An engineered mutant of a host phospholipid synthesis gene inhibits viral replication without compromising host fitness

Viral infections universally rely on numerous hijacked host factors to be successful. It is therefore possible to control viral infections by manipulating host factors that are critical for viral replication. Given that host genes may play essential roles in certain cellular processes, any successful manipulations for virus control should cause no or mild effects on host fitness. We previously showed that a group of positive-strand RNA viruses enrich phosphatidylcholine (PC) at the sites of viral replication. Specifically, brome mosaic virus (BMV) replication protein 1a interacts with and recruits a PC synthesis enzyme, phosphatidylethanolamine methyltransferase, Cho2p, to the viral replication sites that are assembled on the perinuclear endoplasmic reticulum (ER) membrane. Deletion of the CHO2 gene inhibited BMV replication by 5-fold; however, it slowed down host cell growth as well. Here, we show that an engineered Cho2p mutant supports general PC synthesis and normal cell growth but blocks BMV replication. This mutant interacts and colocalizes with BMV 1a but prevents BMV 1a from localizing to the perinuclear ER membrane. The mislocalized BMV 1a fails to induce the formation of viral replication complexes. Our study demonstrates an effective antiviral strategy in which a host lipid synthesis gene is engineered to control viral replication without comprising host growth.

Plant viruses are a major threat to stable agronomic production and cause multibillion-dollar losses each year. The planting of virus-resistant cultivars and application of insecticides to control virus-transmitting insects are common practices for controlling viral diseases in the field. However, pesticides cause pollution and leave chemical residue (1) and become ineffective when insects develop resistance. Breeding for crop cultivars with broad-spectrum and stable viral resistance is, therefore, crucial. Conventional breeding has incorporated many avail-able resistance (R) 4 genes into elite cultivars to control viral diseases (2)(3)(4)(5). R genes can be dominant or recessive. Dominant resistance is governed by a specific interaction between an R gene and a corresponding avirulence (Avr) gene from a pathogen to elicit a hypersensitive response (3,6). However, there are several issues related to dominant resistance. One is that high mutation rates during viral infection can generate Avr gene derivatives that escape recognition and thus overcome the resistance. This is particularly true for RNA viruses whose RNA-dependent RNA polymerases lack proofreading activity. Another issue is strain specificity; although R gene products can recognize Avr gene products from several virus strains, they are unable to control many other strains (5).
Recessive resistance is usually conferred by a mutated form of a host gene whose WT form encodes a protein that is critical for viral infection (7,8). Mutants of these genes do not support viral infection but cause no, or only mild, growth phenotypes in virus-resistant plants as the mutants retain their ability to support normal cellular processes. This resistance mechanism is also likely to work in a wide range of crops. For instance, mutated forms of either eukaryotic translation initiation factor 4E (eIF4E) or eIF4(iso)E have been reported as a recessive R gene in multiple crops, including barley, lettuce, melon, pepper, pea, and tomato (for a review, see Ref. 9). Additionally, recessive R gene-mediated resistance is durable because the mutation rate of host DNA-dependent DNA polymerase is much lower than viral RNA-dependent RNA polymerase (5).
Being the largest among seven viral classes, positive-strand RNA viruses ((ϩ)RNA viruses) include numerous pathogens that infect and cause severe consequences in humans, animals, and plants. Human illness-causing viruses include severe acute respiratory syndrome coronavirus, Zika virus, Dengue virus (DENV), and foot-and-mouth disease virus among many oth-ers. Importantly, the vast majority of plant viruses are (ϩ)RNA viruses. All (ϩ)RNA viruses have a small genome, encode a limited number of viral proteins, and rely heavily on host factors to achieve a successful infection (10 -14). Numerous host factors involved in the replication of hepatitis C virus (HCV) (13,15,16), picornavirus (14), brome mosaic virus (BMV) (12,17), cucumber mosaic virus (18), turnip mosaic virus (10,19,20), and tomato bushy stunt virus (TBSV) (11) have been identified and characterized. In theory, loss or mutation of host genes required for viral replication should provide recessive resistance to (ϩ)RNA viruses. It is thus crucial to identify, characterize, and manipulate host genes required for viral replication to develop virus-resistant hosts.
BMV, which serves as a model system to study (ϩ)RNA viruses, is the type member of the Bromoviridae family and a representative member of the alphavirus-like superfamily (21). The genome of BMV is composed of RNA1, RNA2, and RNA3. RNA1 and RNA2 encode replication proteins 1a and 2a polymerase (2a pol ), respectively (21). RNA3 encodes the movement protein 3a and the coat protein. BMV 1a and 2a pol are necessary and sufficient for its replication in barley, its natural host, and in an alternative host, the bakers' yeast Saccharomyces cerevisiae (21). BMV replication in yeast cells duplicates nearly all major features of its replication in plant cells (21) and enables systematic screenings of host genes that are involved in viral infection (17). About 100 host genes have been identified to be involved in BMV replication (17), including a group of genes involved in lipid metabolism (22)(23)(24)(25)(26). This is consistent with the fact that (ϩ)RNA viruses assemble their viral replication complexes (VRCs) in a tight association with remodeled host intracellular membranes. For example, deleting the ACB1 gene, which encodes acyl-CoA-binding protein that is involved in the trafficking of long-chain acyl-CoA esters among organelles (27,28), affects host lipid homeostasis and results in a more than 10-fold decrease in BMV genome replication (22). Additionally, a single point mutation in OLE1, which encodes a cellular ⌬9 fatty acid desaturase that converts saturated FA to unsaturated FA, causes a mild reduction in host unsaturated FA levels but inhibits BMV RNA replication by 20-fold (23,24). On the contrary, deleting the PAH1 gene, which encodes for Pah1p (phosphatidic acid phosphohydrolase) and is the sole yeast ortholog of human LIPIN genes, results in a 3-fold increase of BMV replication (26). The requirement of and high sensitivity of viral replication to host lipid composition is a common feature shared by numerous other (ϩ)RNA viruses, including HCV (29,30), West Nile virus (31), red clover necrotic virus (32), and TBSV (33)(34)(35). This highly conserved feature of viral genome replication raises the possibility of developing a novel, durable, and broad-spectrum strategy to control (ϩ)RNA viruses by manipulating host genes involved in lipid metabolism.
Phospholipids are major components of cellular membranes. Phosphatidylcholine (PC) in particular is important as it accounts for up to 50% of the total phospholipids in cells (36). A significant increase in PC levels is associated with the replication of multiple (ϩ)RNA viruses, including DENV (37), flock house virus (38), and poliovirus (39) among others. BMV also significantly increases PC content in yeast and barley cells (25). It was further shown that PC was specifically enriched at the sites of BMV replication (25). PC enrichment at the sites of viral replication was also present in HCV-and poliovirus-infected cells, indicating a conserved feature among a group of (ϩ)RNA viruses (25,40). In yeast cells, Cho2p (choline-requiring 2), a phosphatidylethanolamine (PE) methyltransferase (PEMT) that converts PE to monomethyl PE (PMME) (41), is recruited to viral replication sites by an interaction with BMV 1a (25). The increase in PC levels and the recruitment of Cho2p to the viral replication sites suggest that the enhanced PC content is synthesized at the sites of BMV replication. Moreover, disrupting PC synthesis by deleting the CHO2 gene results in a decrease of up to ϳ80% of BMV RNA synthesis (25), highlighting the critical role of PC in viral replication. However, deleting CHO2 also disrupts general PC synthesis and thus affects host cell growth (42), preventing the deletion of CHO2 from being a promising antiviral strategy.
We now report the engineered manipulation of CHO2 that leads to an inhibition of BMV RNA replication without affecting cell fitness. Specifically, we introduced substitutions in CHO2 to make the Cho2p-aia mutant in which glycine 102 and glycine 104 were replaced by an alanine residue. The Cho2p-aia functions as well as WT Cho2p in supporting host cell growth and general PC synthesis but does not support BMV RNA replication. We further demonstrate that the Cho2p-aia mutant disrupts the proper localization of BMV 1a and the formation of VRCs. Our data provide a proof of concept for a novel antiviral strategy by manipulating a host lipid metabolism gene to mistarget the viral replication protein away from the proper viral replication sites and prevent the formation of VRCs.

The Cho2p-aia mutant complements the defect of yeast cell growth in cho2⌬ cells
The biosynthesis of PC in yeast cells in the absence of exogenous free choline proceeds via the cytidine diphosphodiacylglycerol (CDP-DAG) pathway. Cho2p is a key enzyme in the CDP-DAG pathway where it catalyzes the addition of a methyl moiety to PE to produce PMME (Ref. 41) and Fig. 1A). PMME is further converted to PC by Opi3p (overproducer of inositol 3) (Fig. 1A), also known as phospholipid methyltransferase (41). Even though deleting CHO2 inhibits BMV RNA synthesis substantially (25), it also affects host cell growth (25,42). We set out to identify CHO2 mutants that disrupt viral replication without affecting host cell growth. To this end, we made various deletions based on structural analysis of Cho2p using Pfam (43) and Protein Homology/Analogy Recognition Engine V2.0 (PHYRE2) (44) programs. Cho2p has two PEMT domains and a skeletal muscle-and kidney-enriched inositol phosphatase carboxyl homology carboxyl homology (SKICH) domain (Fig.  1B). We found that the N-terminal 466 amino acids or 365 amino acids of Cho2p (F1 and F2 fragments; Fig. 1B), without the second PEMT domain and SKICH domain, were sufficient to support WT level of cell growth. Deleting the first 108 amino acids, however, abolished the ability of F1 to support WT-level cell growth (F3 fragment; Fig. 1B). Searching the first 108 amino acids, a putative GXG motif (GIG; amino acids 102-104) was identified. The GXG motif was reported to play a critical role in Targeting host lipid synthesis gene for virus control binding the methyl donor, N-adenosylmethionine, to PEMT in human cells (45). We constructed a Cho2p mutant with two alanine substitutions for glycines 102 and 104, referred to as the Cho2p-aia mutant (Fig. 1C).
In synthetic defined medium containing galactose as the carbon source without supplemented choline, we found that the doubling time of WT cells was ϳ4.7 h/generation (Fig. 1D). The doubling time of cho2⌬ cells with the expression of a negative control, an ER-resident protein Dpm1p (dolichol phosphate mannose synthase), was 6.5 h/per generation, which was significantly slower than that of WT cells (Fig. 1D, labeled as NC). The presence of HA-tagged WT Cho2p, nevertheless, improved cell growth and led to a doubling time of 5.3 h/generation (Fig. 1D). The presence of Cho2p-aia-HA improved cell growth of cho2⌬ cells. The doubling time became ϳ4.8 h/generation, which was similar to that of WT cells. We also confirmed that both HA-tagged WT Cho2p and Cho2p-aia accumulated to similar levels as determined via Western blotting (Fig. 1D). Pgk1p (phosphoglycerate kinase) was used as a loading control to show the equal loading of total proteins. We concluded that Cho2p-aia functions as well as WT Cho2p in supporting cell growth (Fig. 1D).

The Cho2p-aia mutant largely restores the PC synthesis in cho2⌬ cells
The restored growth rate of cho2⌬ cells in the presence of the Cho2p-aia mutant suggests that Cho2p-aia may function as well as WT Cho2p in supporting PC synthesis. To confirm this assumption, total lipids were extracted from WT and mutant cells and subjected to MS. We quantified phospholipid species by comparing with a defined amount of phospholipid standards and reported the mol % of each of the phospholipid classes in Fig. 2. In cho2⌬ cells with the expressed negative control Dpm1p (labeled as NC in Fig. 2), there was a 2-fold decrease in PC levels compared with WT cells (27% compared with 55%). Expressing CHO2-aia in cho2⌬ cells (cho2⌬ ϩ Cho2p-aia) increased the percentage of PC levels to 46%, which was significantly higher than that of negative control (p Ͻ 0.05) and not statistically significantly different from that in WT cells. The levels (21%) of phosphatidylinositol (PI) in cho2⌬ cells were significantly higher than those of WT (16%). The expressed Cho2p-aia largely restored the PI levels to 17%, which is not significantly different from the WT levels ( Fig. 2). For PE, which is the substrate of PC, deleting CHO2 had a dramatic effect as PE levels increased to 42% of total phospholipids from the 16% in WT cells. Expressing CHO2-aia partially restored PE levels because it decreased PE levels to 27%, which was significantly lower from that measured in cells lacking CHO2 but still higher than WT levels ( Fig. 2, comparing NC and Wt). Similarly, decreased levels of phosphatidylserine (PS) was not restored to WT levels when Cho2p-aia was expressed in cho2⌬ cells. Moreover, there was a significant decrease in phosphatidic acid (PA) levels in the presence of Cho2p-aia over the WT levels.

Targeting host lipid synthesis gene for virus control
Reasons for the decrease of PA levels and the partial restoration of alterations in PE and PS levels by Cho2p-aia are not clear. Nevertheless, our data suggest that Cho2p-aia is able to largely restore general PC synthesis for cell growth. Our data also indicate that cell growth is not noticeably affected by the moderate differences in phospholipid composition ( Figs. 1 and 2).

The Cho2p-aia mutant fails to rescue the BMV replication defect in cho2⌬ cells
Given that the Cho2p-aia mutant supported WT-level cell growth and largely restored PC content to WT levels in cho2⌬ cells, we next checked whether Cho2p-aia is able to complement the viral RNA replication defect. In the engineered BMV yeast system (46), BMV 1a, 2a pol , and RNA3 are expressed from plasmids to provide the necessary components required for BMV RNA replication (Fig. 3A). In WT yeast cells, negativestrand RNA3 and positive-strand RNA4 were detected using BMV RNA strand-specific probes, indicating full replication (Fig. 3B). In cho2⌬ cells with the expressed Dpm1p, which is not known to be involved in BMV replication and is used as a negative control, accumulated positive-strand RNA4 and negativestrand RNA3 were about 17 and 24% of WT levels, respectively. Expressing CHO2-aia did not improve BMV replication as positive-strand RNA4 and negative-strand RNA3 levels were similar to those in cho2⌬ cells expressing DPM1 (Fig. 3B), indicating that Cho2p-aia failed to complement the BMV replication defect in cho2⌬ cells.
To identify specific reasons why BMV replication defects were not complemented by Cho2p-aia, we first checked the accumulation of BMV 1a and 2a pol , which are necessary and sufficient for BMV replication. As shown in Fig. 3C, there was no significant difference in accumulated levels of 1a and 2a pol in these cells, ruling out the possibility that the inhibited BMV genome replication was due to negative effects on expression and/or stability of the viral replication proteins.

The Cho2p-aia mutant interacts with BMV 1a
Because the Cho2p-aia mutant failed to complement the defective BMV RNA synthesis, we wanted to confirm whether Cho2p-aia was stably expressed during viral replication. Western blotting determined that levels of HAtagged Cho2p-aia and WT Cho2p were similar (Fig. 4A), indicating that Cho2p-aia-HA is stably expressed and accumulates in the absence (Fig. 1D) or presence (Fig. 4A) of BMV replication.
Cho2p interacts with and is recruited by BMV 1a to the sites of viral replication to facilitate BMV genome replication (25). It is possible that Cho2p-aia fails to interact with 1a and therefore is not recruited to the VRCs. To confirm or reject this notion, we performed a coimmunoprecipitation (co-IP) assay by coexpressing HA-tagged Cho2p, Cho2p-aia, or Dpm1p with His 6tagged 1a in yeast cells. Similar to Fig. 4A, Cho2p-HA and and 2a pol were expressed from plasmid pB12VG1 under the control of the GAL1 and GAL10 promoters, respectively. BMV RNA3 was launched from plasmid pB3VG128 under the control of the copper-inducible CUP1 promoter, but no copper was added to the growth media. B, BMV replication in WT cells or in cho2⌬ cells with the expressed Dpm1p or Cho2p-aia. BMV positive-and negative-strand RNAs were detected by 32 P-labeled BMV strand-specific probes. 18S rRNA served as a loading control. The blot detecting (Ϫ)RNA3 was exposed for a longer time than that of (ϩ)RNAs. C, BMV 1a and 2a pol accumulation in WT cells or in cho2⌬ cells with the expressed Cho2p-aia or Dpm1p. Pgk1p served as a loading control.

Targeting host lipid synthesis gene for virus control
Cho2p-aia-HA accumulated to similar levels in the presence of BMV 1a (Fig. 4B, Total). As expected from our previous work, precipitating Cho2p-HA with anti-HA antibody pulled down 1a-His 6 (Ref. 25 and Fig. 4B). On the contrary, 1a-His 6 was not coprecipitated with Dpm1p-HA, which was used as a negative control and is not expected to interact with BMV 1a (25). To our surprise, 1a-His 6 was also pulled down along with Cho2paia-HA (Fig. 4B), indicating that Cho2p-aia-HA is still associated or interacts with BMV 1a.

Cho2p-aia prevents the localization of BMV 1a at the perinuclear ER membrane
BMV 1a is associated with the perinuclear ER (nER) membrane where it invaginates the outer nER membrane to form VRCs (12). BMV 1a recruits host proteins such as reticulon homology proteins (47), Snf7 (sucrose non-fermenting 7) (48), and Cho2p (25) to assemble functional VRCs. Like the majority of lipid enzymes, Cho2p is normally localized in ER membranes where lipids are synthesized. In agreement with our previous results (25) and as shown in Fig. 5A, Cho2p-HA was detected by immunofluorescence microscopy and appeared as a two-ring localization pattern in 43.1% of cells (n ϭ 153). The larger ring represents peripheral ER membranes, and the smaller ring surrounding the DAPI-stained nucleus corresponds to the nER membrane (Fig. 5A, upper  left panels). In the presence of 1a, Cho2p is depleted from peripheral ER membranes, enriched in the nER membrane, and colocalized with 1a-His 6 (59 of a total of 178 cells) (Fig.  5A, upper right panels).
To determine whether Cho2p-aia can be relocalized by 1a, we first expressed HA-tagged Cho2p-aia in the absence of BMV replication in cho2⌬ cells and tested its distribution. To our surprise, among all cells (n ϭ 326) in which we can detect Cho2p-aia-HA, only ϳ6 of the total 326 cells (1.8%) had a tworing distribution pattern, and in sharp contrast, Cho2p-aia-HA was found as puncta in about 72% of cells (Fig. 5A, lower left  panels). Intriguingly, when Cho2p-aia-HA and 1a-His 6 were expressed simultaneously, the localization of 1a was altered. BMV 1a-His 6 was localized to the nER membrane in only ϳ3.7% cells (12 of 328 cells) but as puncta in 73.2% cells where we can clearly detect signals of 1a-His 6 , Cho2p-aia-HA, and DAPI (Fig. 5A, lower right panels). It should be noted that even though localization patterns of both Cho2p-aia and 1a changed, they colocalized with each other in all cells where both can be detected, consistent with our co-IP results that Cho2p-aia is associated or interacts with 1a (Fig. 4B).
To better and easily characterize the nature of punctate structures of BMV 1a in the presence of Cho2p-aia, we used an mCherry-tagged 1a, which we have previously shown to have a similar localization pattern as that of 1a-His 6 in WT cells (49).
In cho2⌬ cells, about 35.5% of cells (n ϭ 262) expressing 1a-mCherry have a ring structure in the nER membrane (Fig.  5B, left panel) similar to that in WT cells (25,49). The distribution pattern of 1a-mCherry was also similar to that of 1a-His 6 in the presence of Cho2p-aia as it was associated with punctate structures in 74.6% (n ϭ 386) of cells (Fig. 5, A and B). To determine whether BMV 1a was being redistributed to a different membrane in the presence of Cho2p-aia, we coexpressed 1a-mCherry, Cho2p-aia, and GFP-tagged organelle markers. By comparing 1a-mCherry distributions with those of organelle markers, we found that the majority of puncta did not colocalize with the inclusion body marker von Hippel-Lindau (VHL) (50) or the Golgi marker Sed5p (suppressor of Erd2 deletion 5) (51) (Fig. 5C). In contrast, these dots were colocalized with the ER marker Scs2p (suppressor of choline sensitivity 2) (52) in peripheral ER membranes in about 49% of cells that we counted (n ϭ 123), and only 2.3% of cells have 1a-mCherry localized in the nER membrane (Fig. 5C). Taken together, our results indicate that Cho2p-aia disrupts the normal distribution of 1a to the nER membrane; however, 1a is still primarily associated with ER membranes.

The Cho2p-aia mutant prevents the formation of BMV VRCs
BMV 1a invaginates the outer nER membrane into the lumen to form spherular VRCs during BMV replication (12,53). These spherular compartments are about 60 -80 nm in diameter with an ϳ10-nm neck connecting to the cytoplasm (53,54). The abundance and size of these VRCs can change significantly with deletion or mutation of host genes involved in membrane shaping or lipid metabolism (22,47,48). We previously reported that, in cho2⌬ cells, 1a was detected at the nER membrane, and spherular VRCs were formed, but they were larger than those in WT cells and were not functional in genome replication (25). Because the expressed Cho2p-aia affected 1a's association with

Targeting host lipid synthesis gene for virus control
the nER membrane (Fig. 5), we predicted that VRC formation could be affected. To confirm this, we checked the formation of VRCs by transmission EM in WT cells or cho2⌬ cells expressing CHO2-aia in the presence of 1a, 2a pol , and RNA3. In WT cells with BMV components, the percentage of cells with spherular VRCs was ϳ16% (11 of 67 cells). The average number of spherules was 6 (S.D. ϭ 1.8) with an average diameter of 65 Ϯ 8 nm (Fig. 6A). However, among the 100 cells checked, no spherical structures were observed in cho2⌬ cells expressing the BMV replication components and Cho2p-aia. Of note, the inner and outer nER membranes were clearly observed in these cells (Fig. 6B). The failure to identify spherular VRCs is consistent with our results that Cho2p-aia inhibited the proper targeting of BMV 1a to the nER membrane in cho2⌬ cells (Fig. 5).

Discussion
Positive-strand RNA viruses have limited protein-coding capacity and replicate in a tight association with host intracellular membranes to form the viral replication complexes. As major components of cellular membranes, lipids play critical roles in the replication of multiple (ϩ)RNA viruses (55)(56)(57)(58). For example, PI4P plays a key role in the replication of multiple (ϩ)RNA viruses, including HCV (29, 30, 59 -61) and enteroviruses (29). Poliovirus and several enteroviruses promote an enrichment of free cholesterol to the sites of their replication (62,63). Disrupting sterol biosynthesis and composition inhibits the replication of West Nile virus and DENV (64,65). Among plant viruses, red clover necrotic mosaic virus promotes PA production by hijacking phospholipase D to the sites of its replication (32). TBSV and carnation Italian ringspot virus transport PE to viral replication sites to build the PE-enriched VRCs for their replication (33,66). For BMV, whose replication also requires balanced lipid homeostasis, phospholipids PC and PA are particularly important (22)(23)(24)(25)(26). Thus, dissecting the role of phospholipids in BMV replication and, in turn, manipulating phospholipid metabolism may potentially provide antiviral strategies with a broad range and durable resistance. A major challenge for engineering lipid synthesis genes to control viruses and other pathogens, however, is that lipids also play crucial roles in cellular processes. Any manipulation of lipid synthesis genes will potentially affect host growth. Here, we generated a mutant, Cho2p-aia, which largely complemented PC synthesis and host growth defects but failed to support BMV replication (Figs. 1-3). Although the Cho2p-aia mutant still interacted with BMV 1a (Fig. 4), it disrupted the normal distribution of BMV 1a (Fig. 5) and thus affected the formation of BMV VRCs (Fig. 6).
All well-studied (ϩ)RNA viruses assemble their VRCs in association with various organelle membranes. As such, the trafficking of viral replication proteins to their destination organelles is the first and a critical step for VRC formation. Without transmembrane domains, BMV 1a is associated with the nER membrane via an amphipathic ␣-helix domain, helix A (67, 68). Mutations in helix A result in two major phenotypes: 1a either no longer localizes to the nER, or it induces the formation of smaller, abundant, nonfunctional VRCs compared with WT 1a (69). Besides 1a's helix A domain, host genes also regulate 1a's distribution. For example, coat protein complex II vesicle cargo receptor Erv14p (ER-vesicle protein of 14 kDa) is required for targeting 1a to the nER membrane (49). In cells lacking ERV14, 1a no longer associates with the nER membrane but is distributed in punctate dots and large clusters that are located at peripheral ER membranes (49). However, the mechanism by which Erv14p and coat protein complex II vesicles regulate 1a's localization remains unclear. Expressing CHO2aia resulted in a similar phenotype as the majority of 1a localized in peripheral ER-associated punctate dots (Fig. 5). It is unclear whether the punctate structures present in erv14⌬ cells or cells with the Cho2p-aia mutant are related or formed in a similar fashion.
It should be noted that 1a interacts (Fig. 4B) and completely colocalizes with the Cho2p-aia mutant in puncta (Fig. 5A). Because Cho2p-aia interacts with 1a and localizes to punctate structures when expressed alone, it is most likely that Cho2paia recruits 1a to these dot structures. If this is the case, it is contrary to WT Cho2p, which is recruited to the nER membrane-localized viral replication sites by 1a via their interaction ( Fig. 5A and Ref. 25). Alternatively, although Cho2p-aia largely complemented PC defects in cho2⌬ cells, the phospholipid composition was not identical to that in WT cells: levels of PE were significantly higher and levels of PA and PS were lower than WT levels (Fig. 2). These lipid compositional changes might affect the physical properties of membranes, including thickness, intrinsic curvature, and fluidity. Such changes may affect membrane fusion and fission or the conformation of membrane-associated/integral proteins (70 -72), including ER membrane-localized Cho2p-aia and 1a. It is also remotely possible that both 1a and Cho2p-aia are similarly affected and targeted to punctate structures independently. However, the localization of another ER-resident protein, Scs2p, was not affected, and a two-ring localization pattern was still observed

Targeting host lipid synthesis gene for virus control
in the presence of Cho2p-aia (Fig. 5C), arguing against a similar effect on conformation and/or localization of all ER membrane proteins. Cho2p is involved in converting PE to PC. In the presence of Cho2p-aia, we observed close-to-WT levels of PC, but PE contents were still significantly higher than WT levels (Fig.  2). PE is primarily synthesized by phosphatidylserine decarboxylase from PS in mitochondria and then transported to ER membranes where PE is methylated first by Cho2p (41). It is possible that the mislocalization of Cho2p-aia from entire ER membranes to puncta may affect the access of PE by Cho2p-aia and as such leads to enhanced levels of PE compared with WT levels.
All (ϩ)RNA viruses depend entirely on host factors such as proteins or lipids to complete their life cycles. In theory, host genes required for viral infection (also termed susceptibility genes) (73) can be potentially targeted to develop viral resistance, similar to recessive R genes in nature. The following are some advantages of engineering host genes to achieve virus resistance. 1) Viral resistance caused by the loss of or mutation(s) of a host gene is durable because host genes are less likely to mutate than viral genes under infections. 2) Multiple genes can be targeted in combination to achieve a better or complete control of viral infection. 3) The same or similar manipulation of a specific gene can be applied to several crops. However, these host genes may play critical roles in cellular processes, and in turn, alterations of such host genes may lead to side effects on host survival. This is true for CHO2, whose deletion inhibited BMV replication more than 5-fold but also caused host cell growth defects (Ref. 25 and Figs. 1-3). However, cells expressing CHO2-aia grew as well as WT cells but still inhibited BMV replication by ϳ5-fold. This provides proof-of-principle evidence that host lipid genes can be engineered to control (ϩ)RNA virus genome replication without compromising host fitness. We note that the deletion of or mutations in CHO2 substantially inhibited but did not eliminate BMV replication, suggesting that BMV may acquire PC through the Kennedy pathway, which is a secondary PC synthesis pathway in yeast but the primary pathway in higher eukaryotes. In addition to manipulating CHO2, targeting another component(s) of the Kennedy pathway or other related processes coordinately may provide complete blockage of BMV replication.
In summary, we identified a CHO2 mutant, Cho2p-aia, that supports general PC synthesis required for cell growth but fails to support BMV replication. The Cho2p-aia mutant prevented BMV replication protein 1a from localizing to the perinuclear ER membrane and as such blocked the formation of VRCs and inhibited viral replication. This indicates that engineering host lipid synthesis genes can be an effective way of controlling viral replication without compromising host fitness.

Lipid analysis
Ten OD 600 units of yeast cells were harvested, and total lipids were extracted as described previously (74). Total lipids were reconstituted in chloroform:methanol:water (65: 35:8, v/v/v; 500 ul) with the addition of the phospholipid internal standard mixture purchased from Kansas Lipidomics Research Center. The phospholipid compositions were analyzed by a Waters I-class UPLC interfaced with a Waters Synapt G2-S mass spectrometer (Waters) using conditions as described (75). Data normalization to the internal standards for mol % calculations was performed as described previously (76).

RNA extraction and Northern blotting
Yeast cells were harvested, and total RNA was extracted by a hot phenol method (77). Equal amounts of total RNA were prepared for Northern blot analysis. 32 P-Labeled probes specific to BMV positive-or negative-strand RNAs or 18S rRNA were used for the hybridization. 18S rRNA was used as a loading control to eliminate loading variations. Radioactive signals of BMV positive-, negative-strand, or 18S rRNA were scanned using a Typhoon FLA 7000 phosphoimaging system, and the intensity of signals was quantified using ImageQuant TL software (GE Healthcare).

Protein extraction and Western blotting
Two OD 600 units of yeast cells were harvested, and total proteins were extracted as described previously (49). Equal volumes of total proteins were analyzed using SDS-PAGE and transferred to a polyvinylidene difluoride membrane. Rabbit

Coimmunoprecipitation assay
Ten OD 600 units of yeast cells were harvested, and the co-IP assay was performed as described previously (49). Briefly, harvested cells were lysed in radioimmunoprecipitation assay buffer (50 mM Tris at pH 8.0, 1% Nonidet P-40, 0.1% SDS, 150 mM NaCl, 0.5% sodium deoxycholate, 5 mM EDTA, 10 mM NaF, 10 mM NaPP i , and protease inhibitor mixture). Cell debris was removed, and the supernatant was mixed with Protein A-Sepharose beads and anti-HA pAb overnight at 4°C. Beads were washed three times with radioimmunoprecipitation assay buffer, resuspended in 1ϫ SDS gel-loading buffer, and boiled for 10 min. Samples were resolved by SDS-PAGE followed by Western blotting with anti-His 6 mAb and anti-HA pAb to detect target proteins.

Immunofluorescence assay
Two OD 600 units of yeast cells were harvested and fixed with 4% (v/v) formaldehyde for 30 min at 30°C. After removing the cell wall using lyticase, spheroplasts were permeabilized with 0.1% Triton X-100 and incubated with specified primary antibodies (anti-His 6 mAb or anti-HA pAb at 1:100 dilution) overnight at 4°C followed by an incubation with secondary antibodies (1:100 dilution) for 1 h at room temperature. Secondary antibodies were Alexa Fluor 488 -conjugated anti-mouse (Thermo Fisher Scientific, catalog number A11001) and Alexa Fluor 594 -conjugated anti-rabbit antibodies (Jackson Immu-noResearch Laboratories, catalog number 711-545-152). The nucleus was stained with DAPI (Vector Laboratories, catalog number H-1200) for 10 min. Images were captured using a Zeiss epifluorescence microscope at the Fralin microscopy facility, Virginia Tech.

EM
Ten OD 600 units of yeast cells were harvested. Fixation, dehydration, and embedding were performed as described previously (25). Images were captured using a JEOL JEM 1400 transmission electron microscope at the Virginia-Maryland College of Veterinary Medicine, Virginia Tech.

Statistical analysis
One-way ANOVA analysis was used to compare phospholipids or doubling time in WT and mutant cells. Error bars represent the standard deviation.