ABSTRACT
Viruses are obligatory intracellular parasites and utilize host elements to support key viral processes, including penetration of the plasma membrane, initiation of infection, replication, and suppression of the host's antiviral defenses. In this review, we focus on picornaviruses, a family of positive-strand RNA viruses, and discuss the mechanisms by which these viruses hijack the cellular machinery to form and operate membranous replication complexes. Studies aimed at revealing factors required for the establishment of viral replication structures identified several cellular-membrane-remodeling proteins and led to the development of models in which the virus used a preexisting cellular-membrane-shaping pathway “as is” for generating its replication organelles. However, as more data accumulate, this view is being increasingly questioned, and it is becoming clearer that viruses may utilize cellular factors in ways that are distinct from the normal functions of these proteins in uninfected cells. In addition, the proteincentric view is being supplemented by important new studies showing a previously unappreciated deep remodeling of lipid homeostasis, including extreme changes to phospholipid biosynthesis and cholesterol trafficking. The data on viral modifications of lipid biosynthetic pathways are still rudimentary, but it appears once again that the viruses may rewire existing pathways to generate novel functions. Despite remarkable progress, our understanding of how a handful of viral proteins can completely overrun the multilayered, complex mechanisms that control the membrane organization of a eukaryotic cell remains very limited.
INTRODUCTION
Picornaviruses comprise a group of broadly distributed positive-strand RNA viruses of animals. They include important human and veterinary pathogens such as hepatitis A virus, poliovirus, rhinoviruses, and foot-and-mouth disease virus, among others. Picornaviruses have naked icosahedrical capsids that encapsulate a single strand of genomic ∼6,500- to 8,000-nucleotide (nt)-long RNA of positive polarity containing one open reading frame. Upon internal ribosome entry site-driven translation, the resulting polyprotein undergoes processing by virus-encoded proteases to generate about a dozen intermediate cleavage products and mature peptides. The processing intermediates often have their own role in the viral life cycle, effectively increasing the limited coding capacity of the genome (1). Replication of picornaviruses, like that of all other positive-strand RNA viruses of eukaryotes, is intimately associated with cellular membranes. This association is believed to provide a favorable microenvironment for the replication process by providing a structural scaffold for the replication machinery, by concentrating viral and cellular factors, and by hiding the double-stranded RNA replication intermediates (2).
Most of our knowledge of how picornaviruses hijack the membrane homeostasis machinery comes from studies of just a few viruses, mostly representatives of the genus Enterovirus, such as coxsackieviruses, rhinoviruses, and, notably,poliovirus, one of the best-studied animal viruses. The replication of poliovirus (and some other picornaviruses) is not sensitive to inhibition of nuclear transcription by actinomycin D (3–5), and poliovirus infection rapidly inhibits cap-dependent translation of cellular mRNAs (6), suggesting that the profound remodeling of intracellular membranes in infected cells is not dependent on new expression of cellular genes and is carried out by preexisting cellular factors and viral proteins. How these viruses with their very limited genetic resources manage to completely overrun the complex cellular control of membrane homeostasis in a matter of hours and transform the intricate intracellular membrane architecture into viral replication organelles remains one of the most puzzling questions in virus-cell interactions.
MEMBRANE REMODELING IN PICORNAVIRUS-INFECTED CELLS
The development of membranous replication organelles in poliovirus-infected cells has been extensively studied since the introduction of electron microscopy (EM) imaging in biology. The first virus-induced structures can be detected at about 2 h postinfection as loosely associated clusters of single-membrane vesicles (Fig. 1, early, 3 h p.i.). These initial structures are often located close to the cis-Golgi, and immunostaining of cells at early stages of infection shows the close association of viral antigens with the Golgi marker galactosyl-transferase, suggesting a close relationship between the establishment of viral replication centers and the Golgi membranes. However, as infection progresses, the viral antigens lose their colocalization with Golgi markers, and Golgi structures disappear from EM images of infected cells (7–9). At the midstage of infection (∼4 h postinfection), the replication structures grow considerably in number and complexity and concentrate at the perinuclear area, forming characteristic perinuclear rings. EM images show tightly associated complex assemblages of single-membrane vesicles that appear empty inside (Fig. 1, intermediate, 4 h p.i.). These structures are often found close to dilated endoplasmic reticulum (ER) tubules, but no continuity between them and the ER has been observed (7, 10). Recent electron tomography studies demonstrated that such structures in cells infected with poliovirus or the related coxsackievirus B3 (CVB3) represent conglomerates of convoluted branching membrane tubules rather than clusters of vesicles (7, 11). Further progression of infection results in the transition of the single-membrane structures into double-membrane vesicles (Fig. 1, late, 7 h p.i.). This process is not well understood but is believed to include at least some autophagy-related events resulting in the collapse and wrapping of the single-membrane structures on themselves (7, 11–13). Similar membrane rearrangements are observed in diverse cell types permissive for poliovirus replication (9, 14, 15). The importance of the structural transition from single-membrane to double-membrane replication structures for viral infection is not clear: the exponential phase of viral RNA replication is associated with the single-membrane structures. However, double-membrane structures may also significantly contribute to overall RNA replication output (7). Recently, the double-membrane vesicles were proposed to play a role in virion maturation and non-lytic virus release (16, 17).
Virus-induced membrane remodeling in poliovirus-infected HeLa cells. EM and 3D tomographic reconstructions of membranes at early (3 h postinfection [p.i.]), intermediate (4 h p.i.), and late (7 h p.i.) time points after infection of HeLa cells with poliovirus. (Modified from reference 7.)
THE (UNCERTAIN) ROLE OF THE SECRETORY VESICLE-BUDDING MACHINERY IN PICORNAVIRUS REPLICATION
The origin of the viral replication membranes is not clear. Biochemical fractionation of poliovirus replication complexes revealed that they contain markers of virtually all cellular membranous organelles (15), which does not provide much information on the biogenesis of the replication organelles, but rather reflects the fact that they may represent most of the membranous material in infected cells.
The superficial resemblance of the picornavirus replication structures to clusters of membranous vesicles and their proximity to the Golgi complex early in infection suggested that the replication structures may be generated by the membrane-remodeling system of the secretory pathway, specifically the vesicle-budding machinery. At the ER-Golgi complex interface, vesicular traffic is mediated by COPII and COPI transport carriers (reviewed in reference 18): cargo proteins enter the ER lumen during translation and are exported from the ER in COPII-coated vesicles at specialized ER exit sites called ERES (Fig. 2A). COPII vesicles deliver cargo to the ER-Golgi intermediate compartment (ERGIC), where the COPII coat is disassembled. From the ERGIC, cargo proteins move to the Golgi stacks and, after passage through this organelle, are transported to the cell surface or to the endolysosomal system. ER-Golgi cycling proteins are returned to the ER from the ERGIC and the Golgi cisternae in COPI-coated vesicles. COPI vesicles also play an important role in maintaining the Golgi stacks by facilitating intra-Golgi transport of proteins and membranes (reviewed in references 19 and 20).
(A) Intracellular transport pathways. Diagram depicting the compartments of the secretory and endosomal pathways. Colors indicate the known coats: COPII (brown), COPI (green), and clathrin (red). PM, plasma membrane. Secretory cargos are synthesized in the ER, exit the ER at ERES in COPII-coated vesicles, and are transported to ERGIC. Cargos are sorted from ERGIC into anterograde carriers that move them to the Golgi compartment. After passage through the Golgi compartment, cargos are sorted at TGN for delivery to the PM and early and late endosomes. A COPI-mediated pathway recycles proteins from the Golgi compartment and ERGIC and returns them to the ER. (B) GBF1-mediated Arf activation. Formation of ER-Golgi transport vesicles requires activation of an Arf by GBF1 (step 1), recruitment of a coat (step 2), and cargo sorting/concentration (step 3) into a nascent bud. Additional proteins participate in each step but are not depicted for simplicity. (C) Distinct GBF1 function in viral replication and cellular secretion. The highly conserved domains of GBF1 include a dimerization and cyclophilin binding (DCB) domain, a homology upstream of Sec7d (HUS) domain, the catalytic Sec7d, and three homology downstream of Sec7d (HDS1-3) domains. The region of GBF1 capable of interacting with 3A of poliovirus (bar) is distinct from the Sec7d that binds the substrate Arf (bar). Truncated forms of GBF1 were expressed as a sole functional copy in HeLa cells and tested for their ability to support poliovirus replication and cellular secretion as described in reference 52. The removal of the N-terminal region inhibits both replication and secretion, while deletions of the C terminus have less effect on replication but inhibit secretion. (D) GBF1 interactome. Proteins known to bind mammalian GBF1 are depicted in blue. Proteins shown to bind to the yeast ortholog of GBF1, Gea1/2, are depicted in purple. Cyclophilin is predicted to bind GBF1 based on homology with a plant ortholog and is depicted in black. The 3A poliovirus protein is in green. (E) Model for possible 3A-induced alterations in the GBF1 interactome. The top diagram shows GBF1 interactions in an uninfected cell. The bottom diagram shows inhibition of a subset of GBF1 interactions in infected cells and the establishment of new interactions that are prohibited in uninfected cells, likely resulting in significant changes in the overall GBF1 interactome on the replication membranes compared to that on the membranes in uninfected cells.
The notion that the cellular budding machinery may participate in viral replication was suggested by the well-documented association of components of the COPII and COPI coats with picornavirus replication structures. However, the functional significance of these observations remains controversial. For example, COPII vesicle budding was shown to increase early during poliovirus infection and then to subside later, suggesting that at least during the active replication phase COPII engagement is not important (21). Also, although poliovirus protein 2B colocalizes with the Sec13 and Sec31 COPII coat components at ER exit sites (22), expression of a dominant negative mutant of GTPAse Sar1, which inhibits normal functioning of ER exit sites, only moderately reduced replication of CVB3, a close relative of poliovirus (8).
The requirement for COPI is similarly uncertain: components of the COPI complex were identified in a genetic screen for host factors required for the replication of Drosophila C virus, a picorna-like virus. Further validation experiments showed that only small interfering RNA (siRNA)-mediated knockdown of COPI components, but not COPII components, was detrimental to the replication of both Drosophila C virus and poliovirus (23). However, while COPI proteins were found to partially colocalize with the replication organelles of echovirus, an enterovirus related to poliovirus, COPI subunits were absent from the replication complexes of encephalomyocarditis virus (EMCV), a less closely related cardiovirus (24). Moreover, while the association of COPI components with poliovirus replication membranes can be detected early in infection, it is lost at later time points (8).
These data suggest that the replication membranes for different picornaviruses may be generated by different mechanisms or even that in different cellular systems or at different stages of the replication cycle the same virus may manipulate different sets of cellular factors. However, it is more likely that the association of COPI and/or COPII with replication membranes is a by-product of membrane rearrangement rather than a functional engagement. Indeed, EM studies of different types of cells infected with diverse picorna- and picorna-like viruses reveal very similar if not identical structures, suggesting that the mechanisms of membrane remodeling are most likely shared by different viruses (24–29). Moreover, the complex tubular rather than vesicular three-dimensional (3D) architecture of the replication membranes suggests that the mechanism of their formation does not rely on cellular vesicle-budding processes (7, 11). Furthermore, the limited genome size of these related viruses makes it difficult to envision that they have evolved significantly different strategies for such a fundamental function as the development of replication organelles. Thus, despite the intuitive appeal of viruses simply “hijacking” cellular machineries such as COPI- and COPII-dependent vesicle budding and using them much the same way as a cell does, accruing evidence suggests that viruses utilize some membrane traffic elements for virus-specific functions in different and unique ways, which we discuss below.
GBF1 FUNCTION IN PICORNAVIRUS REPLICATION
The well-documented sensitivity of poliovirus, CVB3, and a few other related viruses to brefeldin A (BFA), a small-molecule inhibitor of secretory traffic, suggested that at least some component(s) of the secretory pathway is required for picornavirus replication (24, 30–34). BFA inhibits three members (GBF1, BIG1, and BIG2) of a family of guanine nucleotide exchange factors (GEFs) that catalyze exchange of a nucleotide bound to the small Arf GTPases from GDP to GTP (35–37). Arf-GTP binds to membranes, where it recruits effector proteins, including coat complexes (Fig. 2B). GBF1 is required for COPI vesicle assembly at the ER-Golgi complex interface, while BIG1 and BIG2 function at the trans-Golgi network (TGN) and regulate the formation of AP/clathrin- and GGA/clathrin-coated vesicles delivering cargo to plasma membranes and endosomal compartments (Fig. 2A) (reviewed in reference 38).
The sensitivity of enterovirus infection to BFA inhibition was found to be dependent on GBF1: GBF1 colocalized with the replication complexes of poliovirus and CVB3 (30, 39), and replication of these viruses in the presence of BFA was rescued by the expression of wild-type or BFA-insensitive variants of GBF1 but not BIG2. Moreover, another inhibitor specific to GBF1 but not BIG1/2, Golgicide A (GCA), was shown to significantly suppress replication of diverse enteroviruses, and this inhibition could be relieved by overexpression of GBF1 (40). Furthermore, siRNA-mediated knockdown of endogenous GBF1 severely inhibited replication of poliovirus and CVB3, showing that GBF1 is required for enterovirus replication (30, 39).
Recruitment of GBF1 to replication complexes.Further support for a GBF1 function in viral replication was provided by studies showing active recruitment of GBF1 to replication complexes through interaction with the viral nonstructural protein 3A (Fig. 2C). Enterovirus 3A is a small protein with a hydrophobic C-terminal domain that mediates its binding to membranes. The strong interaction of 3A proteins of poliovirus and CVB3 with GBF1 was documented in a yeast two-hybrid system in coimmunoprecipitation (co-IP) experiments performed with mammalian cells expressing tagged 3A constructs or in productively infected cells (41–43). Expression of the 3A protein in vitro in crude HeLa cell extracts also resulted in the specific recruitment of GBF1 to membranes (44). Individual expression of poliovirus or CVB3 3A proteins induced relocalization of early Golgi markers into modified ERES structures where 3A proteins were concentrated. The effect of 3A expression on the cellular secretory organelles strongly resembled that of BFA, consistent with the targeting of GBF1 by the viral proteins (41, 45). However, despite the strong and highly specific 3A-GBF1 interaction, the functional significance of this binding in the context of infection remains unclear. Insertion of an additional serine into the N-terminal region of 3A of poliovirus or coxsackievirus B3 reduces the 3A-GBF1 interaction to undetectable levels when tested in the yeast two-hybrid system in co-IP experiments and in an in vitro expression system (39, 41, 42, 46), yet the replication of viruses with this mutation is minimally affected in cell culture under normal conditions. Another poliovirus mutant with a hemagglutinin (HA) tag inserted into the 3A sequence also showed no detectable GBF1-3A interaction in co-IP experiments, yet it also had no apparent replication defects (43). However, the poliovirus and CVB3 mutants with impaired 3A-GBF1 interactions were much more sensitive to BFA inhibition, more strongly inhibited by siRNA-mediated GBF1 knockdown, and much less responsive to rescue in the presence of BFA by GBF1 overexpression, showing that they still required GBF1 for replication (30, 39 and unpublished observations). Thus, while productive 3A-GBF1 interactions appear to promote viral replication, a minimal amount of GBF1 associated with the replication membranes (in a 3A-independent manner, such as nonspecific diffusion) may be sufficient to support viral replication in cell culture.
The problematic role of GBF1 enzymatic activity in replication.GBF1 activates multiple Arfs (Arf1, Arf3, Arf4, and Arf5) (47, 48 and unpublished observations), and viral infection in cell culture or the expression of poliovirus 3A in an in vitro system stimulates accumulation of activated Arf1, -3, and -5 on replication membranes, suggesting that GBF1 retains its catalytic activity when recruited to the replication complexes (44, 49). Moreover, expression of the catalytically inactive GBF1/E794K mutant or a GBF1 mutant lacking the catalytic Sec7 domain failed to rescue poliovirus or CVB3 replication from BFA blockage (30, 39). These findings prompted a model in which GBF1 activated Arfs, which then recruited various effectors to the replication complexes, some of which promoted viral replication (50). However, accruing evidence argues against the importance of GBF1-dependent Arf activation for picornavirus replication. First, a requirement for GBF1-dependent Arf activation in virus-induced membrane remodeling is ruled out by the observation that the expression of poliovirus proteins in the presence of BFA results in membrane remodeling indistinguishable from that observed in control cells (39). Second, activated Arfs do not seem to be required for picornavirus replication, since siRNA-mediated depletion of Arf1 had no effect on CVB3 replication and neither expression of wild-type Arf1, -3, -4, or -5 nor their constitutively activated forms could relieve BFA inhibition of enterovirus infection (30, 39). A report showing siRNA knockdown of Arf1 being strongly inhibitory for CVB3 replication must be interpreted with caution, since the level of Arf1 protein reduction was modest, ∼50%, suggesting possible off-target effects (8). Third, Crotty and colleagues identified two point mutations in the poliovirus 2C and 3A proteins that both confer partial resistance to BFA and had a synergistic effect when combined together in one genome (51). While the replication of this virus in the absence of BFA induced massive accumulation of Arf-GTP on membranes, in the presence of BFA, Arf activation was not detectable, yet the mutant virus exhibited robust replication. Importantly, the BFA-resistant phenotype of these mutants was still dependent on the presence of the GBF1 protein (52). It seems unlikely that the two point mutations conferring BFA resistance to poliovirus would enable a completely different mode of functioning of the replication complexes. Rather, the results suggest that while GBF1 is an indispensable cellular factor for poliovirus replication, Arf activation may be merely a consequence of GBF1 recruitment to the replication membranes. Support for a possible nonenzymatic role for GBF1 also comes from studies showing that the replication of Aichi virus is not sensitive to BFA but is strongly inhibited by a siRNA-mediated knockdown of GBF1 (53).
Models for GBF1 function in replication.How may GBF1 promote viral replication? Experiments with rescue of poliovirus replication from BFA blockage by expression of truncated mutants of GBF1 show that its role in the viral replication complexes may be very different from that in uninfected cells. It was shown that the intact N terminus of GBF1 was absolutely required for poliovirus replication, as deletion of just 37 amino acids from that part of GBF1 completely abolished its ability to support viral replication (Fig. 2C). Importantly, such deletion also compromises GBF1 association with membranes (unpublished observation), suggesting that membrane-bound GBF1 plays an essential role in virus-encoded processes. In contrast, GBF1 with a truncated C-terminal part up to the homology downstream of Sec7d (HDS1) domain was practically as effective in supporting replication as the full-length protein. More extensive C-terminal deletions demonstrated a sharp drop in the ability to support viral replication, but even the N-terminal fragment of GBF1 lacking the catalytic Sec7 domain showed severely reduced but detectable replication rescue in the presence of BFA. How can we reconcile the ability of this mutant to support low levels of replication with the previously mentioned experiments showing the inability of the catalytically inactive GBF1/E794K mutant to rescue poliovirus and CVB3 replication? A possible explanation may reflect the tight association of the GBF1/E794K mutant with membranes (54), making it unavailable for replication complexes, rather than its inability to perform virus-specific functions. A similar stabilization of GBF1 on membranes in the presence of BFA may explain the effect of this drug on virus replication (54, 55).
Importantly, the C-terminally truncated GBF1 mutants capable of supporting viral replication were incompetent in cellular functions of GBF1, as evidenced by their failure to rescue secretion and prevent cell mortality in the presence of BFA (52). Thus, the functional network of GBF1 in uninfected cells is not the same as the one sufficient to support picornavirus replication. This suggests that looking at “normal” GBF1 functions might be superfluous for characterizing GBF1 functions within the context of viral infection. Since the enzymatic activity of GBF1 and the C-terminal region of the protein appear dispensable for viral replication, alternative models that take into account GBF1 interactions with cellular proteins have to be considered.
GBF1 is a multidomain protein known to engage in interactions with multiple cellular partners, such as Rab1 (56), COG4 (57), p115 (58), GGA (59), COPI (60), and ATGL (61) (Fig. 2D). In addition, the yeast ortholog of GBF1, Gea1/2, has been shown to interact with GMH (62), the Trs65 component of TRAPPII (63) and Drs2 (64, 65), and based on the high sequence conservation, it is likely that mammalian GBF1 also interacts with the mammalian orthologs of these proteins. Investigation of two known interactors of GBF1, the small GTPase Rab1b (56) and the tether protein p115 (58), in the context of enterovirus infection demonstrated that the viruses do not require those GBF1 partners for replication. While Rab1b is important for recruitment of GBF1 to membranes in uninfected cells (58), knockdown of Rab1b expression had no effect on poliovirus replication (52). Similarly, overexpression of wild-type or constitutively activated Rab1b could not rescue BFA inhibition of coxsackievirus B3 replication (30). The p115 protein works as a tether to dock COPI vesicles to the target membranes and is both a Rab1B effector and a direct interactor of GBF1, reflecting the intricate coordination of the formation and targeting of cargo vesicles in noninfected cells (58, 66). Poliovirus infection results in a specific degradation of p115, suggesting that the virus destroys important interactors of GBF1 to extract the protein from its normal cellular network and make it available for virally encoded functions (52). The possible role of other GBF1 interactors, especially those binding to the N-terminal region required for replication (such as COG4 and γ-COP), in virus replication has not been tested.
In light of these findings, a model that considers all available data suggests that the GBF1 interactome might be altered in infected cells (Fig. 2E). Recruitment of GBF1 to the replication organelles may prevent its interactions with some of its normal partners and induce GBF1 interactions that never occur within an uninfected cell. It is likely that such novel interactions occur through the N-terminal region of GBF1 required for viral replication. It is also likely that the virus-mandated interactions have an enzymatic component, because minimal amounts of GBF1 are sufficient for replication. The finding that GBF1 without its Sec7 supports replication negates a model in which GBF1 in infected cells would alter its catalytic properties or substrate specificity. A better model posits that GBF1 acquires an enzymatically active partner that would provide an amplification of the limited functionality of the virus genome. If GBF1 interactions are indeed reprogrammed in infected cells, this raises an interesting question of how GBF1 interactions with those partners are blocked in uninfected cells.
The puzzling division of picornaviruses into those that are strongly dependent on GBF1, like enteroviruses, and those that apparently do not require this factor, like cardioviruses (24, 30), raises a question of whether picornaviruses use significantly different strategies for replication or whether cardioviruses substitute a GBF1-provided function with some other mechanism. What is the role of GBF1 in the replication of other positive-strand RNA viruses, like hepatitis C virus and coronaviruses, that also depend on GBF1 (30, 67, 68)? Is GBF1 function in the replication of these viruses analogous to its role in enterovirus replication, suggesting a conserved core of replication requirements shared by diverse positive-strand RNA viruses, or is GBF1 supporting different steps in the replication cycle of different viruses?
Other BFA-sensitive GEFs in picornavirus replication.BIG1, another BFA-sensitive GEF with a domain structure analogous to that of GBF1, also specifically translocates to the replication membranes in poliovirus-infected cells (44). Expression of only the 3CD poliovirus protein was shown to specifically induce association with membranes of both BIG1 and the closely related BIG2 (69). BIG1/2 normally coordinate the formation of clathrin-coated vesicles at the trans-Golgi network membranes (Fig. 2A) (38). The 3CD-induced translocation of these GEFs to membranes resulted in strong activation of Arf1 and the recruitment of the GGA adaptor proteins, the normal Arf1 effectors in the BIG1- and BIG2-regulated pathways (69). The 3CD protein is an uncleaved precursor of the protease 3C and the viral RNA-dependent RNA polymerase 3D. The 3CD-specific recruitment of BIG1/2 was not dependent on 3CD protease activity and could not be reconstituted by expression of 3C and 3D proteins together. Direct 3CD-BIG1 or 3CD-BIG2 interaction was not detected in a yeast two-hybrid screen, but mutations in 3CD that disrupt its ability to recruit the GEFs severely compromised poliovirus replication (69). However, it is impossible to ascertain whether the inhibitory effect on replication is a result of the lack of BIG1/2 recruitment or of other defects in the replication complexes associated with these mutations. Thus, while the recruitment of BIG1/2 correlates with efficient viral replication, the functional role of BIG1/2 in replication is currently unknown.
ROLE OF PHOSPHATIDYLINOSITOL-MODIFYING ENZYMES IN PICORNAVIRUS REPLICATION
Derivatives of phosphatidylinositol play major roles in providing molecular signatures to the distinct membranous domains of a cell. Phosphatidylinositol 4-phosphate [PtdIns(4)P], which constitutes less than one per cent of total cellular phospholipids, is specifically enriched on Golgi membranes, where it plays an important role in maintaining Golgi architecture and coordinating secretory trafficking and sphingolipid synthesis. Golgi-specific PtdIns(4)P synthesis is attributed to phosphatidylinositol kinases PI4KIIα, PI4KIIIα, and PI4KIIIβ (reviewed in reference 70). PI4KIIIα and PI4KIIIβ have recently been documented as important host factors required by diverse viruses, and replication of picornaviruses seems to specifically require PI4KIIIβ.
Recruitment of PI4KIIIβ to replication complexes.PI4KIIIβ, but not PI4KIIIα, is significantly enriched at the replication sites of CVB3 and poliovirus, and immunoprecipitation with anti-PI4KIIIβ antibodies recovered CVB3 nonstructural proteins, suggesting that PI4KIIIβ is a component of the replication complex (8). Two hypothetical mechanisms have been proposed for recruitment of PI4KIIIβ to replication membranes. The first model suggests that recruitment of PI4KIIIβ to replication complexes may be mediated by activated Arf1. PI4KIIIβ is a known Arf1 effector (71), and its redistribution to replication membranes is dependent on the expression of enterovirus protein 3A. Thus, it seemed logical to propose that the 3A-dependent recruitment of GBF1 to replication membranes and the resulting accumulation of activated Arf1 would stimulate PI4KIIIβ recruitment (8). A second model is based on the finding that individually expressed 3A proteins from as diverse picornaviruses as Aichi virus, bovine kobuvirus, human rhinovirus 14, poliovirus, and several coxsackie B viruses copurify with PI4KIIIβ and acyl coenzyme A (acyl-CoA) binding domain protein 3 (ACBD3). ACBD3 in turn can bind PI4KIIIβ independently of 3A, suggesting that ACBD3 may act as an adaptor for 3As to recruit PI4KIIIβ (53, 72). However, recent data cast doubts on the strength of both models. Overall, the data on the role of ACBD3 in picornavirus infection are controversial: ACBD3 depletion was inhibitory to replication of Aichi virus and poliovirus (53, 72), but CVB3 replicated normally in ACBD3-depleted cells (46). Moreover, it appears that recruitment of ACBD3 to replication complexes may exert an antiviral effect in the case of poliovirus infection and is probably a part of a cellular response to infection rather than a mechanism utilized by the virus to facilitate replication (73). A recent report showed that PI4KIIIβ may interact with poliovirus protein 2BC in a mammalian two-hybrid system, but whether this interaction is important during infection and whether it can provide a 3A-independent way of PI4KIIIβ recruitment requires further investigation (74).
Rigorous examination of the mechanism of PI4KIIIβ recruitment by CVB3 protein 3A revealed that it is not dependent on GBF1, Arf1, or ACBD3. Neither siRNA-mediated knockdown of these proteins nor inhibition of GBF1-dependent Arf1 activation by BFA prevented the recruitment of PI4KIIIβ to membranes upon expression of 3A. Even a 3A mutant that lost its ability to bind GBF1 was able to recruit PI4KIIIβ normally in ACBD3-depleted cells and in an in vitro assay (46). Thus, although 3A dependent, the mechanisms that position PI4KIIIβ at replication sites remain to be determined.
Role of PI4KIIIβ and PtdIns(4)P lipid in viral replication.Knockdown or chemical inhibition of PI4KIIIβ is detrimental to the replication of diverse picornaviruses (8, 53, 72, 75, 76), showing that this enzyme may be an essential part of a conserved core replication machinery. The virus-induced recruitment of PI4KIIIβ to replication complexes resulted in a significant accumulation of PtdIns(4)P lipid on membranes. Inhibition of PI4KIIIβ activity by the small-molecule inhibitor PIK93, the expression of a dominant negative mutant of the kinase, or the expression of SacI phosphatase, which removes the phosphate from PtdIns(4)P, all resulted in a noticeable inhibition of viral replication (8). Further support for the importance of the enzymatic activity of PI4KIIIβ for picornavirus replication came from the observations that other compounds inhibiting this enzyme have a strong antiviral effect (76, 77) and that replication can be rescued by the expression of catalytically active, but not inactive, PI4KIIIβ mutants (76).
How may increased PtdIns(4)P content of a membrane contribute to viral replication? Distinct PtdIns(P)s are specifically recognized by select classes of binding domains found in many proteins, and this recognition facilitates spatial compartmentalization of cellular signaling and metabolic networks (reviewed in references 78 and 79). Thus, it is possible that increased concentrations of PtdIns(4)P on the replication membranes may promote recruitment of distinct viral and cellular proteins to support viral replication. It has been proposed that at least in the case of enteroviruses the enrichment of PtdIns(4)P lipid on replication membranes may help to anchor the viral RNA-dependent RNA polymerase 3D, since purified recombinant poliovirus 3D protein preferentially bound to PtdIns(4)P rather than to other lipids in a biochemical spot assay (8). However, arguing against a direct role of PtdIns(4)P in 3D recruitment is the fact that poliovirus and CVB3 mutants selected in the presence of PI4KIIIβ-inhibitng compounds invariably have single-point mutations in the 3A protein but never in the 3D polymerase (77, 80). Three single-amino-acid mutations independently conferring resistance of enterovirus replication to PI4KIIIβ inhibitors were described, one in the hydrophobic, membrane-interacting domain and two in the cytosolic part of 3A (77, 80). CVB3 bearing such mutations replicated with wild-type kinetics in the absence of inhibitors and was no longer dependent on PI4KIIIβ or any other PtdIns(4)P-generating activity, as evidenced by the lack of PtdIns(4)P accumulation on replication membranes (80). Loss of PI4KIIIβ dependence did not decrease replication fitness in cell culture or virulence in CVB3-infected mice (81).
The ease of emergence of mutants resistant to PI4KIIIβ or GBF1 inhibitors (77, 80, 82) raises an important question of how a requirement for presumably indispensable activities of cellular factors can be overcome by a single-amino-acid change in a viral protein. These findings may indicate that the contribution of PI4KIIIβ, just like that of GBF1, to viral replication may be very different from their known enzymatic activity and function in uninfected cells. Elevated levels of PtdIns(4)P as well as activated Arfs represent a background landscape that the replication machinery have adapted to function in because of the needed recruitment of PI4KIIIβ and GBF1. This would explain the strong effect of compounds like BFA or PI4KIIIβ inhibitors on the replication of wild-type viruses, since these compounds alter the membrane microenvironment of the replication complexes. At the same time, the virus can rapidly adapt to such changes by minor adjustments of the membrane-interacting properties of viral replication complex components.
Oxysterol binding proteins modulate the PI4KIIIβ-dependent pathway in picornavirus replication.The importance of specific biochemical properties of membranes for viral replication is further suggested by studies on different compounds that target the PI4KIIIβ-dependent step in picornavirus replication. These diverse molecules are now somewhat misleadingly called enviroxime-like compounds, since the mutation in the 3A protein that rescues viral replication in their presence in picornaviruses resistant to enviroxime, a candidate antirhinovirus agent, was earlier described (77, 80, 82, 83). A genetic screen for factors whose knockdown increased the antiviral potency of some enviroxime-like compounds identified oxysterol binding proteins 1 and 2 (OSBP) (84). These OSBP-dependent inhibitors were designated minor enviroxime-like compounds as opposed to the major enviroxime-like compounds that are presumed to act by directly inhibiting PI4KIIIβ (77).
OSBP-1 and OSBP-2 are members of a multigene protein family containing a C-terminal sterol binding domain and a plecstrin homology (PH) domain that target these proteins to the Golgi complex in a PtdIns4P-dependent manner. OSBP also localizes to the ER via its FFAT domain interacting with the integral ER proteins VAP-A and VAP-B (85). Within these compartments, OSBP responds to levels of cellular cholesterol and oxysterol by modulating sphingomyelin (SM) biosynthesis (86, 87). In addition, OSBP has been shown to work as a lipid transporter to exchange cholesterol synthesized at the ER for PtdIns(4)P generated at the Golgi complex (88). Thus, even a small imbalance in OSBP function would have large consequences in global lipid metabolism. It is likely that the minor enviroxime-like compounds may mimic the natural cholesterol ligand of OSBP and perturb the normal function of OSBP in lipid homeostasis. The data on the involvement of OSBP in poliovirus infection support the important role of cholesterol in the development of the replication organelles (see below). Reported effects of OSBP depletion on poliovirus replication seem to strongly depend on the experimental system. Arita and coworkers observed that while siRNA knockdown of OSBP in HEK 293 cells made replication more sensitive to the minor enviroxime-like compounds, reduction of OSBP expression in the absence of the inhibitors had a minimal effect on viral replication (84). At the same time, short hairpin RNA (shRNA)-mediated inhibition of OSBP expression in HeLa cells seemed to noticeably inhibit poliovirus propagation (89). Likely, compensatory mechanisms exist to protect cells from the loss of a single OSBP protein.
PICORNAVIRUS-INDUCED ALTERATIONS IN LIPID METABOLIC PATHWAYS
The accumulating data show that picornavirus infection results in a major reorganization of lipid biosynthetic and delivery pathways and generates replication membranes with a unique lipid composition not found in uninfected cells. Glycerophospholipids are the major structural components of eukaryotic membranes, with phosphatidylcholine constituting about 50% of the total phospholipid. Picornavirus infection stimulates de novo synthesis of phospholipids, and the newly incorporated labeled precursors of lipid synthesis such as glycerol and choline are found in the replication organelles (28, 90–92). Membrane phospholipids include fatty acids with long (C16 and longer) carbon atom chains (reviewed in references 93 and 94). The long-chain fatty acids necessary for lipid synthesis can originate from intracellular sources or be imported from the extracellular medium. All mammalian cells have a functional pathway for de novo synthesis of long-chain fatty acids by fatty acid synthase (FASN). The activity of cellular lipolytic enzymes also releases free long-chain fatty acids that can be recycled into lipid synthesis pathways. Import of exogenous fatty acids into cells is tightly coupled to the activity of acyl-CoA synthetases. According to the vectorial acylation model, long-chain fatty acids as hydrophobic molecules can freely diffuse through the plasma membrane, while their mobilization into the hydrophilic acyl-CoAs inside the cell prevents their escape (95). Acyl-CoAs serve as acyl donors for downstream metabolic processes, including lipid biosynthesis.
Activation of acyl-CoA synthetases and fatty acid import.Recent findings reveal rapid activation of cellular long-chain acyl-CoA-synthesizing activity in picornavirus-infected cells and the rerouting of the major metabolic destination of the imported fatty acids from triglyceride synthesis and storage in lipid droplets in uninfected cells to highly stimulated phosphatidylcholine production upon infection (Fig. 3). Infected cells import more fatty acids than uninfected cells, and the spectrum of imported fatty acids is different, resulting in the synthesis of an infection-specific pool of phosphatidylcholine, noticeably distinguishing the total membrane composition between infected and noninfected cells (5). It is unclear whether the imported or the endogenously generated fatty acids represent the major source of building blocks for infection-specific phospholipid synthesis. The importance of the imported fatty acids for the development of poliovirus replication membranes is indicated by experiments showing that providing the infected cells with an excess of oleic acid results in inhibition of the viral replication concomitant, with changes in the physical properties of the replication membranes (96). Based mostly on the results of inhibition of the FASN by chemical inhibitors, it was suggested that the de novo synthesis of long-chain fatty acids may be hijacked for membrane development in picornavirus-infected cells (97, 98). Recently, FASN relocalization to the replication membranes and an increase in its enzymatic activity were shown to be important for the replication of some flaviviruses (99, 100), a group of positive-strand RNA viruses evolutionarily rather distant from picornaviruses, suggesting that diverse viruses may require de novo synthesis of long-chain fatty acids.
Virus-induced changes in lipid metabolism. Uninfected cells channel the majority of fatty acids into triglycerides to be stored in lipid droplets. In picornavirus-infected cells, the import of long-chain fatty acids is highly activated, but triglyceride synthesis is inhibited. This fuels the increased production of phosphatidylcholine to generate the membranous scaffold of the replication organelles.
The activation of fatty acid import in poliovirus-infected cells requires the viral protease 2A. Interestingly, the 2A-dependent stimulation of fatty acid import did not depend on the protease activity but required input from other nonstructural poliovirus proteins, since the expression of 2A alone was not sufficient to induce the activation. A similar stimulation of fatty acid import was observed in different cell types infected with diverse picornaviruses, arguing that this cellular response represents a conserved mechanism underlying the stimulation of membrane synthesis and the development of replication organelles (5). However, the detailed mechanism of activation of long-chain acyl-CoA synthetases and fatty acid import in infected cells remains to be established.
Role of new membrane synthesis in the development of replication organelles.Activation of long-chain acyl-CoA synthetases and inhibition of triglyceride synthesis seem to be key steps sufficient to activate infection-specific phospholipid production. The increased supply of long-chain acyl-CoAs in infected cells would result in the accumulation of diacylglycerol, which is the common precursor of both phosphatidylcholine and triglycerides. However, due to inhibited triglyceride synthesis in infected cells, diacylglycerol would be consumed in increased production of phosphatidylcholine (Fig. 3). Interestingly, the initial emergence of poliovirus-specific membranous structures is associated with the Golgi complex (7–9), an organelle enriched in diacylglycerol (101–103), suggesting that establishment of viral replication may benefit from elevated levels of this lipid. The well-documented juxtaposition of the picornavirus replication structures to ER tubules later in infection correlates well with the need for infection-specific activation of the cellular phospholipid-synthesizing machinery which is associated with ER membranes (9, 104, 105). Continuously synthesized phosphatidylcholine would result in the extrusion of the new membrane material. In the aqueous cytoplasmic environment, these new phosphatidylcholine-enriched membranes would spontaneously assemble into characteristic convoluted tubular structures typical of the replication organelles (5, 7, 11), similar to the growth of myelin figures (106), without the requirement for additional input from another membrane-shaping mechanism(s). Thus, it is likely that the activated synthesis of new phospholipids in infected cells may be the major driving force supporting the development of the structural scaffold of the replication organelles.
Enrichment of the replication organelles in cholesterol.The infection-specific rerouting of lipid biosynthetic pathways would disturb normal lipid homeostasis and likely induce activation of compensatory regulatory networks. Indeed, it appears that cholesterol trafficking undergoes dramatic reorganization in enterovirus-infected cells. In uninfected cells, cholesterol is concentrated mostly on the plasma membrane, and intracellular membranes are relatively cholesterol poor, but the development of the CVB3 replication organelles activates reverse trafficking of cholesterol from the plasma membrane to the replication structures, generating a highly unusual enrichment of the intracellular membranes with cholesterol. Interestingly, the intracellular esterified cholesterol stored in lipid droplets apparently is not easily accessible during infection, confirming that metabolism of neutral lipids is severely compromised in infected cells (107).
Collectively, these data strongly suggest that picornavirus membranous replication structures are not the product of remodeling of preexisting organelles by the membrane-shaping machinery borrowed from the secretory pathway, but rather that replication organelles are in large part generated by the de novo synthesis of membranes with unique protein and lipid compositions. The relative concentrations of specific lipids within a membrane regulate its properties such as thickness, permeability, and fluidity, properties that underlie membrane functions as biological barriers and platforms for the assembly of protein complexes. It is likely that the unique composition of the replication membranes is important for the recruitment of viral replication proteins as well as for the association of a specific set of cellular factors that may never be found together on organelles in uninfected cells.
CONCLUDING REMARKS
We are only beginning to appreciate, let alone understand, the ingenious ways picornaviruses evolved to trick and manipulate host membrane homeostasis. Picorna-like viruses are believed to have emerged before the diversification of eukaryotes (108). Over eons of evolution, they apparently perfected the usurping of key elements of membrane metabolism to induce the rearrangement of self-organizing host modules, thus generating novel membranous replication platforms using minimal genome resources. We still cannot answer the fundamental questions of why positive-strand RNA viruses use membranes for replication and whether it is a basic requirement of the replication machinery or a means of escaping antiviral responses or both. We also do not understand the relationship between the various stages of membrane remodeling and the different steps of the viral life cycle, such as genome RNA translation, viral polyprotein processing, RNA replication, and virus assembly and packaging.
It is becoming clear that the role of at least some of the cellular membrane metabolism proteins recruited for viral replication is not directly related to their known activities in normal cellular processes. Emerging data suggest that cellular proteins like GBF1 and PI4KIIIβ may support viral replication independently of their established enzymatic activities; therefore, antiviral strategies based on inhibiting catalytic activities might be counterproductive. Indeed, the antiviral pressure of compounds targeting the “normal” cellular functions like Arf1 activation by GBF1 and PtdIns(4)P production by PI4KIIIβ is easily overcome by picornaviruses with single-amino-acid substitutions. Thus, understanding the virus-specific functions of cellular factors will be critical for developing better antiviral strategies.
Our knowledge of the lipid composition and lipid-lipid and lipid-protein interactions in the replication organelles as well as our understanding of the mechanisms that viruses use to rewire the lipid biosynthetic pathways is rudimentary to nonexistent. The role of sphingolipid metabolism in picornavirus replication is not known at all, while the association of early replication sites with Golgi membranes (7–9), where sphingolpids are synthesized, as well as the emerging role of cholesterol trafficking (107), which is intrinsically connected to sphingolipids, indicates that we may be missing key components that shape the structure and function of replication organelles. Future studies aimed at understanding how the expression of just a few viral proteins overruns the elaborate cellular controls of membrane homeostasis are needed to advance our knowledge of the fundamental mechanisms of cell biology and will be indispensable for the development of a new generation of antiviral therapeutics.
ACKNOWLEDGMENTS
We thank Ellie Ehrenfeld, Carolyn Machamer, Casey Morrow, and unknown referees for their insightful and critical comments.
FOOTNOTES
- Accepted manuscript posted online 11 June 2014.
- Address correspondence to George A. Belov, gbelov{at}umd.edu, or Elizabeth Sztul, esztul{at}uab.edu.
REFERENCES
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.↵
- 96.↵
- 97.↵
- 98.↵
- 99.↵
- 100.↵
- 101.↵
- 102.↵
- 103.↵
- 104.↵
- 105.↵
- 106.↵
- 107.↵
- 108.↵
- Copyright © 2014, American Society for Microbiology. All Rights Reserved.