ABSTRACT
Rhinoviruses (RVs) replicate on cytoplasmic membranes derived from the Golgi apparatus. They encode membrane-targeted proteins 2B, 2C, and 3A, which control trafficking and lipid composition of the replication membrane. The virus recruits host factors for replication, such as phosphatidylinositol 4 (PI4)-kinase 3beta (PI4K3b), which boosts PI4-phosphate (PI4P) levels and drives lipid countercurrent exchange of PI4P against cholesterol at endoplasmic reticulum-Golgi membrane contact sites through the lipid shuttling protein oxysterol binding protein 1 (OSBP1). We identified a PI4K3b inhibitor-resistant RV-A16 variant with a single point mutation in the conserved 2B protein near the cytosolic carboxy terminus, isoleucine 92 to threonine (termed 2B[I92T]). The mutation did not confer resistance to cholesterol-sequestering compounds or OSBP1 inhibition, suggesting invariant dependency on the PI4P/cholesterol lipid countercurrents. In the presence of PI4K3b inhibitor, Golgi reorganization and PI4P lipid induction occurred in RV-A16 2B[I92] but not in wild-type infection. The knockout of PI4K3b abolished the replication of both the 2B[I92T] mutant and the wild type. Doxycycline-inducible expression of PI4K3b in PI4K3b knockout cells efficiently rescued the 2B[I92T] mutant and, less effectively, wild-type virus infection. Ectopic expression of 2B[I92T] or 2B was less efficient than that of 3A in recruiting PI4K3b to perinuclear membranes, suggesting a supportive rather than decisive role of 2B in recruiting PI4K3b. The data suggest that 2B tunes the recruitment of PI4K3b to the replication membrane and allows the virus to adapt to cells with low levels of PI4K3b while still maintaining the PI4P/cholesterol countercurrent for establishing Golgi-derived RV replication membranes.
IMPORTANCE Human rhinoviruses (RVs) are the major cause of the common cold worldwide. They cause asthmatic exacerbations and chronic obstructive pulmonary disease. Despite recent advances, the development of antivirals and vaccines has proven difficult due to the high number and variability of RV types. The identification of critical host factors and their interactions with viral proteins and membrane lipids for the establishment of viral replication is a basis for drug development strategies. Our findings here shed new light on the interactions between nonstructural viral membrane proteins and class III phosphatidylinositol 4 kinases from the host and highlight the importance of phosphatidylinositol 4 phosphate for RV replication.
INTRODUCTION
Human rhinoviruses (RVs) are the most frequent cause of common colds, accounting for about 50% of upper respiratory tract infections (1). Although rarely life threatening, RVs can also replicate in the lower airways, where they play a critical role in causing exacerbations of asthma, chronic obstructive pulmonary disease, and cystic fibrosis (2). RVs are in the genus Enterovirus of the Picornaviridae family and are classified into the three species, RV-A, RV-B, and RV-C (3). While the recently discovered RV-C types use human cadherin-related family member 3 (CDHR3) as a receptor for entry (4), the major types of the RV-A and -B species bind to intercellular adhesion molecule 1 (ICAM1) and the minor types to the low-density lipoprotein (LDL) family receptors (5). Receptor binding leads to viral endocytosis and uncoating of the viral RNA genome (6–8).
The replication of plus-sense RNA viruses in the cytoplasm occurs in close association with membranes of the secretory or the endocytic pathways (reviewed in references 9 and 10–12). Picornavirus infections suppress the early onset of apoptosis and execute viral necrosis (13, 14). They remodel cytoplasmic membranes, which involves host protein recruitment to membranes, synthesis and modification of lipids, and alterations in membrane curvature, flux, and traffic. RVs remodel cytoplasmic membranes where viral and cellular proteins cooperate to replicate viral RNA, so-called replication complexes (15–17). Until recently, the morphology and origin of enterovirus-induced membrane rearrangements remained controversial. Serial electron tomography at different stages of infection revealed that poliovirus (PV) and coxsackievirus (CV) first form convoluted branching membrane tubules and later on process them into double-membrane vesicles (16, 18). Enterovirus replication complexes are established in close association with cis-Golgi membranes, suggesting that Golgi is the initial site of replication complex (RC) formation (19–22). As infection progresses, the Golgi apparatus is disrupted, and replication membrane structures grow in number and complexity in the perinuclear area, near dilated endoplasmic reticulum (ER) tubules.
The membrane-targeted 2B, 2C, and 3A proteins have been implicated in remodeling cytoplasmic membranes into replication membranes (23–25). For example, mutations in 2B and 3A facilitated RV-B39 adaptation to virus growth in murine cells expressing the ICAM1 receptor and enabled the formation of replication membranes, thereby highlighting the importance of viral membrane-interacting proteins (26).
The formation of the replication membranes also critically depends on host factors such as phosphatidylinositol 4 (PI4)-kinase class 3b (PI4K3b) (21, 22, 27). In uninfected cells, PI4K3b is located at the Golgi complex through the small GTP-binding protein Arf1 (28). It is activated by phosphorylation at Ser268 and stabilized by interactions with 14-3-3 proteins. Activated PI4K3b catalyzes the formation of PI4-phosphate (PI4P) lipids, which have key roles in signaling and vesicular trafficking at the Golgi complex (29). PI4K3b is recruited to the replication membranes in enterovirus-infected cells, where it generates high levels of PI4P (21, 22). One model for enrichment of PI4K3b at Golgi membranes suggested that the 3A protein of CVB3 indirectly recruits PI4K3b via the PI4K3b effector Arf1, as 3A recruits GBF1 to the replication membranes, activates Arf1, and thus mimics host recruitment mechanisms of PI4K3b (21). Another model proposed that 3A recruits PI4K3b to the replication membranes by direct interactions. Expression of 3A from various enteroviruses, such as PV, CVB3, and RV-B14, in the absence of other viral proteins showed that 3A copurified with PI4K3b (30). Interestingly, another host protein, acyl-coenzyme A (CoA) binding domain containing 3 (ACBD3), also known as GCP60, copurified with the 3A-PI4K3b complex and was able to bind PI4K3b independently of 3A (30). This suggests that ACBD3 acts as an adaptor for 3A to recruit PI4K3b. However, PI4K3b recruitment by CVB3 3A protein can also occur independently of GBF1, Arf1, or ACBD3, suggesting further mechanisms of lipid kinase recruitment to the RCs (31).
Here, we identified a novel mutation in the 2B protein of RV-A16, which was sufficient to render virus resistant to PI4K3b inhibitors or PI4K3b knockdown. The single point mutation occurred in a site that is highly conserved within species A and B rhinoviruses. Unlike enterovirus 3A mutants, the RV-A16 mutant retained the ability to use the PI4P/cholesterol lipid countercurrents on the replication membranes, akin to native RV-A16. The mutant virus could not replicate in the absence of PI4K3b but replicated more efficiently at limiting levels of PI4K3b, as shown by dose-dependent doxycycline induction of ectopic PI4K3b in PI4K3b knockout (KO) cells. We suggest that the conserved 2B protein near the cytosolic carboxy terminus, isoleucine 92 to threonine (termed 2B[I92T]) mutant protein, facilitates the recruitment of PI4K3b to the replication membrane.
RESULTS
PIK93-resistant RV-A16 is mutated in 2B.PI4K3b is a key host factor for replication of enteroviruses, and chemical inhibitors of PI4K3b block enterovirus infections (12). Mutants of PV, CVB3, and RV-B14 that had been found to be resistant to PI4K3b inhibitors all carried mutations in 3A. However, no RV-A mutants were reported (32–35). RV-A16 is highly sensitive to PIK93, which binds to the ATP binding site of PI4K3b (36). PIK93 has a 50% effective concentration (EC50) of 250 nM (22). We investigated whether PIK93-resistant mutants would also occur in the 3A protein of RV-A16. RV-A16 was serially cultured in the presence of increasing concentrations of PIK93. After 11 passages, a PIK93-selected RV-A16 pool emerged that grew in the presence of 1 μM PIK93 (not shown). When this pool of mutants was inoculated to fresh cells at a multiplicity of infection (MOI) of 20 for 8 h, it gave rise to infection in the presence of up to 4 μM PIK93, whereas wild-type (WT) RV-A16 was strongly inhibited by low concentrations of PIK93 (Fig. 1A). Inoculation with different amounts of virus at 1 μM PIK93 gave rise to dose-dependent infections, in which the PIK93-selected variant replicated better than the WT at comparable MOIs (Fig. 1B). This shows that we selected RV-A16 variants resistant to PIK93.
Identification of a PIK93-resistant RV-A16 2B[I92T] mutant. (A) HeLa cells were infected with RV-A16 WT (white bars) or RV-A16 PIK93-resistant virus pool (black bars) at an MOI of 20 for 8 h. Cells were treated with PIK93 at the indicated concentrations, and infection efficiencies were analyzed by high-throughput immunofluorescence microscopy using anti-VP2 antibodies. The ratio of infected cells was calculated and normalized to control infection without drug. Values represent the means ± standard deviations (SD). n = 2. (B) HeLa cells were infected with RV-A16 WT (white bars) or RV-A16 PIK93-resistant pool (black bars) at the indicated MOI for 8 h in the presence of 1 μM PIK93. The experiment was analyzed as described for panel A. Values represent means ± SD. n = 2. (C) Pairwise alignment of 2B proteins from RV-A1 (formerly known as RV-A1a or RV-A1b), -A2, -A16, -B14, -B37, and CVB3 by ClustalW using a Blosum62 similarity matrix. The hydrophobic regions 1 and 2 are present in all proteins and are separated by a five-amino-acid spacer. Note that the isoleucine located at the 92 position (I92) is near the C-terminal end. (D) ClustalW alignment of 2B amino acid sequences from 99 different RV types (74 RV-A and 25 RV-B; 11 minor RV and 88 major RV) using a Blosum62 similarity matrix. Prevalence of the amino acids isoleucine (I) and valine (V) at position 92 or equivalent were depicted by WebLogo from the aligned sequences. In addition, the prevalence of the amino acid threonine (T) at position 92 or equivalent was depicted for the 12 clones selected in the PIK93-selected mutant pool.
Viruses from the pool were plaque purified in the presence of 1 μM PIK93, and 12 independent clones were picked for genotyping. Sequencing of genomic segments encompassing the nonstructural proteins 2B, 2C, 3A, 3B, 3C, and 3D showed that all 12 clones had a single point mutation in the membrane-associated protein 2B at isoleucine 92 to threonine [I92T]. With the exception of a silent mutation in the 3C protein, no other mutations were observed. 2B is a small protein of about 100 amino acids and contains two hydrophobic domains required for membrane binding and correct localization in ER and Golgi membranes (37). The [I92T] mutation in the cytoplasmic C-terminal tail of 2B is located in a highly conserved region, indicated by amino acid sequence alignment of 99 RV-A and RV-B types (Fig. 1C and D). The [I92T] mutation facing the cytosol alters the side chain properties from a hydrophobic (I) to a polar (T) amino acid and could be in direct contact with a host or viral protein to compensate for the reduction in PI4P.
The 2B[I92T] mutant is resistant to PI4K3b inhibitors and depends on OSBP1.To test if the [I92T] mutation in 2B was necessary and sufficient to confer resistance to PI4K3b inhibitors, we introduced the 2B[I92T] mutation to a plasmid encoding an infectious full-length RV-A16 WT genome (pR16.11) and generated crude virus stocks from RV-A16 WT and RV-A16 2B[I92T] replicons. The PI4K3b inhibitors PIK93 and GSK533A reduce RV-A16 replication at an EC50 of 250 nM and 40 nM, respectively (22). Both compounds were less effective against RV-A16 2B[I92T], as indicated by immunofluorescence infection assays using anti-double-stranded RNA (dsRNA) antibodies (Fig. 2A). Notably, the Arf-GEF inhibitor brefeldin A (BFA) was similarly effective against both WT and 2B[I92T], validating the assay system used here. The calculated EC50 of PIK93 for RV-A16 2B[I92T] was 1.8 μM, about 3-fold higher than that for RV-A16 WT, which was previously shown to be 0.6 μM (22). Similarly, the 2B[I92T] mutant replicated better than RV-A16 WT in the presence of GSK533A, with 2.3-fold higher EC50 than the WT (73.9 nM and 32.5 nM, respectively). The reduced sensitivity of RV-A16 2B[I92T] to PIK93 and GSK533A was in the same range as that for a PI4K3b-resistant CVB3 mutant 3A[H57Y] reported earlier (35), which was as sensitive to BFA as native CVB3 (Fig. 2B). The data show that the 2B[I92T] mutation conferred resistance to PI4K3b inhibitors independent of the chemical mode of inhibition, suggesting a role of the 2B protein in the recruitment or activation of PI4K3b for RV-A16 replication.
Resistance of the RV-A16 2B[I92T] mutant to PI4K3b inhibitors. (A) HeLa cells were infected with RV-A16 WT (white bars) or RV-A16 2B[I92T] (black bars) at an MOI of 20 for 8 h in the presence or absence of BFA, PIK93, or GSK533A. Data were analyzed as described in the legend to Fig. 1, except that anti-dsRNA staining was used here and normalized to the dimethyl sulfoxide (DMSO) control. Values represent means ± SD. n = 2. (B) Same experiment as that described for panel A, except that cells were infected with CVB3 WT (white bars) or CVB3 3A[H57Y] (black bars).
In addition to PI4K3b, enterovirus replication critically depends on PI4P/cholesterol countercurrents at the replication membranes and the lipid exchange protein OSBP1 (22, 38, 39). The RV-A16 2B[I92T] mutant was as dependent on OSBP1 and cholesterol as native RV-A16, indicated by sensitivity to the OSBP1 inhibitor 25-hydroxycholesterol (25-HC), and the cholesterol esterase inhibitor CAY10499, with EC50s of 0.86 μM and 2.16 μM for RV-A16 2B[I92T] and 0.97 μM and 2.62 μM for RV-A16 WT (Fig. 3A and B). 25-HC binds to the cholesterol binding pocket of OSBP1 with higher affinity than cholesterol, and it blocks the lipid exchange activity of OSBP1 (40). This contrasts with the CVB3 3A[H57Y] mutant, which was insensitive to both 25-HC and CAY10499, unlike CVB3 WT, which was sensitive to high concentrations of 25-HC (Fig. 3C and D). The dependency of RV-A16 2B[I92T] on cholesterol was further supported by the finding that it was at least as sensitive to the cholesterol-sequestering compound methyl-β-cyclodextrin (MbCD) as RV-A16, with EC50s of 2.16 μM and 2.62 μM, respectively (Fig. 3E). Both RVs were insensitive to the cholesterol synthesis inhibitor compactin (Fig. 3F) (41), as reported earlier for RV-A16 (22). The data thus far show that RV-A16 2B[I92T] depends on cholesterol, cholesterol esterases, and the PI4P/cholesterol exchange protein OSBP1.
Similar susceptibility of RV-A16 2B[I92T] and WT to inhibitors blocking PI4P/cholesterol countercurrents. HeLa cells were infected with RV-A16 WT (white bars) or RV-A16 2B[I92T] (black bars) at an MOI of 20 for 8 h (A, B, E, and F) or with CVB3 WT or the 3A[H57Y] mutant (C and D) in presence or absence of BFA, the OSBP inhibitor 25-HC (A and C), the cholesterol esterase inhibitor CAY10499 (B and D), the cholesterol-depleting drug MbCD (E), or the cholesterol synthesis inhibitor compactin (F). Infections were analyzed as described in the legend to Fig. 1. Values represent means ± SD. n = 2. EtOH, ethanol.
To further explore how the 2B[I92T] mutant uses the PI4P/cholesterol lipid countercurrents, we transfected cells by short interfering RNA (siRNA) targeting of OSBP1 and PI4K3b, inoculated virus at an MOI of 20, and analyzed infections by immunofluorescence against dsRNA and high-throughput automated microscopy. As expected, RV-A16 2B[I92T] was nearly as sensitive to OSBP1 knockdown as RV-A16 WT and was much less sensitive to PI4K3b knockdown (Fig. 4A). This suggested that the virus can take advantage of PI4K3b if this enzyme is available. The selective PI4K3a inhibitor C23 inhibited neither RV-A16 WT nor RV-A16 2B[I92T] (Fig. 4B). C23 is 100-fold more selective against PI4K3a than PI4K3b, with IC50s of 16 nM and 1.6 μM, respectively (42), which makes it unlikely that PI4K3a is used for RV-A16 2B[I92T] infection.
Dependency of RV-A16 2B[I92T] on PI4Ks. (A) HeLa cells were transfected with siRNA against OSBP1, PI4K3b, or the negative-control siRNA (AllStars; Qiagen). Three days posttransfection, cells were infected with RV-A16 WT (white bars) or RV-A16 2B[I92T] (black bars) at an MOI of 20 for 8 h and analyzed for a fraction of infected cells. Values represent means ± SD. n = 2. b-Tub, β-tubulin. Western blots against OSBP1 and PI4K3b demonstrate the efficiency of protein knockdown by the corresponding siRNAs. (B) HeLa cells infected with RV-A16 WT (white bars) or RV-A16 2B[I92T] (black bars) were treated with DMSO, BFA, or the PI4K3a inhibitor C23 (B) and analyzed as described in the legend to Fig. 1. Values represent means ± SD. n = 2.
RV-A16 2B[I92T] disperses the Golgi apparatus in the presence of PIK93.To further analyze how 2B[I92T] replicates and drives the PI4P/cholesterol countercurrent, we investigated the integrity of the cis-Golgi apparatus by GM130 staining. As expected from previous results with RV-A16 (22), both RV-A16 WT and 2B[I92T] infections lead to dispersion of the Golgi apparatus in the absence of PIK93 (Fig. 5). RV-A16 2B[I92T] but not the WT dispersed the Golgi membrane in the presence of PIK93 (Fig. 5). 2B[I92T] infection increased the overall cytoplasmic levels of PI4P even in the presence of PIK93, indicated by anti-PI4P antibody staining (Fig. 6A). In the absence of PIK93, the perinuclear PI4P levels significantly increased 2.8- and 2.67-fold in RV-A16 WT- and 2B[I92T]-infected cells, respectively, and 2.3-fold in 2B[I92T]-infected cells in the presence of PIK93 (Fig. 6B). PIK93 strongly blocked perinuclear PI4P increase in WT-infected cells.
Golgi disruption in RV-A16 2B[I92T] and WT infection. HeLa cells treated with DMSO or PIK93 (1 μM) were infected with RV-A16 WT or 2B[I92T] at an MOI of 20 for 8 h and stained for viral protein VP1 (green) and GM130 (red). Nuclei stained with DAPI are in blue. Single z-planes from confocal imaging are shown. Scale bar, 10 μm.
RV-A16 2B[I92T] mutant enhances PI4P and recruits PI4P2a and PI4K3b. HeLa cells were treated with DMSO or PIK93 (1 μM), infected with RV-A16 WT or 2B[I92T] at an MOI of 20 for 8 h, and stained for viral protein VP1 and PI4P (A) or PI4K3b (C), and nuclei were stained with DAPI. Images show single z-planes. Quantifications of perinuclear PI4P (B) and PI4K3b (D) show relative signal intensities per area from total projections with number of cells analyzed (n) and mean values and SD. Scale bar, 10 μm.
To explore if class III PI4Ks were involved in the increase of PI4P in 2B[I92T]-infected cells, we immunostained cells with antibodies against PI4K3b. Infected cells were scored by antibodies against the capsid protein VP1. PI4K3b was enhanced in the perinuclear area by both WT and 2B[I92T] virus infections (Fig. 6C and D). In control cells, the recruitment of PI4K3b was higher in cells infected with the 2B[I92T] mutant than in the WT, i.e., 2.44- and 1.86-fold, respectively. PIK93 significantly abrogated the enrichment of PI4K3b to perinuclear areas of WT- but not 2B[I92T] mutant-infected cells, where the perinuclear levels were 2.11-fold higher than those in uninfected cells. This suggests that the 2B[I92T] mutation enhances the recruitment of PI4K3b at the replication membranes, even in the presence of PIK93.
RV-A16 2B[I92T] replicates at low levels of PI4K3b.To scrutinize if the 2B[I92T] mutant takes advantage of very low levels of PI4K3b, we engineered a cell line for doxycycline (Dox)-inducible expression of PI4K3b in the background of PI4K3b KO HeLa-OHIO cells. The PI4K3b knockout (KO) cells were completely resistant to WT and the 2B[I92T] mutant virus, while PI4K2a KO cells were fully susceptible to both viruses (Fig. 7A). Dox induced PI4K3b in a dose-dependent manner (Fig. 7B). Notably, the induced PI4K3b was running slightly faster in the SDS-PAGE than the endogenous HeLa cell protein. The cDNA encoding the induced PI4K3b was missing the exon encoding amino acids 304 to 318, accounting for about 1.5 kDa. Since the difference between the endogenous and the induced PI4K3b is about 10 kDa and a faint upper band running at the position of the endogenous protein is present in the induced cells, we suspect that the two bands differ in the levels of posttranslational modifications, such as sumoylation, phosphorylation, or ubiquitination. Regardless, the expression of PI4K3b enhanced the titers of the WT by 1.67 logs and the 2B[I92T] mutant by 2.33 logs (Fig. 7C). The enhancement of infection was consistently larger for the 2B[I92T] mutant than the WT, as indicated by measuring VP2 production at different dosages of Dox at both 8 and 16 h postinfection (pi) (Fig. 7D). These results indicate that the 2B[I92T] mutant replicates better at low PI4K3b levels than the WT.
RV-A16 2B[I92T] replicates at low levels of PI4K3b. (A) HeLa-OHIO cells were subjected to CRISPR/Cas9 knockout with guide RNAs (gRNAs) targeting PI4K2a and PI4K3b. Knockout was verified by Western blotting. The indicated cell lines were infected at an MOI of 20 for 8 h and analyzed for infected cells. Values representing fractions of infected cells are means ± SD, n = 3. (B) HeLa-OHIO PI3K3b KO cells were stably transfected with a doxycycline-inducible PI4K3b cassette. Expression of PI4K3b was induced by different dosages of doxycycline for 24 h, and PI4K3b levels were analyzed by Western blotting. (C) Cells were infected with RV-A16 WT or RV-A16 2B[I92T] (MOI 20) in the presence or absence of doxycycline. Progeny virus was harvested at 16 h pi and titrated on fresh HeLa-OHIO cells. (D) Cells were infected at an MOI of 20 in the presence or absence of doxycycline, and infection was assessed after 8 h and 16 h, respectively. Values represent means ± SD. n = 2 to 3.
The RV-A16 3A protein recruits PI4K3b.We examined if one of the nonstructural viral proteins recruits the kinase responsible for PI4P generation. HeLa cells were transfected with myc-tagged viral protein 2B WT, 2B[I92T], 3A, or 3AB and analyzed by immunofluorescence staining against PI4P and PI4K3b. None of the viral proteins except 3AB enhanced the levels of PI4P (Fig. 8A), in agreement with results from other enteroviruses (39). Expression of 3A protein was sufficient to recruit PI4K3b at perinuclear sites, while 3AB was less efficient in doing so (Fig. 8B). Notably, neither 2B WT nor 2B[I92T] proteins had an apparent effect on intracellular PI4P levels or PI4K3b recruitment. This suggests that the 3A protein is the predominant recruitment factor for PI4K3b to perinuclear membranes and the 3AB precursor is an enhancer of PI4P levels, perhaps by activating PI4K activity. In sum, RV-A16 2B[I92T] alone is not able to enhance PI4P levels or to recruit PI4K3b to perinuclear membranes. This excludes the 2B protein as a stand-alone recruiting factor for PI4K3b but rather suggests a supporting or stabilizing role of 2B in the recruitment of PI4K3b by 3A and 3AB.
PI4K 3b recruitment by the RV-A16 3A protein. HeLa cells were transfected with plasmid DNA encoding myc-tagged 2B, 2B[I92T], 3A, and 3AB, stained with anti-myc antibodies (green), and stained with antibodies against PI4P (A) or PI4K3b (B). Nuclei were stained with DAPI. Representative single-section confocal images are shown. Scale bar, 10 μm.
DISCUSSION
The best-studied picornaviruses are enteroviruses. Enteroviruses comprise 15 species, Enterovirus A to L and Rhinovirus A, B, and C, and include well-known agents, such as PV, coxsackieviruses, and rhinoviruses. These viruses all induce the formation of single-membrane clusters in proximity to Golgi membranes and ER within just a few hours of infection. The replication membranes grow in size, complexity, and volume as infection progresses and eventually become double-membrane structures. In the case of PV and members of the alphavirus-like superfamily, such as bromovirus, this involves changes in fatty acid metabolism, fatty acid import, and phospholipid synthesis, including phosphatidyl choline (PC) (43, 44). Another key lipid for enterovirus replication membranes is cholesterol. While de novo cholesterol synthesis is not required, exogenous cholesterol from receptor-mediated endocytosis or deesterification of cholesterol esters from lipid droplets is essential for enterovirus replication (for a review, see reference 12). Free cholesterol is delivered to the replication membrane by OSBP1 and related proteins at membrane contact sites (22, 45, 46). This occurs by a lipid countercurrent flux exchanging PI4P and cholesterol and is driven by the formation of PI4P through PI4Ks (28, 47). For replication of enteroviruses, including PV, CVB3, and RV-A and -B, the Golgi-associated PI4K3b is of key importance, although RV-A1A and -A16 can also use PI4K2a and PI4K3a, as suggested by RNA interference (21, 22, 48). Notably, however, the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9-mediated KO of PI4K2a had no effects on RV-A16 infection (Fig. 7A). These data highlight the complexity of virus-tuned lipid modulation.
How PI4Ks are recruited to replication membranes is poorly understood. For enteroviruses, it was initially suggested that PI4K3b is recruited by 3A protein, either indirectly via GBF1 and Arf1 or directly using ACBD3 as an adaptor (reviewed in reference 49). However, none of these options proved to be conclusive. Instead, the expression of CVB3 3A protein alone was sufficient to recruit PI4K3b and generate a high level of PI4P lipids, even in GBF1-, Arf1-, or ACBD3-depleted cells (21, 31). A recent study suggested that the CVB3 mutant 3A[H57Y] can replicate in the presence of PI4K3b inhibitors and without apparent dedicated replication membranes on Golgi membranes (50). Based on our results, we speculate that this mutant grows in the presence of low concentrations of active PI4K3b, perhaps akin to the 2B[I92T] mutant.
We show that the 2B[I92T] mutant confers resistance to PI4K3b inhibitors. Isoleucine 92 is highly conserved among RV-A and RV-B species and is located near the cytosolic carboxy terminus of 2B. 2B comprises about 100 amino acids, forms membrane pores by virtue of two hydrophobic domains and oligomerization, localizes to Golgi membranes, and inhibits protein trafficking (37, 51, 52). Similar to that of 3A, expression of RV-A16 2B protein alone suffices to block protein secretion, although RV-A16 infection apparently did not inhibit the release of a reporter protein, Gaussia luciferase, up to 7 h pi, suggesting that virus infection does not block the entire secretory pathway (53).
Remarkably, RV-A16 2B[I92T] disrupted the Golgi membrane, recruited PI4K3b, and enriched PI4P on replication membranes in the presence of PI4K3b inhibitors. RV-A16 2B[I92T] remained sensitive to other inhibitors of the PI4P/cholesterol countercurrent flux. Remarkably, PI4K3b localizes close to OSBP1 and thereby allows the steady exchange of lipids between the ER and the Golgi stacks, depending on the OSPB1 anchoring protein VAP (54). VAP was also important for RV replication, as shown by RNA interference and the expression of dominant-negative OSPB1 mutants lacking the PI4P binding domain but retaining the VAP binding domain FFAT (22). PI4K2a, which does not locate to the proximity of OSBP1, leads to the oscillating flux of PI4P (54), and this correlates with PI4K2a being unable to support RV-A16 replication (Fig. 7A).
The phenotype of RV-A16 2B[I92T] appears to be different from that of the previously reported 3A mutants, which replicate in a PI4K3b-independent manner when PI4K3b activity is impaired (32–35). This suggests that the A16 2B protein has a different function(s) than 3A protein in the generation of replication membranes. In support of this notion, the expression of 3A but not 2B protein alone recruited PI4K3b, and the precursor 3AB increased PI4P levels in transfected cells. This is different from the CV 3A protein, which recruited PI4K3b and increased PI4P (35). Further, the replication of the CVB3 3A[H57Y] mutant in the presence of PI4K inhibitors no longer depended on PI4K3b or other PI4K isoforms, did not disrupt the Golgi membranes, and did not increase the PI4P levels (35).
In contrast, the RV-A16 2B[I92T] mutant replicated at lower levels of PI4K3b than the WT, as indicated by tunable expression of PI4K3b and two different chemicals targeting the ATP binding site of PI4K3b. It was sensitive to RNA interference against PI4K3b, suggesting either a lower requirement or a nonenzymatic function of PI4K3b for RV-A16 2B[I92T] replication. Notably, PI4K3b recruits small GTPases independent of its catalytic activity and fine tunes vesicular transport (36). Consistent with this, the complete knockout of PI4K3b by CRISPR/Cas9 abrogated both RV-A16 WT and 2B[I92T] infections, which, however, were insensitive to PI4K2a knockout. This strengthens the notion that 2B[I92T] reduces but does not eliminate the requirement for PI4K3b. Low residual functionality of PI4K3b resulting from incomplete inhibition by a compound or siRNA knockdown would therefore be sufficient for the replication of RV-A16 2B[I92T] but not the WT.
It is possible that the transmembrane domains of 2B and 3A or 3AB interact, as suggested for the poliovirus proteins by yeast two-hybrid experiments (55). We speculate that the respiratory RV-A16 uses the 2B protein together with 3A to recruit PI4K3b. This is based on the notion that there are no RV-A types known to have mutations in 3A, which would make them resistant to PI4K3b inhibitors. Since the N terminus of 3A from RV-A types lacks 6 amino acids that are present in RV-B types, we speculate that RV-A types require an additional viral protein for effective recruitment of PI4Ks to the replication membrane. We further suggest that CVB or RV-B can drive the lipid currents that are crucial for establishing the replication membrane by taking advantage of an asymmetry other than those of PI4P lipids. Such asymmetrically distributed lipids could comprise PC. Notably, phosphatidyl inositol transfer protein b (PITPb) shuttles phosphatidylinositol (PI) in exchange to PC between the ER and Golgi membranes and is required for RV-A1A and RV-A16 replication (22). Intriguingly, PC with short acyl chain lengths are enriched in PV-infected cells (43), giving rise to the possibility that PC lipids are involved in tuning the composition of the replication membrane.
In summary, our results show that despite the great plasticity of the host lipid landscape and the readily mutating nonstructural proteins of picornaviruses, the cells provide key metabolic circuits for establishing picornavirus replication membranes. One of these circuits is the PI4P/cholesterol countercurrent flux. Host components that regulate this flux are surgically targeted by at least two nonstructural membrane proteins of RV-A16, 2B and 3A, to drive viral replication on cytoplasmic membranes.
MATERIALS AND METHODS
Chemicals, plasmids, antibodies, and cell lines.PIK93 was purchased from Selleck Chemicals, BFA from LC Laboratories, MbCD and 25-HC from Sigma, and CAY10499 from Cayman Chemical. GSK2998533A (GSK533A) was a kind gift from S. You (GlaxoSmithKline, Infectious Disease R&D, North Carolina, USA), compactin was from L. Rohrer (Institute of Clinical Chemistry, University Hospital Zurich, Switzerland), and AL-9 and C23 were from R. De Francesco (Istituto Nazionale di Genetica Molecolare, Milan, Italy). RV-A16 was used as described in reference 56. The RV-A16 genomic replicon pR16 was a gift from W. M. Lee (Department of Pediatrics, School of Medicine and Public Health, University of Wisconsin, Wisconsin, USA). The CVB3 genomic replicons pRLuc-CB3/T7 and pRLuc-CB3/T7-3A[H57Y] were a gift from F. van Kuppeveld (Department of Infectious Diseases & Immunology, University of Utrecht, Netherlands). cDNAs encoding the RV-A16 nonstructural proteins myc-2B, myc-3A, and myc-3AB were obtained from A. Mousnier (Imperial College London) and expressed from the pRK5-myc plasmid (Clontech) from the cytomegalovirus promoter and with simian virus 40 polyadenylation signal as described previously (53). Mouse monoclonal antibody (MAb) 16-7 directed against VP2 (W. M. Lee, Department of Pediatrics, School of Medicine and Public Health, University of Wisconsin, Wisconsin, USA) and MAb J2 against dsRNA (English & Scientific Consulting) were used as described previously (56), and we also used rabbit polyclonal antibody against VP1 (from K. Niespodziana and R. Valenta, Division of Immunopathology, University of Graz, Austria) (57), MAbs against GM130 and PI4K3b (BD Transduction Laboratories), MAb against PI4P (Echelon), MAb 1C4 against PI4K2a (S. Minogue, Institute of Liver and Digestive Health, University College London, UK), and Alexa Fluor 488- or 594-labeled secondary antibodies against mouse or rabbit IgG or IgM (Invitrogen). HeLa cervical carcinoma cells, strain OHIO (HeLa, from L. Kaiser, Central Laboratory of Virology, University Hospital Geneva, Switzerland), were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1% l-glutamine, called full medium.
Generation of PIK93-resistant viruses.HeLa cells were cultured in 96-well plates overnight at 37°C in full medium and infected with wild-type (WT) RV-A16. PIK93-resistant viruses were obtained by serial passages in the presence of increasing concentrations of PIK93 for 6 days at 33.5°C. Supernatants from cultures with full cytopathic effect (CPE) at the highest concentration of drug were passaged until full CPE was observed at a PIK93 concentration that did not allow replication of the initial inoculum (for example, WT virus at 1 μM PIK93). Resistant virus pools were subjected to plaque purification by culturing HeLa cells in 6-well plates at 37°C in full medium overnight, inoculated with the PIK93-resistant virus pool in 10-fold dilution series in full medium complemented with 1 μM PIK93 at 33.5°C. At 5 h postinfection, the medium was discarded and replaced by fresh full medium complemented with 0.6% agarose (Affymetrix), 1% penicillin-streptomycin solution (Gibco), and 1 μM PIK93. The 6-well plates were incubated for 6 days at 33.5°C. Single plaques were identified and collected for further amplification in HeLa cells on 6-well plates for 2 days at 33.5°C. Total RNA was extracted using TRI Reagent, and cDNAs were obtained by reverse transcription using SuperScript III according to the manufacturer's instructions. PCR fragments covering the nonstructural region of RV-A16 were amplified using Pfu polymerase (Promega) and analyzed by Sanger DNA sequencing.
Site-directed mutagenesis.The site-specific substitution I92T was introduced into the genomic replicon clone of RV-A16 by PCR overlap extension. The first PCR using Pfu polymerase generated two fragments with overlapping ends. The sense primer (5′-AAA GCT TCC TAG GCA GAT CG-3′) and the antisense primer harboring the I92T mutation (5′-CTG ATT CTT TGT GTG TAT AAG TTA ATT-3′) were used to amplify the first fragment with the genomic replicon pR16 as the template. The second fragment was generated with the same template, the antisense primer (5′-TTC ACT GCC CGG GTC AGC AT-3′), and the sense primer harboring the I92T mutation (5′-CAA TTA ACT TAT ACA CAC AAA GAA TCA-3′). The PCR amplification was performed with preheating at 94°C for 2 min, denaturation at 94°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 2 min. The cycle was conducted 25 times, followed by incubation at 72°C for 5 min. The amplified PCR fragments were purified and extracted by electrophoresis from a 1% agarose gel. The second PCR round involved the two overlapping fragments as templates and the two primers located at both ends. The final fragment was amplified with Pfu polymerase with preheating at 94°C for 2 min, denaturation at 94°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 4 min. The cycle was conducted 25 times, followed by incubation at 72°C for 5 min. The final amplified PCR fragment harboring the I92T mutation was gel purified, sequentially digested with AvrII and XmaI restriction enzymes (Promega), and reintroduced into the original pR16 plasmid. Presence of the mutation in the newly generated genomic replicon pR16 2B[I92T] was confirmed by sequencing.
In vitro RNA transcription and crude virus stock production.The plasmids pR16 WT and pR16 2B[I92T] were linearized with SacI (Promega), extracted with phenol-chloroform, in vitro transcribed using T7 RNA polymerase (Fermentas) according to the manufacturer's instructions at 37°C for 2 h, DNase and RNA treated, and extracted by phenol-chloroform. HeLa cells were cultured in 24-well plates overnight at 37°C in full medium and transfected with 250 ng of purified RNA using a TransIT-mRNA transfection kit (Mirus) according to the manufacturer's instructions. When cultures exhibited extensive CPE after 2 days of incubation at 37°C, virus was harvested from cells and supernatant, freeze/thawed three times, and centrifuged at 5,000 × g for 5 min. Supernatants were collected and titrated in a 50% tissue culture infectious dose (TCID50) assay. Equal MOIs from pR16 WT and pR16 2B[I92T] stocks were used in infection assays.
Interference and high-throughput infection.Small interfering RNAs (siRNAs) were reverse transfected to HeLa cells in 96-well plates using serum-free Opti-MEM (Invitrogen) and Lipofectamine RNAiMAX (Invitrogen) (20 nM, 37°C, 72 h) and inoculated with crude virus stocks (MOI of 20) at 33.5°C for 8 h. For chemical interference assay, HeLa cells were treated with drugs at 1 h pi, infected with crude virus stocks (MOI of 20) at 33.5°C for 8 h, fixed, stained with MAb J2, and scored for infection. Images were acquired with an ImageXpress Micro microscope (Molecular Devices) in automated mode using a CoolSNAP HQ 12-bit gray-scale camera (Roper Scientific) and 10×/0.5-numeric-aperture (N.A.) objective (Nikon) and analyzed with a script custom written in Matlab (MathWorks, Inc., Natick, MA, USA) (58–60) or in R. Infection indexes were calculated as the fraction of infected cells per total cell number and plotted with GraphPad Prism software (GraphPad) or with R.
Immunofluorescence and confocal microscopy.HeLa cells on coverslips were treated with PIK93 and infected with crude virus stocks at 33.5°C for 8 h. Alternatively, HeLa cells were transfected with plasmid DNA using Lipofectamine 2000 (Invitrogen) for 24 h. For PI4P staining, plasma membrane and cytoplasmic PI4P pools were stained and quantitated by expanding the 4′,6-diamidino-2-phenylindole (DAPI) mask as described previously (22, 61). Other immunofluorescence staining was done with cells fixed with 4% paraformaldehyde (PFA) and permeabilized with 0.2% Triton X-100 blocked for 1 h in phosphate-buffered saline (PBS) supplemented with 1% bovine serum albumin (BSA), followed by treatment with primary antibodies in blocking buffer overnight at 4°C. Alexa Fluor 488 or 594 secondary antibodies were used in blocking buffer for 1 h. Coverslips were mounted in mounting medium (Dako) and analyzed with an upright Leica TCS SP8 scanning laser confocal microscope with an HCX PL APO 63×/1.4-N.A. oil immersion objective. Images were acquired using LAS AF software (Leica), processed with ImageJ (National Institutes of Health), and quantitated as described earlier (22). Quantification of PI4P signal in single cells was carried out by maximal projection of eight z-stacked images 2 μm in width for each channel. An outline extending 10 pixels from the nuclear periphery based on DAPI staining was used to define a perinuclear area for PIP4 intensity quantification. The average score, normalized by area, was plotted as relative fold change.
CRISPR/Cas9 knockout and inducible expression of PI4K3b.Knockout HeLa-OHIO cell lines were generated with a modified version of the lentiCRISPRv2 one vector system described in reference 62. The Cas9 cassette was exchanged with the high-fidelity Cas9 variant by using the XbaI and BamHI restriction sites (63). For tetracycline inducibility, the promoter region was extended at the EcoRI, XbaI, and XhoI sites by the thyroid hormone response element (TRE), TetR on pCW57.1 (41393; Addgene), and EF1a from Lenti-eCas9 (52962; Addgene). The sequences introduced into the plasmid as templates for the guide RNA were 5′-GCACGGCAGTTACACCACTG-3′ for the PI4K3b KO and 5′-GGATCCTGAGTTCGAGGCGG-3′ for the PI4K2a KO. Clonal knockout cell lines were raised by limiting dilution of parental polyclonal cells. Individual clones were analyzed for expression of PI4K2a or PI4K3b by Western blotting.
For the generation of a cell line expressing tetracycline-inducible PI4K3b, full-length cDNA of Homo sapiens PI4K3b (GenBank accession no. XM_005245264.3) was cloned into the lentiviral expression vector pLVX-tet-BSD using the XhoI and NotI restriction sites. pLVX-tet-BSD was constructed by replacing the CMV promoter, IRES, and puromycin resistance gene in pLVX-IRES-Puro (Clontech) with a tetracycline response element (TRE) followed by a multiple cloning site and an expression cassette containing a blasticidine deaminase (BSD), a P2A sequence, and an rTetR under a constitutive EF-1a core promoter.
ACKNOWLEDGMENTS
We thank D. Seiler and M. Suomalainen for comments on the manuscript and M. Bauer for the construction of the vectors pLVX-tet-BSD and the modified lentiCRISPRv2 as well as his advice and protocols. We also thank the following colleagues for kindly providing reagents: R. De Francesco and P. Neddermann for AL-9 and C23, L. Rohrer for compactin, K. Niespodziana and R. Valenta for antibody against VP1, S. Minogue for MAb 1C4 against PI4K2a, A. Mousnier for cDNAs encoding myc-tagged RV-A16 2B, 3A, and 3AB, F. van Kuppeveld for CVB3 genomic replicons, W. M. Lee for MAb 16-7 directed against VP2, and L. Kaiser for HeLa cervical carcinoma cell strain OHIO.
This work was supported by the Swiss National Science Foundation (grant number 310030B_160316), a Medical Research and Development Project from the Swiss initiative for systems biology SystemsX.ch (2014/264 Project VirX, evaluated by the Swiss National Science Foundation), and the University of Zurich Research Priority Program Evolution in Action to U.F.G.
U.F.G. conceived of and coordinated the study, P.S.R. and L.P.M. performed experiments, and P.S.R., L.P.M., and U.F.G. interpreted results and wrote the manuscript.
FOOTNOTES
- Received 23 August 2018.
- Accepted 31 August 2018.
- Accepted manuscript posted online 12 September 2018.
- Copyright © 2018 American Society for Microbiology.
