The Domains Required to Direct Core Proteins of Hepatitis C Virus and GB Virus-B to Lipid Droplets Share Common Features with Plant Oleosin Proteins*

In mammalian tissue culture cells, the core protein of hepatitis C virus (HCV) is located at the surface of lipid droplets, which are cytoplasmic structures that store lipid. The critical amino acid sequences necessary for this localization are in a region of core protein that is absent in flavi- and pestiviruses, which are related to HCV. From our sequence comparisons, this region in HCV core was present in the corresponding protein of GBV-B, another virus whose genomic sequence has significant similarity to HCV. Expression of the putative GBV-B core protein revealed that it also was directed to lipid droplets. By extending the comparisons to cellular proteins, there were amino acid sequence similarities between the domains for lipid droplet association in HCV core and plant oleosin proteins. To determine whether these similarities were related functionally, an oleosin encoded by the Brassica napus bniii gene was expressed in different mammalian cell lines, where it retained the capacity to bind to lipid droplets. Analysis of deletion mutants indicated that the critical region within the protein required for this localization was the same for both plant and mammalian cells. A common feature in the viral and plant sequences was a motif containing proline residues. Mutagenesis of these residues in HCV core and plant oleosin abolished lipid droplet association. Finally, the domain within HCV core required for binding to lipid droplets could substitute for the equivalent domain in oleosin, further indicating the functional relatedness between the viral and plant

Lipid droplets are intracellular storage organelles that are found in all eukaryotic organisms and some prokaryotes (reviewed in Refs. [1][2][3]. They consist of a core of neutral lipid, comprising mainly triacylglycerols and/or cholesterol esters, surrounded by a monolayer of phospholipids. Bounding the phospholipid layer is a proteinaceous coat. The best characterized family of proteins that associate with lipid droplets (referred to as lipid bodies or oil bodies in plant cells) are the plant oleosins that accumulate in desiccation tolerant seeds (3,4). Oleosins consist of a central, hydrophobic domain flanked by amphipathic N-and C-terminal regions. The hydrophobic do-main is critical for association with lipid bodies (5), and all such domains within oleosins share a conserved motif referred to as a "proline knot" (4). The proline knot is composed of three closely spaced proline residues, and mutation of these residues impairs localization to lipid bodies (6). In mammalian cells, the principal lipid droplet binding proteins that have been identified are adipophilin (also called adipocyte differentiation-related protein; see Refs. 7 and 8) and a related family of proteins termed the perilipins (9 -11). Adipophilin is present in a wide range of cell types and in increased quantities in certain diseases where intracellular lipid accumulation is evident (12)(13)(14). By contrast, expression of the perilipins is restricted to adipocytes and steroidogenic cells (15,16).
In addition to adipophilin and the perilipins, the core protein encoded by HCV 1 also associates with lipid droplets in mammalian cells (17)(18)(19). HCV is the sole member of the hepacivirus genus that belongs to the Flaviviridae family along with two other genera, the flavi-and pestiviruses. All of the Flaviviridae are positive-sense, single-stranded RNA viruses that have similar genome arrangements and share sequence similarities. HCV core is a structural component of the virus particle, and by analogy with flavi-and pestiviruses, it is likely to be the sole component of the capsid (20,21). The protein is generated from a polyprotein encoded by the viral genome by cleavage at the endoplasmic reticulum (ER) (22)(23)(24)(25)(26). Two processing events have been postulated for maturation of core, both of which are performed by cellular proteases. One activity is signal peptidase that cleaves the polyprotein at position 191, downstream of the signal peptide, to generate the N-terminal end of glycoprotein E1 (24). There is considerable evidence that the second processing event removes either a portion of or the entire hydrophobic domain containing the signal peptide for E1 (17,19,24,26). Expression of core can lead to the genesis of lipid droplets in tissue culture cells (18) and the development of steatosis in transgenic mice (27). Moreover, interaction between core and apolipoprotein AII, a component of lipid droplets, has been demonstrated (28). These data indicate that not only does core associate with lipid droplets, but it also may have the capacity to influence metabolic events within the cell involving the storage of lipid.
As part of our studies to understand the significance of the interaction between HCV core and lipid droplets, we had identified previously a region within the viral protein that is essential for its association with these storage organelles (19). In this report, we have examined whether there is any similarity be-tween the sequences within this region and those of a virus called GB virus-B (GBV-B), which has significant sequence identity to HCV, although it still remains unclassified among the genera of the Flaviviridae (29). GBV-B shares a tropism for liver hepatocytes with HCV and is infectious in tamarins (30,31). Thus, it has been suggested that GBV-B infection of tamarins may be a surrogate model system for HCV infection of humans (31). However, few comparative studies between equivalent proteins encoded by the two viruses have been conducted. Our sequence comparisons were extended also to eukaryotic cellular proteins that attach to lipid droplets. Thus, we have studied the behavior of the putative core protein of GBV-B and a plant oleosin protein in mammalian cells, and we examined their ability to associate with lipid droplets. The results demonstrate that these disparate plant and viral proteins share similar sequence motifs that play important roles in enabling them to bind to lipid droplets in different cell types.

EXPERIMENTAL PROCEDURES
Construction of Plasmids-Construction of plasmids pSFV/1-195 and pSFV/1-169 has been described previously (19). The codons for the proline residues in these constructs were mutated to encode alanine by insertion of an oligonucleotide (HCV1 in Table I). This oligonucleotide was inserted between BspHI and BstXI sites in construct pgHCV/⌬135-144 (19) to give plasmid pgHCV/1-195 (P3 A) ; the BspHI site is not a natural site in the sequence for HCV strain Glasgow and was introduced during the construction of pgHCV/⌬135-144. To generate pSFV/ 1-195 (P3 A) and pSFV/1-169 (P3 A) , a 502-bp fragment from pgHCV/1-195 (P3 A) , produced by cleavage by BglII and BstEII, was inserted into pSFV/1-195 and pSFV/1-169 digested with the same enzymes. A tagged version of HCV core was generated by first converting the nucleotide sequence in pgHCV/1-195 that encodes amino acids 116 and 117 from TCG CGC to TCT AGA; this introduced a novel XbaI site into the HCV core-coding region without changing the encoded amino acids. An oligonucleotide (HCV2 in Table I) that encoded the epitope tag (32) was introduced into the XbaI site to give plasmid pgHCV/1-195tag. The tagged version of core was transferred as a BglII fragment from pgHCV/ 1-195tag into Semliki Forest virus (SFV) expression vector pSFV1 (33) to give construct pSFV/1-195tag.
To express portions of the GBV-B polyprotein, relevant regions were amplified by PCR from a construct pGBB (31). pGBB contains the consensus sequence for an infectious molecular clone of GBV-B (31). Primers for PCR amplification were derived from the viral sequences in pGBB and were used in the following pairs to produce DNA fragments that encoded N-terminal regions of the GBV-B polyprotein: residues 1-141, GBV-B primers 1 and 2; residues 1-194, GBV-B primers 1 and 3; and residues 1-398, GBV-B primers 1 and 4. Amplified fragments were introduced initially into plasmid pGEM1 and thereafter into pSFV1 by standard cloning techniques. To permit detection of GBV-B core, an epitope tag (32) was introduced into the coding region immediately following amino acid residue 85 (Fig. 2B). This was accomplished by first introducing a novel XbaI site at nucleotide residue 695 by converting the sequence from TCT CGC to TCT AGA; this did not alter the encoded amino acid sequence. An oligonucleotide encoding the epitope tag (GBV-B5 in Table I) was inserted between this XbaI site and a TfiI site (position 708 in the native GBV-B nucleotide sequence). The final SFV constructs that contained tagged versions of GBV-B core were termed pSFV/GB1-141, pSFV/GB1-194, and pSFV/GB1-398.
Constructs containing oleosin sequences (Brassica napus oleosin gene bniii, GenBank TM accession number X61937) were derived from plasmid pBnIII that contained the cDNA sequence for the oleosin transcript (34). A 580-bp DNA fragment ( Fig. 1), produced by Tsp509I cleavage of pBnIII, was inserted into the EcoRI site of pGEM1. The inserted fragment was flanked by BglII (at the 5Ј end) and BamHI sites (at the 3Ј end) by standard cloning methods, and the resultant construct was termed pg/Oln. Inserting oligonucleotide Oln1 (Table I) Table I) into pg/Oln-(⌬89 -111) that is not present in the native sequence. To generate the oleosin proline mutant pg/Oln (P3 A) , Oln 5 was inserted between HpaI and BseRI sites in pg/Oln-(⌬89 -111). DNA fragments carrying the various forms of oleosin were transferred into pSFV1 by standard cloning methods and termed the pSFV/Oln series.
The construct expressing the chimeric oleosin/HCV core protein was made by replacing a 350-bp BglII/XbaI fragment in pgHCV/1-195tag that encoded HCV core amino acids 1-116 with an 180-bp BglII/XbaI fragment from pg/Oln encoding oleosin amino acids 1-54. The resultant construct, pg/Olnϩcore, consisted of amino acids 1-55 of oleosin, followed by 11 amino acids that constituted the epitope tag and thereafter amino acids 117-195 of the HCV polyprotein. A BglII fragment containing the chimeric gene was transferred into pSFV1 to give plasmid pSFV/Olnϩcore.
Maintenance of Tissue Culture Cells and Treatment with MG132-Baby hamster kidney (BHK) C13 cells were grown and maintained in Glasgow minimal Eagle's medium supplemented with 10% newborn calf serum (CS), 4% tryptose phosphate broth, and 100 IU/ml penicillin/ streptomycin (ETC10). Huh7 cells were propagated in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, non-essential amino acids, and 100 IU/ml penicillin/streptomycin. To treat BHK cells with MG132 (supplied by Boston Biomedica), cells were incubated for 5 h after electroporation at 37°C, and the media were replaced with fresh media containing the protease inhibitor at a final concentration of 2.5 g/ml. Incubation was continued at 37°C for a further 12 h before the cells were either harvested for Western blot analysis or fixed for indirect immunofluorescence studies.
Immunological Reagents-The monoclonal antibodies used to detect HCV core protein (mAb JM122) and the epitope tag (mAb 9220) have been described previously (Refs. 19 and 32, respectively). Rabbit antiserum to detect oleosin protein was prepared and used as described previously (35).
In Vitro Transcription and Electroporation of SFV RNA into Cells-RNA was transcribed in vitro from recombinant pSFV constructs linearized with SpeI. BHK and Huh7 cells were electroporated with in vitro transcribed RNA as described previously (19,36). Cells were incubated at 37°C for 15 h and then harvested.

Protein Domains Required for Lipid Droplet Association
Preparation of Cell Extracts, Polyacrylamide Gel Electrophoresis, and Western Blot Analysis-Extracts were prepared, and polyacrylamide gel electrophoresis performed as described previously (19,36). Oil body extracts were isolated from cells by flotation through a sucrose cushion following centrifugation, as outlined previously (35).
Indirect Immunofluorescence and Staining of Lipids-Cells on 13-mm coverslips were fixed for 30 min in 4% paraformaldehyde, 0.1% Triton X-100 (prepared in PBS) at 4°C. Following washing with PBS and blocking with PBS/CS (PBS containing 1% newborn calf serum), cells were incubated with primary antibody (diluted in PBS/CS at 1/200 for JM122 and 9220 antibodies and 1/10,000 for oleosin antisera) for 2 h at room temperature. Cells were washed extensively with PBS/CS and then incubated with conjugated secondary antibody (either anti-mouse or anti-rabbit IgG raised in goat) for 2 h at room temperature. After washing with PBS/CS and PBS, staining of lipid droplets by oil red O was performed as described previously (19). Cells were rinsed finally with H 2 O before mounting on slides using Citifluor (Citifluor Ltd., UK). Samples were analyzed using a Zeiss LSM confocal microscope.
Computing-Sequences were aligned using the ClustalW alignment program (37) and hydropathicity plots generated by ProtScale (38).

A Domain in the HCV Core Protein That Is Absent in Flaviand Pestiviruses but Present in GBV-B-Previously
, we proposed (19,39) that the HCV core protein consisted of three domains and that the second of these domains was absent in related pesti-and flaviviruses. Although there is a lack of sequence identity between viral sequences in the Flaviviridae, each of the capsid proteins in members of this virus family have a high content of basic amino acids (40). Therefore, our comparisons were based on the proportion of positively charged residues in predicted coding regions accompanied by hydropathicity plot studies. We observed that the mature capsid (C) protein of yellow fever virus (YF), a flavivirus, had a high proportion of basic residues (27%). In contrast, the signal peptide sequence of YF that is removed from C protein upon maturation did not contain any basic amino acids. Hydropath-icity analysis of immature C protein revealed that sequences at the site cleaved by the NS2B/3 protease, which removes the signal peptide, corresponded to a hydrophilic region of the protein ( Fig. 2A) that contained a stretch of basic amino acids (RKRR; see Refs. 41 and 42). These amino acids were immediately N-terminal to the NS2B/3 cleavage site, and mutations in this region can abolish both cleavage by the viral protease and virus production (42). Immediately following this segment, the hydrophobicity rose sharply due to the nature of the residues that composed the signal peptide sequence for the downstream glycoprotein, prM ( Fig. 2A). Similar features were observed for the coding region of the capsid protein of bovine viral diarrhea virus, a pestivirus (data not shown). Analysis of the hydropathicity of the HCV core protein also revealed a highly hydrophilic region containing a similar segment of basic residues (RRRSR) between residues 113 and 117 (Fig. 2, A and B). However, there is no evidence for cleavage in this region in tissue culture cells (17,19,25,26). Indeed, the predominant core protein detected extends to about residue 173. This suggests that there are additional motifs within the HCV core protein approximately from residues 118 to 173. The presence of an additional domain(s), which does not have a counterpart in the flavi-and pestiviruses, is further supported by an analysis of the proportion of basic residues in HCV core. Up to amino acid 117, 23% of residues were basic, and this dropped to 7% between residues 118 and 173 (Fig. 2C). Thus, the N-terminal 117 amino acids of HCV core have a similar character to those in the YF C protein, but there are no sequences corresponding to the region between 118 and 173 of the HCV polypeptide. As in previous reports (19,39), this region is referred to as domain 2 (Fig. 2C).
The most closely related virus to HCV is GBV-B, a virus that was isolated from tamarins but whose natural host is not known. To date, the proteolytic events to generate the mature proteins of GBV-B have been assumed from comparison with HCV polypeptide processing (29). Comparing the putative GBV-B core sequence with that of HCV did not identify any stretches of similarity until residue 75 of GBV-B (Fig. 2B). According to the sequence alignment, domain 1 for GBV-B core ended at residue 85 (corresponding to residue 117 in HCV core; Fig. 2B), and thus, for this domain, the GBV-B sequence was shorter than that for HCV. The overall sequence identity between these domains in HCV and GBV-B was about 22% (Fig.  2C). From residue 86 and up to the start of the putative signal peptide sequence (residue 140) of GBV-B, sequence identity between the two viral sequences increased to 41%, and apart from one additional residue in GBV-B, the sequences were co-linear (Fig. 2, B and C). In HCV core, this segment is composed of domain 2 and indicated that sequences corresponding to this region are present in GBV-B. Sequence identity between the signal peptides (residues 140 -156 for GBV-B) was slightly lower at 38% and reduced further in the equivalent E1 sequences (ϳ25%; data not shown). Distribution of basic residues in the putative GBV-B core protein revealed a high lysine/ arginine content (21%) in the N-terminal region up to amino acid 85, a reduced percentage beyond this point (3.6% between amino acids 86 and 139), and no positively charged amino acids in the signal peptide sequence. This pattern of distribution corresponded to that present in HCV core. From these data, we concluded that the putative GBV-B core protein shares the same domain arrangement as its counterpart in HCV.
Intracellular Localization of the GBV-B Core Protein-Our previous studies (19) on the intracellular localization of HCV core showed that the protein was directed to lipid droplets, and the primary sequence determinants for this localization were present in domain 2. Since the region of highest sequence identity between the core proteins of GBV-B and HCV encom- passed this domain, we expressed an N-terminal region of the GBV-B polyprotein using pSFV/GB1-398 in which the coding region extended beyond the predicted C terminus of E1 to analyze the intracellular localization of GBV-B core. As immunological reagents against GBV-B proteins were not available, we placed a short epitope tag (32) at residue 85 to detect the protein. Placing this tag into the corresponding region of the HCV core protein did not affect its intracellular localization, and the protein was present on lipid droplets (data not shown). Indirect immunofluorescence of cells electroporated with RNA from pSFV/GB1-398 using an antibody, 9220, that recognizes the epitope tag revealed staining around lipid droplets stained with oil red O (Fig. 3A, panels i-iii). Western blot analysis of cell extracts indicated that the size of core protein detected was ϳ17 kDa (Fig. 3B, lane 1). This is consistent with a product of about 155 amino acids. By taking into account the additional 12 amino acids for inclusion of the epitope tag, this placed the processing events involved in cleavage of the polyprotein in the proximity of the signal peptide sequence, which begins at position 140. Analysis of the GBV-B coding sequences predicted a signal peptidase cleavage site at residue 156 (29,43). In the HCV core protein, there is considerable evidence that protein maturation involves cleavage by signal peptidase (at position 191) and a second cellular protease at the signal peptide to give a product that is about 173 amino acids in length (24,26). Thus, GBV-B may be processed in the same manner as HCV to produce core protein.
To verify further the intracellular localization of GBV-B core and the likely events involved in its maturation, two other constructs containing GBV-B sequences were examined. First, a construct pSFV/GB1-141 that would correspond approxi-mately in size to the product was detected with pSFV/GB1-398. Analysis of cells expressing the tagged product produced by pSFV/GB1-141 again revealed localization of the protein around lipid droplets (Fig. 3A, panels iv-vi). The size of the protein made by pSFV/GB1-141 was only slightly smaller than that made by pSFV/GB1-398 (Fig. 3B, lanes 1 and 2). Another construct that expressed the N-terminal 194 residues of GBV-B gave a major product that was identical in size to that produced by pSFV/GB1-398; a second minor band of about 20 kDa represented the uncleaved protein (Fig. 3B, lane 3). We concluded that, in common with HCV core, the equivalent GBV-B protein is directed to lipid droplets in tissue culture cells, and cleavage of the GBV-B polyprotein to produce the mature form of core is directed by cellular peptidases. Our studies on HCV core revealed that domain 2 contained the sequences essential for lipid droplet association. Based on our sequence comparisons, this domain is present also in the equivalent GBV-B protein.
Hence, we propose that the essential sequences for association with lipid droplets reside within the corresponding region of GBV-B core.
The Lipid Droplet Binding Domains of HCV and GBV-B Core Proteins Share Similar Features to Plant Oleosins-There are a limited number of mammalian proteins that have been shown to associate with lipid droplets. The best characterized of these proteins are the perilipins and adipophilin that share sequence similarity with each other (11). However, comparison of the domain 2 sequences from HCV and GBV-B core proteins with adipophilin and perilipins did not reveal any obvious sequence similarity, and the general hydrophobic nature of the viral sequences in this domain had no counterpart in these proteins (data not shown). The other family of well characterized pro- teins that associate with lipid storage structures are oleosins, which are found at the surface of plant lipid bodies. These proteins are characterized by three domains, a central hydrophobic domain flanked by N-and C-terminal amphipathic regions. It is known that the central domain is essential for binding to lipid bodies and, in the middle of this region, a conserved motif containing three proline residues, termed a proline knot, plays a key role in targeting (6). Comparing  Huh7 (lanes 2 and 3) and BHK C13 cells (lanes 4 -9) were electroporated with RNA, and extracts were prepared after incubation at 37°C. Aliquots of extracts containing the same number of cell equivalents were examined by Western blot analysis with oleosin antisera. Samples were from cells electroporated with RNA from pSFV1 (lanes 2 and 9), pSFV/Oln (lanes 3 and 8), pSFV/Oln-(⌬89 -111) (lane 4), particularly in those surrounding the proline residues. We noted also the presence of leucine and valine residues between the prolines in the viral and plant proteins. The third feature was the motif "YATG" that was found in the oleosin and HCV sequences toward the C terminus of the central domain (positions 133-136, Fig. 1) and domain 2 (positions 164 -167 in HCV, Fig. 2B), respectively. In GBV-B, this motif has a single substitution (Tyr to Trp; position 133 in GBV-B, Fig. 2B). Such similarities suggested that these domains in the viral and plant sequences could have related properties in relation to their role in directing proteins to lipid droplets.
Targeting of Plant Oleosin to Lipid Droplets in Mammalian Cells-To examine further the possible relatedness between the sequences responsible for directing the viral core proteins and oleosin to lipid droplets, the plant protein was expressed in mammalian cells using the SFV system. Indirect immunofluorescence of cells electroporated with RNA from pSFV/Oln, combined with staining with oil red O, revealed that the plant oleosin was detected at the surface of lipid droplets in BHK C13 cells (Fig. 4A, panels i-iii). In Huh7 cells, which originate from human hepatoma, oleosin again was present at the surface of lipid droplets (Fig. 4A, panels iv-vi). Western blot analysis of extracts from both cell types showed that the major species of 19 kDa produced by pSFV/Oln was identical in size to oleosin protein present in an extract from plant seeds (Fig. 4C, compare  lanes 1, 3, and 8). Lower molecular weight species that were less abundant were found also in the extracts from the mammalian cells (indicated in Fig. 4C). We presumed that these products were generated by proteolytic cleavage of the full-length oleosin protein, and from analysis of deletion mutants (see below), digestion appeared to occur in the C-terminal portion of the protein.
Previous reports (5) have identified the central domain of oleosin as critical for association with lipid bodies in plants. To determine whether the same region was necessary for association with lipid droplets in mammalian cells, mutants were constructed that removed portions of the N-and C-terminal domains and central domains, and the intracellular localizations of the variant oleosin proteins were analyzed. In all cases, the deletion mutants were still recognized by our anti-oleosin antibodies (35). The N-terminal mutant lacked residues 6 -53 (pSFV/Oln-(⌬6 -53)), and two C-terminal truncation variants expressed 153 and 164 amino acids of oleosin (pSFV/Oln-(1-153) and pSFV/Oln-(1-164), respectively). For the central domain mutant, we removed residues 89 -111 (pSFV/Oln-(⌬89 -111)), thus deleting the proline knot motif. In BHK C13 cells, the N-and both C-terminal mutant proteins remained associated with lipid droplets (Fig. 4B, panels i, iii, and iv). However, the central domain deletion mutant produced by pSFV/Oln-(⌬89 -111) was distributed throughout the cytoplasm and showed no specific localization at the surface of lipid droplets (Fig. 4B, panel ii). From Western blot analysis, the major oleosin products detected corresponded to the predicted sizes of the mutants (Fig. 4C, lanes 4 -7). Similar to the extracts containing the full-length protein, a lower molecular weight species was found in the extract with the central domain mutant (lane 4) but not in any of the other samples. The most likely explanation for the appearance of this species was cleavage in the C-terminal region of the protein since the extracts for the C-terminal mutants lacked this product. We would predict that any C-terminally truncated product for the N-terminal mutant pSFV/Oln-(⌬6 -53) would not resolve on our gel system because of its low molecular weight (Fig. 4C, lane 7).
We concluded from our results that a plant oleosin retained the capacity to locate to the surface of lipid droplets in mammalian cells, and this property was not cell type-dependent. Moreover, in common with data obtained in plant cells, the primary sequence determinants for lipid droplet association reside within the central domain. Hence, plant oleosin appears to have similar characteristics in relation to association with the surface of lipid storage structures in both plant and mammalian cells.
Mutation of Proline Residues in the Lipid Droplet Binding Domains of HCV Core and Plant Oleosin Proteins Abolishes Association with Lipid Droplets-As stated above, the proline knot motif, which contains three closely spaced proline residues, is conserved within the central domain of plant oleosin proteins. Mutation of these residues impairs association with lipid bodies (6). In the corresponding part of domain 2 of the HCV and GBV-B core proteins, there are two proline residues, which are separated by 4 and 3 amino acids, respectively. We have suggested above that the proline residues in the viral and plant proteins may be functionally related. To test this, the three prolines in oleosin and two prolines in HCV core were mutated to alanine residues. In the case of HCV core, the initial construct, pSFV/1-195 (P3 A) , expressing the proline mutant was truncated at amino acid 195 of the HCV polyprotein. The wild-type form of this construct gave a product that was cleaved by cellular signalases at the signal peptide to give a protein of about 173 amino acids, which was directed to lipid droplets (Fig. 5, A, panel i, and B, lane 1). The protein produced by pSFV/1-195 (P3 A) did not associate with lipid droplets (Fig.  5A, panel iv); however, the size of the major product detected in Western blot analysis indicated that most of the protein had not been efficiently processed at the signal peptide sequence (Fig. 5B, lane 2). Hence, the inability of the protein to locate to lipid droplets may have resulted from the continued presence of the signal peptide sequences, anchoring it to the ER. To overcome this problem, a second proline mutant was made (pSFV/ 1-169 (P3 A) ) in which the HCV coding sequences were truncated at residue 169. Our previous studies (19) had shown that expression of the N-terminal 169 amino acids of the core protein was sufficient to direct the protein to lipid droplets (Fig.  5A, panel ii). By contrast, the protein made by pSFV/1-169 (P3 A) was distributed throughout the cytoplasm and did not specifically locate to the surface of lipid droplets (Fig. 5A, panel  v). In the case of the oleosin protein, mutation of the three prolines within the proline knot motif abolished localization to lipid droplets (Fig. 5A, compare panels iii and vi). We also noted in the case of the oleosin mutant that there was a considerable reduction in the amount of protein detected in extracts compared with the wild-type form of the protein (Fig. 5C, compare  lanes 1 and 3). The reduced amount of mutant protein appeared to result from enhanced degradation since incubation of cells in the presence of the proteasomal inhibitor, MG132, increased its abundance (Fig. 5C, compare lanes 1 and 3 with lanes 2 and 4). Our results suggested that the proline residues at the center of the lipid droplet binding domains of both HCV core and oleosin were essential for association with these structures. Hence, we postulate that they may perform similar functions in these domains.
A Chimeric Protein Composed of Regions of Oleosin and HCV Core Is Directed to Lipid Droplets-To directly address whether sequences in HCV core can functionally substitute for those in oleosin that directed the protein to lipid droplets, we made a pSFV/Oln-(1-153) (lane 5), pSFV/Oln- (1-164) (lane 6), and pSFV/Oln- (⌬6 -53) (lane 7). An oil extract from plant seeds was used as a control (lane 1) to show the correct size of oleosin in plants. Monomeric (oln), dimeric (2x), and trimeric (3x) forms of oleosin are indicated as noted previously (51). Asterisks show the oleosin breakdown products referred to in the text. construct, pSFV/Olnϩcore, that expressed a chimeric protein composed of regions of the two proteins. In pSFV/Olnϩcore, the central and C-terminal domains of oleosin were replaced with domains 2 and 3 of HCV core. Immunolocalization of the protein in BHK cells indicated that it could be detected at the surface of lipid droplets (Fig. 5A, panel vii). Since the N-terminal domain of oleosin and domain 3 of HCV core are not essential for association with lipid droplets, we concluded that domain 2 of HCV core was primarily responsible for directing the chimeric protein to these structures. This provided direct evidence that domain 2 of HCV core and the central domain of oleosin performed similar functions in relation to lipid droplet association. DISCUSSION From previous analysis (19,39) and data presented in this study, it was proposed that the HCV core protein consisted of three domains (Fig. 2C). Domain 1 corresponded to the mature core protein of flaviviruses, whereas no sequences equivalent to domain 2 were present in either pesti-or flaviviruses. Domain 3 contained the signal peptide sequence that directs the HCV E1 glycoprotein to the ER lumen. Here, we have extended our sequence comparisons to include GBV-B, a virus that is the closest known related virus to HCV in terms of sequence identity (29). From comparisons of the two viral sequences, domain 2 in HCV core was found also in the corresponding GBV-B protein. Sequence identity between the predicted amino acid sequences of HCV and GBV-B in the region containing domain 2 was higher (ϳ41%) than that in the N-and C-terminal flanking regions (Fig. 2C). This degree of similarity would imply that these domains in HCV and GBV-B might perform similar functions. From our extensive mutational analysis and immunolocalization studies of the HCV core protein, domain 2 contained the key elements for directing the protein to lipid droplets (19). The data presented in this report indicated that the core protein of GBV-B also could associate with these lipid storage structures. Thus, we conclude that a function of these domains is to direct the core protein of the two viruses to lipid storage organelles.
It is perhaps noteworthy that, unlike flaviviruses, there is no evidence that maturation of the HCV core protein involves a virus-encoded protease, and from the preliminary evidence in this report, a similar situation may occur for maturation of GBV-B core. In flaviviruses, maturation of core requires cleavage of the polyprotein not only by signal peptidase but also a complex of the NS2B and NS3 proteins that removes the hydrophobic region containing the signal peptide for glycoprotein prM (41,42). Interestingly, the NS2B/3 complex cleaves the polyprotein of flaviviruses within a stretch of positively charged residues that is immediately followed by the segment containing the signal peptide. In HCV, there is a similar run of basic residues that appears to define the junction between domains 1 and 2, but no product corresponding to cleavage in this region has been identified (17,19,25,26). Hence, there are two fundamental differences between HCV and the flaviviruses with respect to their core proteins. The HCV protein has an additional domain and is processed by cellular proteases, and in the flaviviruses, maturation requires a virus-encoded protease. These differences presumably reflect distinct evolutionary pathways that may indicate variations in the replication of HCV and the flaviviruses.
From the above evidence, it is reasonable to conclude that the HCV and GBV-B core proteins have similar domain configurations. Based on our data (Ref. 19 and this paper), domain 2 in both proteins represented sequences that directed them to lipid droplets. Comparison with mammalian proteins that associate with these structures did not identify any region in such proteins with features similar to those in domain 2. However, we did observe similarities with the oleosin family of proteins that are surface components of lipid bodies in plant cells. Expression of an oleosin from B. napus in mammalian cells indicated that it did associate with lipid droplets, and this association was not cell type-dependent. Moreover, from our studies with deletion mutants, the central domain of oleosin was essential for this localization, whereas the N-and C-terminal domains were dispensable. This agrees with other data identifying the central domain as necessary for association with lipid bodies in plant cells (5). Therefore, we conclude that the signals and processes that recruit proteins to lipid droplets are related in mammalian and plant cells. It has been shown also that oleosins can target lipid bodies in yeast cells (44). Hence, the requirements for directing proteins to lipid storage compartments may well be related in all eukaryotes. The relative ease with which localization of wild-type and mutant forms of oleosin can be examined in mammalian cells clearly identifies our system as a model for studying the targeting of proteins to lipid droplets.
One of the key features within the central domain of plant oleosins is a hydrophobic segment containing three closely spaced proline residues that has been termed a proline knot (4). This motif is highly conserved in all oleosins and is found also in a related family of lipid-associated proteins termed caleosins (3,45,46). Point mutations introduced at each of the proline residues reduce lipid body targeting in plant cells (6). The function of the proline knot motif is not clear except that it is likely to introduce a 180 o turn in the peptide backbone (47). This would suggest that it might facilitate intra-molecular interactions within individual oleosin molecules or topology of the protein at membranes. In addition, mutation of the residues that constitute the proline knot considerably reduces the amount of oleosin that accumulates in plant seeds and embryos (6). Our data in mammalian cells provides an identical set of observations upon mutation of the prolines in this motif. Thus, in mammalian cells, mutant proteins containing Pro to Ala substitutions were unable to target to lipid droplets and were subject to degradation by a pathway involving the proteasome. This illustrates further the relatedness in the behavior of oleosin in plant and mammalian cells and underlines the utility of our system for more general studies into the targeting of proteins to lipid droplets.
The HCV and GBV-B core proteins each contained two closely spaced proline residues within domain 2. Given that they resided within an hydrophobic region, they could correspond to two of the prolines in the proline knot motif of plant oleosins. Substitution of these prolines in HCV core abolished lipid droplet association. Elsewhere within domain 2, HCV core  1 and 2) and pSFV/Oln (P3 A) (lanes  3 and 4). Samples in lanes 2 and 4 of C are extracts from cells that had been treated with the proteasome inhibitor MG132. Mature (mat) and immature (imm) forms of core produced by pSFV/1-195 and pSFV/1-195 (P3 A) are indicated, as is the position of oleosin protein (oln). and oleosin proteins shared a common motif (YATG) that was almost completely conserved in GBV-B (WATG). We have shown previously (19) that removal of this motif from HCV core prevented lipid droplet association. Moreover, replacement of the central and C-terminal domains of oleosin with domain 2 of HCV core generated a chimeric protein that retained the capacity to localize to lipid droplets. Based on these findings, we suggest that the HCV and GBV-B core and plant oleosins may be members of a family of proteins that share similar sequence characteristics for targeting to lipid droplets.
More recently, it has been shown (48 -50) that caveolins also can accumulate on lipid droplets. Caveolins consist of a hydrophobic domain flanked by hydrophilic regions, and the hydrophobic domain is critical for lipid droplet association (48). Based on sequence analysis, there is no identifiable proline knot within the hydrophobic domain of caveolin and no sequence identity with either the oleosin or core proteins. Nonetheless, a common feature of all of these proteins is that they are directed to the ER, and this is followed by trafficking to the surface of lipid droplets. By contrast, the other characterized mammalian proteins that associate with lipid droplets, adipophilin, and the perilipins do not contain a recognizable hydrophobic domain, and unpublished observations suggest that they are neither synthesized on membrane-bound ribosomes nor co-sediment with ER membranes (11). Therefore, there appear to be at least two distinct classes of protein that can associate with lipid droplets. One class contains a hydrophobic domain and is directed to lipid droplets via the ER; the second class has neither of these features and the process of trafficking to lipid droplets is unknown. Identifying the processes that are necessary for association of adipophilin and perilipins with lipid droplets would further elucidate similarities and differences with the oleosin, caveolin, and viral core proteins.
To conclude, we have identified a conserved domain that was present in equivalent structural proteins encoded by HCV and GBV-B, which directed these proteins to lipid storage compartments. The sequence composition of this domain had similar characteristics to a region in plant oleosins that also was essential for association with lipid droplets. In the context of viral infection, the relevance of this distribution is unclear and directly addressing its importance is confounded by the inability to efficiently grow these viruses in tissue culture systems. However, the high degree of conservation between HCV and GBV-B in the domain required for lipid droplet association may signify that interaction of the viruses with lipid storage organelles is an important stage in their life cycles. Additional studies that focus on the behavior of lipid droplets accompanied by the development of relevant model systems that mimic processes involved in virus assembly should shed light on the significance of trafficking of core to these structures.