesis ( Fig. 6B and C; Table 1), but neither effect was statistically significant. Despite the lack of statistical significance, the observed trend suggests that both of these mutations activate RNA synthesis, likely by reducing the speed of P123 processing (Fig. 4A). C483Y substitution in nsP4 also resulted in an increase in the transcriptional activity of the SFV CP2 replicase, although the effect was minor (Fig. 6C) and statistically not significant. In contrast, R611K substitution in nsP4 failed to increase any of the activities of the SFV CP2 replicase (Fig. 6B and C; Table 1), indicating that the adaptation of the viruses with this substitution was not due to changes in their replicase activities.
Mutation resulting in the R611K substitution is cis-active and results in increased activity of the SFV SG promoter. R611K was the only adaptive mutation found in two different swapped viruses ( Fig. 2A). It was also the only mutation that rescued the infectivity of the swapped genomes but did not affect ns-polyprotein processing and/ or increase the activity of RNA replicase (Fig. 4 and 6). In viruses with the R611K substitution, the nucleotide located at position 21 of the SFV SG promoter is changed from guanine to adenine. A corresponding variation has been found in natural SFV isolates (45), and it was demonstrated that in the context of recombinant SINV, an inserted SFV SG promoter with adenine at position 21 was ;7-fold more efficient than its counterpart with guanine at that position (55). As nsP2 is a protein that recognizes the SG promoter (56), it was logical to assume that the 21 guanine-to-adenine substitution in the SG promoter found in viruses with swapped nsP2 or nsP2p region represents an adaptation that increases the activity of the SFV SG promoter.
To test this hypothesis, position 21 of the SFV SG promoter in the template construct was changed from guanine to adenine (Fig. 7A). The abilities of WT SFV replicase and of the replicases of SFV CP2 , SFV CP214C483Y , SFV CP214R611K , SFV CP , SFV CP12G460S , and SFV CP14R611K to transcribe templates encoded by HSPolI-FG-SFV and HSPolI-FG-SFV-SG 21G/A were compared at 16, 20, and 24 h posttransfection (p.t.). It was observed that the template with the 21 adenine residue always outperformed its counterpart with the 21 guanine residue; at 20 h p.t., these differences were 3-to 7-fold and were statistically significant (Fig. 7B). These results confirm that the R611K mutation or, from the functional point of view, the 21 guanine-to-adenine substitution is a cis-active mutation that increases the activity of the SG promoter of SFV. For this reason, here, the R611K mutation is referred to as the SG(-1)G/A mutation, SFV CP2-4R611K is designated SFV CP2-SG(-1)G/A , and SFV CP2-4R611K is designated SFV CP-SG(-1)G/A . Unlike SFV4, which was used as the backbone in the swapped genomes, CHIKV has a 21 adenine residue in its SG promoter. Therefore, it appeared plausible that the 21 guanine-to-adenine substitution may also act by increasing the affinity of the nsP2 region of CHIKV for the SG promoter in the template RNA. Consistent with this hypothesis, the smallest increase in Gluc activity (from 2-fold at 16 h p.t. to 3-fold at 20 h p.t.) was observed for WT SFV replicase, and the largest increases (from 3-fold at 16 h p.t. to 12-fold at 24 h p.t.) were observed for SFV CP , SFV CP2 , and their variants (Fig. 7B). Thus, the transcription activities of replicases harboring nsP2 or nsP2p of CHIKV indeed benefited more from the SG(-1)G/A mutation than did the transcription activity of the WT SFV replicase. However, although this may indeed reflect improvement in the nsP2/SG promoter interaction, the observed differences were relatively small and could also have resulted from other factors such as large differences in the baseline values of Gluc activation (expression from HSPolI-FG-SFV).
Swapping of nsP2p of SFV with nsP2p of CHIKV affects SG promoter activity in SFV replicon vectors and inhibits the formation of virus replicon particles. Swapping of nsP2p of SFV with that of CHIKV had minimal impact on the ability of trans-replicase to replicate the RNA template (Fig. 6B) but severely reduced its ability to transcribe the template (Fig. 6C; Table 1). The adaptive V2F and G460S substitutions found in nsP2 of SFV CP only caused small and statistically nonsignificant increases in transcription in the trans-replication assay (Fig. 6C; Table 1). At the same time, these mutations markedly increased the rescue efficiency of the swapped viruses (Fig. 2B), indicating that their impact on self-replicating RNA may be different from their effect on trans-replicase. Furthermore, as nsP2 is likely involved in the packaging of alphavirus genomes and is even included in virions (57), it was hypothesized that the V2F and G460S substitutions may increase the efficiency of virion formation.
To analyze these possibilities, advantage was taken of the availability of self-replicating (replicon) vectors and packaging systems of SFV making it possible to obtain virus replicon particles (VRPs). nsP2p swaps without or with V2F and G460S substitutions were introduced into an SFV replicon vector containing a sequence encoding nanoluciferase (NLuc) under the control of the SG promoter (Fig. 8A). Transfection of BHK-21 cells with the corresponding RNA transcripts revealed that the nsP2p swap drastically (;900-fold) reduced NLuc expression. The presence of V2F and G460S substitutions resulted in ;40-fold and ;3-fold increases, respectively, in NLuc expression (Fig. 8B). Thus, SG promoter-mediated reporter expression was affected in the same manner as was seen in the trans-replicase system (compare Fig. 6C and 8B). Furthermore, similar to the expression of Fluc by the trans-replicase, the nsP2p swap and adaptive mutations did not have a prominent effect on the expression of nsP1, which is translated directly from replicon RNA (Fig. 8C).
To analyze whether nsP2p swaps and adaptive mutations affect VRP formation, BHK-21 cells were cotransfected with replicon and SFV helper RNAs (Fig. 8A). The presence of SFV helper RNAs did not significantly alter NLuc or nsP1 expression ( Fig. 8B and C). Infection of naive BHK-21 cells with supernatant harvested from cells cotransfected with SFV helper RNAs and SFV1-NLuc RNA resulted in NLuc activities ;20,000fold greater than background levels, indicating effective formation of VRPs. In contrast, no expression of NLuc above the background level was observed for cells infected using supernatants harvested from cells cotransfected with SFV helper-RNAs and transcripts from pSP6-SFV1 CP -NLuc, pSP6-SFV1 CP12V2F -NLuc, or pSP6-SFV1 CP12G460S -NLuc. The lack of VRP formation was likely due to the inability of the mutant replicons to activate SG promoter-mediated expression of capsid protein (Fig. 8C) and presumably also the expression of envelope proteins. Due to the lack of expression of structural proteins, it remains unknown whether there was also any specific effect of the mutations on VRP formation. To exclude the possibility that the observed inability to activate the SG promoter in helper RNAs was due to a mismatch between the CHIKV nsP2p region of the chimeric replicon and the SG promoter in SFV helper RNAs, the experiment was repeated using CHIKV helper RNAs. Again, the presence of these helper RNAs had no impact on NLuc reporter expression. No expression of NLuc above the background level was detected in BHK-21 cells infected with any of the harvested supernatants. The reason for this was a complete lack of CHIKV capsid protein expression (data not shown), indicating that none of the replicons used, including SFV1-NLuc, were able to replicate and transcribe CHIKV helper RNAs.

DISCUSSION
In this study, swapping of homologous domains and proteins was used to study the poorly understood functional interactions of alphavirus RC components. Complementary experiments, including analysis of the infectivity and adaptation of the obtained viruses and replicons, processing of ns-polyproteins, and the activities of RNA replicases and the SG promoter, enabled us to obtain unique insights into the process of alphavirus replication.
Sequences obtained from CHIKV were better tolerated by SFV than were their counterparts from SINV (Fig. 1A), indicating a better match between domains originating from more closely related viruses. Thus, replacing the entire nsP2 or nsP2p region with CHIKV sequences resulted in chimeric genomes with reduced infectivity (Fig. 1A) but did not prevent virus rescue and subsequent adaptation. In contrast, swapping of the nsP2 or nsP2p region with the corresponding regions of SINV resulted in noninfectious genomes and nonfunctional replicases (Fig. 1A and 6B and C). However, in a parallel study, an opposite trend was observed; swapping nsP4 of SFV with nsP4 of SINV (sequence identity 74%) increased the activity of the corresponding trans-replicase, whereas the use of nsP4 of CHIKV (sequence identity, 79%) reduced its activity, and replicase harboring P123 of SFV and nsP4 of Ross River virus (RRV; sequence identity, 81%) almost entirely lacked the ability to synthesize SG RNA. These phenotypes were also reflected in the properties of the corresponding recombinant viruses (see the companion article [38]). Thus, the similarity/identity of swapped sequences per se is not the main factor that determines the activity of chimeric RNA replicases of alphaviruses. It is likely that the most important factor is functional matching of the swapped regions; thus, inactivity of obtained chimeric replicases or genomes serves as an indication of the disturbance of crucial function(s). Here, for SFV SP and SFV SP2 , the affected function was most likely processing of the 3/4 site (Fig. 4C). We previously observed that in trans, SINV nsP2 can cleave the 2/3 site of SFV but not the 1/2 or 3/4 sites (31). Current data indicate that cleavage of the SFV 3/4 site by SINV protease in cis is also not possible and/or that SINV protease requires the assistance of other SINV protein(s) to cleave at this site.
Chimeric viruses with highly infectious genomes, such as SFV CN , SFV CM , and SFV CA , can replicate without acquiring adaptive mutations. Surprisingly, however, no mutation with beneficial effects was found for any isolate of SFV SN . The P1234 processing of SFV SN was similar to that of WT SFV (Fig. 4C), and its replicase displayed activities higher than those of the replicases of some adapted viruses ( Fig. 6B and C). Thus, compromised RNA synthesis was unlikely to be the reason for the low infectivity of SFV SN RNAs (Fig. 1A). Because evidence suggests that there is a functional connection between the nsP2 NTD and capsid protein (4,44,58), the capsid protein regions of SFV SN isolates were also sequenced. No mutation was found, leading to the hypothesis that the effect of defects in SFV SN is probably limited to reduced rescue efficiency.
Adaptive changes were found in the genomes of evolved SFV CP2 , SFV CP , and SFV SM (Fig. 2A). Each of these substitutions was found in at least two virus isolates; the extreme case was SFV SM , for which all four isolates contained the V515E substitution ( Fig. 2A). Other substitutions in nsP2 were located close to its termini, in the RecA1 domain (E228K), or in the connector domain (G460S) of nsP2h (11). The functional defect compensated for by the adaptive mutations was evident for the E228K, A796V, and V515E substitutions, all of which improved the RNA synthesis abilities of the corresponding replicases ( Fig. 6B and C). In all three cases, this effect can be connected to the conversion of P1234 processing from abnormally fast to more like that of WT SFV ( Fig. 4A and C). Interestingly, all these mutations appear to act by increasing the stability of both P123 and smaller polyprotein(s), presumably including P23. Indeed, in SFV CP , the 2/3 site has been made more favorable for cleavage by a change in the P4 residue of the site from unfavorable threonine (SFV) to favorable arginine (CHIKV). Single substitutions resulting in similar effects are well tolerated by SFV; however, when coupled with accelerated processing at the 1/2 site, they reduce the infectivity of the virus RNA and trigger the accumulation of adaptive mutations (4). It is likely that the SINV macro domain, which is present in SFV SM , also facilitates the processing of the 2/3 site. Thus, these adaptive mutations have likely been selected for to counteract the accelerated processing at these cleavage sites.
Substitution of nsP2p of SFV with that of CHIKV resulted in diminished synthesis of SG RNAs (Fig. 6B and C) and a major defect in P123 processing (Fig. 4B). Surprisingly, however, the adaptive mutations found in nsP2 of SFV CP had no effect on ns-polyprotein processing (Fig. 4A), and their beneficial effect on the synthesis of SG RNAs was minor and statistically not significant (Fig. 6C). It is possible that the ability of chimeric nsP2 to process the 1/2 site cannot be efficiently increased by a single adaptive substitution. However, when substitutions were introduced into both the P5 and P4 positions of the 1/2 site, the WT-like processing of P123 was restored, and the infectivity of the mutant genome was increased (Fig. 5). It is therefore possible that genomes such as that of SFV CP11Y533D11H534R remained attenuated and were eliminated during in cellulo adaptation. However, the observation that SFV CP11Y533D11H534R was capable of replicating to titers similar to those produced by in cellulo-adapted SFV CP , SFV CP12V2F , SFV CP12G460S , and SFV CP-SG(-1)G/A (compare Fig. 2B and 5B) argues against this possibility.
The mechanism of action of the V2F and G460S substitutions apparently includes activation of SG RNA synthesis. This trend was observed for trans-replicase (Fig. 6C) and was more prominent in the context of self-replicating RNA replicons (Fig. 8B). However, the impact of these mutations was relatively mild, and the observed increase in SG RNA-mediated expression did not reach statistical significance. Therefore, additional mechanism(s), such as more efficient use of a limited supply of structural proteins to increase virion production, remain possible. The V2 residue is located in the region of nsP2 that has been shown to be functionally linked to the capsid protein (4). The G460 residue is located in the connector region, and mutations in this region may affect the interaction of the nsP2 domains with each other or with other proteins. As nsP2 is likely involved in the packaging of the alphavirus genome and is even included in virions (57), it is possible that the V2F and G460S substitutions may increase the efficiency of virion formation. Our attempts to investigate this possibility using SFV replicon vectors and their packaging system were, however, unsuccessful, as none of the mutant replicons proved capable of triggering replication and transcription of the provided helper RNAs that are required for the expression of SFV structural proteins (Fig.  8C). Interestingly, SFV CP12V2F and SFV CP12G460S were the only mutant viruses that were specifically attenuated in mosquito cells (Fig. 2B). The defect is unlikely to be caused by inefficient processing, as increased P123 stability activates alphavirus RNA replication in mosquito cells (48). Therefore, it is possible that the V2F and G460S substitutions, which were selected in mammalian cells and increase the propagation of SFV CP in these cells, may have different effects in mosquito cells.
The C483Y substitution in nsP4 was first discovered upon passaging CHIKV in the presence of ribavirin or 5-fluorouracil; it increases the fidelity of CHIKV replication and reduces virus fitness in vivo (46). The same selection conditions also selected for the G641D substitution in nsP2, which, in combination with the C483Y substitution, resulted in complete resistance to the mutagenic activity of ribavirin. Interestingly, position 641 of nsP2 of SFV is occupied by a glutamate residue, which is similar to the aspartate residue found in the high-fidelity CHIKV mutant. Furthermore, it was found that RCs isolated from cells infected by CHIKV harboring the C483Y substitution displayed elevated SG RNA synthesis (47). C483Y substitution also resulted in a modest increase in SG RNA synthesis by trans-replicase (Fig. 5C), and this may be one of the mechanisms responsible for the adaptive effect of the C483Y substitution. Even if there are additional mechanism(s) through which the C483Y mutation increases the infectivity of SFV CP2 genomes, our data unequivocally confirm the functional connection between the nsP2 and nsP4 regions.
The majority of isolates of SFV CP2 and all isolates of SFV CP harbor combinations of adaptive mutations (Fig. 2A). Each of these substitutions alone restored the infectivity of the corresponding genome (Fig. 2B); furthermore, SG(-1)G/A was sufficient to increase the replication of SFV CP to a level exceeding that of the evolved stock (Fig. 3B). Why, then, were the combinations of mutations needed? Interestingly, most isolates harbored combinations of mutations that facilitate virus replication by different mechanisms; mutations affecting protease activity (E228K and A796V) were combined with cis-active SG(-1)G/A mutations in the SG promoter and/or with mutations that increase the fidelity and activity of RCs (C483Y). Similarly, a cis-active SG(-1)G/A mutation was also found in combination with mutation(s) that affect the ability of replicase to synthesize SG RNAs ( Fig. 2A). Most likely, such combinations confer the greatest advantages for internal and external competition and were therefore selected. A simultaneous rise of several beneficial mutations is improbable. Sequential accumulation of mutations is also unlikely, because the advantages offered by additional mutations are relatively small. It is therefore possible that most of the combinations of adaptive mutations were created by homologous recombination. If this is the case, additional adaptive changes ( Table 2) that had negative effects on individual viruses may represent a source for successive adaptation. On its own, SFV CP212E228K12C798F has considerable growth disadvantages. However, viruses harboring such mutations typically lack superinfection exclusion (59,60). Thus, cells infected by SFV CP212E228K12C798F or similar viruses can be superinfected by other viruses, resulting in mixed infection and providing new possibilities for subsequent adaptation and recombination.
Although the initial aim of this study was to use swapping of functional modules in the SFV replicase to analyze interactions among protein surfaces by discovering compensatory mutations that occur at the interaction hot spots and adjust protein complementarity at the interfaces, we can now conclude that for the studied alphaviruses, such an experimental approach does not lead to this goal. Instead, taken together, the results of this study provide new insight into how alphavirus replicase proteins work together. Our data are consistent with a model in which functional modules in alphavirus replicase function relatively independently and thus are interchangeable. Such an organization of replicases is likely responsible for the robustness of alphavirus genomes. Therefore, even the replacement of the entire multifunctional nsP2, which accounts for ;32% of the length of ns-polyprotein, was tolerated, and only a few amino acid changes were sufficient to adjust proteolytic homeostasis. Interestingly, we repeatedly detected the same adaptive mutations (e.g., nsP2 V515E, nsP4 C483Y) that were discovered previously through completely different experiments with alphaviruses, leading to the conclusion that these residues are true functionally crucial hot spots. We previously obtained a similar result for the P-side residues of the 1/2 site, which are involved in controlling replicase formation, the sensitivity of alphaviruses to G3BP depletion, and replication efficiency in mosquito cells as well as activation of innate immune responses and pathogenicity in mice (4,45,48,53,61,62). Logically, such functional hot spots represent the best starting points for attenuation of alphaviruses (e.g., Mayaro virus) that have not yet been studied at the same level of detail as SFV and CHIKV. In addition, swapping of homologous domains of replicase proteins did result in viruses that remained attenuated, even in the presence of adaptive mutations. Thus, the approach can be used to obtain chimeric attenuated alphaviruses and possibly other positive-strand viruses that represent candidates for live attenuated vaccines.
It has also not escaped our notice that the chimeric virus genomes described in this study and in the companion article (38) may potentially find an application, e.g., in screening for inhibitors that target the replicase components of pathogenic alphaviruses under lower biosafety level (BSL) conditions. For example, instead of using CHIKV to test anti-CHIKV protease inhibitors, an experiment that would require BSL3 conditions, such experiments can be performed using newly developed SFV CP2 or SFV CP variants with viral fitness-improving mutations outside the region of interest [e.g., C483Y in nsP4 or SG(-1)G/A] because the swapped viruses are sufficiently stable and can undergo a full replication cycle; thus, they would faithfully present a target for inhibition and might even allow for the selection of resistance mutations that alter the binding pockets in response to selective pressure from inhibitors. We propose that similar chimeric viruses carrying foreign replication modules can be relatively easily created for other medically important alphaviruses such as RRV, Mayaro virus, and o'nyong'nyong virus, thus avoiding the need to establish a reverse genetics system for each alphavirus. In this case, because the genetic background of SFV is similar to that of these viruses, these chimeras can be used in parallel in screening for potential inhibitors.
Finally, our experiments revealed that alphaviruses use a rather limited set of strategies to survive and adapt; among these, the main strategy is likely downregulation of the efficiency of P1234 processing. In this case, when the corrupted replication process becomes evident, the processing rate decreases, and this likely promotes the production of an excessive number of copies of full-length genomes for prolonged replication instead of the usual switch to transcription of SG RNA culminating in virion production. It is not precisely known whether the fidelity of viral RNA polymerase in the context of