Determination of the Transmembrane Topology of Herpes Simplex Virus Type 1 Glycoprotein K (gK)*

Herpes simplex virus type 1 glycoprotein K (gK) plays an essential role in viral replication and cell fusion. gK is a very hydrophobic membrane protein that contains a signal sequence and several hydrophobic regions. It has been shown that mutations inducing cell fusion map to two distinct domains of gK, suggesting that these domains are functionally important. To understand the transmembrane topology of gK and the localization of these functional domains, we constructed a set of gK deletion, insertion, and truncation mutants and expressed these by in vitro translation in the presence of microsomal membranes. The transmembrane topology of gK was determined by examination of the post-translational processing and protease sensitivity of the mutant proteins. Our data demonstrate that gK contains three transmembrane domains (amino acids 125–139, 226–239, and 311–325). Another hydrophobic domain (amino acids 241–265), which is relatively less hydrophobic and much longer compared with the transmembrane sequences, is located in the extracellular loop. The analysis showed that the domains containing syncytial mutations are both ectodomains. They may interact with each other to form a complex tertiary structure that is critical for the biological function of gK.

Herpes simplex virus type 1 (HSV-1) 1 enters host cells by fusion between the virion envelope and the cell plasma membrane (1). Although many viruses that penetrate by direct fusion also cause syncytia formation (cell fusion), extensive cell fusion is caused by very few wild-type HSV strains. The majority of cells infected by wild-type HSV-1 round up and clump (2). However, mutants of HSV-1 that cause extensive cell fusion have been found. This phenotype has been designated the syncytial (Syn) phenotype (2)(3)(4)(5)(6). Mutations that give rise to the Syn phenotype have been mapped to at least four genes in the HSV viral genome: the gK gene (UL53) (7,8), the gB gene (UL27) (9,10), the UL24 gene (11,12) and the UL20 gene (13). A large percentage of Syn mutations have been mapped to the gK gene (1, 7, 8, 14 -16). Glycoprotein K has several notable characteristics compared with other HSV-1 glycoproteins. It contains multiple hydrophobic domains and may traverse the plasma membrane several times, giving gK a complex trans-membrane structure (8,16). Also in infected cells, gK is expressed at low levels relative to other glycoproteins of HSV-1 and does not reach the plasma membrane (17,18). It appears that gK is essential for viral replication since mutant viruses with insertions or deletions in the UL53 open reading frame cannot grow in tissue culture (18,19). Hutchinson et al. (20) constructed HSV-1 mutant virus FgK␤, in which a lacZ gene cassette was inserted downstream of the amino-terminal portion of gK. They found that FgK␤ was defective in virus egress. Recent studies by Jayachandra et al. (19) on the mutant ⌬gK, which lacks the entire gK gene, indicated that gK is required for both capsid envelopment and virus egress.
gK is a highly hydrophobic protein with 338 amino acids. It has characteristics of a membrane protein: a cleavable 30amino acid NH 2 -terminal signal sequence, two sites for Nlinked glycosylation, and several hydrophobic domains ( Fig. 1) (8). Cleavage of the NH 2 -terminal signal sequence and addition of carbohydrates on asparagine residues at 48 and 58 of the gK amino acid sequence suggest that the NH 2 -terminal domain of gK is an ectodomain (21,22). A truncated gK protein containing hydrophobic domain 1 is membrane-bound, indicating that this domain is a transmembrane or membrane-binding domain (22). However, the topology of the remainder of gK is difficult to predict. Eleven Syn mutants from KOS have been found to have mutations mapping to the gK gene (6,14,21,23). Sequence analysis of these mutants indicated that syncytial mutations can occur in two distinct domains of gK: the NH 2terminal ectodomain of gK and the region between hydrophobic domains 2 and 4 (21). These two regions of gK are likely to be functional domains that play a role in regulating cell fusion. Determination of the transmembrane topology of gK, especially the localization of Syn mutation domains, is essential for the understanding of the structure and function of gK.
In this paper, we constructed a set of gK deletion, insertion, and truncation mutants. These were expressed by in vitro translation in the presence of microsomal membranes. The transmembrane topology of gK was determined by examination of their post-translational processing and protease sensitivity. This provided evidence for a gK structure with three transmembrane domains in which the syncytial mutations occur in the gK ectodomains.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Growth Media, and Conditions-E. coli strains INV␣FЈ (Invitrogen) and DH5␣ (Life Technologies) were routinely cultured in Luria broth at 37°C. These strains were host for the recombinant plasmids used in deletion and truncation mutagenesis experiments. Cells containing recombinant plasmids were selected by addition of ampicillin at a final concentration of 100 g/ml.
Genetic Manipulations: Plasmid DNA Construction-Wild-type gK was amplified from pBS-KBLN (22) by PCR. The T7 promoter sequence was added to the 5Ј-primer and BamHI sites were put at the ends of 5Ј and 3Ј primers ( Table I). The plasmid pTAgK was constructed by amplifying gK using the primers shown in Table I, followed by cloning of the product in the pCRII TA cloning vector (Invitrogen). The BamHI DNA fragment containing the T7 promoter and the gK open reading frame from pTAgK was ligated into the BamHI site of pUC13, forming pgKwt. Restriction analysis was used to determine the orientation of inserts.
Construction of pgK Mutants-Recombinant plasmids pgK⌬1 and pgK⌬2 were made by the primer-mediated mutagenesis method (24). Two PCR products that overlap in sequence were made to produce each mutant. These primary PCR products were denatured and allowed to anneal together by their overlaps. The heteroduplex products containing recessed 3Ј ends were extended by Taq DNA polymerase to produce a fragment that was the sum of the two overlapping fragments. These were amplified further by additional PCR cycles. PCR products with deletion mutations were digested by BamHI and subcloned into the BamHI site of pUC13.
Introduction of the MRGSH4 Epitope into gK-Two backward primers (Table I) were designed to add a sequence coding for the MRGSH4 epitope (26) and a stop codon downstream to the gK COOH terminus (gKMRGSH6) and the region upstream of hydrophobic domain 4 (gKMRGSH4) by PCR. The forward primer used in these reactions was the same as that described above for amplification of wild-type gK. The PCR products were gel-purified and used as DNA templates for in vitro transcription.
Introduction of Termination Codons into Truncated gK Genes-Truncated gK genes containing stop codons immediately downstream of the BstEII site at codon 205 or at codon 310, immediately upstream of hydrophobic domain 4, were produced by PCR. These were designated gKwtB2S, gK⌬1B2S, KR256, KR270, and KR281, respectively. These PCR products were gel-purified and used as DNA templates for in vitro transcription. Primer sequences are shown in Table I.
Biochemical Techniques: Preparation of Canine Pancreatic Rough Microsome Membranes-Rough microsomes (RM, a kind gift from P. Walter, University of California at San Francisco) were treated with staphylococcal nuclease (Boehringer Mannheim) to digest endogenous mRNAs (27). To 300 l of RM, 3 l of 100 mM CaCl 2 buffer were added followed by 4.8 l of staphylococcal nuclease S7 (5,000 units/l) for a final concentration of 80 units/l. Digestion was carried out for 10 min at 23°C. The digestion was stopped by adding 6 l of a 100 mM EGTA solution. The nuclease-treated RM were transferred to a 13 ϫ 51-mm centrifuge tube (Beckman). The tube was topped off by layering 100 mM CaCl 2 buffer on top of the RM. Membranes were pelleted in a Beckman SW55Ti rotor at 33,000 ϫ g for 30 min. The supernatant was removed and the membranes were resuspended in 300 l of 100 mM CaCl 2 buffer. The nuclease-treated RM were aliquoted and stored at Ϫ70°C.
In Vitro Transcription-Plasmid DNAs were digested with BamHI, and the fragments containing the T7 promoter and gK open reading frame were gel-purified. An in vitro T7 transcription kit (MEGA-script™, Ambion) was used to transcribe the purified DNA fragments. Briefly, each transcription reaction contains 1 g of DNA template, 1 ϫ transcription buffer (Ambion), 7.5 mM of each NTP, 1 ϫ T7 RNA polymerase mix, and nuclease-free water in a total volume of 20 l. The transcription mixture was incubated at 37°C for 2 h. Then mRNAs were precipitated by 10 l of LiCl at Ϫ70°C. RNA pellets were washed with 50 l of 80% ethanol and redissolved in nuclease-free water. RNA transcripts were analyzed by agarose gel electrophoresis before being used for in vitro translation reactions. Some DNA templates were produced directly by PCR. They were gel-purified and transcribed to make RNAs encoding the truncated gK proteins.
In Vitro Translation-RNAs were translated in vitro using a reticulocyte lysate system (Promega). Typically, a 25-l translational mixture contained 1-2 g of RNA, 2.5 Ci of [ 35 S]methionine, 0.125 mM MgAc 2 , 12.5 mM KCl, 0.01 mM amino acid mixture (without methionine), and 12.5 l of rabbit reticulocyte lysate (28). Nuclease-treated microsome membranes (3.5 l) were added to certain reactions where post-translational processing was needed. After 2 h of incubation at 30°C, 2 l of a 10 g/l ribonuclease A solution was added to digest RNAs.
Proteinase K Protection Assay-In protease protection experiments, in vitro translation reactions were divided into several aliquots. These were left untreated or were digested by adding proteinase K to a final concentration of 2 or 0.5 g/l, either with or without addition of Triton X-100 to a final concentration of 1.0%. After 1 h of digestion at 4°C, proteolysis was stopped by adding excess phenylmethylsulfonyl fluoride (PMSF, final concentration 4 mM). These samples were analyzed on SDS-polyacrylamide gels.
N-Glycosidase Treatment-Endoglycosidase H (Endo H) and peptide N-glycosidase F (PNGase) from New England Biolabs were used for carbohydrate digestion of gK proteins. To test for Endo H sensitivity, samples of translation reactions were denatured by addition of 0.1 volume of 10 ϫ Endo H denaturation buffer, followed by incubation for 30 min at 37°C. Then, 0.1 volume of 10 ϫ Endo H reaction buffer was added to the denatured sample, followed by incubation with 2 l of 1,000 units/l Endo H for 1 h at 37°C. To test for PNGase sensitivity, samples of translation reactions were first denatured by addition of 0.1 volume of 10 ϫ PNGase denaturation buffer. Then 0.1 volume of 10 ϫ PNGase reaction buffer and 0.1 volume of 10% Nonidet P-40 were mixed with the denatured samples. Finally, the mixtures were incubated with 2 l of 500 units/l PNGase for 1 h at 37°C.
Immunoprecipitation-To immunoprecipitate MRGSH4-tagged gK proteins, 3-5 l of in vitro translation reaction was added to 200 l lysis buffer (1% Triton X-100, 1% bovine serum albumin, 10 mM Tris-Cl, pH 8.0, 150 mM NaCl, 0.025% NaN 3 freshly supplemented with 1.0 mM PMSF, and 1 mM iodoacetamide) for 1 h at 4°C. Anti-MRGSH4 antibody (Qiagen) (0.4 g) was added, and the samples were mixed for 2 h at 4°C on a rotator. A volume of 25 l of 50% protein A-Sepharose suspension (Life Technologies, Inc.) was added, and the samples were mixed overnight at 4°C on a rotator to precipitate antigen-antibody complexes. The beads were washed twice with 1 ml of lysis buffer. Finally, 12.5 l of 2 ϫ ESS loading buffer was used to release proteins from the beads at room temperature.
GCATGGATCCTAATACGACTCACTATAG GGATCCTCAATGATGATGATGATGATGG GCACGCCATGCTCGCCG CTACCACGCATTACATCAAACAGGCGCC a A start codon is bold and underlined. Nucleotides in bold type represent MRGSH4 epitope and stop codon added to the gK COOH-terminus and the region upstream of hydrophobic domain 4.
b Oligonucleotides used for primer-mediated mutagenesis. The overlapping sequences were underlined. c Primers applied in the one-tube PCR-based mutagenesis protocal.
SDS-PAGE Analysis-In vitro-translated proteins were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Polyacrylamide slab gels (15.5%, 3% cross-linker) were made using the method of Laemmli (29). N,NЈ-diallyltartardiamide was used as the cross-linker instead of methylenebisacrylamide (bis). Since in vitro-translated gK aggregates when heated, even in the presence of SDS (22), gK protein samples were mixed with an equal volume of 2 ϫ electrophoresis sample buffer and loaded directly onto gels without heating.
DNA Sequencing-DNA sequencing reactions were performed using the Taq DyeDeoxy terminator cycle sequencing system and were analyzed by automated sequence analysis (Applied Biosystems, Inc.). Three primers in each direction (Wayne State University Macromolecular Core Facility) were used for sequencing of the whole gene (21). The nucleotide sequence data were analyzed by AssemblyLign software program (Oxford Molecular Group).

RESULTS
Hydropathy Analysis of Glycoprotein K-The gK sequence consists of 338 amino acids (Fig. 1A), and its hydropathic profile is shown in Fig. 1B. A strongly hydrophobic region at the NH 2 terminus has been shown to be part of a cleavable signal sequence (22). The four major hydrophobic domains (HD) are indicated. HD 1 , HD 2 , and HD 4 are relatively short (14 -15 amino acids), but are highly hydrophobic. HD 3 meets the criteria for transmembrane domains proposed by Eisenberg et al. (30), but it does not meet those suggested by Kyte and Doolittle (31).
Protease Protection Assay for Analysis of gK Topology-Membrane and secreted proteins translated in vitro in the presence of canine pancreatic microsomal membranes undergo those steps of post-translational processing normally occurring in the endoplasmic reticulum (ER): membrane insertion, signal sequence cleavage, and N-linked glycosylation. Topologically, the interior of the pancreatic microsomes corresponds to the lumen of the ER. The exterior surface of the microsomes is derived from the cytoplasmic side of the ER (27,32). Protease protection assays have been used to determine the topology of a number of membrane proteins (33)(34)(35)(36)(37). Domains protected from protease digestion by microsomes are located in the interior of the microsome or within the microsomal membrane itself and thus are ectodomains or transmembrane domains, respectively. Protease-sensitive domains are cytoplasmic domains that are exposed on the outside of the microsome.
To verify reaction conditions, a well characterized secreted protein, yeast ␣-mating factor, was studied. Post-translational processing of ␣-mating factor involves cleavage of its NH 2terminal signal sequence and addition of N-linked carbohydrate chains (38). After in vitro translation in the presence of microsomes, yeast ␣-mating factor should be located on the interior of microsomal membranes and should be protected from protease degradation. As shown in Fig. 3A, in the absence of pancreatic microsomes, yeast ␣-mating factor was synthesized as an 18-kDa protein (Fig. 3A, lane 1). In the presence of microsomes, the majority of this protein was processed to a form with an apparent molecular mass of 28 kDa (Fig. 3A, lane  2). A partially glycosylated form of 25 kDa was also observed, as well as a small amount of the unprocessed 18-kDa form. Treatment of in vitro translated and microsomally processed ␣-mating factor with proteinase K in the presence of Triton X-100 resulted in complete degradation of the protein (Fig. 3A,  lane 3). This was inhibited by the protease inhibitor PMSF (Fig. 3A, lane 4). When the microsomes were treated with proteinase K alone at the concentration of 2 or 0.5 g/l, the 28-kDa and 25-kDa forms of ␣-mating factor were protected, but the 18-kDa form was completely degraded (Fig. 3A, lanes 5  and 6). Protection of the 28-kDa and 25-kDa forms by the microsomal membranes indicated that these forms were on the interior of the microsomes, whereas the unprocessed 18-kDa form remained outside the microsomes. In summary, the microsomes protected ␣-mating factor proteins that were translocated into the microsomal vesicles.
To use protease protection analysis to determine gK topology, the BamHI DNA fragment from pgKwt containing the T7 promoter and the gK open reading frame (Fig. 2B) was used as a template for in vitro transcription. Transcripts of wild-type gK were translated in vitro in the presence or absence of microsomal membranes. Microsomally processed samples were aliquoted, and proteinase K and Triton X-100 were added as indicated in Fig. 3B. The samples were run on a 15.5% SDS-PAGE gel. Without post-translational processing, gK migrated with an apparent molecular mass of 28 kDa (Fig. 3B, lane 1). This is considerably smaller than the predicted molecular mass of unprocessed gK (38 kDa). This is due to incomplete denaturation of gK in these samples, which have not been heated. Previous studies have shown that heating of in vitro translated gK, even in SDS sample buffer, leads to complete aggregation of the protein (22). A few bands larger than 28 kDa were observed in the absence of microsomes (Fig. 3B, lane 1). These appear to be ubiquitin-conjugated gK proteins. The bands form a ladder with approximately 8-kDa spacing, consistent with addition of multiple ubiquitins (molecular mass 8.5 kDa). Rab-bit reticulocyte lysates are known to be able to carry out ubiquitination reactions (39,40). Unprocessed gK protein translated in the absence of microsomes is misfolded, and this abnormal protein is apparently recognized by ubiquitinylating enzymes to add ubiquitin to one or more of the six lysines in gK.
In the presence of microsomes, wild-type gK protein was mostly processed to a form with an apparent molecular mass of 38 kDa. A partially glycosylated form of approximately 35 kDa was also observed, as well as a small amount of the unprocessed 28-kDa form (Fig. 3B, lane 2). Ramaswamy and Holland (22) showed that 38-kDa and 35-kDa forms have undergone addition of N-linked carbohydrates and signal sequence cleavage. Treatment of in vitro translated and microsomally processed gK with proteinase K in the presence of Triton X-100 resulted in complete degradation of all forms of gK (Fig. 3B, lane 5). When microsomally processed gK samples were digested with proteinase K alone at concentrations of 2 or 0.5 g/l, protected bands of 26 and 23 kDa were observed (Fig. 3B,  lanes 3 and 4). Since the NH 2 -terminal domain of gK was expected to be an ectodomain (22), this domain should be protected from protease digestion. This was verified by treating samples of microsomally processed, proteinase K-treated gK with endoglycosidases. After PNGase and Endo H treatment, the 26-and 23-kDa forms of gK were reduced in apparent size to 22 and 19 kDa, respectively (data not shown). Since the N-glycosylation sites are located in the NH 2 -terminal domain of gK, this confirmed that the NH 2 -terminal domain is an ectodomain.
To study the transmembrane topology of gK, truncated DNA templates were made either by cleavage at restriction sites within the gK open reading frame or by primer-mediated PCR mutagenesis (Fig. 2C). After in vitro translation of RNAs transcribed from these DNA templates, the proteins were separated and analyzed by SDS-PAGE. To analyze the function of hydrophobic domain 1 of gK, this domain was deleted by PCR mutagenesis to produce gK⌬1. Truncated forms of wild-type gK and gK⌬1 lacking hydrophobic domains 2, 3, and 4 were produced by two methods. In the first method, gK and gK⌬1 templates were digested with BstEII prior to transcription and  3. Protease protection assay. A, processed and unprocessed forms of yeast ␣-mating factor were synthesized by in vitro translation in the presence or absence of microsomes. The processed forms of ␣-mating factor were treated with proteinase K, Triton X-100, and PMSF as indicated. B, gKwt transcripts were translated in vitro in the presence or absence of microsomes, as indicated. The processed forms of gKwt were treated with proteinase K and Triton X-100 as indicated. Samples were analyzed on 15.5% SDS-PAGE gels.
translation. The proteins produced were designated gKwtB2 and gK⌬1B2 (Fig. 2C). Since we were concerned that proteins translated from mRNA lacking stop codons might be processed abnormally, we also constructed gK and gK⌬1 templates with stop codons added at the BstEII site (gKwtB2S and gK⌬1B2S) (Fig. 2C). As shown in Fig. 4A, both gKwtB2 and gKwtB2S were synthesized as 22-kDa proteins in the absence of pancreatic microsomes (Fig. 4A, lanes 1 and 5). In the presence of microsomes, the majority of this protein was processed to a form with an apparent molecular mass of 28 kDa. A partially glycosylated form of approximately 26 kDa was also observed, as well as a small amount of the unprocessed 22-kDa form (Fig.  4A, lanes 2 and 6). When the microsomally processed samples were subjected to proteinase K digestion, the apparent molecular masses of these fully and partially glycosylated forms were decreased by 3 kDa. This decrease in mass was consistent with cleavage within the 66-amino acid cytoplasmic tail of gKwtB2/ 2S. On the other hand, the small amount of unprocessed 22-kDa form was completely degraded (Fig. 4A, lanes 3 and 7). In contrast to the case for gKwtB2/2S, the microsomally processed 27-kDa and 24-kDa forms of gK⌬1B2/2S were unaffected by proteinase K treatment, but the unprocessed 20-kDa form was degraded (Fig. 4B, lanes 3 and 7). This suggests that in the absence of HD 1 , the gK⌬1B2/2S proteins were released into the interior of the microsomes, indicating that HD 1 is a transmembrane domain.
To determine the function of hydrophobic domains 2 and 3, three additional gK mutants (KR256, KR270, and KR281) were produced by PCR mutagenesis. To eliminate the effects of hydrophobic domain 4, PCR was used to produce transcription templates truncated upstream of this domain. These truncated open reading frames were terminated with stop codons. The mutant KR270, lacking hydrophobic domain 1 (amino acids 125-139) and hydrophobic domain 3 (amino acids 241-265), contains hydrophobic domain 2 only and terminates after residue 310 of wild-type gK (arginine). KR281, lacking hydropho-bic domain 1 and hydrophobic domain 2 (amino acids 226 -239), has hydrophobic domain 3 only and was truncated after residue 310 of wild-type gK (arginine). KR256, lacking hydrophobic domains 1-3, was terminated after residue 310 of wild-type gK (arginine). gK⌬123 contains hydrophobic domain 4 but lacks hydrophobic domains 1-3 (Fig. 2C). After in vitro translation and proteinase K treatment, the processed forms of KR270 were reduced in apparent size from 34 and 30 kDa to 31 and 27 kDa (Fig. 5, lanes 2 and 3 of panels A and B). This decrease in mass was consistent with digestion of most of the peptide on the COOH-terminal side of HD 2 (46 amino acids). In contrast, no reduction in size of KR281 was observed (Fig. 5A, lanes 6  and 7). These results indicate that hydrophobic domain 2 is a transmembrane domain, but that hydrophobic domain 3 is not. KR256, which lacks all hydrophobic domains, was completely protected from proteinase K degradation (Fig. 5B, lanes 5 and  6). This is consistent with secretion into the interior of the microsomes. gK⌬123 also appeared to be unaffected by proteinase K treatment (Fig. 5B, lanes 8 and 9), but due to the small size of the domain COOH-terminal to hydrophobic domain 4 (13 amino acids), it could not be determined whether this domain was degraded.
Analysis of the COOH-terminal Domain of gK by Immunoprecipitation-To study the function of hydrophobic domain 4, two additional gK mutants were made by PCR mutagenesis. The MRGSH4 epitope was added to the COOH terminus of KR281 to generate gK⌬1,2RGSH4 and to gK⌬1,2 to generate gK⌬1,2RGSH6. gK⌬1,2RGSH4 contains hydrophobic domain 3 only, and gK⌬1,2RGSH6 has hydrophobic domains 3 and 4 (Fig. 2C). gK⌬1,2RGSH4 and gK⌬1,2RGSH6 were translated in vitro in the presence or absence of microsomal membranes, as indicated in Fig. 6. After translation, proteinase K and Triton X-100 were added as indicated. Proteins reacting with anti-RGSH4 IgG1 were immunoprecipitated and analyzed by SDS-PAGE (26). This antibody precipitated both the unprocessed (29 kDa, 31 kDa) forms and microsomally processed (34 kDa, 36 kDa) forms of both gK⌬1,2RGSH4 and gK⌬1,2RGSH6, respectively (Fig. 6, lanes 1, 2, 5, and 6). After proteinase K digestion of microsomally processed gK, the 34-kDa form of gK⌬1,2RGSH4 was immunoprecipitated, whereas the 36-kDa form of gK⌬1,2RGSH6 was not (Fig. 6, lanes 3 and 7). This indicated that microsomally processed gK⌬1,2RGSH4 was completely protected by the microsomes, and that the MRGSH4 epitope of gK⌬1,2RGSH6 remained outside the microsomes. Together, this suggested that hydrophobic domain 4 is a transmembrane domain.
Transmembrane Topology of Glycoprotein K-These studies indicated that gK is a membrane-bound glycoprotein with three transmembrane domains (amino acids 125-139, 226 -239, and 311-325). These correspond to hydrophobic domains 1, 2, and 4, which had been identified by hydrophobicity analysis. Hydrophobic domain 3 (amino acids 241-265) is located in the extracellular loop. Therefore, gK has two ectodomains and two cytoplasmic domains. The amino-terminal ectodomain (amino acids 31-124) contains two N-linked glycosylation sites (N48, N58). The entire domain between residues 141 and 225 is located in the cytoplasm and does not traverse the membrane. The COOH terminus (amino acids 326 -338) is also a cytoplasmic domain. The topological arrangement of gK based on the data presented in this study is shown in Fig. 7. DISCUSSION We have determined the transmembrane topology of gK by examination of the post-translational processing and protease sensitivity of mutant forms of gK. In this study, we identified hydrophobic domains 1, 2, and 4 as transmembrane domains. Therefore, all known Syn mutations in gK are located in either the amino-terminal ectodomain or the ectodomain between hydrophobic domains 2 and 4, suggesting that these ectodomains are important functional domains of the gK glycoprotein.
In contrast to other HSV-1 glycoproteins, gK appears not to be expressed on plasma membranes of infected cells. This conclusion is based on the studies of Hutchinson et al. (18). First, immunofluorescence experiments showed that gK was localized almost exclusively to the perinuclear and nuclear membranes, whereas gD, an HSV-1 envelope protein used as a control, was found in the plasma membrane and Golgi, as well as the perinuclear and nuclear membranes. Second, the forms of gK detected in infected cells were sensitive to Endo H digestion, indicating that gK does not reach the Golgi, the site where N-linked oligosaccharides are processed to an Endo H-resistant form. Third, they found that gK expressed by an adenovirus vector in the absence of other HSV-1 proteins was localized to perinuclear and nuclear membranes. Our results also suggest that gK is likely to be sequestered to intracellular membranes. As we have demonstrated, the transmembrane domains of gK are relatively short (14 -15 amino acids). Bretscher et al. (41) have noted that the transmembrane domains of proteins re-tained in intracellular membranes, such as the endoplasmic reticulum, are typically about 15 residues in length, compared with transmembrane domain lengths of 20 -25 residues for plasma membrane proteins. The shorter intracellular transmembrane domains are also richer in the bulky residue phenylalanine (13.7% versus 5.2%) (41). The first two transmembrane domains of gK each contain two phenylalanines, for an average phenylalanine content of 13.8%. Taken together, these data suggest that gK is likely to be an ER protein. This is consistent with the reported role of gK in envelopment and virion maturation (19,20).
If gK is localized to intracellular membranes, its role in cell fusion must be indirect. It is possible that gK may influence cell fusion by regulating the conformation or interactions of other HSV glycoproteins, specifically those involved in cell fusion (i.e. gB, gD, and gH:gL). Precedent for control of the conformation of a viral fusion protein by another viral membrane protein exists; fowl plague virus hemagglutinin assumes an abnormal conformation unless coexpressed with the fowl plague virus M2 protein (42), although the mechanism of action of M2 (transmembrane proton transport) is likely to be different from that of gK. It is possible that gK acts as an HSV-specific chaperonelike protein. Cellular chaperonins exist in both the cytoplasm and lumen of the endoplasmic reticulum and assist in the folding of nascent proteins (43). Both membrane-bound and soluble chaperonins have been found in the ER lumen (44, 45). The fact that gK Syn mutations map to the gK ectodomains suggests that it may act on the ectodomains of other HSV glycoproteins. It is not clear at this time whether the gK function affecting cell fusion is directly related to its roles in virion envelopment and egress or is a separate function.
Homologs of HSV-1 gK have been reported in a number of other members of the alphaherpesvirinae subfamily, including HSV-2, bovine herpesvirus 1, gallid herpesvirus 1, pseudorabiesvirus, equine herpesvirus 1, Marek's disease virus, and varicella zoster virus (46 -52). Alignment of these proteins using the CLUSTALW program (53) shows substantial similarity among them. 2 Structural features conserved among the gK homologs of these viruses include the amino-terminal signal sequence, the presence of N-linked glycosylation sites in the amino-terminal domains, and 7 of the 13 cysteine residues found in HSV-1 gK. Of the conserved cysteines, all but one are located in gK ectodomains, suggesting that they may participate in disulfide bridges important to the structure of the gK ectodomains. In addition, these proteins contain hydrophobic 2 C. Mo and T. C. Holland, unpublished data. domains similar in structure and position to those of HSV-1 gK, indicating similar transmembrane structure.