A novel viral strategy for host factor recruitment: The co-opted proteasomal Rpn11 protein interaction hub in cooperation with subverted actin filaments are targeted to deliver cytosolic host factors for viral replication

Positive-strand (+)RNA viruses take advantage of the host cells by subverting a long list of host protein factors and transport vesicles and cellular organelles to build membranous viral replication organelles (VROs) that support robust RNA replication. How RNA viruses accomplish major recruitment tasks of a large number of cellular proteins are intensively studied. In case of tomato bushy stunt virus (TBSV), a single viral replication protein, named p33, carries out most of the recruitment duties. Yet, it is currently unknown how the viral p33 replication protein, which is membrane associated, is capable of the rapid and efficient recruitment of numerous cytosolic host proteins to facilitate the formation of large VROs. In this paper, we show that, TBSV p33 molecules do not recruit each cytosolic host factor one-by-one into VROs, but p33 targets a cytosolic protein interaction hub, namely Rpn11, which interacts with numerous other cytosolic proteins. The highly conserved Rpn11, called POH1 in humans, is the metalloprotease subunit of the proteasome, which couples deubiquitination and degradation of proteasome substrates. However, TBSV takes advantage of a noncanonical function of Rpn11 by exploiting Rpn11’s interaction with highly abundant cytosolic proteins and the actin network. We provide supporting evidence that the co-opted Rpn11 in coordination with the subverted actin network is used for delivering cytosolic proteins, such as glycolytic and fermentation enzymes, which are readily subverted into VROs to produce ATP locally in support of VRO formation, viral replicase complex assembly and viral RNA replication. Using several approaches, including knockdown of Rpn11 level, sequestering Rpn11 from the cytosol into the nucleus in plants or temperature-sensitive mutation in Rpn11 in yeast, we show the inhibition of recruitment of glycolytic and fermentation enzymes into VROs. The Rpn11-assisted recruitment of the cytosolic enzymes by p33, however, also requires the combined and coordinated role of the subverted actin network. Accordingly, stabilization of the actin filaments by expression of the Legionella VipA effector in yeast and plant, or via a mutation of ACT1 in yeast resulted in more efficient and rapid recruitment of Rpn11 and the selected glycolytic and fermentation enzymes into VROs. On the contrary, destruction of the actin filaments via expression of the Legionella RavK effector led to poor recruitment of Rpn11 and glycolytic and fermentation enzymes. Finally, we confirmed the key roles of Rpn11 and the actin filaments in situ ATP production within TBSV VROs via using a FRET-based ATP-biosensor. The novel emerging theme is that TBSV targets Rpn11 cytosolic protein interaction hub driven by the p33 replication protein and aided by the subverted actin filaments to deliver several co-opted cytosolic pro-viral factors for robust replication within VROs.

Tombusviruses co-opt many cytosolic host proteins into VROs via unknown mechanisms [8,40]. It would take a large number of p33 molecules, which are membrane associated, to be involved in the rapid and efficient recruitment of all these cytosolic proteins one-by-one to facilitate the formation of large VROs. It is likely that TBSV p33 molecules will need help accomplishing such major tasks. A possible way for p33 to do the major recruitment task of the numerous host cytosolic proteins into VROs is to target putative cytosolic "protein interaction hub" proteins, which associate with many other cytosolic proteins. Among the putative cytosolic hub proteins known to interact with p33 replication protein is the Rpn11 proteasomal protein, which was originally identified in a systematic screen with TBSV based on a temperature-sensitive library of yeast mutants [41,42]. The highly conserved Rpn11 (Regulatory Particle Non-ATPase, called POH1 or PSMD14 in humans) metalloprotease is part of the 19S regulatory particle, which constitutes the 26S proteasome lid [43]. Rpn11 essential function is to couple deubiquitination and degradation of proteasome substrates. In the presence of mutated Rpn11, polyubiquitinated proteins accumulate in yeast [44]. Proteasomes get degraded in the absence of Rpn11, making it essential for maintaining cellular protein homeostasis. An important function of Rpn11 is the formation of proteasome storage granules under certain cellular conditions, such quiescent stage in yeast [45]. However, Rpn11 is a major contributor to several pathways that are independent of its catalytic activity [44,46]. The N-terminal part of Rpn11 contains the deubiquitinase (DUB, the catalytically active JAMM/MPN+) domain, whereas the C-terminal domain regulates the stability of the proteasomal lid, cellcycle progression and mitochondrial fission and peroxisomal division [46,47]. Thus, mutations in the multifunctional Rpn11 might have pleiotropic effects on the cell and its essential role makes it more challenging to dissect its pro-viral function in viral replication. Importantly, previous works with TBSV revealed that the canonical function of Rpn11 in the proteasome is not required for its pro-viral function [42]. Rpn11 is recruited into the VROs and it is important in facilitating the subversion of the pro-viral cytosolic DDX3-like Ded1 RNA helicase (called RH20 in plants) [42].
Because Rpn11 is a major protein interaction hub in cells and it is known to interact with the actin network [48,49], in this work we explored the possible function of Rpn11 to facilitate the subversion of other cellular proteins into the TBSV VROs. This proposed function of Rpn11 may depend on the actin filaments, which are known to participate in the formation of TBSV VROs in yeast and plant cells [50]. The actin filaments are stabilized by TBSV via p33-based blocking of Cof1 (cofilin, also called actin depolymerization factor) function in disassembling actin filaments. The p33-mediated stabilized actin filaments then are used by TBSV to deliver vesicle cargoes, such as Rab5-decorated early endosomes and retromer tubular carriers into VROs to provide lipids/membranes and lipid enzymes for the biogenesis of VROs [51,52]. In this work we provide supporting evidence that the stabilized actin network is used by TBSV for delivering cytosolic proteins, such as glycolytic and fermentation enzymes needed for local ATP generation into the large VROs. Altogether, we propose that p33 replication protein does not deliver each subverted cytosolic host factor one-by-one into VROs. Instead, TBSV targets Rpn11, which then serves as a major cytosolic protein interaction hub by facilitating the recruitment of other cellular cytosolic factors with the contribution of subverted actin filaments into the VROs. Therefore, the current work shows that Rpn11 is not only acting as a "matchmaker" between the viral p92 pol and the co-opted cellular DDX3-like Ded1p (RH20 in plants) DEAD-box helicase [42]. The emerging model is that TBSV targets Rpn11 cytosolic protein interaction hub driven by the p33 replication protein and aided by the stabilized and subverted actin filaments to deliver several co-opted cytosolic pro-viral factors for robust replication within VROs.

Critical role of the cytosolic Rpn11 in assisting tombusviruses during recruitment of pro-viral glycolytic and fermentation enzymes into VROs
To test the concept that the subversion of several different cytosolic proteins by TBSV into the VROs might be facilitated by p33 replication protein via targeting a putative "cellular cytosolic protein interaction hub", we decided to decipher the pro-viral function of Rpn11 proteasomal deubiquinase factor in subversion of other host factors. Because Rpn11 physically interacts with a large number (~1,000) of yeast proteins [49,53], we decided to focus on the possible connection between Rpn11 and the host cytosolic glycolytic and fermentation enzymes, which are known to interact with Rpn11. The host cytosolic glycolytic and fermentation enzymes are readily subverted into VROs via p33 to produce ATP locally in support of VRO formation, VRCs assembly and viral RNA replication [19][20][21]54]. However, the recruitment mechanism/ pathway of various cytosolic proteins by a membrane-bound p33 into VROs is currently unknown.
Rpn11 is an essential protein, and therefore, we applied different approaches to manipulate Rpn11 availability for pro-viral functions. First, we knocked down Rpn11 mRNA level via VIGS in N. benthamiana (S1A Fig) followed by transient expression of p33 replication protein and three glycolytic enzymes and two fermentation enzymes, which are known pro-viral host factors [19][20][21]55]. The glycolytic enzymes included the ATP generating Pgk1 (phosphoglycerate kinase 1) and PK (pyruvate kinase, Cdc19 in yeast and PKM2/PKLR in humans) as well as Fba2 (fructose 1,6-bisphosphate aldolase), whereas the fermentation enzymes included Pdc1 (pyruvate decarboxylase 1) and Adh1 (alcohol dehydrogenase 1). These fermentation enzymes are required for the replenishing of NAD + , which is critical regulatory compound in sustaining aerobic glycolysis pathway [56,57].
We performed BiFC assays, which are suitable to determine protein-protein interactions and the subcellular location of the interactions if cellular markers are also co-expressed [58]. Based on the BiFC experiments, we did not observe interaction between p33 and Pgk1, PK1 and Fba2 glycolytic enzymes within the p33-induced VRO-like structures in N. benthamiana after VIGS treatment that knocked down Rpn11 mRNA level (Fig 1A and 1B and 1C). The BiFC results were comparable when the plants were also infected with cucumber necrosis virus (CNV, closely related to TBSV) to induce functional VROs (Fig 1). This is in contrast with the efficient p33-Pgk1, p33-PK1 and p33-Fba2 interactions within the large VROs decorated with the RFP-SKL peroxisomal marker in the control plants (Fig 1). We also observed the reduced level of peroxisome aggregation in the Rpn11 knockdown plants in comparison with the high level peroxisomal aggregation in the control plants, which is a characteristic feature of tombusvirus VROs (Fig 1) [59,60]. Pgk1, PK1 and Fba2 proteins were expressed in Rpn11 knockdown plants (S1B Fig). In addition, we observed the lack of detectable interaction via BiFC between p33-Pdc1 and p33-Adh1 fermentation enzymes in Rpn11 knockdown plants infected with CNV or mock-infected (Fig 2A and 2B). This is in contrast with the robust interactions between p33 and the fermentation enzymes within VROs in control plants (Fig 2A and  2B). Pdc1 and Adh1 proteins were expressed in Rpn11 knockdown plants (S1B Fig). Altogether, these data indicated that reduced expression of Rpn11 led to poor recruitment of selected glycolytic and fermentation enzymes by p33 into tombusvirus VROs in N. benthamiana plants.
To provide further evidence on the role of Rpn11 as the key regulator of recruitment of pro-viral cytosolic enzymes into VROs, we retargeted a bulk fraction of Rpn11 from the cytosol into the nucleus in N. benthamiana. This was achieved through incorporating a nuclear retention signal (NRS) into Rpn8 proteasomal protein, which is a strong interactor with Rpn11 in the proteasomal lid [61][62][63]. Originally, Rpn8 and Rpn11 are distributed in both cytosol and the nucleus (S2 Fig). GFP-NRS-Rpn8, however, was exclusively retained in the level affects the interaction between TBSV p33 replication protein and the cellular Pgk1. N. benthamiana plants were silenced with VIGS vector targeting a region of Rpn11 for 8.5 days. Then, the leaves were co-agroinfiltrated with BiFC plasmids pGD-nYFP-Pgk1 and pGD-p33-cYFP as well as pGD-RFP-SKL to express peroxisomal matrix marker to indicate VROs. Confocal images were taken 1.5 days after agroinfiltration. The control experiments included the TRV2-nMBP VIGS plasmid. The merged images show the colocalization of the BiFC signal with the peroxisomal marker, indicating the interaction between TBSV p33 replication protein and Pgk1 within VROs. The plants were also agroinfiltrated with pGD-CNV or pGD vector (as a control). (B-C) VIGS of Rpn11 mRNA level affects the interaction between p33 and the glycolytic PK1 or Fba2 enzymes. See more details in panel A. Scale bar is 10 μm. Each experiment was repeated. Expression of GFP-NRS-Rpn8 inhibited CNV replication by~3-fold, likely due to cosequestration of Rpn11 into the nucleus (Fig 3A). On the contrary, ectopic expression of GFP-Rpn8 did not affect CNV replication (Fig 3A). Confocal microscopy analysis revealed that RFP-Rpn11 was inefficiently recruited into VROs in N. benthamiana expressing GFP-NRS-Rpn8 and infected with CNV ( Fig 3B). This is in contrast with the efficient recruitment of RFP-Rpn11 into VROs expressing GFP-Rpn8 and infected with CNV ( Fig 3B). Next, we tested the recruitment of the glycolytic Pgk1 and
We observed similar inhibition of recruitment of the fermentation enzymes Pdc1 (Fig 4A), and Adh1 ( Fig 4D) into VROs in N. benthamiana expressing GFP-NRS-Rpn8 and infected with CNV. Both these fermentation enzymes are efficiently recruited into VROs in N. benthamiana expressing GFP-Rpn8 and infected with CNV ( Fig 4B and 4E). Overall, these data revealed that sequestration of Rpn11 into the nucleus via GFP-NRS-Rpn8 inhibited the recruitment of selected glycolytic and fermentation enzymes by p33 into tombusvirus VROs in N. benthamiana plants.
To confirm the above findings, we utilized a temperature-sensitive (ts) Rpn11 yeast strain [43,64]. The yeast His 6 -tagged Fba1 (a homolog of the plant Fba2, Fig 5B and 5C), His 6 -Pdc1 (Fig 5D and 5E) and His 6 -Adh1 (Fig 5F and 5G) were poorly co-purified with the Flag-tagged p33 and Flag-p92 pol , representing the tombusvirus replicase from detergent-solubilized membrane fraction of rpn11 ts yeast cultured at the semi-permissive temperature (i.e., 32˚C) in comparison with the WT yeast. However, the above host proteins were as efficiently co-purified with the tombusvirus replicase from rpn11 ts yeast cultured at the permissive temperature (i.e., 23˚C) as from WT yeast (Fig 5), albeit the amount of Flag-p33 expressed was slightly lower in the rpn11 ts yeast. Co-purification of His 6 -Pgk1 with the tombusvirus replicase from rpn11 ts yeast cultured at the semi-permissive temperature was also lower than from WT yeast ( Fig  5A). This was also observed with Tdh2 and Tdh3 NADH-producing glyceraldehyde-3-phosphate dehydrogenese (GAPDH, called Tdh2/3 in yeast) (S3 Fig). The enhanced co-purification of the pro-viral His 6 -RH2 DEAD-box helicase [12] with the tombusvirus replicase from rpn11 ts yeast cultured at the semi-permissive temperature shows that p33-based recruitment of not all cytosolic host proteins is dependent on Rpn11 (S3B Fig). Altogether, both the plantand yeast-based data strongly support the critical role of the cytosolic Rpn11 in assisting tombusviruses during recruitment of pro-viral glycolytic and fermentation enzymes from the cytosol into VROs.

The co-opted Rpn11 facilitates ATP production within tombusvirus VROs
The recruited glycolytic and fermentation enzymes are exploited by TBSV to produce ample amount of ATP locally within the VROs [19][20][21]54]. To confirm the key role of Rpn11 in ATP production in situ in TBSV VROs, we used a FRET-based ATP-biosensor [65], which was previously adapted to estimate ATP levels within VROs [20,21]. The ATP-biosensor is based on a fusion protein, linking the ATP-sensor module with p33 replication protein (called p33-ATeam) (Fig 6A). Upon binding to ATP, p33-ATeam increases FRET signal, which is measured by confocal laser microscopy. The localization of p33-ATeam to the VROs allows PLOS PATHOGENS for the estimation of ATP level within the VROs. We found that the TBSV VROs in Rpn11 knockdown plants produced~3 times less ATP within VROs than in the control N. benthamiana plants (Fig 6C and 6D versus 6B). These results connected the role of Rpn11 in recruitment of glycolytic and fermentation enzymes with ATP production within tombusvirus VROs in plant cells. The actin filaments play a key role in co-opting Rpn11 into tombusvirus replication and VRO formation Because Rpn11 interacts with the actin network in yeast cells [48,49], and the actin network is co-opted by tombusviruses to build the VROs [50], we decided to test the possible combined and coordinated role of Rpn11 and the subverted actin network in VRO biogenesis.
First, we applied a new approach to manipulate the actin network in plant cells infected with TBSV. This was based on two Legionella bacterium effectors, namely VipA and RavK, which alter the actin filaments differently. VipA is an actin nucleator, which promotes stable actin filaments [66,67]. On the other hand RavK is a protease, which cleaves off actin monomers from the actin filaments [68]. However, the cleavage by RavK results in a nonfunctional actin monomer that cannot be reused to build new actin filaments. This process thus leads to the destruction of most of the actin filaments in cells [68].
Transient expression of Legionella VipA in N. benthamiana leaves infected with TBSV resulted in the formation of the characteristic VROs decorated with p33-BFP and consisting of aggregated peroxisomes (decorated with RFP-SKL) ( Fig 7A). The sizes of VROs frequently looked larger than those VROs formed in plants infected with TBSV in the absence of VipA expression (Figs 7B, Z-stack images, and S4A). The actin filaments were abundant in TBSV-the same samples as above with anti-Flag antibody. Third panel: Western blot analysis shows the levels of Flag-p33 in total protein extracts detected with anti-Flag antibody. Fourth panel and Fifth panel: Western blots of His 6 -Pgk1 and His 6 -p33 in total protein extracts detected with anti-His antibody. Sixth panel: Coomassie-blue stained gel SDS gel of the total protein extracts as loading controls. (B-C) Co-purification of the glycolytic Fba1 enzyme with the viral replicase. WT and rpn11-14 ts yeasts co-expressed Flag-p33 and Flag-p92 replication proteins and the TBSV repRNA together with His 6 -Fba1. See further details in panel A. (D-G) Co-purification of yeast His 6 -Pdc1 and His 6 -Adh1 with the viral replicase complex from WT and rpn11-14 ts yeasts cultured at permissive and semi-permissive temperatures. See further details in panel A. Each experiment was repeated three times and standard error is calculated. https://doi.org/10.1371/journal.ppat.1009680.g005

PLOS PATHOGENS
actin filaments (Fig 7A). Interestingly, the p33 replication protein was still localized to the peroxisomes, which were not intensively aggregated in TBSV-infected cells, when RavK was expressed (Fig 7A). VipA and RavK expression affected the actin filament formation in the absence of viral components (Fig 7C and 7D). Based on these results, we suggest that RavK expression inhibits TBSV VRO formation via destruction of the actin filaments.
Testing the mitochondria-associated CIRV, we observed similar phenomenon, including (i) that the VipA-driven stabilization of actin filaments did not inhibit CIRV VRO formation and mitochondrial aggregation within VROs (Fig 7E), but frequently resulted in enlargedsized CIRV VROs when VipA is co-expressed in plants (S4C and S4D Fig). The control image is shown in Fig 7E, top image panel; (ii) the RavK-based destruction of the actin filaments greatly inhibited VRO formation and the p36 replication protein-driven aggregation of mitochondria in CIRV-infected plant cells (Fig 7E). Therefore, it seems that affecting the actin filaments by the Legionella VipA and RavK, respectively, influenced TBSV and CIRV VRO biogenesis in N. benthamiana.
To test the effectiveness of the Legionella VipA and RavK effectors in modulation of the actin filaments on tombusvirus replication, we measured TBSV and CIRV genomic (g)RNA accumulation in N. benthamiana leaves transiently expressing the effectors. Northern blot analysis revealed 2-to-6-fold increased accumulation of tombusviruses in N. benthamiana expressing VipA effector (Fig 8A and 8B). The severity of symptoms caused by tombusviruses was also enhanced in N. benthamiana expressing VipA (Fig 8A and 8B). VipA expression in yeast also increased TBSV and CIRV repRNA accumulation by~2-3-fold (Fig 8C and 8D). VipA mutant (i.e., N-VipA) lacking the C-terminal acting-binding domain [69] was ineffective in enhancement of TBSV and CIRV replication (Fig 8C and 8D). On the contrary, transient expression of RavK inhibited TBSV accumulation by~5-fold in N. benthamiana leaves ( Fig  8E) and~3-fold in yeast (Fig 8F). The leaves expressing RavK and used for the studies looked normal at the time of sampling (Figs 8E and S5A and S5B).
Next, using the above effector protein tools, we studied if the actin network is involved in facilitating the subversion of Rpn11 into tombusvirus replication complexes. First, we Flagaffinity purified the p33 and p92 pol replication proteins, which are the major components of the TBSV VRCs [31], from the detergent-solubilized membrane fraction of yeast replicating TBSV repRNA and co-expressing His 6 -tagged Rpn11 and 3xHA-tagged VipA or His 6 -RavK. Western blot analysis revealed a 2-fold increase in the amount of co-purified Rpn11 in yeast co-expressing VipA ( Fig 9A). Whereas, Rpn11 was barely detectable in the purified replicase preparation from yeast expressing RavK ( Fig 9B). Second, we confirmed that Rpn11 co-localized with the p33 and the actin filaments in plant cells ( Fig 9C). Third, we showed that expression of RavK blocked the recruitment of GFP-Rpn11 into TBSV VROs in N. benthamiana ( Fig  9D), whereas GFP-Rpn11 was efficiently recruited into VROs in the control plants ( Fig 9E). Expression of RavK did not affect GFP-Rpn11 distribution between the nucleus and cytosol in the absence of tombusviruses ( Fig 9F). All these results establish the critical role of the actin network in subversion of Rpn11 for pro-viral functions into tombusvirus VROs.

The actin filaments play a critical role in subversion of the cytosolic glycolytic and fermentation enzymes into tombusvirus VROs
Because recruitment of Rpn11 cytosolic protein interaction hub protein by tombusviruses depends on the actin network, as established above, and Rpn11 affects the recruitment of select group of cytosolic host factors into VROs, we assumed that modulating the activities of the actin network would have major effects on the subversion of cytosolic host factors into TBSV replication. Again, we decided to focus on the glycolytic and fermentation enzymes due their First, we transiently expressed RavK in N. benthamiana leaves and tested the interaction with p33 replication protein and the recruitment of glycolytic enzymes into VROs using BiFC. Interestingly, RavK expression inhibited the interaction between the glycolytic ATP-generating Pgk1 and PK1 (Fig 10) and Fba2 (Fig 11A and 11B) with the p33 replication protein within the VROs of CNV or TBSV. RavK expression also inhibited the recruitment of Pgk1 and PK1 into the CIRV VROs and the interaction with the p36 replication protein (Fig 10D and 10H). RavK expression led to reduction in sizes of VROs consisting of aggregated peroxisomes (marked with RFP-SKL) in case of TBSV and CNV infections as well as VROs from aggregated mitochondria in case of CIRV infection (Fig 10). Second, the transient expression of RavK inhibited the interaction with the replication proteins and subversion of Pdc1 and Adh1 fermentation enzymes into TBSV and CIRV VROs in N. benthamiana leaves (Fig 11). The Pgk1, PK, Fba2, Pdc1 and Adh1 proteins were expressed in N. benthamiana leaves also co-expressing RavK (S5C Fig). Third, we used yeast to purify the tombusvirus replicase from membrane fraction of yeast expressing the VipA effector. Western blot analysis of the co-purified host proteins revealed 2-3-fold increased levels of the glycolytic Pgk1, Cdc19 (PK), Tdh3 (GAPDH), and Pdc1 and Adh1 fermentation enzymes in the purified tombusvirus replicase preparation when yeast expressed VipA (Fig 12). On the contrary, low-level expression of RavK effector in yeast led tõ 2-fold reduction of Pdc1 and Adh1 levels in the purified tombusvirus replicase preparations (Fig 12F and 12G). We confirmed the increased recruitment of Cdc19 (PK) into TBSV replicase using a yeast actin mutant (act1 ts ), which results in stabilized actin filaments (Fig 12H) [50,70]. All these data demonstrated the key role of the actin filaments in recruitment of glycolytic and fermentation enzymes by tombusviruses.

The actin filaments affect ATP production within tombusvirus VROs
To confirm that the subverted actin filaments indeed important to deliver glycolytic and fermentation enzymes to produce ATP locally within the VROs, we again used the above a FRET-based ATP-biosensor approach [19][20][21]54]. The ATP-biosensor module was fused with p92 pol (called ATeam-p92 pol ) (Fig 13A). We found previously [20,21] that the ATeam-tagged p92 pol is a fully functional RdRp. ATeam-p92 pol localizes to the VROs allowing the estimation of ATP level within the VROs. We found that the TBSV VROs in act1 ts mutant yeast produced 4 times more ATP than in WT yeast at the semi-permissive temperature (Figs 13B and S6). In addition, a cofilin mutant yeast (cof1 ts ), which is partially deficient in actin filament depolymerization at semi-permissive temperature, also supported~3-fold increased ATP production within the TBSV VROs (Figs 13B and S6). Time-point experiment also showed the faster ATP production within TBSV VROs in act1 ts yeast at the semi-permissive temperature than in WT yeast (Fig 13). On the contrary, expression of the RavK effector in WT yeast inhibited ATP production within the TBSV VROs ( Fig 13G) and the CIRV VROs (Fig 13H) by~3-to-4-fold.

PLOS PATHOGENS
The role of the actin filaments in ATP production within tombusvirus VROs is also tested in N. benthamiana plants. Expression of RavK reduced ATP production within TBSV and CIRV VROs by~3-to-4-fold in N. benthamiana (Fig 14). Altogether, these results confirmed

Discussion
Biogenesis of VROs is a major activity within infected cells. VRO formation requires subversion of several membranous compartments and vesicles as well as numerous co-opted cytosolic host proteins [2,3,[5][6][7][8][71][72][73][74]. Indeed, tombusviruses co-opt cellular membranous carriers, such as retromer-based tubular carriers, COPII vesicles and their selected cargoes via p33-based targeting of cellular membrane proteins, such as Rab1 and Rab5 small GTPases or the retromer complex and delivering them to the VROs for various functions and membrane modifications and membrane proliferation [51,52,[75][76][77]. Another major function of p33 is the selection of the TBSV (+)RNA for replication and recruitment into VROs [32,78]. However, the efficient subversion of several dozens of different cytosolic host proteins by TBSV into the VROs by a single viral protein, p33 [79,80], raises the question: how can p33 perform so many recruitment tasks? This is in addition to performing major replication function by p33 within VROs as a structural component as well as RNA chaperone function, and the role in subversion of various membranes and membrane-bound host factors into the VROs [8,11,34]. One-by-one and rapid recruitment of every single cytosolic protein by the p33 replication protein would likely require a vast number of p33 molecules. Therefore, we wanted to explore if cellular cytosolic proteins might be recruited by p33 targeting of a protein interaction hub, namely the proteasomal Rpn11. Based on the presented data, we propose that Rpn11 acts as a cellular "cytosolic protein interaction hub", which is targeted by TBSV via p33-based direct binding to subvert numerous cytosolic proteins associated with Rpn11 into VROs. In addition to the previously characterized role of Rpn11 to facilitate the recruitment of DDX3-like Ded1/ RH20 DEAD-box RNA helicase into VROs [42], the current work expands the list of Rpn11-dependent co-opted host factors to four glycolytic [Pgk1, PK1 (Cdc19 in yeast), GAPDH (Tdh2/3) and Fba2 (Fba1)] and two fermentation enzymes (Pdc1 and Adh1). Using co-localization, BiFC and co-purification with the tombusvirus replicase complexes, we show the subversion of these cytosolic metabolic enzymes greatly depends on the cellular Rpn11 level or subcellular distribution or location of Rpn11, in addition to p33 replication protein.
Knocking down Rpn11 level via VIGS, or sequestering of Rpn11 away from the cytosol into the nucleus via a modified retargeted Rpn8 cellular interactor protein in plants, or using a temperature-sensitive Rpn11 mutant in yeast, all provided evidence on the key role of Rpn11 in subversion of the metabolic enzymes into VROs. Rpn11 physically interacts with the above metabolic enzymes in yeast in the absence of tombusviral components [53,81], possibly within proteasome storage granules, which form with the help of Rpn11 and predictably contain many Rpn11-interacting cytosolic proteins [45]. The glycolytic enzymes are also known to form "glycolytic or G" bodies [82,83], which might also contribute to their efficient recruitment by TBSV into VROs. It is important to note that Rpn11 is among the most involved glycolytic His 6 -Cdc19 (PK), His 6 -Tdh3 (GAPDH) and the His 6 -Pdc1 and His 6 -Adh1 fermentation enzymes with the viral replicase from yeast expressing 3xHA-VipA. See further details in panel A. (F-G). Co-purification of His 6 -Pdc1 and His 6 -Adh1 fermentation enzymes with the viral replicase from yeast expressing His 6 -RavK. See further details in panel A. (H) Flag-p33 and Flag-p92 replication proteins were expressed in WT and act1 ts yeasts together with His 6 -Cdc19 (PK) glycolytic enzyme. First panel: Western blot analysis of co-purified His 6 -Cdc19 with TBSV replicase from detergent-solubilized membrane fraction of yeast cultured at either permissive (23˚C) or semi-permissive (27˚C) temperatures. Co-purified His 6 -Cdc19 was detected with anti-His antibody. Second panel: Western blot shows Flag-affinity purified p33 in the same samples as above with anti-Flag antibody. Third panel: Western blots of His 6 -Cdc19 in total protein extracts detected with anti-His antibody. Fourth panel: Coomassie-blue stained gel SDS gel of the total protein extracts as loading controls. Each experiment was repeated three times and standard error is calculated. protein interaction hubs with documented physical interaction with~1,000 yeast proteins (~16% of the entire yeast proteome), many of them are pro-viral host factors [8,40,48,49]. We predict that the most abundant host proteins, which interact with Rpn11, will have the best chances to be recruited by p33 into VROs. The productions of many host proteins, including glycolytic and fermentation enzymes, are greatly induced by TBSV infection [19][20][21]55]. The increased amounts of the induced host proteins would likely favor their associations with Rpn11. Thus, they could have the more favorable circumstances for recruitment into VROs via assistance from Rpn11 than the less abundant host proteins. Altogether, we predict that not only glycolytic and fermentation enzymes, but several more abundant cytosolic pro-viral factors might be co-opted by p33 with the help of the Rpn11 cytosolic protein interaction hub protein. The pro-viral function of Rpn11 is likely independent from its proteasomal function. Indeed, blocking proteasomal activities with MG132 inhibitor led to increased p33 replication protein level in yeast under some conditions [84]. The putative role of the proteasome system in tombusvirus replication will require future studies.
The tested metabolic enzymes are components of the aerobic glycolysis pathway, which regulates the balance between fast ATP production and biosynthesis of new biomass, including ribonucleotides, lipids and several amino acids. Tombusviruses hijack these enzymes into VROs to support local and efficient production of ATP within VROs [19][20][21]54]. Their recruitments into VROs, however, apparently are affected by Rpn11.
Another major finding of this work is the involvement of the actin filaments in facilitating the TBSV p33-driven recruitment of Rpn11 cytosolic hub protein with the associated cytosolic proteins. Stabilization of the actin filaments by expression of the Legionella VipA effector in yeast and plant, or via mutation of ACT1 in yeast resulted in more efficient and rapid recruitment of Rpn11 and the selected glycolytic and fermentation enzymes into VROs. On the contrary, destruction of the actin filaments via expression of the Legionella RavK effector led to poor recruitment of Rpn11 and glycolytic and fermentation enzymes into VROs. Ultimately, the subverted actin filaments, stabilized by TBSV p33 replication protein or VipA effector in collaboration with Rpn11 were needed for the efficient and local production of ATP by the coopted glycolytic and fermentation enzymes within VROs representing the sites of TBSV replication in yeast and plant. Interestingly, the mitochondria associated CIRV utilizes a comparable mechanism of replication protein-driven targeting of the cytosolic Rpn11 and the actin network to deliver the glycolytic and fermentation enzymes into the VROs.
Altogether, the proposed recruitment concept for cytosolic proteins by TBSV somewhat resembles to the previously established mechanism of p33-driven subversion of targeted subcellular membranes or membrane subdomains by TBSV. For example, the small replication protein targets two membrane-associated cellular hubs, which include (i) the syntaxin18-like Ufe1 SNARE protein and Sac1 PI4P phosphatase-positive subdomains within the ER membrane, and (ii) the Rab5-positive early endosomes [51,[85][86][87]. The above TBSV-targeted cellular hub proteins are key parts of highly-active centers for various membrane trafficking in cells.
In summary, we present results that support a novel viral recruitment strategy for cytosolic host factors based on TBSV and CIRV. These viruses target via the small viral replication protein the cytosolic Rpn11 protein interaction hub protein and the co-opted and stabilized actin filaments. The combined and coordinated subversion of Rpn11 and the actin network allows tombusviruses to gain access to abundant cytosolic proteins, such as the glycolytic and fermentation enzymes, which are then efficiently delivered to perform pro-viral functions into the  Fig 6A). Comparison of relative ATP levels produced within the tombusvirus VROs in N. benthamiana expressing the Legionella RavK effector or the empty vector and p33-ATeam YEMK . The plants were infected with TBSV or mock-inoculated 16 h after agroinfiltration. FRET analysis was performed 1.5 dpi. The more intense FRET signals are white and red (between 0.5 to 1.0 ratio), whereas the low FRET signals (0.1 and below) are light blue and dark blue. We show the quantitative FRET values for a number of samples in the graph on the right. C-D). Comparable experiments in N. benthamiana expressing the Legionella RavK effector, but using the CIRV p36-ATeam YEMK for ATP measurement within CIRV VROs. FRET analysis was performed 2 dpi. See details in panel B-C. Scale bars represent 10 μm. Each experiment was repeated three times. https://doi.org/10.1371/journal.ppat.1009680.g014

PLOS PATHOGENS
VROs, which represent the site of viral replication. Accordingly, knockdown of Rpn11 or destruction of actin filaments diminishes VRO biogenesis and tombusvirus replication in yeast and plant cells. Other (+)RNA viruses with small number of genes might also exploit similar strategies to maximize the recruitment of host factors with the involvement of limited number of viral proteins into VROs [88].

Materials and methods
Additional Materials and Methods are presented in the supplementary information (S1 Text).

Plant and yeast expression plasmids
Plasmids used in this study are described in S1 Text.

Tombusvirus replication assays in yeast
To test the effect of the Legionella VipA effector on tombusvirus replication, BY4741 yeast was co-transformed with pGBK-CUP-Flagp33/Gal-DI72, pGAD-Cup-Flagp92 or pGBK-CUP-Flagp36/Gal-DI72, pGAD-Cup-Flagp95 and one of the following plasmids: pYC-NT vector, pYC-NT-VipA [75], pAG416GAL-ccdB-VipA and pYC-N-VipA. Transformed yeasts were pre-grown in SC-ULH − media supplemented with 2% glucose and BCS (from VWR) at 29˚C for 16 h. Yeast cultures were washed and grown in SC-ULH − supplemented with 2% galactose at 23˚C for 10 h. TBSV repRNA replication was induced by adding 50 μM of CuSO 4 for 24 h at 23˚C. To test the effect of Legionella RavK effector on tombusvirus replication, BY4741 yeast was co-transformed with pGBK-CUP-Flagp33/Gal-DI72, pGAD-Cup-Flagp92 and one of the following plasmids: pYC-NT vector, pYC-NT-RavK, pAG416GAL-ccdB-RavK [75]. Yeast cultures were grown as above. RNA samples were analyzed by northern blot using the 32 P-labeled DI72 RI/IV as a radioactive probe [90]. Total protein was extracted as described before [91] and protein samples were analyzed by western blot using anti-FLAG antibody to detect the viral replication proteins and anti-His antibody to detect His 6 -VipA and His 6 -RavK proteins.
Transformed BY4747-ADH-His92 yeast cultures were plated in SC-ULH − media, whereas transformed Sc1 yeast cultures were plated in SC-ULHT − media. Single colonies were streaked and grown in 20 ml of SC-ULH − (or SC-ULHT − in the case of Sc1 yeast) supplemented with 2% glucose and 100 μM BCS at 23˚C for 16 h. Yeast cultures were washed with sterile water and grown in 40 ml of SC-ULH − (or SC-ULHT − ) supplemented with 2% galactose and 50 μM of CuSO 4 at 23˚C for 24 h. Yeast cultures were harvested and incubated in 35 ml of Phosphate Buffer Saline (PBS) containing 1% formaldehyde for 1 h on ice [93]. Formaldehyde was quenched with 0.1 M of glycine and incubated for 5 min on ice. Yeast pellets were collected and washed with PBS buffer. Flag-p33 replication proteins were purified from detergent-solubilized membrane fraction using anti-FLAG M2 agarose as described before [92]. Briefly, 0.2 g of yeast pellet was re-suspended in 200 μl of High Salt TG Buffer (50 mM Tris-HCl, pH 7.5, 10% glycerol, 15 mM MgCl 2 , 10 mM KCl) with 0.1% of yeast protease inhibitor. Yeast cells were broken and centrifuged to discard the supernatant. Yeast pellet was solubilized in High Salt TG buffer with 2% Triton X100 and 0.1% of yeast protease inhibitor. The tubes were rotated for 8 h in the cold room. Then, the tubes were centrifuged at 35,000 g for 20 min and the supernatant was transferred to an equilibrated Bio-Rad Bio-Spin chromatography column with 20 μl of FLAG M2 resin (Sigma). The column was rotated for 16 h at 4˚C. The column was washed with High Salt TG Buffer and the Flag-p33 protein preparation was recovered from the column with 30 μl of SDS loading buffer. β-Mercaptoethanol was added to the eluted sample and boiled for 40 minutes to reverse the crosslink. Purified Flag-p33 preparations were analyzed by western blot and Flag-p33 protein was detected with anti-FLAG antibody. The copurified proteins were detected with anti-His antibody. Detection with NBIP-BCIP has been described previously [91]. Protein amounts were quantified with ImageQuant software. The quantifications were analyzed in excel and standard error was calculated.