In a previous study, we found that a chimeric poliovirus in which the predicted N-terminal amphipathic helix in protein 2C was replaced with the corresponding region from HRV14 produced very small plaques on HeLa cell monolayers and displayed delayed kinetics and low yields of RNA in a single-step growth cycle. Among the small plaques, large plaque variants appeared with relatively high frequency. Sequencing through the region encoding protein 2C identified several mutations that accumulated in genomes with improved growth properties. However, the majority of viruses that emerged with improved growth properties (approximately 60%) contained no reversions within the HRV14 amphipathic helix-encoding sequence or anywhere else in the region encoding 2C. In the present study, we performed sequence analyses on those pseudorevertant viral genomes that revealed nucleotide changes in sequences encoding protein 2B or 3A. The sequence changes in the 2B or 3A protein, as well as previously reported changes in protein 2C (31
), were not conserved among the isolates. Thus, there was no common reversion strategy; rather, independent mutations could compensate for the poor replication of the poliovirus chimeric genome encoding the HRV14 amphipathic helix in protein 2C. When representative mutations found in the RNA sequences coding for either the 2B or the 3A protein were reconstructed in the original chimeric genome, they were shown to restore growth to normal or near-normal levels. These data suggested that proteins containing 2B or 3A sequences physically interact with 2C (or its precursors) within the replication complex and that such interaction was perturbed when the N-terminal amphipathic helix in 2C was replaced with the HRV14 sequence but was restored by the compensatory changes in either 2B or 3A. Interestingly, a similar conclusion was reached based on the analysis of chimeric PV genomes in which the hydrophobic region in protein 3A was exchanged for the orthologous element from HRV14. Those chimeric viruses also were viable but demonstrated defects in replication that were improved by secondary mutations in 2B-encoding sequences (34
). The postulated interaction between 2C and/or 2BC protein and 3A protein seems quite plausible, since all these proteins are localized together in membranous replication complexes. Moreover, it has been shown previously that when 3A is expressed together with 2BC protein, 3A localizes to a lower-density membrane fraction, characteristic of 2BC-containing membranes, than when expressed alone. These data prompted the suggestion that the ensemble of proteins 2BC and 3A stimulates the formation of membranous vesicles most closely resembling the vesicles induced during poliovirus infection (28
). Taken together, the results reported from our and others' laboratories indicate a network of functional interactions among 2B, 2C, and 3A protein sequences likely involving the induced membrane structures that form the PV replication complex.
To test whether the identification of compensating mutations resulted from direct physical interactions among the viral proteins, we utilized a mammalian two-hybrid system to demonstrate specific interactions. We performed analyses of homo- and heteromultimerization of PV proteins 2B, 2C, 2BC, 3A, and 3AB in COS cells. This study demonstrated previously unidentified protein-protein interactions between proteins from the P2 and P3 regions of the poliovirus genome. The yeast two-hybrid system was applied previously to catalogue interactions between the nonstructural proteins within the P2 or P3 regions (5
), but interactions between proteins from the P2 and P3 regions were not examined. A mammalian two-hybrid analysis was performed for the closely related coxsackie B3 virus proteins 2B, 2C, and 2BC (6
). More recently, protein-protein interactions for nonstructural proteins of porcine teschovirus 1, a representative of a different genus of picornavirus, were studied in a yeast two-hybrid system (43
). None of the porcine teschovirus P2 proteins were shown to interact with any P3 proteins tested; in that study, however, analysis was performed only with mutant forms of proteins 2BC, 2C, and 3A, which may have affected the results. Table shows a summary and comparison of the results obtained in all these studies. The majority of interactions observed in our current study are consistent with those subsets examined by others. Major differences were seen in interactions of 3AB, especially in homodimerization of this protein, which gave a very strong positive signal in the previous yeast two-hybrid analysis (42
). The failure to detect such interaction in our experiments may have been due to relatively poor expression of Act-3AB in our studies. On the other hand, evidence for 2C-2C multimerization was not detected by others in two-hybrid analyses, and although signals in the present study were relatively low, they were reproducibly observed. 2C homodimerization has been demonstrated previously by glutathione S
-transferase pulldown assay (5
); multimerization was proposed on the basis of genetic studies of resistance to guanidine-HCl determined by 2C (33
); and 2C and 2B clusters were visualized by immunoelectron microscopy of 2B and 2C protein sequences on isolated vesicular membranes of the replication complex (12
). A limitation of the two-hybrid methodology for identifying protein interactions, regardless of whether yeast or mammalian cell systems are utilized, is the probability of missing weak interactions, which may be extremely important, especially in regulatory pathways (41
Comparison of picornavirus P2 and P3 protein-protein interactions observed in different studies
Analysis of the effect of replacement of the PV N-terminal amphipathic helix on the demonstrated interactions of 2C-containing proteins revealed an apparent increase in chimeric 2C dimerization. This homodimerization was restored to wild-type levels of binding by a single point mutation at residue 28 (L28F) within the HRV14 sequence, which we had shown previously to also restore the wild-type replication phenotype. In addition, the interactions between PV 3A with the chimeric 2BC protein containing the amphipathic helix from HRV14, or with the PV 2BC protein harboring the single F28L mutation in the 2C sequence, were both reproducibly decreased compared to 3A interactions with the wild-type PV 2BC protein. Again, reversion at 2C-L28F in the chimeric 2BC protein, which rescued the growth of chimeric virus, significantly increased the 2BC-3A interactions. It has also been reported that mutations in protein 2C of HRV16 were sufficient for adaptation of this virus to grow in mouse cells. Interestingly, some of these mutations were shown to affect the dimerization of 2BC protein, as measured by yeast two-hybrid analysis (15
It remains unclear whether amino acid residues in protein 3A that are changed in viruses with improved growth properties participate directly in interactions with the N-terminal amphipathic helix in protein 2C or if the effect of these mutations is indirect. The amino acid substitutions found in pseudorevertant viruses generally represent conservative changes, and they are sometimes found in corresponding positions of other picornaviruses; however, they are not specific for HRV14. The structure of the soluble domain of the PV protein 3A has been determined by nuclear magnetic resonance spectroscopy (27
). The compensating mutation that we found in virus R9 (K9R) is located in the unstructured N-terminal portion of 3A. Other mutations in this region affect 3A's ability to inhibit ER-to-Golgi traffic (7
) and 3AB's stimulation of 3CD protease activity (20
). The amino acid substitution found in R10 (K39E) is in a solvent-exposed region shown to participate in homodimerization of 3A in solution. Compensating mutation S48R in R11 is located at the junction between a central structured region and an unstructured region linking to the membrane-binding domain. Thus, it is difficult to predict how these mutations would affect the structure of 3A.
For enterovirus 2B proteins, two hydrophobic regions have been identified (3
). One of these regions (PV 2B residues 32 to 55) is predicted to form a partially amphipathic cationic helix; the second region (PV 2B residues 61 to 81) is thought to be a transmembrane domain. Mutations that alter either the amphipathic character of the first domain or the hydrophobicity of the second domain have been shown to interfere with the ability of 2B to increase membrane permeability and with viral RNA replication (2
). Two of the three mutations in the 2B sequence described in this study are located in the first hydrophobic region; each causes substitution of a polar amino acid by a more hydrophobic one. More detailed structural information will be necessary to understand the effects of the described mutations on the architecture of poliovirus RNA replication complexes.
Although replication of all positive-strand RNA viruses depends upon the reorganization of host intracellular membranes, different viruses utilize different organelle membranes and generate different morphological structures upon which replication complexes assemble. It has been shown that the N-terminal region of nodavirus flock house virus protein A, which normally targets the protein to mitochondrial membranes (18
), can be readily replaced by viral or cellular ER-targeting sequences and thus establish replication complexes of greatly differing morphologies without affecting the ability of the virus to replicate. It is not known whether or how the specific morphology of membrane structures induced by PV proteins affects their function in viral RNA replication. In the studies reported here and in our previous publication, the chimeric viruses that generated functional viral replication complexes also formed vesicles with the morphological characteristics of those induced by wild-type PV proteins. Nonfunctional or viable but defective replication complexes (e.g., those formed by PV-2Cah-HRV14) appeared altered in their morphological aspects but exhibited the typical PV vesicle clusters when function was restored by mutations in the same or other proteins that restored replication activity. Thus, for the replication-defective viruses analyzed in these studies, we observed a positive correlation between the induced membrane structures and the capacity to replicate the virus genome.