Determinants for Membrane Association of the Hepatitis C Virus RNA-dependent RNA Polymerase*

The hepatitis C virus (HCV) RNA-dependent RNA polymerase (RdRp), represented by nonstructural protein 5B (NS5B), is believed to form a membrane-associated RNA replication complex together with other nonstructural proteins and as yet unidentified host components. However, the determinants for membrane association of this essential viral enzyme have not been defined. By double label immunofluorescence analyses, NS5B was found in the endoplasmic reticulum (ER) or an ER-like modified compartment both when expressed alone or in the context of the entire HCV polyprotein. The carboxyl-terminal 21 amino acid residues were necessary and sufficient to target NS5B or a heterologous protein to the cytosolic side of the ER membrane. This hydrophobic domain is highly conserved among 269 HCV isolates analyzed and predicted to form a transmembrane (cid:1) -he-lix. Association of NS5B with the ER membrane oc-curred by a posttranslational mechanism that was ATP-independent. These features define the HCV RdRp as a new member of the tail-anchored protein family, a class of integral membrane proteins that are membrane-tar-geted posttranslationally via a carboxyl-terminal insertion and incubated with primary antibodies in phosphate- buffered saline containing 3% bovine serum albumin and 0.05% sapo-nin. Bound primary antibody was revealed with fluorescein isothiocya- nate (FITC)-conjugated goat F(ab (cid:3) ) 2 fragment to mouse IgG F(ab (cid:3) ) 2 (Cappel, Durham, NC) or sheep F(ab (cid:3) ) 2 fragment to rabbit IgG (Roche Molecular Biochemicals). For co-localization experiments, Texas Red (TXR)-conjugated sheep F(ab (cid:3) ) 2 to mouse IgG or goat antibody to rabbit IgG (ICN/Cappel, Aurora, OH) was used as secondary antibody. For co-localization experiments involving two mAbs of murine origin, the mAb 5B-12B7 was biotinylated using the FluoReporter Biotin-XX la-beling kit and revealed with TXR-conjugated streptavidin (both from Molecular Probes, Eugene, OR). Coverslips were mounted in SlowFade (Molecular Probes, Eugene, OR) and examined with a Zeiss Axiovert photomicroscope equipped with an epifluorescence attachment. Confocal laser scanning microscopy was performed using a Zeiss LSM 410 microscope, and images were processed with the Adobe Photoshop 3.0.5 program. Western Blot Analysis— Western blot analysis was performed as de- scribed previously (21, 22).

With an estimated 170 million chronically infected individuals the hepatitis C virus (HCV) 1 is a major cause of chronic hepatitis, liver cirrhosis, and hepatocellular carcinoma worldwide (1). A protective vaccine does not exist to date, and therapeutic options are still limited (2). HCV has been classified in the Hepacivirus genus within the Flaviviridae family that includes the classical flaviviruses, such as yellow fever virus, and the animal pestiviruses (3). HCV contains a single-stranded RNA genome of positive polarity and ϳ9600 nucleotides (nt) in length that encodes a polyprotein precursor of about 3000 amino acids (aa) (see Refs. 4 and 5 for recent reviews) (Fig. 1A). The polyprotein precursor is co-and posttranslationally processed by cellular and viral proteases to yield the mature structural and nonstructural proteins. HCV replication proceeds via synthesis of a complementary minus strand RNA using the genome as a template and the subsequent synthesis of genomic plus strand RNA from this minus strand RNA template. The key enzyme responsible for both of these steps is the RNA-dependent RNA polymerase (RdRp), represented by nonstructural protein 5B (NS5B).
The HCV RdRp has been shown to be essential for viral replication in vitro (6) and in vivo (7). It has recently been characterized both biochemically (8 -12) and with respect to its three-dimensional structure (13)(14)(15). The HCV NS5B protein contains motifs shared by all RdRps and possesses the classical fingers, palm, and thumb subdomains. As a unique feature of the HCV RdRp extensive interactions between the fingers and thumb subdomains result in a completely encircled active site. Interestingly, deletion of the highly hydrophobic carboxyl-terminal domain of NS5B has been found to increase solubility of the protein in Escherichia coli (10,11) and to alter the subcellular localization in mammalian cells (10). Three-dimensional structures of NS5B reported thus far lack this carboxyl-terminal domain.
Compared with the detailed knowledge of its biochemical and structural features, much less is known about the characteristics of NS5B in a cellular context. This is due in part to the lack of an efficient cell culture system permissive for HCV infection and replication and the difficulty to reliably detect viral proteins in naturally infected liver tissue. In a preliminary study, NS5B was found to be present as fine speckles in the cytoplasm of transiently transfected COS 7 cells, with accumulation in the perinuclear region. The subcellular localization of this protein was not further defined, however (16). Membrane flotation analyses from recombinant baculovirusinfected insect cells revealed NS5B in both membrane and cytosolic fractions, and several NS5B species with slightly different electrophoretic mobility were detected by immunoblot using sera from patients with chronic hepatitis C as a source of primary antibody (16). In a more recent study, a green fluorescent protein (GFP)-NS5B fusion protein was found to be distributed throughout the cytoplasm in a "mesh-like pattern" (17).
Here, we used monoclonal antibodies (mAbs) and continuous human cell lines inducibly expressing NS5B either alone or in the context of the entire HCV polyprotein to define the subcellular localization of the HCV RdRp. In addition, a comprehensive set of deletion mutants and GFP fusion constructs as well as an in vitro transcription-translation (IVTT) system were employed to examine the mechanism of NS5B membrane association.
To construct the expression vector pUHDHCV(H)con, plasmid pBRTM/HCV1-3011con was linearized with AflII downstream of the HCV stop codon. The recessed 3Ј terminus was filled in with Klenow polymerase, followed by digestion of the plasmid with EcoRI which cuts immediately upstream of the HCV start codon and at nt position 8205 of the HCV cDNA. The 7803-base pair EcoRI-EcoRI fragment and the 1200-base pair EcoRI-AflII fragment were ligated into the EcoRI-XbaI sites (XbaI site blunted with Klenow polymerase) of pUHD10-3 to yield plasmid pUHDHCV(H)con. This construct allows expression of the entire open reading frame derived from a functional HCV consensus cDNA with authentic translation initiation and stop codons under the transcriptional control of a tTA-dependent promoter.
Plasmid pUHDEGFP was constructed by ligation of the EcoRI-XbaI fragment of pEGFP-N1 (CLONTECH, Palo Alto, CA), coding for an enhanced GFP, into the EcoRI-XbaI sites of pUHD10-3.
The NheI-EcoRI fragment of pEGFP-C1 (CLONTECH, Palo Alto, CA), comprising the coding region for an enhanced GFP and convenient restriction sites for carboxyl-terminal fusions, was subcloned into the NheI-EcoRI sites of pcDNA3.1 to yield plasmid pCMVGFP. Plasmids pCMVGFPNS5BconC12, pCMVGFPNS5BconC16, pCMVGFPNS5BconC21, and pCMVGFPNS5BconC26 ( Fig. 3A) were constructed by ligation of the preannealed primer pairs NS5B580-591fwd-NS5B580-591rev, NS5B576-591fwd-NS5B576-591rev, NS5B571-591fwd-NS5B571-591rev, and NS5B566-591fwd-NS5B566-591rev, respectively (Table I), into the BspEI-EcoRI sites of pCMVGFP. These constructs allow the expression of the last 12, 16, 21, or 26 aa of NS5B fused in frame to the carboxyl terminus of GFP. pCMVGFPNS5Bcon63, which allows expression of GFP with the last 63 aa of NS5B fused to its carboxyl terminus, was constructed by PCR amplification of the corresponding NS5B fragment from pBRTM/ HCV1-3011con using primers NS5B529fwd and NS5B591rev, followed by digestion of the amplification product with BspEI and EcoRI, and ligation into the BspEI-EcoRI sites of pCMVGFP. All expression constructs were verified by sequencing.

PCR primers and oligonucleotides
Restriction enzyme recognition sites are underlined, and start and stop codons are bold.

BspEI
Membrane Association of the HCV RdRp derived founder cell line UTA-6 (25) was co-transfected with pUHDNS5Bcon, pUHDHCV(H)con, or pUHDEGFP, respectively, and pBabepuro (26). G418 and puromycin double-resistant clones were isolated and screened for tightly regulated HCV protein or GFP expression, respectively, by immunofluorescence microscopy and immunoblot analyses.
Antibodies-The NS5B-specific mAbs 5B-3B1 and 5B-12B7 will be described elsewhere. 2 Briefly, mAb 5B-3B1 recognizes a linear epitope at the palm-thumb subdomain boundary of the HCV RdRp and functions well in immunoblot applications, whereas mAb 5B-12B7 recognizes a conformational epitope and functions well in immunofluorescence and immunoprecipitation analyses. A polyclonal rabbit antiserum against protein disulfide isomerase was obtained from StressGen (Victoria, British Columbia, Canada). The mAb G1/93 against human ER-GIC-53 (28) was kindly provided by Hans-Peter Hauri, University of Basel, Switzerland. A polyclonal rabbit antiserum to mannosidase II (29) was kindly provided by Kelley Moremen, University of Georgia, Athens, GA. The mAbs JL-8 against GFP and C23 (MS-3) against nucleolin were obtained from CLONTECH (Palo Alto, CA) and Santa Cruz Biotechnology (Santa Cruz, CA), respectively.
Indirect Immunofluorescence and Confocal Laser Scanning Microscopy-Indirect immunofluorescence microscopy was performed as described previously (21,22). In brief, cells grown as monolayers on glass coverslips were fixed with 2% paraformaldehyde, permeabilized with 0.05% saponin, and incubated with primary antibodies in phosphatebuffered saline containing 3% bovine serum albumin and 0.05% saponin. Bound primary antibody was revealed with fluorescein isothiocyanate (FITC)-conjugated goat F(abЈ) 2 fragment to mouse IgG F(abЈ) 2 (Cappel, Durham, NC) or sheep F(abЈ) 2 fragment to rabbit IgG (Roche Molecular Biochemicals). For co-localization experiments, Texas Red (TXR)-conjugated sheep F(abЈ) 2 to mouse IgG or goat antibody to rabbit IgG (ICN/Cappel, Aurora, OH) was used as secondary antibody. For co-localization experiments involving two mAbs of murine origin, the mAb 5B-12B7 was biotinylated using the FluoReporter Biotin-XX labeling kit and revealed with TXR-conjugated streptavidin (both from Molecular Probes, Eugene, OR). Coverslips were mounted in SlowFade (Molecular Probes, Eugene, OR) and examined with a Zeiss Axiovert photomicroscope equipped with an epifluorescence attachment. Confocal laser scanning microscopy was performed using a Zeiss LSM 410 microscope, and images were processed with the Adobe Photoshop 3.0.5 program.
Western Blot Analysis-Western blot analysis was performed as described previously (21,22). Subcellular Fractionation-Subcellular fractionation was performed essentially as described previously (21). In brief, 5 ϫ 10 7 cells were homogenized in a hypotonic buffer containing 10 mM Tris⅐HCl, pH 7.5, and 2 mM MgCl 2 , followed by centrifugation at 1000 ϫ g for 5 min to yield a nuclear pellet. The supernatant fraction was adjusted to 0.25 M sucrose, and a mitochondrial pellet was obtained by centrifugation at 9,000 ϫ g for 10 min. Finally, a microsomal pellet was separated from the cytosolic supernatant by centrifugation at 100,000 ϫ g for 40 min.
In Vitro Transcription-Translation (IVTT)-The TNT T7-and SP6coupled reticulocyte lysate systems (Promega, Madison, WI) were used essentially following the manufacturer's recommendations. IVTT was routinely performed for 90 min at 30°C in the presence of 0.8 mCi/ml [ 35 S]methionine (Amersham Pharmacia Biotech) in a volume of 25 l. Where indicated, 1.5 l of canine pancreatic microsomes (kindly provided by Martin Spiess, Biozentrum, University of Basel, Switzerland, and Matthias Mü ller, Department of Biochemistry, University of Freiburg, Germany) were added.
For membrane sedimentation analyses, 15 l of NTE buffer (100 mM NaCl, 10 mM Tris⅐HCl, pH 8.0, 1 mM EDTA) were added after completion of the IVTT reaction, followed by centrifugation at 12,000 ϫ g for 15 min. Supernatants were collected, and pellets were resuspended in 40 l of NTE buffer. Subsequently, pellet and supernatant fractions were analyzed by SDS-PAGE, followed by autoradiography. Gels were scanned on a Fuji BAS1000 PhosphorImager and analyzed using the Fuji MacBAS version 2.4 software.
For analyses of co-and posttranslational membrane association, microsomal membranes were added to the reaction either during or for 45 min at 30°C after completion of IVTT. In the latter setting transla-tion was stopped by 1.25 mM puromycin prior to the addition of microsomal membranes.
For membrane extraction experiments, microsomal membranes were posttranslationally added to IVTT reactions, followed by sedimentation of membrane-associated material as described above. Subsequently, membrane pellets were resuspended in NTE buffer, 1 M NaCl, 100 mM sodium carbonate, pH 11.5, 2, 4, or 6 M urea, or 1% Triton X-100 and incubated for 20 min at 4°C. Finally, membrane sedimentation analyses were performed and fractions analyzed by SDS-PAGE.
For protease protection assays, nuclease-free Pronase from Streptomyces griseus (Roche Molecular Biochemicals) was added to a final concentration of 0.7 mg/ml to IVTT reactions performed in the presence of microsomal membranes. Triton X-100 at a final concentration of 0.5% was added to some of the reactions to disrupt microsomal membranes. After 15 min of incubation at 35°C, proteolysis was terminated by the addition of protease inhibitors (Complete® Protease Inhibitor Mixture, Roche Molecular Biochemicals), followed by SDS-PAGE of the samples.
ATP was depleted from IVTT reactions after the addition of puromycin by incubation with 10 units/ml apyrase (Sigma) for 15 min at 30°C.
Sequence Analyses and Structure Predictions-The NS5B aa 561-591 sequence of the HCV H strain consensus cDNA (18) (GenBank TM accession number AF009606) was used to retrieve all reported isolates from the EMBL data base using the FASTA homology search program (30). Incomplete sequences were removed from the list of matching sequences. A final set of 269 sequences of all genotypes was analyzed to construct Fig. 8, A and B. Multiple sequence alignments and the consensus sequence determination were carried out with the ClustalW program (31). All analyses were made using the IBCP HCV data base website facilities (hepatitis.ibcp.fr). Visualization of sequence alignments and plotting of the most frequently represented aa residues at each position were done with the MPSA program (32). At each aa sequence position, the residue types and their respective frequencies were computed using a program developed at the IBCP. 3 Various methods were combined for the prediction of transmembrane sequences as follows: PHDhtm (33) (www.embl-heidelberg.de/predictprotein/), TM-HMM (34) (www.cbs.dtu.dk/services/TMHMM-1.0/), DAS (35) (www. sbc.su.se/ϳmiklos/DAS/), and TopPred2 (36) (bioweb.pasteur.fr/seqanal/interfaces/toppred.html). Sequences homologous to the NS5B aa 561-591 segment were searched in the Protein Data Bank of threedimensional structures with SSEARCH program (37) using IBCP website facilities (npsa-pbil.ibcp.fr/).

RESULTS
Tetracycline-regulated Cell Lines-A tetracycline-regulated gene expression system was used to establish U-2 OS human osteosarcoma-derived cell lines inducibly expressing NS5B either alone (UNS5Bcon) or in the context of the entire HCV polyprotein (UHCVcon). In addition, cell lines inducibly expressing GFP (UGFP) were generated as a control for the subcellular fractionation experiments. Screening of 50 antibiotic double-resistant clones each resulting from transfections of the tTA-expressing founder cell line UTA-6 with the constructs pUHDNS5Bcon, pUHDHCV(H)con, and pUHDEGFP allowed the isolation of several tightly regulated UNS5Bcon, UHCVcon, and UGFP cell lines, respectively. Detailed characteristics of the cell lines are available from the authors upon request. In the following, data obtained with the cell lines UNS5Bcon-5, UHCVcon-57.3, and UGFP-9.22 will be presented. In addition, all results were confirmed in at least one independent cell clone. These cell lines were maintained in continuous culture for more than 12 months and over 50 passages with stable characteristics and without loss of tightly regulated protein expression.
NS5B Is Localized in the ER-The subcellular localization of NS5B was determined by indirect immunofluorescence microscopy. A representative analysis of UNS5Bcon-5 cells is shown in Fig. 1B. Virtually no immunoreactivity was detected when the cells were cultured in the presence of tetracycline. NS5B expression became clearly detectable 6 h following tetracycline withdrawal (data not shown) and increased to reach a steady-state level after 24 h. At this time point, the mAb 5B-12B7 revealed a reticular staining pattern, which surrounded the nucleus, extended through the cytoplasm, and appeared to include the nuclear membrane. No nuclear or plasma membrane staining was observed. In UHCVcon-57.3 cells, which inducibly express NS5B in the context of the entire HCV polyprotein, the cytoplasmic reticular staining pattern was very similar to that observed in UNS5Bcon-5 cells (Fig. 1B). Taken together, the NS5B staining pattern observed in these cell lines was typical of a membrane-associated protein and highly suggestive of a localization of the protein in the ER. Co-expression of other HCV structural and nonstructural proteins in the context of the entire polyprotein did not appreciably alter the subcellular localization of NS5B. Finally, control UGFP-9.22 cells showed the typical diffuse cytoplasmic and nuclear GFP fluorescence upon tetracycline withdrawal (Fig. 1B).
Subcellular fractionation experiments were performed to confirm the membrane association of NS5B suggested by the immunofluorescence data. For this purpose, cells were lysed in a hypotonic buffer and separated roughly into nuclear, mitochondrial, microsomal, and cytosolic fractions by differential centrifugation (Fig. 1C). When equal amounts of protein from each fraction were analyzed by immunoblot, NS5B was detected only in the membrane-containing fractions, i.e. the nuclear pellet (which contains the outer nuclear membrane (contiguous with the ER) and membranes adsorbed to the nucleus), the mitochondrial pellet (containing ER membranes, which in U-2 OS cells are often wrapped around mitochondria (21)), and the microsomal pellet, whereas it was not detected in the cytosolic fraction. By contrast, GFP was found in all cell fractions (Fig. 1C). Taken together, these results clearly demonstrate the membrane association of NS5B.
The staining pattern and the subcellular fractionation data were highly suggestive of an association of NS5B with the ER. To explore further the subcellular localization of NS5B, double label immunofluorescence experiments with antibodies to cellular marker proteins were performed. As shown in Fig. 1D, the presence (ϩtet) or for 24 h in the absence of tetracycline (Ϫtet) and subsequently processed for indirect immunofluorescence microscopy using the mAb 5B-12B7 or, in the case of UGFP-9.22 cells, viewed directly by fluorescence microscopy. C, subcellular fractionation. UNS5Bcon-5 (right panel) and UGFP-9.22 cells (left panel) were cultured for 24 h in the absence of tetracycline and subsequently subjected to subcellular fractionation as described under "Experimental Procedures." About 60 g of protein per lane was separated by 12% SDS-PAGE and analyzed by immunoblot using the mAb 5B-3B1 against NS5B or the mAb JL-8 against GFP, respectively. nuc, nuclear; mit, mitochondrial; mic, microsomal; cyt, cytosolic fraction. D, UNS5Bcon-5 cells were cultured for 24 h in the absence of tetracycline and subsequently processed for double immunolabeling with polyclonal rabbit antisera against protein disulfide isomerase (PDI), mannosidase II (ManII), and mAb 5B-12B7 against NS5B. Bound primary antibodies were revealed with FITC-and TXR-conjugated secondary antibodies, respectively. Slides were analyzed by confocal laser scanning microscopy as described under "Experimental Procedures." Horizontal sections taken through the center of the nuclei are shown. Images recorded in red (TXR) and green (FITC) channels are presented separately on the left and on the right, respectively, and composite images are shown in the middle.
NS5B co-localized perfectly with protein disulfide isomerase, a marker for the ER. The NS5B staining pattern observed in these cells was different, however, from that revealed by antibodies directed against ERGIC-53, a marker of the ER-to-Golgi intermediate compartment (data not illustrated), and mannosidase II, a marker of the Golgi apparatus (Fig. 1D). A minor association of NS5B with these related compartments cannot be completely excluded by this technique. Taken together, however, these results clearly demonstrate that the major localization of NS5B is the ER or an ER-like modified compartment.
As previously shown by us (21)(22)(23)(24) and others (38) using the tetracycline-regulated gene expression system, there was some heterogeneity in expression levels among individual cells of a given monoclonal cell line. This feature inherent to the expression system explains the observation that not all cells stained with antibodies against marker proteins also stained for NS5B in the double immunolabeling experiments.
The Carboxyl-terminal 21 aa of NS5B Serve as a Membrane Anchor-Evidence obtained in E. coli (10,11) and mammalian cells (10) suggests that the highly hydrophobic carboxyl-terminal domain of NS5B serves as a membrane anchor. To explore systematically the role of this domain in determining the subcellular localization of NS5B, we generated a panel of carboxylterminal deletion constructs shown in Fig. 2A. The NS5B⌬C12 construct encodes at its 3Ј end the four leucine residues that are conserved in all HCV genotypes and were found in E. coli to be an important determinant of protein solubility (11). NS5B⌬C16 lacks these four leucine residues, NS5B⌬C21 the entire highly hydrophobic carboxyl-terminal domain, and NS5B⌬C26 the absolutely conserved positively charged aa residues flanking the hydrophobic domain. Finally, the NS5B⌬C63 construct represents the minimal domain required for polymerase activity of NS5B (11). These constructs were transiently transfected into U-2 OS human osteosarcoma and HuH-7 human hepatocellular carcinoma cells (39), followed by immunofluorescence analyses using the NS5B-specific mAb 5B-12B7. Representative data obtained in U-2 OS cells are shown in Fig. 2B. Identical results were found in HuH-7 cells (data not illustrated). Interestingly, deletion of the carboxylterminal 12 aa of NS5B (NS5B⌬C12) abolished the typical ER staining pattern, resulting in a diffuse cytoplasmic and nuclear staining. Deletion of 16 aa (NS5B⌬C16) led in addition to a concentration in nuclear globular structures, corresponding to the nucleoli (Fig. 2C), as well as occasional large cytoplasmic dots (particularly in cells expressing high levels of the truncated protein). The staining pattern of NS5B⌬C21 and NS5B⌬C26 was very similar to NS5B⌬C16. NS5B⌬C63, however, did not accumulate in the nucleoli but rather appeared to spare these. A shared structural feature of nucleolar proteins is the presence of an RNA recognition motif that binds either ribosomal RNAs synthesized in the nucleolus or small nucleolar RNAs (40). Our observation, therefore, suggests the presence of an RNA binding domain located between aa positions 529 and 566 of NS5B. Indeed, based on modeling of the HCV RdRp with the template and primer RNA (15), aa residues 558 -563 are probably involved in RNA contact. Alternatively, Arg-531, Lys-533, and Lys-535 (see Fig. 2A) are basic residues with intrinsic nucleic acid-binding properties and thus could also play a role in nucleolar localization of NS5B⌬C16, NS5B⌬C21, and NS5B⌬C26.
Confocal laser scanning microscopy was performed to confirm the findings obtained by conventional immunofluorescence analyses. As representatively shown for the full-length NS5B and the NS5B⌬C21 constructs (Fig. 2C), deletion of the carboxyl-terminal 21 aa led to nuclear redistribution of the protein with accumulation in the nucleoli. A double label im-munofluorescence analysis demonstrating co-localization of NS5B⌬C21 with the nucleolar marker protein nucleolin is shown in the inset.
The Carboxyl-terminal 21 aa of NS5B Are Necessary and Sufficient to Target a Heterologous Protein to the ER Membrane-To assess whether the membrane anchor of NS5B can target a heterologous protein to the ER, we generated a panel of fusion constructs. As shown in Fig. 3A, the last 12, 16, 21, 26, or 63 aa residues of NS5B were fused in frame to the carboxyl terminus of GFP. These constructs were transiently transfected into U-2 OS cells and examined by fluorescence microscopy. GFP as well as the GFPC12 and GFPC16 fusion constructs were diffusely distributed in the cytoplasm and nucleus. Interestingly, fusion of the 21 carboxyl-terminal aa of NS5B to the carboxyl terminus of GFP led to a dramatic change in the subcellular distribution. In this case, the GFPC21 fusion construct showed the same staining pattern as NS5B with a fine reticular network involving the nuclear membrane and extending into the cytoplasm. GFPC26 and GFPC63 were very similar to GFPC21. Identical results were obtained in transiently transfected HuH-7 cells (data not shown).
These fluorescence microscopy results were confirmed by subcellular fractionation of transiently transfected U-2 OS cells. As shown in Fig. 3C, the amount of cytosolic GFP dramatically decreased with addition of 21 or 26 carboxyl-terminal aa of NS5B, whereas addition of 16 aa had no significant effect.
Since a lysine residue located at position Ϫ24 of the GFPC21 fusion construct (. . . . .ELYKSG2WFWFCLLLLAAGVG-IYLLPNR) could functionally substitute the conserved arginine residues flanking the hydrophobic carboxyl-terminal domain of NS5B (. . . . .SHARPR2WFWFCLLLLAAGVGIYLLPNR), we mutated this lysine to a serine. However, the subcellular localization of the resulting GFPC21-K24S construct (. . . . .ELYSSG2WFWFCLLLLAAGVGIYLLPNR) was identical to that of the original GFPC21 construct (data not illustrated), confirming that the carboxyl-terminal 21 aa of NS5B are sufficient to target a heterologous protein to the ER membrane. Nevertheless, in the subcellular fractionation experiments shown in Fig. 3C, the amount of cytosolic GFP was lower for GFPC26 as compared with GFPC21, suggesting that membrane association may be stabilized by the positively charged aa residues amino-terminal to the hydrophobic core sequence.
Taken together, these experiments unequivocally demonstrate that the carboxyl-terminal 21 aa of NS5B are necessary and sufficient to target a heterologous protein to the ER membrane.
Membrane Association of NS5B Occurs Post-translationally-IVTT and membrane sedimentation analyses were performed to characterize further the membrane association of NS5B. NS5B was translated in a coupled rabbit reticulocyte lysate system in the presence or absence of microsomal membranes. Subsequently, membrane-associated material was separated by centrifugation, and NS5B was quantified in both fractions. To elucidate the mechanism of membrane association, we first examined whether membrane targeting of NS5B occurs co-or posttranslationally. In eukaryotic cells, ER transport of membrane proteins is generally mediated by a signal sequence that is recognized by the signal recognition particle (SRP). The SRP interacts with the signal sequence of nascent polypeptide chains during translation and directs the translation complex to the ER membrane. SRP-mediated ER transport, therefore, occurs only co-translationally. NS5B, however, would be expected to be inserted into membranes posttranslationally because the membrane anchor will be buried within the translating ribosome when the termination codon is reached. To distinguish between these two possibilities, microsomal membranes were added to the reaction either during or after completion of IVTT. Puromycin was added to the reaction mixture in the posttranslational setting to stop translation and to ensure that polypeptides were released from ribosomes. As represented in Fig. 4, when NS5B was translated in the ab- (NS5B⌬C21), pCMVNS5Bcon⌬C26 (NS5B⌬C26), or pCMVNS5Bcon⌬C63 (NS5B⌬C63), as indicated by the captions. Cells were subsequently processed for indirect immunofluorescence microscopy using the mAb 5B-12B7 as described under "Experimental Procedures." C, confocal laser scanning microscopy. U-2 OS cells were transiently transfected with pCMVNS5Bcon or pCMVNS5Bcon⌬C21 and processed 36 h later for confocal laser scanning microscopy using the mAb 5B-12B7 as described under "Experimental Procedures." Cells were counterstained with propidium iodide (PI) to visualize nuclei. Horizontal sections taken through the center of the nuclei are shown. Images recorded in green (FITC) and red (PI) channels are presented separately on the left and on the right, respectively, and composite images are shown in the middle. The inset in the lower left panel shows a double label immunofluorescence analysis with the mAb C23 (MS-3) against nucleolin and biotinylated mAb 5B-12B7 against NS5B. In this case, reactivity of the anti-nucleolin mAb was revealed with a FITC-labeled secondary antibody and reactivity of the biotinylated anti-NS5B mAb with TXR-conjugated streptavidin.
sence of microsomal membranes only 7% was subsequently found in the pellet fraction. By contrast, 88% of the protein pelleted when translation was performed in the presence of microsomal membranes. These results demonstrate that membrane association of NS5B occurs very efficiently also in vitro. Interestingly, 89% of NS5B was found in the pellet when the A, schematic representation of carboxyl-terminal GFP fusion constructs. The aa sequence of the HCV H consensus clone is shown at the top. B, U-2 OS cells were transiently transfected with pCMVGFP (GFP), pCMVGFPNS5BconC12 (GFPC12), pCMVGFPNS5BconC16 (GFPC16), pCMVGFPNS5BconC21 (GFPC21), pCMVGFPNS5BconC26 (GFPC26), or pCMVGFPNS5BconC63 (GFPC63), as indicated by the captions. Cells were subsequently viewed by fluorescence microscopy. C, subcellular fractionation. U-2 OS cells were transiently transfected with pCMVGFP (GFP), pCMVGFPNS5BconC16 (GFPC16), pCMVGFPNS5BconC21 (GFPC21), or pCMVGFPNS5BconC26 (GFPC26) and subsequently subjected to subcellular fractionation as described under "Experimental Procedures." In this case, the 9000 ϫ g centrifugation step was omitted, and the combined mitochondrial and microsomal pellet resulted from a 100,000 ϫ g centrifugation. Per lane 15 g of protein was separated by 12% SDS-PAGE and analyzed by immunoblot using the mAb JL-8 against GFP. nuc, nuclear; cyt, cytosolic; mit/mic, mitochondrial and microsomal fraction. membranes were added posttranslationally. This demonstrates that membrane association of NS5B can occur by a posttranslational mechanism.
NS5B Is a Cytoplasmically Oriented Integral ER Membrane Protein-The presence of a carboxyl-terminal membrane anchor mediating posttranslational membrane association defines the HCV RdRp as a member of the so-called tail-anchored protein family. As such, NS5B would be expected to behave as a cytoplasmically oriented integral membrane protein. This possibility was explored by membrane extraction and protease protection experiments. NS5B was translated in vitro and posttranslationally incubated with microsomal membranes, followed by differential extraction of the pellet. High salt extraction (1 M NaCl) shields charges and weakens ionic interactions that bind peripheral proteins to membranes either directly or indirectly through other membrane proteins (41). Treatment with 100 mM sodium carbonate, pH 11.5, releases peripheral proteins by transforming microsomes into membrane sheets (42). As shown in Fig. 5, NS5B remained predominantly associated with microsomal membranes under both conditions. In addition, membrane pellets were extracted with 2, 4, or 6 M urea. As shown in Fig. 5, about 80% of the in vitro translated protein remained in the pellet fraction following extraction with 4 M urea. Finally, membranes were disrupted with 1% Triton X-100, resulting in the release of NS5B into the supernatant fraction. Taken together, these results demonstrate that NS5B behaves as an integral membrane protein and thus fulfills the criteria of a typical tail-anchored protein. In this respect, the extraction profile paralleled that of Cb5 and Vamp1 (data not illustrated).
Protease protection experiments were performed to determine the orientation of NS5B in the ER membrane. As shown in Fig. 6, Pronase treatment of IVTT reactions performed in the presence of microsomal membranes resulted in the complete disappearance of the NS5B signal. This protease sensitivity of NS5B indicates that the majority of the protein is localized on the cytoplasmic side of the ER membrane. In these experiments, ppl was used as control for the integrity of microsomal membranes. Ppl is directed to the ER membrane by interaction of its signal sequence with the SRP. Signal sequence cleavage is performed by the signal peptidase located at the luminal side of the ER membrane, followed by release of prolactin into the ER lumen (43). As expected, prolactin was protected from Pronase digestion in the presence of microsomal membranes. It became accessible to proteolysis only after disruption of the membranes by detergent. Most importantly, addition of the NS5B insertion sequence to the carboxyl terminus of GFP targeted the fusion protein to the cytosolic side of microsomal membranes, as shown for GFPC26 in Fig. 6.
Posttranslational Membrane Association of NS5B Occurs by an ATP-independent Mechanism-Tail-anchored proteins fall into two major categories, proteins whose membrane targeting depends on ATP (exemplified by the VAMPs) and those whose membrane targeting occurs by an ATP-independent mechanism (exemplified by Cb5). Therefore, we next examined the ATP dependence using these proteins as controls. ATP was depleted from IVTT reactions by the adenosine 5Ј-tri-and -phosphatase activity of apyrase. As shown in Fig. 7, posttranslational targeting of NS5B to microsomal membranes in vitro was not affected by ATP depletion, indicating that it occurs by an ATP-independent mechanism.
Sequence Comparisons and Structure Predictions-The data shown above define the carboxyl-terminal domain of NS5B as a new membrane insertion sequence. Sequence comparisons were performed to assess the degree of conservation of this sequence among different HCV isolates and to identify motifs potentially involved in membrane targeting and insertion. Analysis of the conservation of this sequence among 269 HCV isolates of various genotypes revealed a high degree of aa sequence similarity (Fig. 8A). Fifty percent of residues are fully conserved, and most of the positions showing apparent variability are in fact occupied by aa residues with similar hydropathic character (Fig. 8B). Overall, despite the presence of conserved neutral residues at positions 580, 582, and 584, the carboxyl-terminal sequence appears as a highly hydrophobic core (segment 572-589) that is predicted to be a transmembrane segment by all analyses performed (PHDhtm, TMHMM, DAS, and TopPred2, see under "Experimental Procedures"). The length of this transmembrane segment (18 aa) is consistent with the typical length of transmembrane ␣-helices. The transmembrane segment is flanked by two (or three) positively charged residues on the amino-terminal side (aa positions 566, 568 and 570) and a positively charged arginine residue at the very carboxyl terminus (aa position 591). Arg-568 and Arg-570 were found to be absolutely conserved, indicating that they are essential. Position 569 is always occupied by a small residue (Pro, Thr or Ser), as well as position 567 (Ala or Val). Hence, the 566 -570 segment appears flexible, positively charged, and is probably involved in membrane surface binding via electrostatic interactions with the polar head of phospholipids. Position 571 is clearly variable but either occupied by a hydrophobic residue or histidine, i.e. residues that are likely to be located at the membrane interface. Positions 572-579 (except 575) as well as positions 585-588 are occupied by large hydrophobic residues that indicate an ␣-helical folding of this region. By contrast, the connecting segment 580 -584 (SVGVG) between these two hydrophobic stretches exhibits flexible properties, in particular at the level of the fully conserved Gly-582 and Gly-584. Because glycine residues are known to act as helix breakers, one can wonder whether this connecting segment can adopt an ␣-helical fold. To address this issue, we searched for sequence homologies between the NS5B aa 561-591 segment and proteins of known three-dimensional structure. Interestingly, 40.9% aa identity was found between the NS5B aa 571-588 segment and the first transmembrane ␣-helix of bacteriorhodopsin (Fig. 8C). In addition, 41.2% identity was found between the NS5B aa 571-587 segment and the photosynthetic reaction center, chain M, whose structure has been identified as a transmembrane ␣-helix as well (Protein Data Bank entry code 6PRM; data not shown). The presence of similar GLG and SASVG segments in these known ␣-helices confirms that GVG in NS5B can adopt an ␣-helical fold in a membrane environment. Finally, residues at positions 589 and 590 are characteristic of the carboxyl-terminal end of an ␣-helix (proline is a helix breaker and asparagine is often involved in carboxylterminal helix capping). In conclusion, the carboxyl-terminal domain of NS5B can be reliably predicted to form a transmembrane ␣-helix starting at residue 570 and ending at residue 589. An ␣-helix projection of the NS5B aa 571-588 segment is shown in Fig. 8D. DISCUSSION Formation of a membrane-associated replication complex is a characteristic feature of positive-strand RNA viruses (44 -49). In this context, physical interactions between NS4A and a NS4B-5A cleavage substrate on the one hand (50) and between NS5B and NS3 as well as NS4A on the other hand have been described (51). The mechanisms of membrane association and the protein-protein interactions involved in formation of the HCV replication complex, however, are poorly understood. A highly complex and subtly regulated scenario is likely, not the least in view of recent data on membrane targeting of the HCV NS3-4A complex by the NS4A polypeptide (23) and the conformational changes of this complex predicted for cis-and transprocessing events (52).
The best characterized mechanism of membrane insertion in mammalian cells is the SRP-mediated pathway (53). Here, membrane targeting is initiated co-translationally by a signal sequence encoded near the amino terminus of the nascent peptide. The signal sequence in the ribosome-bound nascent chain interacts with the SRP that then docks at the ER. For these proteins integration occurs via a complex multistep process ending with release of the polypeptide into the lipid membrane coincident with the completion of protein synthesis. Another small but rapidly growing class of membrane proteins lacks an amino-terminal signal sequence and instead is targeted via a carboxyl-terminal hydrophobic domain termed insertion sequence (reviewed in Ref. 54). The prototype of this class of integral membrane proteins, termed tail-anchored proteins, is Cb5. Other examples include members of the soluble Subsequently, reaction mixtures were centrifuged for 15 min at 12,000 ϫ g to sediment microsomal membranes containing associated NS5B protein. The supernatants were removed, and the pellets were resuspended in NTE buffer, 1 M NaCl, 100 mM sodium carbonate, pH 11.5, 2, 4, or 6 M urea, or 1% Triton X-100, and incubated for 20 min at 4°C. Subsequently, membrane sedimentation analyses were performed as described under "Experimental Procedures." Supernatant (S) and pellet (P) fractions were applied in equivalent amounts and separated by 12% SDS-PAGE. N-ethylmaleimide-sensitive factor attachment protein receptor proteins, such as the VAMPs (20,55,56), Bcl-2 (57), polyoma virus middle T antigen (58,59), vaccinia virus H3L envelope protein (60), and pseudorabies virus Us9 protein (61). The carboxyl-terminal location of insertion sequences implies that these proteins are targeted to and integrated into the bilayer of membranes posttranslationally. Therefore, neither SRP nor SRP receptor is involved in the membrane association of these proteins.
The features described here, namely posttranslational membrane association via a carboxyl-terminal insertion sequence, behavior as an integral membrane protein, and cytosolic orientation, define the HCV RdRp as a new member of the tailanchored protein family. NS5B represents the first polymerase within this family. The NS5B insertion sequence was mapped to the highly hydrophobic carboxyl-terminal 21 aa of the protein. This domain was necessary and sufficient to target NS5B or a heterologous protein to the cytosolic side of the ER membrane. Sequence analyses reliably predicted that this segment can form a single ␣-helix despite the presence of flexible glycine residues within the center of the transmembrane domain. It is well documented that glycine residues can be involved in specific dimerization of transmembrane segments in membrane proteins such as glycophorin A (62) and phage M13 coat proteins (63). In these cases, however, the glycine residues are located on the same side of the ␣-helix, as observed with the typical GXXXG motif (64,65). In the case of NS5B, however, Gly-582 and Gly-584 occupy opposite sides in the putative transmembrane helix (Fig. 8D). Although it could not be ruled out that these glycines might be involved in transmembrane helix-helix interactions, it is tempting to speculate that these residues are essential to allow some flexibility to the polypeptide chain during the membrane insertion process.
The mechanism of membrane association of NS5B is by current knowledge unique among HCV proteins. In this context, membrane targeting of the structural proteins appears to be mediated by the classical SRP-dependent pathway (66,67), whereas NS3 has been shown to be targeted to the ER or an ER-like modified compartment via interaction with its cofactor NS4A (23), and NS4B is co-translationally targeted to the ER where it behaves as an integral membrane protein (24).
Very little is known about the mechanisms involved in insertion sequence-mediated membrane integration and the mechanisms that regulate membrane selectivity (56,68). Membrane integration of Cb5 in vitro is promiscuous, spontaneous, and independent of membrane proteins (20). Nevertheless, when expressed in cells, both Cb5 and fusion proteins containing the Cb5 insertion sequence associate specifically with the ER membrane (69). Therefore, Cb5 insertion sequence-mediated subcellular localization appears to be regulated by targeting of the molecule to the ER membrane. After correct targeting, membrane integration probably occurs spontaneously (20). Membrane binding of VAMPs, on the other hand, requires ATP and a trypsin-sensitive component of the ER membrane (20,  (77). Struc., secondary structure of the transmembrane segment deduced from the bacteriorhodopsin three-dimensional structure: h, helix; s, bend; t, hydrogen-bonded turn. D, ideal ␣-helix projection of the NS5B[571-588] segment. The variability of residues at each position is included, according to A. The larger characters indicate the most frequently observed residues. Outline, italic, and bold letters correspond to neutral (G, S, and T), hydrophilic, and hydrophobic residues, respectively. 55). Thus, there are at least two different mechanisms for correct membrane integration of proteins with insertion sequences, one mediated primarily by targeting and one relying on putative receptors in the target membrane to mediate selective integration (20). Here, we showed that posttranslational membrane association of NS5B occurs by an ATP-independent mechanism. Further studies will be aimed at identifying the determinants for membrane selectivity and the mechanism of membrane insertion. In this context, it will be interesting to systematically mutate conserved aa residues within the NS5B insertion sequence and to analyze the phenotype of these mutants in vitro and in transfected cells. The most complete threedimensional structure of NS5B composes the structure up to aa residue 563 (15). Ultimately, therefore, resolution of the threedimensional structure of the very hydrophobic carboxyl-terminal domain of the HCV RdRp will provide a framework for a molecular understanding of the insertion mechanism.
The Saccharomyces cerevisiae ubiquitin-conjugating enzyme UBC6 is a tail-anchored protein found in the ER. Interestingly, lengthening of the insertion sequence from 17 to 21 aa resulted in retargeting to the Golgi complex and a further increase in length to 26 aa allowed the modified protein to traverse the secretory pathway and gain expression at the plasma membrane (70). Similar observations were made with the yeast ER t-soluble N-ethylmaleimide-sensitive factor attachment protein receptor Ufe1p, where lengthening of the transmembrane domain allows transport along the secretory pathway (71) or, in mammalian cells, in the case of Cb5, where targeting to the ER membrane was found to be defined by the length of the insertion sequence (72,73). In the case of NS5B, however, extension of the insertion sequence by 4 or 8 hydrophobic residues did not alter the subcellular localization of the protein. 4 Membrane association of NS5B was independent of the expression of other HCV proteins. In this context, co-transfection experiments with the NS5B⌬C21 construct and the NS3-4A complex, NS4B, or NS5A, which by themselves are membraneassociated (23, 24), 5 did not alter the subcellular localization of the carboxyl-terminally truncated NS5B protein. Protein-protein interactions within the presumed HCV replication complex, therefore, do not seem to be sufficient to target NS5B to the ER membrane. By contrast, poliovirus three-dimensional polymerase expressed by itself, for example, is not membraneassociated. In this case, membrane association is believed to be mediated by interactions with other components of the poliovirus replication complex, possibly the viral protein 3AB (74,75). With respect to related members of the Flaviviridae family, analyses of GB virus sequences by various transmembrane prediction methods clearly indicate that the carboxyl terminus of GBV-B NS5B contains a putative transmembrane domain similar to that observed for HCV (data not shown). By contrast, in the case of GBV-A and GBV-C/HGV NS5B the presence of a carboxyl-terminal membrane anchor is ambiguous and should be investigated experimentally. Interestingly, no transmembrane domain was predicted for flavi-and pestivirus NS5 and NS5B proteins, respectively. This is a major difference when compared with hepaciviruses, suggesting that the RdRps of flavi-and pestiviruses are membrane-targeted by different mechanisms, if at all.
In the context of the HCV polyprotein, membrane anchoring by the carboxyl-terminal end, presumably occurring rapidly after release from the ribosome, could represent a strategy to hold together the components of the replication complex during polyprotein processing. Proteolytic cleavage by the NS3-4A complex has been shown both in heterologous expression systems as well as in cell lines harboring subgenomic HCV replicons to occur rapidly between NS5A and NS5B, resulting in a rather stable NS4A-4B-5A precursor that is processed slowly into the individual products (76). If NS5B has a loose association with other replicase components anchoring of NS5B to the ER membrane in a well defined orientation could facilitate low affinity interactions with the other ER-associated replicase components. Such a loose association of NS5B with other replicase components may be important for its multiple roles in initiation of minus and plus strand RNA synthesis and chain elongation.
Finally, identification of the carboxyl-terminal 21 aa of NS5B as a signal for targeting and insertion of heterologous proteins to the cytosolic side of the ER membrane may have a number of interesting applications. For example, this insertion sequence may be used to target antiviral effector molecules to the HCV replication complex.
In conclusion, the results presented here define the HCV RdRp as a new member of the tail-anchored protein family. Elucidation of the determinants for membrane selectivity and the membrane insertion mechanism of the HCV RdRp as well as its involvement in formation of the membrane-associated replication complex may lead to new insights into fundamental cellular processes and define novel targets for antiviral intervention.