Role reversal of functional identity in host factors: Dissecting features affecting pro-viral versus antiviral functions of cellular DEAD-box helicases in tombusvirus replication

Positive-stranded (+)RNA viruses greatly exploit host cells to support viral replication. However, unlike many other pathogens, (+)RNA viruses code for only a limited number of genes, making them highly dependent on numerous co-opted host factors for supporting viral replication and other viral processes during their infections. This excessive dependence on subverted host factors, however, renders (+)RNA viruses vulnerable to host restriction factors that could block virus replication. Interestingly, cellular ATP-dependent DEAD-box RNA helicases could promote or inhibit the replication of Tomato bushy stunt virus (TBSV) replication. However, it is currently unknown what features make a particular DEAD-box helicase either pro-viral or antiviral. In this work, we succeeded in reversing the viral function of the antiviral DDX17-like RH30 DEAD-box helicase by converting it to a pro-viral helicase. We also turned the pro-viral DDX3-like RH20 helicase into an antiviral helicase through deletion of a unique N-terminal domain. We demonstrate that in the absence of the N-terminal domain, the core helicase domain becomes unhinged, showing altered specificity in unwinding viral RNA duplexes containing cis-acting replication elements. The discovery of the sequence plasticity of DEAD-box helicases that can alter recognition of different cis-acting RNA elements in the viral genome illustrates the evolutionary potential of RNA helicases in the arms race between viruses and their hosts, including key roles of RNA helicases in plant innate immunity. Overall, these findings open up the possibility to turn the pro-viral host factors into antiviral factors, thus increasing the potential antiviral arsenal of the host for the benefit of agriculture and health science.


Introduction
Positive-stranded (+)RNA viruses exploit host cells by co-opting many cellular factors to facilitate viral replication. In addition, many (+)RNA viruses take advantage of metabolic pathways, subcellular membranes and intracellular trafficking to build elaborate membranous viral replication compartments (also called viral replication organelles, VROs), which are the sites of viral replication, in the cytosol of the infected cells [1][2][3][4][5][6][7][8]. The emerging picture is that the coopted host factors affect many steps of RNA virus replication. For example, the assembly of membrane-bound viral replicase complexes (VRCs) is assisted by host translation initiation and elongation factors, protein chaperones, RNA-modifying enzymes, SNARE and ESCRT proteins, the actin network, and lipids [1][2][3]7,9,10].
The mechanistic roles of host factors in viral RNA replication is intensively studied with Tomato bushy stunt virus (TBSV) and other tombusviruses infecting plants. [7,8,11]. Expression of the two TBSV replication proteins, termed p33 and p92 pol , and a replicon (rep)RNA leads to efficient viral replication in yeast (Saccharomyces cerevisiae) surrogate host [9,12,13]. The abundant p33, which overlaps with the N-terminal region of p92 pol , is an RNA chaperone and a master regulator of replication, whereas p92 pol is the RdRp [7]. Several cis-acting replication elements have been defined in the TBSV (+) and (-) RNAs, which have critical functions in RNA template selection, RNA recruitment, in the assembly of VRCs, activation of the p92 RdRp and in viral RNA synthesis [13][14][15][16][17][18][19].
Among the largest family of cellular proteins affecting TBSV replication is the ATP-dependent DEAD-box RNA helicases. Multiple global screens in yeast with TBSV have identified 11 yeast helicases (10 DEAD-box and 1 DEAH-box helicases), which are involved in viral replication or viral RNA recombination [20][21][22][23][24][25][26]. Moreover, TBSV, which lacks its own helicase, usurps several plant ATP-dependent DEAD-box RNA helicases to promote TBSV RNA replication. For example, the plant DDX3-like RH20, DDX5-like RH5 and the eIF4AIII-like RH2 DEAD-box proteins affect plus-strand synthesis, viral genome integrity and viral RNA recombination [27][28][29][30]. On the other hand, the antiviral DDX17-like RH30 DEAD-box protein is relocalized from the nucleus to the site of tombusvirus replication via interacting with the TBSV p33 and p92 pol replication proteins. RH30 inhibits tombusvirus replication through blocking several steps in the replication process. The action of RH30 DEAD-box helicase interferes with VRC assembly, viral RdRp activation and the specific interaction between p33 replication protein and the viral (+)RNA [31]. DEAD-box helicases are a large family of RNA helicases and they are involved in cellular metabolism [32][33][34][35], and in responses to abiotic stress and pathogen infections [36][37][38][39][40][41]. DEAD-box helicases unwind RNA duplexes, affect RNA folding and remodel RNA-protein complexes [42]. Interestingly, DEAD-box helicases can open up completely double-stranded RNA regions by directly loading on duplexes and opening up a limited number of base pairs. This unwinding mode is termed local strand separation [34,42]. Ever increasing data suggest that cellular DEAD-box helicases affect translation and replication of many RNA viruses [43][44][45][46]. For example, turnip mosaic virus and brome mosaic virus co-opt cellular DEADbox helicases for pro-viral functions in translation or replication [44,47,48]. HIV infections are affected by several helicases providing either pro-viral or antiviral functions [49,50]. Thus, the emerging idea is that cellular DEAD-box helicases are important co-opted host factors for several viruses, whereas other cellular DEAD-box helicases play active roles in restricting RNA virus replication. Overall, cellular DEAD-box helicases perform critical roles in virus-host interactions.
In this work, we characterized domains that specify the viral restriction function of the cellular DDX17-like RH30 DEAD-box RNA helicase and the pro-viral function of the DDX3-like RH20 helicase in tombusvirus replication. Expression of deletion mutants of RH30 and RH20 in Nicotiana benthamiana plants revealed that the unique N-terminal domain is important in determining the viral functions of these helicases. To gain insights into what makes a particular DEAD-box helicase either pro-viral or antiviral, we converted the antiviral plant RH30 helicase into a pro-viral helicase through modifying the N-terminal sequences. We also turned the originally pro-viral plant RH20 helicase into an antiviral helicase using a similar strategy. We found that the antiviral helicases target a cis-acting element (namely RII(+)-SL) in the viral (+) RNA, which is critical during the early steps of replication. Therefore, we suggest that the antiviral DEAD-box helicase must act at the earliest steps in the replication process to inhibit TBSV replication. On the contrary, the pro-viral helicases in this work targeted a different cisacting element (namely RI that includes the 5' noncoding region) in the viral RNA, which is needed for (+)RNA synthesis, a latter step in the replication process. This surprising outcome suggests that the antiviral or pro-viral roles of host DEAD-box helicases could be reversed when regulatory/specifying domains are removed from or exchanged between the helicase proteins. These findings on cellular DEAD-box helicases could have major implications in the arms race between the virus and plant hosts to influence the outcome of viral infections.

Dissecting domains specifying the viral restriction function of the cellular RH30 DEAD-box RNA helicase in tombusvirus replication
Sequence comparison of the DDX17-like RH30 DEAD-box RNA helicase from Arabidopsis, which shows strong restriction function against tombusviruses [31], with the pro-viral DDX3-like RH20 DEAD-box RNA helicase [51] suggests the existence of three domains. The highly conserved core DEAD-box helicase domain is present in the center of these two helicases, whereas the N-terminal and short C-terminal domains are more divergent from one another (S1 Fig). To identify what domains are responsible for the antiviral function of RH30 DEAD-box RNA helicase, we expressed deletion mutants of RH30 helicase in N. benthamiana plants via agroinfiltration. Interestingly, similar to the helicase functional mutant RH30 F416L , the N-domain deletion mutant of RH30 also lost its antiviral effect against TBSV replication in the inoculated leaves (RH30 ΔN2-162 , S1 Table,

PLOS PATHOGENS
Role reversal of functional identity in host factors tombusvirus, carnation Italian ringspot virus (CIRV), which utilizes the outer membranes of mitochondria to build VROs, was also not inhibited in N. benthamiana by the transient expression of the RH30 ΔN2-162 missing the N-terminal domain (S1 Table). Taken together, we suggest that RH30 requires the unique N-terminal domain to be an active restriction factor against TBSV and CIRV replication.
As expected, deletion of the core DEAD-box helicase domain in RH30 nullified its antiviral activity against both TBSV and CIRV in N. benthamiana (RH30 ΔHel , S1 Table). In contrast, deletion of the short C-terminal domain in RH30 did not interfere with the restriction function in N. benthamiana (RH30 ΔC547-592 , S1 Table). Combined deletion of the core DEADbox helicase and C-terminal domains in RH30 nullified the antiviral activity of RH30 in N. benthamiana (RH30 ΔHel/ΔC , S1 Table) and in yeast (Fig 1A, lanes 13-16). However, combined Northern blot analysis of TBSV repRNA using a 3' end specific probe shows the level of accumulation of repRNA in wt yeast strain expressing truncation mutants of RH30. Viral proteins His 6 -p33 and His 6 -p92 pol were expressed from plasmids from the CUP1 promoter, while DI-72(+) repRNA was expressed from the GAL1 promoter. His 6 -RH30 derivatives were expressed from a plasmid. Northern blots were evaluated with a phosphoimager using Imagequant software. Standard error was calculated based on three repeats. Bottom panel: Northern blot with 18S ribosomal RNA specific probe was used as a loading control. (B) Western blot analysis of the level of His 6 -p33, and His 6 -RH30 with anti-His antibody. The samples were from wt yeast strain expressing RH30 derivatives. (C) N. benthamiana plants expressing RH30 ΔN/ΔC helicase were inoculated with (C-D) TBSV, (E) CNV, (F) CIRV, respectively. Top panel: Northern blot analyses of tombusvirus gRNA using a 3' end specific probe shows increased accumulation of gRNA and subgenomic RNAs in plants expressing RH30 ΔN/ΔC helicase when compared with plants expressing RH30 or in control plants (treatment with the empty vector). Bottom panel: Ethidium-bromide stained gel shows 18S ribosomal RNA as a loading control. Note that the northern blot analyses of TBSV (+)gRNA is shown in panel C and (-)gRNA using a 3' end specific probe for (-)RNA detection is shown in panel D. Each experiment was repeated at least three times. deletion of both N-and C-terminal domains, which left only the core DEAD-box helicase domain in RH30, converted this helicase into a pro-viral factor by increasing TBSV replication by~2-to-3-fold in N. benthamiana (RH30 ΔN/ΔC , S1 Table and Fig 1C, lanes 1-4). Expression of RH30 ΔN/ΔC , also enhanced the accumulation of the closely-related cucumber necrosis virus (CNV) and CIRV accumulation by 2-to-3-fold in N. benthamiana (Fig 1E and 1F). This surprising outcome suggests that an antiviral DEAD-box helicase could become a pro-viral factor when inhibitory (or regulatory) domains are removed from the protein.

The pro-viral RH30 ΔN/ΔC DEAD-box helicase is re-localized into the tombusvirus replication compartment in plants
The antiviral RH30 DEAD-box helicase has to re-localize into the large p33-containing replication compartment in order to restrict tombusvirus replication [31]. To test if the pro-viral RH30 ΔN/ΔC helicase is also present in VROs, we co-expressed GFP-tagged RH30 ΔN/ΔC with p33-BFP replication protein and either H2B-RFP (nuclear marker) or RFP-SKL (a peroxisomal matrix marker) in N. bentamiana infected with TBSV or CNV (Fig 2A and 2B). Confocal microscopy analysis of N. bentamiana leaves showed the partial re-localization of GFP-RH30 ΔN/ΔC into the large TBSV or CNV VROs, which consist of aggregated peroxisomes. GFP-RH30 ΔN/ΔC helicase was also partially re-targeted in CIRV-infected N. benthamiana cells into the CIRV p36 and p95 pol -containing VROs (Fig 2B, top panel), which consist of aggregated mitochondria decorated by CoxIV-RFP marker [52,53]. Based on these experiments, we propose that the nuclear and cytosolic-localized pro-viral RH30 ΔN/ΔC helicase is re-targeted efficiently into the tombusvirus replication compartment, similar to the full-length RH30 helicase with antiviral function.
We also analyzed the subcellular distribution of GFP-RH30 ΔN(ΔN2-162) , which showed no significant antiviral activity when expressed in N. bentamiana leaves (Fig 1A, lanes 17-20, and S1 Table) This unusually strong nuclear localization of GFP-RH30 ΔN might explain its lack of activity during TBSV replication. Accordingly, we have shown that the antiviral activity of the full-length RH30 depended on its efficient redistribution from the nucleus to the VROs [31]. Future detailed experiments are needed to determine the factors affecting the subcellular localization of these cellular helicases and mutants.
To test if the RH30 ΔN/ΔC helicase interacts with the tombusvirus replication protein, we have conducted bimolecular fluorescence complementation (BiFC) experiments in N. benthamiana leaves. Comparison with the antiviral full-length RH30 helicase, RH30 ΔN/ΔC helicase also showed interaction with the TBSV p33 replication protein within the viral replication compartment, marked by the peroxisomal matrix marker RFP-SKL (Fig 2C-2E).

The pro-viral RH30 ΔN/ΔC DEAD-box helicase promotes tombusvirus RNA synthesis in vitro
To gain deeper insights into the acquired pro-viral function of RH30 ΔN/ΔC helicase, we affinity-purified the recombinant RH30 ΔN/ΔC helicase from E. coli, followed by testing its activity in vitro in a TBSV replicase reconstitution assay with purified recombinant TBSV p33 and p92 pol replication proteins [18,54]. Addition of the purified RH30 ΔN/ΔC helicase to the replicase reconstitution assay, which is based on yeast cell-free extract (CFE), led to increased TBSV repRNA replication by~50% (Fig 3, lanes 1-2). The double-stranded repRNA replication Confocal microscopy images show that the re-targeted GFP-RH30 ΔN/ΔC helicase is present in the VRO, marked by the CIRV p36-BFP replication protein and CoxIV-RFP mitochondrial marker. Arrows point at the large VRO. Expression of the above proteins from the 35S promoter was done after co-agroinfiltration into N. benthamiana leaves. The agro-infiltrated leaves were collected for intermediate was also produced by~40% more efficiently in the presence of RH30 ΔN/ΔC helicase. These in vitro data suggest that RH30 ΔN/ΔC helicase likely promotes RNA synthesis or the VRC assembly during TBSV replication. This idea is also supported by the increased level of both (-)RNAs and (+)RNAs observed when RH30 ΔN/ΔC helicase was expressed in N. benthamiana ( Fig 1C, lanes 1-4 and 1D, lanes 1-3). This is in contrast with the strong inhibitory effect of the full-length RH30 DEAD-box helicase on TBSV RNA replication (Fig 3, lanes 5-6) [31]. Please note that the pro-viral activity of RH30 ΔN/ΔC helicase is not fully identical with that of confocal imaging at 2.5 days post agro-infiltration. Bottom panel: The TBSV p33-BFP replication protein and RFP-SKL peroxisomal matrix marker co-localize with the GFP-RH30 ΔN/ΔC helicase within the large VRO. Scale bars represent 5 μm. Each experiment was repeated twice. (C) Interactions between TBSV p33 replication protein and the RH30 ΔN/ΔC helicase were detected by BiFC. The TBSV p33-cYFP replication protein and the nYFP-RH30 ΔN/ΔC and the RFP-SKL peroxisomal marker protein were expressed via agroinfiltration. The merged image shows the efficient colocalization of the peroxisomal RFP-SKL with the BiFC signals, indicating that the interaction between the tombusvirus replication protein and the recruited RH30 ΔN/ΔC helicase takes place in VROs, which consist of aggregated peroxisomes. Scale bars represent 5 μm.

PLOS PATHOGENS
Role reversal of functional identity in host factors the previously shown DDX3-like yeast Ded1 or the plant RH20 helicases, which mostly enhanced (+)RNA synthesis [29]. The enhanced activity of RH30 ΔN/ΔC helicase in (-)-strand synthesis is similar to some Ded1 mutants [27] and it is likely due to a yet undefined helicase function during VRC assembly or the use of (+)RNA template during (-)RNA synthesis.

The altered specificity of pro-viral RH30 ΔN/ΔC DEAD-box helicase in unwinding critical cis-acting elements in the viral RNA in comparison with the antiviral RH30 helicase
The antiviral activity of the full-length RH30 DEAD-box helicase is attributed to the efficient unwinding of the secondary structure of the RII(+) region in the TBSV (+)RNA [31]. RII(+) contains a critical cis-acting stem-loop element, termed RII(+)SL, which is required for the p33-mediated recruitment of the TBSV (+)RNA template into replication [16], and for the activation of the p92 RdRp [13]. Therefore, we tested if the purified RH30 ΔN/ΔC helicase could unwind partial dsRNA substrates in a dsRNA-strand separation assay [29,31], where parts of the TBSV repRNA were double-stranded as shown schematically in Fig 4. Interestingly, unlike the full-length RH30 helicase, the RH30 ΔN/ΔC helicase was found to inefficiently separate the partial dsRNA template, involving the RII sequence, in the presence of ATP ( https://doi.org/10.1371/journal.ppat.1008990.g004 versus 9-12). These findings suggest that, in contrast with the full-length RH30 helicase, the RH30 ΔN/ΔC helicase cannot change the RII(+)SL hairpin structure, thus it might not block the binding of TBSV p33 replication protein to the critical RII(+)-SL RNA recognition element required for template recruitment into replication. This could explain the lost antiviral activity of the RH30 ΔN/ΔC helicase.
However, RH30 ΔN/ΔC helicase was able to unwind the RI-containing partial dsRNA in vitro ( Fig 4A, lanes 1-4 versus 5-8). Helicase-driven unwinding of the dsRNA structure within RI sequence of the dsRNA replication intermediate is critical during initiation of (+)-strand RNA synthesis [29,51]. The RI(-) region contains the promoter for (+)RNA synthesis, including loading of the p92 pol on the (-)RNA template [55]. We note that the unwinding activity of RH30 ΔN/ΔC helicase was comparable to that of the full-length RH30 helicase when tested with RIII and RIV (Fig 4C and 4D). Altogether, this ability of RH30 ΔN/ΔC helicase to unwind RI dsRNA structure, while inefficient on the RII dsRNA structure could explain its acquired proviral function by stimulating viral (+)-strand synthesis without inhibiting (+)RNA template selection/recruitment by p33 replication protein (see discussion).
To further characterize the lost antiviral function of RH30 ΔN/ΔC helicase during tombusvirus replication, we tested if RH30 ΔN/ΔC helicase could inhibit the selective binding of p33 replication protein to the viral RNA template in vitro. We used our recently developed p33-RNA capturing assay [31] based on a biotin-labeled RII(+) sequence of the TBSV (+) RNA, which represents RII(+)-SL RNA recognition element required for template recruitment into replication by the p33 replication protein [16]. RII(+)-SL of the TBSV RNA is also an essential assembly platform for the replicase complex [56]. Briefly, the biotin-labeled RII(+) RNA was pre-incubated either with purified RH30 ΔN/ΔC or RH30 helicases ( Fig 5A). Then, we added the purified recombinant TBSV p33C (the soluble C-terminal region, including the RNA-binding and p33:p33/p92 interaction region of the p33 replication protein), which can bind specifically to RII(+)-SL carrying the C•C mismatch in the internal loop [16]. Subsequently, the biotin-labeled RII(+) RNA was captured on streptavidin-coated magnetic beads. The helicase and p33C proteins bound to the RII(+) RNA were eluted from the beads, followed by western blot analysis with anti-MBP antibody. This assay revealed that RH30 ΔN/ΔC did not inhibit the binding (thus capturing) of p33C to RII(+) RNA ( Based on these data, we suggest that RH30 ΔN/ΔC has lost its ability to interfere with the selective binding of p33 replication protein to the (+)RNA template. This lost ability of RH30 ΔN/ΔC is likely due to its inability to change RII(+) RNA structure. Indeed, this result is in agreement with the dsRNA unwinding data, which showed that RH30 ΔN/ΔC cannot efficiently separate the RII duplex ( Fig 4B). Altogether, these observations could explain why RH30 ΔN/ΔC has no viral restriction function, while the full-length RH30 likely prevents the selective recognition/ recruitment of TBSV (+)RNA by the p33 replication protein [31]. In addition, we propose that RH30 ΔN/ΔC acquired pro-viral function by selectively unwinding dsRNA structures within the RI segment required for TBSV (+)RNA initiation [28,57].

The pro-viral role of the N-terminal domain of the cellular RH20 DEADbox RNA helicase in tombusvirus replication
To learn if the above findings of reversing the viral function of an Arabidopsis DEADbox RNA helicase could be generalized, we also tested the pro-viral DDX3-like RH20 DEADbox RNA helicase [29,51]. To identify what domains are responsible for the pro-viral function of RH20 DEAD-box RNA helicase, we expressed deletion mutants of RH20 in N. benthamiana

PLOS PATHOGENS
Role reversal of functional identity in host factors plants via agroinfiltration. Expression of a short N-domain truncation mutant of RH20 led to elimination of the pro-viral activity of the RH20 helicase in TBSV replication in the inoculated leaves (RH20 ΔN2-36 , S2 Table). More importantly, expression of the entire N-terminal domain deletion mutant of RH20 helicase in N. benthamiana plants inhibited TBSV RNA accumulation by almost 3-fold (RH20 ΔN2-96 , S2 Table and Table) was also inhibited by 2-to-3-fold in N. benthamiana by the transient expression of the RH20 ΔN2-96 missing the N-terminal domain (this mutant will be named RH20 ΔN below and in the figures). In contrast, deletion of the short C-terminal domain in RH20 did not interfere with the pro-viral function in N. benthamiana (RH20 ΔC , S2 Table). Overall, in the absence of its N-terminal domain, the pro-viral RH20 DEAD-box helicase gains a new antiviral function. Thus, similar to the case with RH30 helicase, the functional identity of pro-viral RH20 helicase could be reversed by removing a regulatory domain.

The antiviral RH20 ΔN DEAD-box helicase inhibits tombusvirus RNA synthesis in vitro
To learn what features of RH20 ΔN helicase caused it to become antiviral, we affinity-purified the recombinant RH20 ΔN helicase from E. coli, and tested its activity in vitro in a TBSV replicase reconstitution assay [18,54]. Addition of the purified RH20 ΔN helicase to the replicase reconstitution assay resulted in decreased TBSV repRNA replication by~3-fold ( Fig 6D, lanes 1-2 versus lanes 5-6). The double-stranded repRNA replication intermediate was also reduced by~40% in the presence of RH20 ΔN helicase in comparison with the full-length RH20 helicase. Based on these data, we suggest that RH20 ΔN helicase likely inhibits both (-) and (+)-strand RNA synthesis during TBSV replication. This is in agreement with in planta data on TBSV accumulation, which also showed decreased levels of both (-)RNAs and (+)RNAs when RH20 ΔN helicase was expressed in N. benthamiana (Fig 6A, lanes 13-15 and Fig 6B, lanes  4-6).

The antiviral RH20 ΔN DEAD-box helicase is re-targeted into the tombusvirus replication compartment in plants
To test if the antiviral RH20 ΔN helicase is present in the replication compartment, we conducted BiFC experiments in N. benthamiana leaves. The antiviral RH20 ΔN helicase, similar to the pro-viral full-length RH20 helicase, also showed interaction with the TBSV p33 replication protein within the viral replication compartment, marked by the peroxisomal matrix marker RFP-SKL (Fig 7A and 7B). Thus, both the pro-viral full-length RH20 and antiviral truncated RH20 ΔN helicases are re-targeted to the VROs during TBSV replication. were allowed to form an RNP complex in the presence or absence of 1 mM ATP for 15 min, followed by addition of MBP-p33C protein (1.9 μM), and additional incubation for 15 min. Then, the biotin-labeled RII(+) RNA-protein complex was captured on streptavidin-coated magnetic beads, which were sequentially washed with a buffer. (B-C) The bound proteins were eluted from the beads, followed by measuring the amounts of MBP-p33C in the eluates by western blotting using anti-MBP antibody. Reduced amounts of MBP-p33C in the eluates mean that the given helicase inhibited the binding of p33C to the viral RNA, likely due to remodeling the RNA structure that could not be recognized by p33 replication protein. Nonbiotinylated RNA (panel B, lane 5) was used as a control. Bottom panels show SDS-PAGE analysis of the purified recombinant proteins stained by coomassie-blue. Each experiment was repeated three times. https://doi.org/10.1371/journal.ppat.1008990.g005

Role reversal of functional identity in host factors
The antiviral RH20 ΔN DEAD-box helicase efficiently unwinds critical cisacting elements in the TBSV RNA Here, we again used the in vitro strand-separation assay to test the antiviral activity of the RH20 ΔN helicase. Surprisingly, we observed a dramatically enhanced unwinding activity of RH20 ΔN helicase in comparison with the pro-viral full-length RH20 helicase. For, example, the purified RH20 ΔN helicase, unlike the full-length RH20 helicase, could efficiently unwind the partial dsRNA template, involving the critical RII sequence, in the presence of ATP (Fig 8B,  lanes 1-5 versus 11-15). These findings suggest that, in contrast with the full-length RH20 helicase, the RH20 ΔN helicase can 'destroy" the RII(+)SL hairpin structure, thus it might be able to block the binding of TBSV p33 replication protein to the critical RII(+)-SL RNA recognition element, which event is required for template recruitment into replication. Actually, RH20 ΔN helicase was able to efficiently unwind all the TBSV dsRNA templates provided (Fig  8). This efficient unwinding of viral RNA structures by RH20 ΔN helicase likely is the reason for its antiviral function. The pro-viral full-length RH20 is more selectively targets RI sequence, which is critical during initiation of (+)-strand RNA synthesis, but not that of RII of the

PLOS PATHOGENS
Role reversal of functional identity in host factors dsRNA replication intermediate [29,51]. Therefore, we suggest that the RH20 ΔN helicase becomes antiviral based on its lost RNA sequence selectivity in comparison with the more selective pro-viral full-length RH20 helicase.
To further characterize the acquired antiviral function of RH20 ΔN helicase during tombusvirus replication, we again used our p33-RNA capturing assay [31] based on a biotin-labeled RII(+) sequence. The biotin-labeled RII(+) RNA was pre-incubated either with purified RH20 ΔN or RH20 and subsequently, with the purified TBSV p33C as described above (Fig 5A). This assay revealed that RH20 ΔN strongly inhibited the binding of p33C to the RII(+) RNA in the presence of ATP (Fig 5C, lane 4 versus lane 1). On the contrary, the full-length RH20 did not inhibit efficiently the binding of p33C to the RII(+) RNA (Fig 5C, lanes 2-3 versus lane 1). Based on these data, we suggest that RH20 ΔN acquired a new ability to interfere with the selective binding of p33 replication protein to the (+)RNA template. This observation could explain why RH20 ΔN obtained a new viral restriction function, while the full-length RH20 does not likely prevent the selective recognition/recruitment of TBSV (+)RNA by the p33 replication protein [31]. These data also invite the attention to the possible roles of cellular helicases to unwind the hairpin structure of a cis-replication element in TBSV (+)RNA to prevent its selective recognition and recruitment into VRC by the p33 replication protein, which is an early step in replication (see discussion).

Chimeric DEAD-box helicases confirm the critical roles of the N-terminal domains in DEAD-box helicases in TBSV replication
To obtain further evidence on the critical roles of the N-terminal domains of RH20 and RH30 DEAD-box helicases in specifying their functions in TBSV replication, we constructed two chimeric DEAD-box helicases by switching the N-terminal domains between RH20 and RH30 (Fig 9). The chimeric helicase containing the N-terminal domain from RH30 and the remaining core helicase and C-terminal sequences from RH20 (i.e., named as RH 322 helicase) showed antiviral activity against TBSV when expressed in N. benthamiana (Fig 9, lanes 13-15 versus  1-3). Thus, the chimeric RH 322 has anti-TBSV activity similar to the full-length RH30 helicase (Fig 9, lanes 13-15 versus 7-9). On the contrary, the other chimeric helicase containing the Nterminal domain from RH20 and the remaining sequences from RH30 (i.e., named as RH 233 helicase) showed pro-viral activity on TBSV replication when expressed in N. benthamiana (Fig 9, lanes 10-12 versus 1-3). These results support the notion that the N-terminal domain in the DEAD-box helicases might be an important determinant of the specific pro-viral or antiviral roles of these DEAD-box helicases in TBSV replication.

The N-terminal-domain in the antiviral RH30 DEAD-box helicase is required to interfere with the interaction between RH20 helicase and the p33 replication protein
An important feature of both antiviral RH30 and the pro-viral RH20 helicases is to bind to the TBSV p33 replication protein [29,31,51]. Therefore, it is possible that the antiviral DEADbox helicases are inhibitory to TBSV replication due to their predicted abilities to out compete the pro-viral RH20 helicase for the interaction with the p33 replication protein. This model was tested in co-purification experiments from yeast. Co-expression of the truncated antiviral RH20 ΔN helicase with the pro-viral full-length RH20 resulted in similar amounts of co-purified RH20 with the p33 replication protein from yeast membranes (Fig 10A, lanes 3 versus 1). Coexpression of the truncated pro-viral RH30 ΔN/ΔC helicase with the pro-viral full-length RH20 also resulted in similar amounts of co-purified RH20 with the p33 replication protein from yeast membranes (Fig 10B). Thus, the co-expression of these truncated DEAD-box helicases did not change the amount of co-purified RH20 with the p33 replication protein from yeast membranes. Therefore, it is unlikely that the acquired antiviral function of RH20 ΔN helicase or the pro-viral RH30 ΔN/ΔC helicase is based on altering the interaction between the pro-viral RH20 helicase and the TBSV p33 replication protein.
On the contrary, we found that co-expression of the full-length antiviral RH30 helicase with the pro-viral RH20 resulted in reduced amounts of co-purified RH20 with the p33 replication protein from yeast membranes (Fig 10C, lanes 2 versus 1). Similarly, co-expression of the full-length antiviral RH30 with the pro-viral eIF4AIII-like RH2 helicase or the yeast DDX3-like Ded1 helicase [an ortholog of the plant RH20 DEAD-box helicase with similar pro-viral function [29,51] resulted in reduced amounts of co-purified RH2 or Ded1 helicases with the p33 replication protein from yeast membranes ( Fig 10C). Thus, this is in contrast with the results obtained when we co-expressed the truncated antiviral RH20 ΔN with the pro-viral RH20 in yeasts. Based on these results, it seems that the truncated DEAD-box helicases lack the ability to compete with the pro-viral RH20 helicase to interact with the p33 replication protein, whereas the full-length antiviral RH30 helicase has this ability. This could explain why the full-length RH30 has a much more potent inhibitory activity than RH20 ΔN on TBSV replication in N. benthamiana (see below).

Dominant effects of the antiviral DEAD-box helicases over the pro-viral RH20 DEAD-box helicase in TBSV replication
Through reversing the pro-viral role of RH20 DEAD-box helicase into an antiviral role in the RH20 ΔN helicase, we created a novel antiviral tool. To test how effective is this new antiviral tool, we asked the interesting question whether RH20 ΔN helicase could compete with the proviral RH20 to affect TBSV replication. Interestingly, co-expression of RH20 with RH20 ΔN helicase resulted in reduced TBSV RNA accumulation in N. benthamiana (Fig 11A, lanes 7

PLOS PATHOGENS
Role reversal of functional identity in host factors versus 1-3). Because these data suggested that the antiviral function of a DEAD-box helicase could be a dominant feature over the pro-viral RH20 helicase, we also co-expressed the fulllength antiviral RH30 with the pro-viral RH20 in N. benthamiana infected with TBSV. We found that TBSV replication was reduced to a similar level as it was observed with the expression of RH30 helicase alone (Fig 11B, lanes 14-16 versus 5-7 and 8-10). Co-expression of the full-length antiviral RH30 with another pro-viral DEAD-box helicase, namely eIF4AIII-like RH2 helicase, which function differently from RH20 by enhancing the activity of a replication enhancer element in the viral (-)RNA [28], also resulted in TBSV replication at a highly reduced level (Fig 11B. lanes 17-19). Thus, it seems that the antiviral DEAD-box helicases, both the full-length RH30 and truncated version (RH20 ΔN helicase), acted in a dominant manner over the pro-viral RH20 and RH2 DEAD-box helicases in plants. The dominant effect over the pro-viral RH20 DEAD-box helicase by the antiviral RH30 and RH20 ΔN in TBSV replication is likely due to their abilities to unwind the critical RII(+)-SL cis-acting RNA element, which is an early step in replication (see discussion). Nevertheless, the dominant nature of antiviral helicases over pro-viral RNA helicases shows a great potential to develop new antiviral factors. This aspect of cellular DEAD-box helicases will need to be further studied after the functions of more cellular DEAD-box helicases in TBSV replication are characterized in the future.

Discussion
Unlike many other pathogens, (+)RNA viruses code for only a rather limited number of genes, making them highly dependent on numerous co-opted host factors for supporting viral replication and other viral processes during their infections. This extreme dependence on subverted host factors, however, renders (+)RNA viruses vulnerable to host restriction factors that could block virus replication [58]. Accordingly, tombusviruses are dependent on the cellular DDX3-like RH20 and eIF4AIII-like RH2 DEAD-box helicases during their replication in plants [27][28][29]51]. However, there are over a hundred RNA helicases coded in plant genomes [36,59], which might not assist tombusvirus replication, but instead, might block the viral replication process. Indeed, we have previously shown that the DDX17-like RH30 helicase, which is rather similar to the pro-viral DDX3-like RH20 helicase, has strong anti-tombusvirus activity in plants [31]. However, it is currently incompletely defined what features make particular DEAD-box helicases either pro-viral or antiviral.
In this work, we have unraveled that the functional identity/orientation of the co-opted DEAD-box helicases could be altered by changing the pro-viral RH20 helicase into antiviral and the antiviral RH30 helicase into a pro-viral helicase via deletion of regulatory sequences of these cellular helicases or making helicase chimeras. Notably, we find that the unique N-terminal domains in RH20 and RH30 DEAD-box helicases seem to be really important to specify the activity of the given cellular helicase. In the absence of the N-terminal domain, the core His 6 -RH30 ΔN/ΔC and His 6 -RH20 were detected with anti-His antibody, while Flag-p33 replication protein was detected with anti-FLAG antibody. Bottom panels: blot of total His 6 -p33, His 6 -RH30 ΔN/ΔC and His 6 -RH20 and Flag-p33 in the total yeast extracts detected with anti-His and anti-Flag antibodies, respectively. (B) Co-purification of His 6 -tagged RH20 and His 6 -RH20 ΔN with Flag-p33 replication protein from subcellular membranes of yeasts. See further details in panel A. (C) Co-purification of His 6 -tagged RH20, His 6 -RH2 and His 6 -Ded1 with Flag-p33 replication protein from subcellular membranes of yeasts. Yeasts were co-expressing His 6

PLOS PATHOGENS
Role reversal of functional identity in host factors helicase domain becomes unhinged, showing altered specificity in RNA duplex unwinding activities. We note that the RNA helicase activity, however, seems to be important for both pro-viral and antiviral activities, because deletion or active site mutation of the core helicase domain eliminated these activities (S1 and S2 Tables). We also observed that the changed functional identity of the cellular helicases manifested with both the peroxisomal-replicating TBSV and the mitochondrial-replicating CIRV in plants.
The tombusvirus genome is loaded with cis-acting RNA replication and translation elements to facilitate well-orchestrated and efficient viral replication [19]. The emerging picture from recent studies with TBSV is that cellular helicases could target various cis-acting elements in TBSV RNAs. Accordingly, in vitro studies revealed that one of the most important features of the cellular helicases is their ability to unwind the RNA structures within particular regions of the TBSV RNA. We find that the critical feature for the pro-viral DEAD-box helicases, such as RH30 ΔN/ΔC helicase (this work) and the full-length RH20, is to selectively target and unwind the RI sequence (5'UTR) of the dsRNA replication intermediate, which is critical during initiation of (+)-strand RNA synthesis [27][28][29]51]. On the contrary, the critical feature of the antiviral DEAD-box helicases, such as RH20 ΔN helicase (this work) and the full-length RH30 [31], is to efficiently target and "destroy" the RII(+)SL hairpin structure, and thus block the selective binding of TBSV p33 replication protein to the critical RII(+)-SL RNA recognition element, which is absolutely required for RNA template recruitment into replication, VRC assembly and viral RdRp activation (Fig 12) [13,14,16,60]. Indeed, we demonstrated that the antiviral RH20 ΔN (this work) and RH30 helicases [31] interfere with the binding of the p33 replication protein to the viral RII(+)RNA template in vitro, whereas the pro-viral RH30 ΔN/ΔC and RH20 helicases did not inhibit this process efficiently (Fig 5).
One of our interesting findings in this work is the observation of the dominant nature of the antiviral helicases over the pro-viral RH20 helicase. For example, although RH20 ΔN helicase facilitates the unwinding of RI-containing sequences, which are important for initiation of (+)RNA synthesis by the TBSV RdRp [29,51], this "pro-viral" feature of RH20 ΔN helicase might not be important in vivo. This is because the antiviral activity of RH20 ΔN helicase, involving unwinding of RII(+), which remodels RII(+)-SL and inhibits the recruitment of the viral RNA into replication, is an early step during TBSV replication (Fig 12). Thus, by unwinding the RII(+) RNA secondary structure and thus, inhibiting viral (+)RNA recognition by the p33 replication protein, both RH20 ΔN and the full-length RH30 helicases block this early step in tombusvirus replication. This makes this early event a dominant feature via preventing the predicted pro-viral potential of RH20 ΔN helicase or the pro-viral RH20 helicase (when RH30 is co-expressed) to facilitate the initiation of (+)RNA synthesis, which is a late step and depends on the (+)RNA requirement during the early steps.
The truncated cellular helicases with altered functional identity were different from the fulllength helicases in biochemical features. For example, in contrast with the full-length RH30 helicase, the truncated RH30 ΔN/ΔC or RH20 ΔN helicases were not found to affect the important interaction between the pro-viral RH20 helicase and the p33 replication protein. Taken together, we propose that the major difference is that the antiviral and pro-viral helicases target different critical cis-acting regions within the viral RNA as discussed above.
Overall, the evolving features of the unique N-terminal domains of cellular DEADbox helicases seem to be key components of the arms race between tombusviruses and their hosts. We propose that deletions/mutations in critical positions within the N-terminal domains of cellular DEAD-box helicases might render them either antiviral or pro-viral. The modular nature of the regulatory/specificity domains in cellular DEAD-box helicases might allow plants to gain new antiviral functions during evolution. Moreover, these findings will open up the possibility for science to turn pro-viral host factors into antiviral factors, for the benefit of agriculture and health science. We envision that the modular structure of regulatory/specificity domains in RNA helicases make them highly suitable to create numerous restriction factors from known pro-viral factors by altering existing regulatory sequences in these proteins. Alternatively, by using synthetic biology tools, future research could increase our arsenal of antiviral factors.
It will be interesting to learn if functions of helicases could be reversed in case of other RNA viruses and retroviruses. Those viruses also usurp several cellular DEAD-box helicases to facilitate their replication and other viral processes during infection [49,61]. Moreover, the host also deploys DEAD-box helicases to inhibit the replication of the above viruses [41,61]. Many of the identified DEAD-box helicases with restriction functions are conserved in plants and mammals.
In summary: We have demonstrated that the viral functions of the plant DDX17-like RH30 DEAD-box helicase and DDX3-like RH20 DEAD-box helicase could be reversed in tombusvirus replication by deletion of unique N-terminal domain or by making helicase chimeras. The discovery of the sequence plasticity within regulatory domains of DEAD-box helicases, which can alter recognition of different cis-acting elements in the viral genome, illustrates the evolutionary potential of RNA helicases in the arms race between viruses and their hosts, including key roles of RNA helicases in plant innate immunity.

Materials and methods
Please find several basic approaches described in the supplementary S1 Text. Based on our current and previous data, we propose that the RH20 ΔN helicase becomes a novel antiviral restriction factor by inhibiting the recruitment of the viral (+)RNA and, thus, blocking VRC formation. This inhibitory effect of RH20 ΔN helicase is likely through remodeling the structure of the RII(+)-SL cis-acting RNA element, which might block the specific recognition by p33 replication protein, which requires the RII(+)-SL stem-loop structure. RH20 ΔN helicase inhibits VRC assembly as well, because the stem-loop structure in RII(+)-SL is essential part of the VRC assembly platform. In contrast, RH30 ΔN/ΔC helicase becomes a novel pro-viral factor by opening up the RI-containing dsRNA structure within the dsRNA replication intermediate. This, in turn, RH30 ΔN/ΔC helicase likely facilitates the initiation of (+)-strand synthesis by the TBSV RdRp. We propose that unlike the full-length RH30, the RH30 ΔN/ΔC helicase has lost its ability to remodel RII(+)-SL structure, which is required for antiviral activity of RH30. https://doi.org/10.1371/journal.ppat.1008990.g012

PLOS PATHOGENS
Role reversal of functional identity in host factors
To The total RNAs were extracted and analyzed by northern blot assay as previously described [31].

Confocal laser microscopy
The subcellular localization of AtRH30 ΔN/ΔC was observed in N. benthamiana plant epidermal cells after expression of an N-terminal fusion of AtRH30 ΔN/ΔC to GFP. In the experiments with TBSV and CNV, the transgenic N. benthamiana (expressing H2B fused to RFP, as a nuclear marker) leaves were infiltrated with agrobacteria carrying expression plasmids pGD-p33-BFP (OD 600 0.4), pGD-GFP-AtRH30 ΔN/ΔC (OD 600 0.4) and pGD-P19 (OD 600 0.2). The leaves were also co-infiltrated with agrobacterium to express CNV 20Kstop gRNA for CNV infections. For TBSV infections, the infiltrated leaves were inoculated with TBSV sap inocula 16 h post agroinfiltration. To observe the subcellular localization of GFP-AtRH30 ΔN/ΔC with cellular markers, the wild-type N. benthamiana leaves were agroinfiltrated to express p36-BFP and CoxIV-RFP (a mitochondria marker) in the case of CIRV infections or p33-BFP and RFP-SKL (peroxisome luminar marker) in the case of TBSV infections together with GFP-AtRH30 ΔN/ΔC and P19 (as a gene silencing suppressor). Approximately 2 days post-virus inoculations, imaging of infiltrated leaves was performed on an Olympus FV1200 confocal microscopy using 60X water-immersion objective equipped lasers as described [31]. BFP was excited by 405 nm laser. GFP was excited by 488 nm laser and RFP was excited by 543 nm laser. Images were obtained and merged using Olympus FLUOVIEW 1.5 software.

BiFC assay
To

Analysis of TBSV replication with in vitro reconstituted TBSV replicase in yeast cell-free extract
The yeast cell-free extracts (CFE) that support in vitro TBSV replicase reconstitution were prepared from yeast strain BY4741 as described [18,54]. The in vitro reconstituted TBSV replicase assays were performed with the mixture of 2 μl of CFE, 0.5 μg DI-72 (+)repRNA, 0.2 μg affinity-purified MBP-p33 as well as MBP-p92 pol (both recombinant proteins were purified from E.

Biotinylated RNA-protein interaction assay
To test if the RNA helicases and their mutants could interfere with viral template recognition by TBSV p33, we performed a biotinylated RNA-protein interaction assay as previously described [31] with a minor modification. The biotinylated RNA composed of the RII(+) of DI-72 was synthesized by in vitro T7 transcription supplemented with 1.
The co-purification assays were performed as previously described [31]. Briefly, the transformed yeast cells were grown in SC-ULHmedia supplemented with 2% glucose at 23˚C for 16 h. After that, the cell cultures were transferred to SC-ULHmedia supplemented with 2% galactose for 24 h at 23˚C, followed by the incubation with additional 50 μM CuSO 4 for 6 h at 23˚C. The obtained yeast cells were resuspended in a high-salt TG buffer (50 mM Tris-HCl [pH 7.5], 10% glycerol, 0.5 M NaCl, 10 mM KCl, 15 mM MgCl 2, 1% [V/V] yeast protease inhibitor cocktail [Ypic]) and were broken using a FastPrep Homogenizer (MP Biomedicals) and glass beads. Centrifugation at 500 g for 5 min at 4˚C was performed to remove the cell debris. The membrane fraction containing the viral replicase complexes was collected by centrifugation at 35,000 g for 20 min at 4˚C, followed by solubilization of membranes in a highsalt TG buffer containing 2% Triton X-100, 1% [V/V] Ypic for 3 h at 4˚C. The supernatant of detergent-solubilized membranes was obtained by centrifugation at 35,000 g for 20 min at 4˚C and then was incubated with anti-FLAG M2-agarose affinity resin (Sigma) in columns for 16 h at 4˚C. The columns were washed with high-salt buffer for three times. For elution, the protein samples in the columns were incubated with SDS-PAGE loading dye at 85˚C for 6 min and were collected by centrifugation at 500 g for 3 min. The collected samples were then boiled with β-mercaptoethanol for 20 min. Affinity-purified Flag-p33 was analyzed by western blot with anti-FLAG antibody, and co-purified 6xHis-tagged proteins was analyzed by western blot with anti-His antibody [31].

Pull-down assay
To test direct binding between proteins, we performed an in vitro pull-down assay as previously described [63,64]. The cell lysates of E. coli strain BL21 (DE3) CodonPlus (Stratagene) expressing recombinant proteins, including MBP, MBP-AtRH30, MBP-AtRH30 ΔN/ΔC , GSThis-T33C, GST, GST-AtRH20, GST-AtRH20 ΔN2-96 , and MBP-T33C were obtained as described above. The E. coli lysates expressing MBP or GST fusion proteins were separately incubated with amylose resin (NEB) or GST bind resin (EMD Millipore) for 4 h at 4˚C, followed by washing with cold column buffer (20 mM HEPES [pH7.4], 25 mM NaCl, 1 mM EDTA [pH 8.0]) for four times. The amylose or GST bind resin was then incubated with E. coli lysates containing GST or MBP fusion proteins for 4 h at 4˚C in the presence of 0.5% NP-40 and 0.1% [V/V] yeast protease inhibitor cocktail, followed by washing with cold column buffer containing 0.5% NP-40 for four times. The protein complexes bound to MBP resins were eluted with cold column buffer containing 0.36% [W/V] maltose and 1 mM DTT. For those bound to the GST bind resin were eluted with a column buffer containing 10 mM glutathione and 1 mM DTT in pH 7.5. The amounts of proteins were analyzed by western blot assay with anti-MBP or anti-His antibodies [31].

Gel mobility shift assay (EMSA) and dsRNA strand-separation assay
The conditions for the EMSA experiments were described previously [16]. Briefly, the EMSA assay was performed with 0.1 pmol of 32 P-labeled RNA probes along with different concentrations (0.4, 1.9, and 5.7 μM) of purified recombinant MBP-fusion proteins or MBP in the presence of RNA binding buffer (10 mM HEPES [pH7.4], 50 mM NaCl, 1 mM DTT, 1 mM EDTA, 5% Glycerol, 2.5 mM MgCl 2 ), 2 U of RNase inhibitor, as well as 0.1 μg of tRNA in a total of 10 μl reaction volume. Two different amounts (2 and 4 pmol) of unlabeled RNAs together with 5.7 μM of either MBP-RH30 or MBP were used for template competition.