Determinants of membrane association for poliovirus protein 3AB.

Poliovirus protein 3AB may serve as the lipophilic carrier of a protein primer (VPg or 3B) used for the initiation of genomic viral RNA synthesis. In order to study the membrane-protein interactions of 3AB required for its role in poliovirus RNA replication, we have developed an in vitro membrane association assay capable of distinguishing membrane-bound from non-membrane-bound proteins that are cotranslated together in the presence of canine microsomal membranes. This assay utilizes equilibrium sedimentation analysis in high density sucrose gradients to measure membrane association of both wild type and mutated forms of 3AB. Using this assay and other biochemical assays, we have identified the following properties of the 3AB-membrane interaction: (a) 3AB is able to post-translationally associate with microsomal membranes, (b) 3AB is able to associate with membranes in a manner consistent with that of an integral membrane protein, (c) 3AB contains a critical hydrophobic sequence within the carboxyl-terminal half of the protein that is required for membrane association, and (d) the introduction of charged residues into this hydrophobic sequence disrupts the 3AB membrane-protein interaction. Taken together, these studies indicate that poliovirus protein 3AB associates tightly with biological membranes de novo in a manner that would allow it to serve as a lipophilic anchor for the assembly of the poliovirus RNA replication complex.

Cellular membranes are complex structures composed of lipid and protein that together compartmentalize the intracellular space and serve as the physical matrix upon which numerous biosynthetic events are performed. Intracellular membranes are also utilized by some positive strand RNA viruses to organize and facilitate their viral RNA replication processes. This has been shown for members of the picornavirus family, such as poliovirus (1), as well as members of the alphavirus family (2) and potyvirus family (3). Poliovirus RNA synthesis takes place in tight replication complexes that are associated with virus-induced smooth membrane vesicles that bud from the rough endoplasmic reticulum (1, 4 -8). Membrane association of this viral replication complex is presumably mediated by the poliovirus proteins 2C and 3AB (9,10,11), possibly in the form of larger 2C-and 3AB-containing precursor proteins. An interesting feature of poliovirus RNA replication is that all newly synthesized RNA chains contain a small, highly basic protein, VPg (or 3B), covalently attached to their 5Ј end. It has therefore been postulated that one function of 3AB is to serve as the lipophilic carrier of VPg (3B) to the viral RNA replication machinery. The primary candidate region for the 3AB membrane association determinant is a highly conserved 22-amino acid hydrophobic domain present in the COOH-terminal half of 3A ( Fig. 1) (9,12). The presence of this hydrophobic domain corresponds with the ability to isolate the COOH-terminal amino acids of 3A, along with 3B, as part of a larger protease-protected fragment from extracts of poliovirus-infected cells (13).
Membrane-protein interactions can be typically separated into two categories, peripheral and integral (14,15). Peripheral membrane proteins generally do not interact significantly with the hydrophobic interior of the lipid bilayer, while integral membrane proteins do interact, either through a transmembrane region or a hairpin loop that does not pass through the bilayer (16 -18). Peripheral membrane proteins associate with membranes through electrostatic interactions, typically between positively charged amino acids and negatively charged phospholipid headgroups (16). These interactions can be further strengthened by the presence of covalently bound fatty acids or phospholipids such as myristic acid or glycosylphosphatidylinositol, which provide additional hydrophobic interactions (16).
The goal of this study is to biochemically define the 3ABmembrane interaction. Previous experiments that examined the membrane association of poliovirus replication complexes indicated that the viral protein 3AB was tightly associated with smooth membranes of infected cells in a manner resistant to treatment with 0.5 M salt or 4.0 M urea (19). However, these studies were not able to demonstrate that this tight membrane interaction was the sole result of molecular determinants contained within the 3AB protein. An in-depth biochemical analysis of the determinants responsible for the 3AB membrane interaction in the absence of the viral replication complex has not been performed, nor has the inherent strength of the 3ABmembrane interaction been examined. In this study, we address the following questions. (a) Can 3AB associate with membranes when expressed in the absence of functional poliovirus replication complexes? (b) Can 3AB associate with membranes post-translationally? (c) What are the minimum molecular determinants for 3AB membrane association? (d) Is 3AB-membrane association consistent with that of a peripheral or integral membrane-protein interaction?

MATERIALS AND METHODS
Plasmids and Cloning-All amino acid substitutions and deletions were ultimately cloned into pTM1 (20) for generation of RNA transcripts to be used for in vitro translations. Amino acid substitutions in 3AB were initially generated by cassette mutagenesis as described by Giachetti et al. (21,22), and cloned into a poliovirus subgenomic cDNA termed pKO⌬ (PV nucleotides 4154 -7053 with nucleotides 6056 -6516 deleted). Using the polymerase chain reaction (PCR) 1 and the PCR primers JT3A5106ϩ (5Ј-TTCAAGGACCACTCCAGTATAAAG-3Ј) and * This work was supported by Public Health Service Grant AI22693 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. Tel.: 714-824-7573; Fax: 714-824-8598; E-mail: blsemler@uci.edu. JT3BBAMH1-(5Ј-AGTTGGATCCTATTGTACCTTTGCTG-3Ј), the coding region for 3AB was amplified such that an in-frame stop codon was placed after the ultimate glutamine residue in 3B (VPg). The amplified 3AB sequence was then digested with AvaII and blunt-ended using the Klenow fragment of Escherichia coli DNA polymerase. The 3AB DNAcontaining fragments were then digested with BamHI, gel-purified, and ligated into purified pTM1 (digested with EcoRI and BamHI) along with the annealed oligonucleotides encoding the IBI-Kodak FLAG epitope (5Ј-AATTCTTGACTACAAGGACGACGATGACAAGG-3Ј) and (5Ј-CCT-TGTCATCGTCGTCCTTGTAGTCAAG-3Ј) using T4 DNA ligase. When annealed, the FLAG-encoding oligonucleotides contained a 5Ј-overhang compatible with an EcoRI site. The resulting pTM1FG3AB construct encoded Met-Gly-Ile-Leu followed by the FLAG epitope fused in frame and 5Ј-proximal to the entire 3AB coding sequence (see Fig. 2, top). When the wild type 3AB sequence was cloned into pTM1 as described above, the resulting plasmid was called pTM1FG3AB-wt. Deletions of domains I and IϩII within 3AB were initiated by digesting pTM1FG3AB with either BglI or HindIII. The ends of the digested DNA were then repaired using T4 DNA polymerase, followed by digestion with EagI (EagI cuts once in the pTM1 vector sequence). The BglI to EagI and HindIII to EagI fragments were isolated and ligated into a purified PstI to EagI pTM1FG3AB fragment using T4 DNA ligase. This procedure was used to generate pTM1FG3AB-DI⌬ and pTM1FG3AB-D(IϩII)⌬, respectively. The plasmid pTM1FG3AB-DII⌬ was generated by digesting pKO⌬ with BglI (blunt-ended by incubation with the Klenow fragment of E. coli DNA polymerase) and BamHI, and the resulting fragment was ligated into purified BamHI-and HindIII-cut pKO⌬ DNA (the HindIII site was blunt-ended). The resulting pKO⌬ was used as the DNA template for the PCR amplification and subsequent cloning of 3AB-DII⌬ into pTM1 (along with the FLAG epitope sequence) as described above, resulting in the construct pTM1FG3AB-DII⌬.
The plasmid encoding rabbit cytochrome b 5 was generously provided by Dr. Alan Steggles. The cytochrome b 5 cDNA was first digested with EcoRI and HincII to release the entire cytochrome b 5 coding region including the UGA termination codon. This DNA fragment was then digested with HinP1I and blunt-ended. The cytochrome b 5 cDNA fragment was ligated, along with oligonucleotides encoding the FLAG epitope described above, into gel-purified pTM1 vector sequence cut with EcoRI and StuI. This strategy removes the nucleotides encoding the first two amino acids of cytochrome b 5 and places the sequences encoding the FLAG epitope in frame and 5Ј proximal to the cytochrome b 5 cDNA. The resulting plasmid was designated pTM1FGCytb5.
The plasmid encoding human ␤-globin was the kind gift of Drs. Michael Green and Maria Zapp. The human ␤-globin cDNA was first digested with ApaLI and MseI and then blunt-ended. The ApaLI to MseI fragment was gel-purified and cloned with the annealed oligonucleotides encoding the FLAG epitope into purified pTM1 that had to be digested with both EcoRI and StuI. This strategy removed the nucleotides encoding the first amino acid of ␤-globin and placed the nucleotides encoding the FLAG epitope in-frame and 5Ј proximal to the sequences encoding human ␤-globin. The resulting plasmid was designated pTM1FG␤-Gl.
In Vitro Transcriptions-Before transcription, all plasmids were linearized (using StuI for pTM1FG3AB and HincII for pTM1FG␤-Gl and pTM1FGCytb 5 ) and gel-purified. RNA was synthesized as described by Charini et al.,(23) with the following exceptions: (a) the reaction mixtures were in 100-l reactions, (b) the final concentration of all four NTPs was 0.5 mM, (c) the amount of transcription template was 1-2 g of appropriate transcription template, (d) the incubations were carried out at 37°C for 1.5-2.0 h, and (e) the RNA transcripts were not subsequently treated with DNase I. Following transcription, the RNAs were extracted twice with phenol/chloroform and ethanol-precipitated two times using ammonium acetate followed by one time using sodium acetate. RNAs were then resuspended in diethyl pyrocarbonate-treated water. Finally, the RNAs were quantitated by either ethidium staining next to known quantities of similar sized RNAs or by trace labeling using 12.5 Ci of [␣-32 P]UTP/100-l transcription.
Gradient Membrane Association Assay-For each form of 3AB to be analyzed in the presence and absence of canine microsomal membranes, a 200-l in vitro translation was set up on ice to contain the following: 0.8 volume (160 l) of rabbit reticulocyte lysate (Promega; precentrifuged at 50,000 ϫ g for 30 min to remove any protein aggregates), 0.075 volume (15 l) of potassium buffer that yields 15 mM KCl and 5 mM KSCN at final concentration, 0.05 volume (10 l) of [ 35 S]methionine (Amersham, Ͼ1000 Ci/mmol), 0.025 volume (5 l) of amino acids minus methionine (Promega), 0.05-0.1 volume of RNA (10 -20 l) (ϳ1.5-2 g of mRNA (ϳ3-4 pmol) for cytochrome b 5 , ϳ3-4 g (ϳ6 -8 pmol) of mRNA for ␤-globin, and ϳ8.0 g (ϳ12 pmol) of mRNA for each form of 3AB). The translation mixture was then divided in half (100 l each), and to one half 0.25 eq/l canine microsomal membranes (Promega) were added. An equivalent volume of membrane diluent (buffer B; Ref. 26) was added to the other half. The translations were allowed to proceed for 30 min at 30°C, followed by the addition of cycloheximide to a final concentration of 300 M to stop all protein synthesis. Each sample was then subjected to equilibrium centrifugation in high density sucrose gradients essentially as described by Caliguiri and Tamm (1), with the volumes proportionately scaled down to be used in a 2.2-ml (total volume) gradient. Briefly, the in vitro translations were diluted to 0.27 volume (600 l) with 36% w/w sucrose that was 1. . Asterisks (*) indicate conserved positively charged residues that flank the hydrophobic domain, while periods (.) denote a space used solely to maintain the alignment. and 1.5 mM MgCl 2 ) ultimately leaving the in vitro translation in 30% sucrose and 1 ϫ RSB. Each sample was then layered on a 60%-45%-40% (w/w) sucrose step gradient generated by adding 0.27 volume (600 l) of 60% sucrose followed successively by 0.18 volume (400 l) of 45% sucrose and 0.18 volume (400 l) of 40% sucrose. Each sucrose layer was also 1 ϫ RSB. Finally, the sample layer was overlaid by 0.09 volume (200 l) of RSB alone and the sucrose gradient was centrifuged in a TLS-55 rotor (Beckman) at 86,000 ϫ g for 16 -18 h at 4°C. Following centrifugation, each gradient was harvested from the bottom by piercing the polyallomer tube with an 18-gauge needle and pumping out the gradient using a Pharmacia peristaltic pump. Each ϳ120 -130-l fraction was then subjected to immunoprecipitation analysis as described above using the anti-FLAG M2 monoclonal antibody.
Pellet/Supernatant Membrane Association Assay-[ 35 S]Methioninelabeled in vitro translations were set up exactly as described above only proportionately scaled such that each corresponding pellet and supernatant fraction is derived from a ϳ23-25-l translation reaction. Following the 30-min incubation at 30°C, each translation was then centrifuged in a TLA-100.3 rotor (Beckman) as described in the appropriate figure legends. Following the centrifugation step, the supernatant was removed and the pellet was washed once with 50 l of cold TE (pH 7.6). The TE wash volume was then added to the corresponding supernatant fraction. 20 l of 2 ϫ LSB was added to each supernatant fraction, and 40 l of 1 ϫ LSB was added to each pellet fraction. All samples were then boiled, vortexed, and subjected to immunoprecipitation analysis described above.

RESULTS
In Vitro 3AB Membrane Association Assay-In order to analyze the molecular determinants for membrane association, poliovirus protein 3AB was generated by in vitro translation in the absence of other poliovirus proteins. The mRNAs encoding 3AB and the model proteins (discussed below) contained an internal ribosome entry site from encephalomyocarditis virus genomic RNA and coded for an epitope tag (DYKDDDDK; termed FLAG) at the amino terminus of each protein (Fig. 2). The membrane association properties of the wild type and mutated forms of poliovirus protein 3AB were measured by sucrose density gradient analysis as described under "Materials and Methods" and outlined in Fig. 2. Membrane-protein interactions were detected as the differential association of 3AB with the rough endoplasmic reticulum (RER). The rationale for using gradient analysis was to separate aggregated protein from that which is truly associated with the RER. The conditions for centrifugation were similar to those described by Caliguiri and Tamm (1), in which the authors showed by phospholipid analysis and [ 3 H]uridine labeling that the rough endoplasmic reticulum reaches equilibrium in the 45%-60% sucrose interface (density Ͼ1.2 g/ml) following a 16 -18-h spin at 86,000 ϫ g. A direct interaction of a protein with the RER would be indicated by protein fractionation into the high density sucrose portion of the gradient following centrifugation when the translation was carried out in the presence of microsomal membranes.
To validate the assay, our analyses included two "model" proteins with known membrane association characteristics. As a positive control, rabbit cytochrome b 5 was chosen for its characteristics as an integral membrane protein that spontaneously associates with microsomal membranes via a monotopic anchor sequence (18,27,28). For the negative control, the non-membrane-associated protein human ␤-globin was used. Unless otherwise indicated, the model proteins were cotranslated along with wild type and mutated forms of 3AB. Each gradient fraction was subjected to immunoprecipitation analysis as described under "Materials and Methods" using a monoclonal antibody directed toward the FLAG epitope present at the amino terminus of each in vitro translated protein.
The results of a typical sucrose density gradient analysis are shown in Fig. 3. In the absence of microsomal membranes, all three in vitro translated proteins were found in the top half of the sucrose gradient following centrifugation (Fig. 3, top panel, lanes 8 -16), with minimal radiolabeled protein found toward the bottom. However, when the wild type form of 3AB, cytochrome b 5 , and ␤-globin were translated in the presence of canine microsomal membranes, cytochrome b 5 and 3AB were found in the bottom half of the gradient (Fig. 3, middle panel,  lanes 3-7). In contrast, the model non-membrane protein, ␤-globin, localized within the gradient in a manner identical to that when it was translated in the absence of membranes (lanes 8 -16). This result indicated that this assay allowed the discrimination between membrane and non-membrane proteins when translated together in vitro, and that poliovirus protein 3AB behaved like the model integral membrane protein cytochrome b 5 and associated with the RER. Additionally, this analysis demonstrated biochemically that 3AB is able to associate with microsomal membranes in the absence of other poliovirus proteins.
One additional question that was addressed using this assay was whether the wild type form of 3AB could associate with microsomal membranes post-translationally. All three proteins, 3AB, cytochrome b 5 , and ␤-globin, were first translated in the absence of microsomal membranes, and, following the addition of cycloheximide to inhibit further translation, equivalent amounts of microsomal membranes were added to the in vitro translation mixture. As shown in Fig. 3 (bottom panel,  lanes 4 -7), both the cytochrome b 5 and poliovirus protein 3AB were found in the high density sucrose fractions, indi-FIG. 2. Protocol outlining the membrane association assay using high density sucrose gradient centrifugation. The line diagram at the top represents the mRNAs used for in vitro translations followed by a brief flow chart of the gradient membrane association assay. The box adjacent to the FLAG sequence corresponds to the coding sequence for 3AB (and its mutated derivatives), cytochrome b 5 , or ␤-globin. Following centrifugation, the gradient was harvested from the bottom as described under "Materials and Methods" and each fraction was subjected to immunoprecipitation using the anti-FLAG monoclonal antibody. The immunoprecipitations were then analyzed on 15% polyacrylamide-SDS gels, and the gels were fluorographed and visualized by autoradiography.
cating that 3AB was able to associate with membranes post-translationally.
Deletions and Substitutions within Poliovirus Protein 3AB-The phylogenetically conserved hydrophobic domain present in the carboxyl-terminal half of 3AB is the most probable determinant for membrane association (9,12,29). In order to address the role of this domain in directing 3AB-membrane association, we engineered multiple amino acid substitutions to place charged residues within the hydrophobic domain. These substitutions were centered around a predicted amphipathic helix thought to consist of at least the first 15 amino acids of the hydrophobic domain (aa 59 -73) 2 (30). Unexpectedly, when multiple charged residues were placed together on the predicted hydrophobic side of this amphipathic helix (Phe-69 3 Lys and Val-66 3 Glu and Ala-65 3 Ser), no decrease in membrane association was observed when tested in the gradient membrane association assay described above. 3 As a result, more dramatic sequence changes were engineered to define the regions of the hydrophobic domain required for membrane association. Deletions and substitutions within the hydrophobic domain of 3AB that were tested for their effects on membrane association are shown in Fig. 4. Based on restriction sites engineered into the poliovirus cDNA (21,22) that lie within the nucleotide sequence encoding amino acids 59 -81, the hydrophobic domain of 3AB has been divided into two subdomains, domains I and II, corresponding to amino acids 64 -72 and 73-80, respectively. Removal of domain I (DI⌬) removes most of the amino acids predicted to form the amphipathic helix, while deletion of domains I and II (D(IϩII)⌬) effectively removes the entire hydrophobic domain except for the first five amino acids.
Characterization of the 3AB Domain I Deletion-We analyzed the deletion of domain I in a mutated form of 3AB (DI⌬), which has a faster electrophoretic mobility than that of wild type 3AB, in the gradient membrane association assay. As shown in Fig. 5, both 3AB-DI⌬ and 3AB-wt behave similarly to cytochrome b 5 in that each is found in nearly identical proportions in the high density sucrose fractions when the translations were carried out in the presence of microsomal membranes (lanes 3-8). This result indicated that the 3AB-DI⌬ deletion does not result in a decrease in membrane association, consistent with the previous observations that this region of the hydrophobic domain could tolerate the addition of multiple charged residues without diminishing 3AB membrane association. 3 We cannot, however, rule out the possibility that the mutations in domain I result in reductions of 3AB membrane association that are not detectable by this assay.
One possibility that could explain the presence of 3AB-DI⌬ in the high density sucrose fractions when translated in the presence of membranes (Fig. 5) was that the mutated form of 3AB was able to assemble heteromultimers with the wild type form of 3AB causing 3AB-DI⌬ to appear to be membrane-associated. To test this possibility, 3AB-DI⌬ was translated both in the presence and absence of microsomal membranes and in the presence and absence of 3AB-wt. The results of this analysis are shown in Fig. 6. In this experiment, the in vitro translation reactions were sedimented through a sucrose cushion, and the pellet (membrane fraction) and supernatant (non-membrane fraction) fractions were analyzed. This type of analysis, while less definitive than the gradient assays described above, is much faster for analyzing membrane association under multiple conditions. The presence of 3AB-DI⌬ in lane 4, in which membranes but not 3AB-wt were present, indicated that 3AB-DI⌬ was able to associate with microsomal membranes independently of 3AB-wt. Furthermore, 3AB-DI⌬ did not demonstrate an increase in membrane association when translated in the presence of 3AB-wt (compare lanes 4 and 6). These results strongly suggest that domain I is not required for membrane association in this assay and that 3AB membrane association does not occur via a predicted amphipathic helix within aa 59 -73 of the viral protein.
Characterization of Deletions of Domains I and II-Membrane association of 3AB is dramatically different when both domains I and II of the hydrophobic domain are removed. The results of this analysis are shown in Fig. 7. The top panel demonstrates that none of the in vitro translated proteins were found in the bottom half of the gradient in significant amounts when translated in the absence of microsomal membranes, as expected. However, when 3AB-wt and 3AB-D(IϩII)⌬ were translated along with the model proteins in the presence of membranes, only 3AB-wt and cytochrome b 5 were found in the lower half of the gradient (lanes 3-7). 3AB-D(IϩII)⌬ behaved in a fashion nearly identical to that of the non-membrane protein ␤-globin, indicating that removal of nearly all of the hydropho- . Note that we consistently found that the mRNAs for 3AB produced weaker protein signals following in vitro translation than those for both cytochrome b 5 and ␤-globin despite the similar numbers of methionines in each protein.
bic domain abolished 3AB membrane association. This result is consistent with the results of Datta et al. (29), in which they removed the first 15 amino acids of the hydrophobic domain and saw a partial decrease in membrane association as well as a more diffuse immunofluorescence pattern in transfected cells. Those studies, however, could not discriminate between membrane-associated protein and particulate matter.
Role of Domain II in Membrane Association-Based upon the inability of 3AB-D(IϩII)⌬ to associate with microsomal membranes while 3AB-DI⌬ could, we concluded that the amino acid residues most critical for membrane association were likely to be those in domain II. In order to test this hypothesis, the amino acids in this region were either deleted (3AB-DII⌬) or substituted with charged residues (3AB-DII-3E). When domain II was deleted from in vitro translated 3AB, the mutated poliovirus protein showed a decreased ability to associate with microsomal membranes (Fig. 8). In this experiment, little to no 3AB (wild type or mutant) was seen in the bottom portion of the sucrose gradient when translated in the absence of membranes (top panel, lanes 1-8). Furthermore, when these proteins were translated in the presence of microsomal membranes, only 3AB-wt and cytochrome b 5 could associate with membranes, as indicated by their presence in the bottom half of the gradient (bottom panel, lanes 3-9). The amount of 3AB-DII⌬ found in the lower fractions of the gradient following translation in the presence of membranes was similar to that seen when translated in the absence of membranes, indicating that the presence of the hydrophobic amino acids in domain II is crucial for 3AB membrane association.
A logical prediction that stems from the diminished ability of 3AB-DII⌬ to associate with microsomal membranes is that a major determinant for 3AB membrane association is through hydrophobic interactions between the lipid bilayer and the hydrophobic residues present in domain II. Therefore, substitution of charged residues within this region for the highly conserved and most hydrophobic residues (valines 75, 76, and 78) should render this mutant form of 3AB (3AB-DII-3E) deficient in membrane association. When 3AB-DII-3E was examined in the membrane association assay (Fig. 9), membrane association was severely decreased when compared to that of 3AB-wt (bottom panel). As was the case for 3AB-DII⌬ (Fig. 8), 3AB-DII-3E behaved essentially like the non-membrane control protein while the wild type form of 3AB behaved like the model integral membrane protein cytochrome b 5 . These results suggest that maintenance of hydrophobicity within domain II of 3AB is necessary for 3AB to associate in vitro with microsomal membranes.
Strength of 3AB-wt Membrane Association-Previous stud- ies by Takegami et al. (13) and Tershak (19) indicated that the association of 3AB within the membranous crude replication complexes was a strong interaction. However, these studies did not address whether this interaction was mediated through membrane and/or protein contacts within the replication complex, or if 3AB was capable of this tight membrane association in the absence of other poliovirus proteins. After demonstrating that 3AB-wt is capable of interacting with microsomal membranes de novo, we then wanted to determine if this interaction is consistent with that of an integral or peripheral membrane protein. The distinction between a peripheral verses an integral membrane protein can usually be determined by different biochemical treatments. Treatments with high salt, high pH (Ն pH 11.0), or chaotropic reagents such as guanidine or urea will dissociate a peripheral membrane protein from the lipid bilayer (16,31,32). In contrast, the interaction of an integral membrane protein with the lipid bilayer is much stronger than that of a peripheral membrane protein and will generally not be dislodged by such treatments. Such proteins nearly always require the use of a detergent to extract the protein from the membrane (33). We translated 3AB-wt along with the two model proteins in either the presence or absence of microsomal membranes. Following translation, equal fractions were subjected to various biochemical treatments aimed at disrupting either electrostatic or hydrophobic interactions. Since each of the various biochemical treatments changed the sedimentation properties of the microsomal membranes (32), neither gradient analysis nor sucrose cushions were used (see "Materials and Methods"). Instead, following the biochemical treatment, each in vitro translation reaction was centrifuged at a relative centrifugal force of 120,000 ϫ g to pellet any membranous material regardless of ultrastructure. The results of this analysis are shown in Fig. 10. Under physiological conditions, pH 7.4, approximately 50% of both cytochrome b 5 and 3AB-wt were found in the pellet fraction when translated in the presence of microsomal membranes (lanes 2 and 3). This result was consistent with that seen in Fig. 5 and likely reflects a limiting amount of microsomal membranes added to each in vitro translation. 3 When parallel reactions were treated with either 4.0 M urea or adjusted to pH 11.0 (lanes 4 and 5 and lanes 6 and 7, respectively), cytochrome b 5 and 3AB-wt were only partially ex-tracted, indicating that these agents had slight effects on the membrane association of these proteins. Similarly, treatment with high salt up to 1.5 M NaCl also had little or no effect on 3AB-wt and cytochrome b 5 membrane association (lanes 10 -13). Taken together, these results suggest that electrostatic interactions are not the primary 3AB membrane association determinant. Our data do not rule out stabilizing or orientation functions for the charged residues that flank the hydrophobic domain. In contrast, treatment with a nonionic detergent that should disrupt hydrophobic interactions abolished both 3AB-wt and cytochrome b 5 membrane association (lanes 8 and 9). This latter result is consistent with a model in which the hydrophobic residues in domain II of 3AB are the crucial residues necessary to allow 3AB to interact with the lipid environment. DISCUSSION In this study, we have presented evidence that poliovirus protein 3AB behaves in a manner consistent with that of an integral membrane protein. Furthermore, deletion and substitution analysis of in vitro translated 3AB indicates that a crucial domain within 3AB required for membrane association lies within amino acids 73-80 and that introduction of charge into this domain abrogates membrane association. While we have not demonstrated this directly, we speculate that domain II interacts substantially and directly with the hydrophobic core of the lipid bilayer and is able to do so post-translationally. The lines of evidence which support this interpretation of the data are as follows: (a) within the 22-amino acid hydrophobic domain of 3AB, the most hydrophobic (and most "hydrophobically conserved") region corresponds to that of domain II (aa 73-80), (b) the introduction of negative charge into domain II abrogates membrane association, while the introduction of multiple charged residues into domain I, or its removal entirely, has no measurable effect on 3AB membrane association, and (c) the wild type form of 3AB is extracted from microsomal membranes only when biochemical treatments aimed at disrupting hydrophobic interactions (i.e. nonionic detergent) are used; treatments with 4 M urea, high pH, or high salt have only minimal effects. We recognize that the microsomal membranes used in these experiments contain significant amounts of cellular integral membrane proteins. It is therefore possible that a mechanism for 3AB membrane association is to tightly interact with unidentified cellular integral membrane protein(s) (16). Studies aimed at identifying possible cellular and/or viral protein binding partners of 3AB are ongoing.
One interesting question that remains to be answered is how can such a small hydrophobic domain facilitate such a tight interaction with the microsomal membranes. Removal of domain I leaves only 13 amino acids of the hydrophobic region, which, by itself, is unlikely to contain enough hydrophobic character to anchor 3AB into the membrane in a manner consistent with that of an integral membrane protein. At least 20 amino acids are required to span a lipid bilayer in an ␣-helical structure (34), more amino acids than remain in the hydrophobic domain of 3AB-DI⌬. Recently a characterization of synaptobrevin, a member of a class of proteins that utilize COOHterminal membrane anchors, revealed that a minimum of 12 consecutive hydrophobic residues were required for post-translational membrane insertion in a manner resistant to pH 11.5 (35,36). This mechanism of synaptobrevin insertion into membranes, while post-translational, is both ATP-and protein-dependent and results in a complete spanning of the membrane by the COOH-terminal anchor. Given that 1) 3AB requires a carboxyl-terminal hydrophobic sequence for membrane association, 2) 3AB can associate with microsomal membranes posttranslationally, and 3) 3AB behaves biochemically like an integral membrane protein, it is conceivable that 3AB uses a FIG. 6. Effect of the wild type form of 3AB on the membrane association of 3AB-DI⌬. Three 25-l in vitro translation reactions carried out in the presence or absence of microsomal membranes were set up using mRNAs coding for the mutated form of 3AB (3AB-DI⌬), cytochrome b 5 , and ␤-globin. In one reaction, an additional mRNA coding for the wild type form of 3AB was included (lanes 6 and 7). Following translation, each reaction was centrifuged at 50,000 ϫ g through a 20-l 15% w/w sucrose cushion (sucrose was 1 ϫ RSB) for 30 min. The corresponding pellet and supernatant fractions of each sample were harvested and immunoprecipitated as described under "Materials and Methods." The samples were resolved by electrophoresis on 15% polyacrylamide-SDS gels. similar post-translational mechanism to insert into membranes. However, a complete spanning of the membrane by the hydrophobic domain of 3AB is not predicted. If 3AB contains a transmembrane helix, this would predict the existence of two membrane-spanning domains since the NH 2 and COOH termini of 3AB should be on the same side of the membrane to be recognized by the viral proteinase and mediate potential RNA binding functions. Protein structures that would require fewer amino acids to span the lipid bilayer consist of ␤-sheets and random coils but, unless in the form of a ␤-barrel, these are energetically unfavorable in a lipid environment (34). Therefore, we speculate that the positively charged residues flanking the hydrophobic domain (Arg-54, Arg-58, and Lys-81 denoted by asterisks (*) in Fig. 1) contribute to membrane binding via strong electrostatic interactions with the negatively charged phospholipid head groups. The role of electrostatic interactions is consistent with the partial membrane extraction by the 4 M urea and the high pH treatments shown in Fig. 10.
It is probable that domain I contributes to membrane association of the wild type form of 3AB, but its deletion is a tolerable one under the conditions tested. In the studies with 3AB DI⌬, membrane binding only under physiological conditions was examined. This deleted form of 3AB may be more susceptible to biochemical treatments such as high pH or high ionic strength, results that would suggest the existence of additional contributions to membrane binding by flanking electrostatic interactions. In addition, it is possible that determinants for membrane association lie outside of the hydrophobic domain. Theoretical predictions of membrane interactive domains have been previously contradicted by experimental data (37). Our current structural model for how 3AB associates with biological membranes is that the hydrophobic domain forms an ␣-helical insertion sequence minimally consisting of the amino acids present in domain II (and the first five residues at the beginning of the hydrophobic domain), and this sequence further utilizes the flanking arginine and lysine residues as stabilizing forces.
Our studies are consistent with the hypothesis that poliovirus protein 3AB associates tightly with biological membranes in a manner that would allow it to serve as a lipophilic anchor for the poliovirus RNA replication complex (9). A potential biological advantage for the use of such a membrane-protein interaction would be to limit diffusion of protein and RNA replication components to two dimensions, resulting in increased local concentrations (22). There are a number of recent reports identifying multiple 3AB interactions with components (both RNA and protein) of the RNA replication machinery (38 -42) that would substantiate a role for 3AB as the anchor for the replication complex and a primary component of the RNA replication process. 3AB association with membranes may also cause a conformational change in 3AB structure that is necessary for recognition by the viral encoded proteinase (43) and possibly other functions of 3AB. One intriguing observation with respect to the amino acid composition of the 3AB hydrophobic domain (especially domain II) is the high degree of conservation of ␤-branched hydrophobic amino acids (V and I). These ␤-branched amino acids are hydrophobic residues that destabilize ␣-helices and promote ␤-sheet formation in globular proteins. However, when these ␤-branched amino acids are placed in a membrane environment, they can be readily accommodated into ␣-helices, suggesting an environment-dependent modulation of protein conformation (44,45). Given the multiple functions proposed for 3AB combined with the regulated asymmetric synthesis of viral RNA within the replication reactions, a conformational switch dictated by degrees of 3AB membrane association is indeed an attractive one. We are currently studying the role of membrane association on the multiple functions of 3AB and how this protein may exert its effects within the poliovirus RNA replication complex. Likewise, we are examining determinants of membrane association in addition to those contained within domain II of poliovirus protein 3AB. Three mRNAs encoding the wild type form of 3AB, cytochrome b 5 , and ␤-globin were included together in a rabbit reticulocyte lysate in vitro translation. The translation was carried out either in the presence or absence of canine microsomal membranes, and an equal fraction of each translation (23 l) was added to an equal volume of 2 ϫ buffer containing the indicated biochemical reagent. The biochemical reagents were (at 1 ϫ): (a) pH 7.4 RSB, (b) 4 M urea in 1 ϫ RSB, (c) 50 mM CAPS (pH 11.0), 10 mM KCl, and 1.5 mM MgCl 2 , (d) 0.5% Nonidet P-40 that was in 1 ϫ RSB, (e) 0.5 M NaCl, (f) 1.5 M NaCl. Each reaction was incubated 10 min on ice and then centrifuged at 120,000 ϫ g for 30 min in the absence of a sucrose cushion. The corresponding pellet and supernatant fractions of each sample were harvested and immunoprecipitated as described under "Materials and Methods." The samples were resolved by SDS-polyacrylamide gel electrophoresis, and the gels were fluorographed and subjected to autoradiography.