Function of the KKXX Motif in Endoplasmic Reticulum Retrieval of a Transmembrane Protein Depends on the Length and Structure of the Cytoplasmic Domain*

Transmembrane glycoproteins with type 1 topology can be retrieved to the endoplasmic reticulum (ER) by a retrieval signal containing a di-lysine (KK) motif near the C terminus. To investigate the structural requirements for ER retrieval, we have constructed mutants of the simian immunodeficiency virus (SIV) envelope (Env) protein with cytoplasmic tails of different lengths and containing a KK motif at the −3 and −4 positions. Such proteins were found to be retained intracellularly when the signal was located 18 amino acids or more away from the membrane spanning domain. The retrieval signal was found to be functional even when placed at the distal end of the wild-type SIV Env protein with 164 amino acids in the cytoplasmic tail, as shown by the lack of proteolytic processing and lack of cell surface expression of the mutant proteins. However, proteins with a cytoplasmic tail length of 13 amino acids or less having the di-lysine motif at the −3 and −4 positions were not retrieved to the ER since they were found to be processed and transported to the cell surface. The surface-expressed proteins were found to be functional in inducing cell fusion, whereas the proteins retained intracellularly were defective in fusion activity. We also found that the KK motif introduced near an amphipathic helical region in the cytoplasmic tail was not functional. These results demonstrate that the ability of the KK motif to cause protein retrieval and retention in the endoplasmic reticulum depends on the length and structure of the cytoplasmic domain. The ER retrieval of the mutant proteins was found to correlate with increased intracellular binding to β COP proteins.

For type 1 integral membrane proteins, a KKXX or XKXX motif (K residues at the Ϫ3 or Ϫ3 and Ϫ4 positions) functions as an efficient ER 1 retrieval and retention signal (1,2). The E3/19k protein of adenovirus was found to contain the sequence DEKKMP at the carboxyl terminus, which determined its ER localization (3) and caused a block in cell surface expression when transferred to chimeric proteins (1). Subsequent studies have demonstrated a requirement for lysine residues at the Ϫ3 and Ϫ4 positions from the C terminus for ER retrieval (4 -8), which appears to be mediated by a 7-subunit receptor called the coatomer complex (9 -12).
The cytoplasmic tail length for type 1 transmembrane glycoproteins ranges from a few to several hundred amino acids. It has not been established whether there is a minimum or maximum length of intracytoplasmic amino acids required for the efficient ER retrieval of proteins containing the KK motif. To investigate the possible role of the position of such a signal within the cytoplasmic domain, we have used the envelope glycoprotein of simian immunodeficiency virus (SIV). This envelope glycoprotein was chosen because of the following properties: (a) it is a type 1 transmembrane protein with an unusually long cytoplasmic tail of 164 amino acids; (b) mutants with various truncations in the cytoplasmic tail have been shown to be efficiently transported to the cell surface (13,14) which enabled us to investigate the effect of placing a KK motif at different positions in the cytoplasmic tail; (c) the cytoplasmic tail includes two amphipathic helical regions which enabled us to determine the effect of placing the KK motif near such a helical region; and (d) the effect of the KK motif on surface expression can be evaluated by functional analysis of membrane fusion activity. The SIV Env protein is synthesized as a precursor (gp160); during its transport to the cell surface, about 10 -15% of gp160 is cleaved by a cellular protease resulting in the generation of surface (SU) and transmembrane (TM) subunits (13,15). In the present study, we have constructed a series of mutant proteins with lysine residues at Ϫ3 and Ϫ4 positions in the cytoplasmic domain and analyzed their intracellular processing, cellular localization, and ability to induce membrane fusion.

EXPERIMENTAL PROCEDURES
Cells, Virus, and Reagents-HeLa T4 (16) cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. HUT 78 cells were maintained in RPMI 1640 supplemented with 10% fetal calf serum. The recombinant vaccinia virus vTF7-3, which expresses bacteriophage T7 RNA polymerase in infected cells (17), was obtained from the National Institutes of Health AIDS Research and Reference Reagent Program. The rhesus antiserum to SIV was kindly provided by Dr. P. Marx (Aaron Diamond AIDS Research Center, New York, NY). The cDNA encoding ␤ COP was kindly provided by Dr. T. Kreis (University of Geneva, Geneva, Switzerland), and the antiserum to ␤ COP was purchased from Sigma.
Cloning of SIV Envelope Mutant Genes Carrying ER Retention Signals-Polymerase chain reaction was employed to construct SIV239 env mutants. Briefly, oligonucleotide primers which incorporated coding sequence for lysines at Ϫ3 and Ϫ4 positions followed by a stop codon were synthesized. The 5Ј-primer AAT ACG ACT CAC TAT AGG GCG AA was used with a panel of 3Ј-primers to obtain the desired mutants. Primers 3Ј-AGA GAA GAA TTC TTA ATA CCC CTT CTT TAA CTT  AGC TAG CAT TTG T (SIVenv6RS), 3Ј-CTG GAA GAA TTC TTA TGG  GGA TTT TTT CAC TGG CCT ATA CCC CTG (SIVenv13RS), 3Ј-TTG  GAT GAA TTC CTA CTG GAA TTT TTT GGG TGG GGA AGA GAA  CAC (SIVenv18RS), 3Ј-ACC GCC GAA TTC TTA GTC TCT TTT TTT  GCC TTC TCT GGT TGG CAG T (SIVenv37RS), 3Ј-TAG GTA GAA TTC  TTA AGT CCT TTT TTT TTC TCG AAT CCT CTG TAG (SIVenv103  RS), and 3Ј-TAT TTC GAA TTC TCA CAA GAG TTT TTT CTC AAG  CCC TTG TCT AAT (SIVenv164RS) were used. In addition to having the altered coding sequence for lysines and stop codon, the 3Ј-primers also incorporated an EcoRI restriction site. Polymerase chain reaction was performed using pBSKS ϩ vector (Stratagene) containing the SIV-mac239 env gene as the template. The sense and antisense primers were mixed with the template and appropriate buffer. Vent polymerase (New England Biolabs) was added, and polymerase chain reaction was performed for 30 cycles. After amplification, the products were purified, cut with EcoRI and XbaI, and cloned into the pBSKS ϩ vector that was cut with appropriate enzymes. The introduced mutations were confirmed by sequencing the plasmid DNA using Sequenase Version 2 (U. S. Biochemical Corp.) following the protocol recommended by the manufacturer.
Transfection, Radioimmunoprecipitation, and Protein Analysis-Transfection and protein analyses were done as described previously (5). Briefly, HeLa T4 cells (5 ϫ 10 5 ) were infected with vaccinia virus vTF7-3 (multiplicity of infection 10), and DNA (5 g) and Lipofectin (10 g) were added to the cells. At 7 h post-transfection, the cells were starved in medium lacking methionine and cysteine and then labeled with 100 Ci of [ 35 S]methionine and -cysteine (Amersham Life Sciences, Inc.) for 30 min. For better detection of the TM proteins, the cells were labeled with 100 Ci of [ 3 H]leucine after preincubation in medium lacking leucine. At the end of the labeling, the label was removed, and Dulbecco's modified Eagle's medium with fetal calf serum was added and chased for different times. Cells were lysed in radioimmune precipitation buffer and clarified, and proteins were immunoprecipitated with SIV-specific antiserum from an infected rhesus monkey. The immunoprecipitated proteins were extensively washed and analyzed by SDS-PAGE and autoradiography. For separation of the precursor and gp120, and for detection of the TM proteins, aliquots were analyzed on both 7.0 and 10.0% SDS-PAGE. The media collected at different chase times were immunoprecipitated and analyzed similarly.
Immunofluorescence-At 7 h post-transfection, the expression and cellular localization of proteins were analyzed after fixing cells for 10 min with paraformaldehyde (3.6%), permeabilizing for 5 min with Nonidet P-40, incubating with primary antibodies specific to the Env protein of SIV, and then using secondary mouse anti-monkey antibodies conjugated with fluorescein isothiocyanate. The coverslips were washed with phosphate-buffered saline, mounted on glass slides, and viewed in a Nikon fluorescence microscope.
Cell Fusion Assay-HeLa T4 cells were infected with vTF7-3 and transfected with plasmids as described above. At 8 h post-transfection, the cells were detached from the dishes using versene and cocultured with an equal number of CD4 ϩ HUT 78 cells. The cells were photographed using a modulation contrast imaging system (18) attached to a Nikon Diaphot microscope.
Assay for Intracellular Interaction of ␤ COP and SIV Env Proteins-HeLa T4 cells transfected with plasmid DNA were labeled with [ 35 S]methionine and -cysteine for 3 h, washed with phosphate-buffered saline, permeabilized with saponin (19), and incubated with ␤ COP antibody (Sigma) for 30 min at 4°C. The supernatant was collected, and it was determined that the SIV Env proteins did not leak out due to permeabilization. The cells were then washed in phosphate-buffered saline to remove the ␤ COP antibody and lysed in CHAPS buffer (20). The lysate of the transfected cells was divided into two equal portions; one portion was reacted with SIV antiserum to analyze the cell-associated Env proteins, and protein A was added to the other portion to detect the ␤ COP-Env protein complexes. The proteins were analyzed using SDS-PAGE and a PhosphorImager (Molecular Dynamics).

SIVmac239 Env Protein Mutants with KK Residues at Ϫ3
and Ϫ4 Positions Are Efficiently Retrieved to the ER-We have used the SIVmac239 Env protein to investigate the function of the KKXX retrieval signal in proteins with cytoplasmic domains of various lengths. Fig. 1 depicts the mutations introduced in the cytoplasmic tail, which all contained two lysine residues as the Ϫ3 and Ϫ4 amino acids from the C-terminal end. Using the recombinant vaccinia virus-based transient T7 expression system (19), cells transfected with plasmid DNA expressing wt or mutant SIV Env proteins were labeled with [ 35 S]methionine and -cysteine for 30 min and chased for up to 6 h. With the wt SIVmac239 Env protein, the precursor protein of 160 kDa was synthesized during the 30-min pulse (Fig. 2, lane 1). During 3-and 6-h chases, a fraction of the precursor protein was cleaved into the SU component gp120 and the TM component gp41 (Figs. 2A, lanes 2 and 3, and 2B, lane 1). These results are consistent with earlier observations (13,21,22).
To investigate the structural features necessary for a functional ER retrieval signal, we placed KK residues at the Ϫ3 and Ϫ4 positions of SIV Env proteins having cytoplasmic tail lengths of 18, 37, or 164 amino acids (Fig. 1). Immunoprecipitation of the protein encoded by construct SIVenv18RS showed the presence of a precursor protein ( Fig. 2A, lane 4) which was not proteolytically cleaved during the chase periods, as evidenced by the lack of SU or TM subunits (Figs. 2A, lanes 5 and 6, and 2B, lane 2). With the mutant protein encoded by construct SIVenv37RS, which has 37 amino acids in the cytoplasmic tail and lysines at Ϫ3 and Ϫ4 positions, similar results were obtained (Figs. 2A, lanes 7-9 and 2B, lane 3). To exclude the possibility that the lack of cleavage of proteins encoded by the constructs SIVenv18RS and SIVenv37RS was due to the deletions in the cytoplasmic tail, we also analyzed truncated proteins with cytoplasmic tails of 18 and 37 amino acids but which lacked the di-lysine motif. These proteins were efficiently cleaved and transported to the cell surface, as evidenced by cell surface immunofluorescence staining and secretion of gp120 into the media (data not shown). This demonstrates that the lack of proteolytic processing was a consequence of dilysine-mediated ER retrieval and retention.
To determine whether the signal is functional when placed in a protein with 164 amino acids in the cytoplasmic tail, we mutated the coding sequence of full length SIV Env to introduce di-lysines at Ϫ3 and Ϫ4 positions. When this construct (SIVenv164RS) was analyzed for protein expression, a precursor protein similar in size to wt SIV Env was synthesized during the 30-min pulse. With chases up to 6 h, this protein did not undergo proteolytic processing (Figs. 2A, lanes 10 -12 and 2B, lane 4), indicating its retrieval and retention in the ER.
To further investigate the cellular localization of these proteins, their site of expression was analyzed by indirect immunofluorescence. In permeabilized cells, the wild-type SIV Env protein exhibited a reticular staining pattern that extended throughout the cells (Fig. 3, panel A). As expected, the cells also showed surface expression of the envelope protein (panel E).
The intracellular staining pattern of the mutant proteins resembled that of the wild-type protein (Fig. 3, panels B-D). However, in contrast to the wild-type Env protein, no cell surface staining was detected for any of the three mutants (panels F-H), confirming that those proteins are efficiently retrieved and retained in the ER. Thus a di-lysine motif func- tions as a retrieval signal in proteins with a cytoplasmic tail 164 amino acids in length.
SIV Envelope Proteins with a Di-lysine Motif near the Membrane Spanning Domain Are Transported to the Cell Surface-To determine if ER retrieval signals placed closer to the membrane spanning domain are functional, we analyzed the mutant proteins encoded by constructs SIVenv6RS and 13RS and having 6 or 13 amino acids in the cytoplasmic tail and lysines at Ϫ3 and Ϫ4 positions from the C terminus. The proteins encoded by both constructs were found as a precursor during the 30-min pulse, and chases for 3 and 6 h resulted in cleavage into SU and TM components (Fig. 4). These data indicate that the mutant proteins are not retained in the ER but undergo proteolytic processing in a post-ER-Golgi compartment (15,23).
For both the mutant proteins, the intracellular immunofluorescence pattern resembled that of the wild-type protein (Fig.  5, A and B). Although the mutants contained the di-lysine motif, cell surface expression was observed on unpermeabilized cells (panels D and E). Thus, these results support the conclusion that a cytoplasmic domain of a minimum length is needed for efficient retrieval and retention in the ER by a di-lysine containing signal.

A Di-lysine Motif near an Amphipathic Helical Region of the Cytoplasmic Tail of SIV Env Does Not Function as a Retrieval
Signal-The cytoplasmic domain of HIV is postulated to form amphipathic helical structures between amino acids 770 -794 and 824 -856 (24), which may associate with the inner surface of the plasma membrane (25). Both of these regions are conserved in HIV and SIV genomes (26). To determine whether a retrieval signal positioned near such an amphipathic region could be functional, we analyzed the construct SIVenv103RS, which encoded a protein truncated after 103 amino acids in the cytoplasmic domain. During a 30-min pulse, a precursor protein of approximately 155 kDa was synthesized in transfected cells (Fig. 6A). After a 3-or 6-h chase, the precursor was cleaved into SU and a TM subunit of approximately 36 Kda, indicating that the retrieval signal was not functional (Figs.  6A, lanes 1-3 and 2B, lane 7).
The lack of recognition of the retrieval signal was confirmed by detection of gp120 in the extracellular medium from cells transfected with plasmid SIVenv103RS (Fig. 6B, lanes 1 and 2).
The protein exhibited a reticular immunofluorescence pattern within the cell (Fig. 6C, panel a) and was readily detected on the cell surface (panel b). These results indicate that the dilysine motif did not function as an effective ER retrieval signal when positioned near an amphipathic helical region.
Membrane Fusion Properties of SIV env Mutants-In cells transfected with expression vectors or SIV-infected cells, membrane fusion occurs when the cell surface-expressed Env protein interacts with the CD4 receptor on neighboring cells (13,27). Hence, we determined if the mutants with the di-lysine motif in different length cytoplasmic tails were functional in inducing the formation of syncytia. Coculture of HUT 78 cells with HeLa T4 cells expressing the wt SIV Env protein revealed syncytium formation (Fig. 7A). In contrast, the mutant proteins encoded by plasmids SIVenv18RS, SIVenv37RS, and SIVenv164RS failed to induce membrane fusion (panels B-D). These results are consistent with the observed defect in proteolytic processing and cell surface transport. In contrast, the mutant proteins expressed from plasmids SIVenv6RS and SIVenv13RS with the KK motif close to the membrane spanning domain were able to cause extensive membrane fusion (panels E and F), and there was no apparent difference between fusion activity of these mutants and truncated proteins with tail length of 6 and 13 amino acids which lacked the retrieval signal. These results are consistent with earlier data indicating increased fusogenic properties of cytoplasmic tail truncation mutants (13). The mutant protein encoded by construct SIVenv103RS, which has the retrieval signal near the amphipathic region, also induced membrane fusion (panel G). Thus, the results with the syncytium assay are consistent with the other results showing cell surface expression versus ER retrieval of the Env mutants.
Interaction of ␤ COP with SIV Env Proteins-To investigate the mechanism involved in the observed differences in retrieval of SIV mutant proteins, we analyzed the interaction of the Env proteins with ␤ COP proteins. ␤ COP is a 110-kDa protein that is associated with non-clathrin-coated vesicles and the Golgi complex (28) and has been implicated in the transfer of proteins from the ER to the Golgi complex (29). We developed an assay using saponin permeabilization which enabled us to determine binding of the retrieved or retained proteins to ␤ COP. Quantitation of the amount of ␤ COP-associated Env proteins showed that proteins encoded by plasmids 164RS, 37RS, and 18RS bound to ␤ COP at least at 3-to 5-fold higher levels than the wt Env protein (Figs. 8A, lanes 2-4, and 8B). In contrast, the KK motif-containing proteins that were not retained showed lower binding to ␤ COP (Figs. 8A, lanes 1 and 5-7, and 8B). The amount of radiolabeled ␤ COP detected in Env-expressing cells was less than that of the coprecipitated Env proteins, presumably because most of the ␤ COP is unlabeled due to the 3-h labeling time used. To confirm the identity of ␤ COP, we transfected cells with a plasmid DNA expressing ␤ COP, and the lysate was immunoprecipitated using ␤ COP antibodies (Fig. 8A, lane 8). The intracellular expression levels of the Env proteins were similar with all the constructs (Fig.  8C). These results demonstrate that an increased binding to ␤ COP is observed in the proteins with a functional ER retention motif.

DISCUSSION
Most previous studies of ER retrieval and retention have analyzed proteins with cytoplasmic tails of more than 15 amino acids although there are examples of retained proteins having only 10 amino acids in the cytoplasmic tail (10,30,31). Chimeric proteins used to analyze the di-lysine motif had at least 15 amino acids in the cytoplasmic tail, and attempts to place the lysines closer to the membrane were not reported. The present results provide evidence that potential retrieval signals in the SIV Env proteins with a tail length of 6 or 13 amino acids are not functional, whereas such a signal is functional in proteins with longer cytoplasmic domains. Structural modeling (24) and topogenic analyses of the HIV-1 envelope protein (25) indicated that the cytoplasmic domain has the propensity to form two amphipathic helical structures. Peptides which correspond to the amphipathic region were found to interact with membranes (32,33), indicating that such amphipathic regions may promote membrane association. The mutant protein encoded by plasmid SIVenv103RS contains a cysteine residue that is the site for palmitoylation (34), and this modification was postulated to enable the amphipathic regions to tightly associate with cellular membranes. Our observation that a retrieval signal placed near the predicted amphipathic helical region of the SIV Env protein is not recognized by the cellular retention machinery may be explained by the close association of this region with cellular membranes.
The finding that di-lysines introduced at different positions in the SIV envelope protein result in proteins that have different cellular localization phenotypes raises several possibilities. When lysines are positioned very near to the membrane spanning domain, they may not be recognized because of the lack of some structural features. Alternatively, when the signal is located near the membrane, a cellular protein may sterically mask the signal and thereby result in lack of recognition. A masking of a retention signal has been reported to occur during the assembly of the ␣ and ␥ chains of human high affinity receptor for immunoglobulin E (35). The retention or retrieval of proteins in the ER is a consequence of the interaction of the di-lysine motif with the members of the 7-subunit coatomer complex (9 -11). It has been shown that di-lysine-containing proteins bind to ␣, ␤Ј and ⑀ (10), ␥ COP (36), and ␦ and (37) subunits of the coatomer complex. Our data indicate that the Env proteins which are retained intracellularly exhibited higher binding to ␤ COP than the proteins which were transported to cell surfaces, suggesting that this interaction is involved in the mechanism for the ER retrieval of SIV Env proteins.
The cell surface transport of several viral proteins can be modulated by the length of the cytoplasmic tail. Deletions or changes in the cytoplasmic tails of VSV G (38), paramyxovirus HN (39), and paramyxovirus F (40) proteins have been shown to inhibit or delay the cell surface delivery of those proteins. VSV G mutants that were retained in the ER had highly charged C termini and a lysine occupying the Ϫ3 position (38). In contrast, for HIV and SIV, the truncation of the cytoplasmic tail does not have a significant effect on cell surface transport (13,(41)(42)(43). Furthermore, we have confirmed that the lack of cleavage and cell surface transport of proteins expressed from plasmids SIVenv18RS and SIVenv37RS were not due to the truncations in the cytoplasmic tail. We found that all mutants in which the di-lysine motif was changed to other amino acids (Ϫ3 and Ϫ4 FF or PF) were processed and transported to the cell surface, confirming the strict requirement for lysines in the retrieval signal. Several studies have indicated that the amino acids surrounding the KK motif did not contribute significantly to the retrieval phenomenon (2,5,44). Although the mutants we have described have different amino acids at the Ϫ1 and Ϫ2 positions, these or similar amino acids are present in other proteins that were found to be retrieved to the ER by the KK motif at the Ϫ3 and Ϫ4 positions (2,44). Hence, we conclude that the lack of retention of the mutant proteins expressed by plasmids SIVenv6RS, SIVenv13RS, and SIVenv103RS is due to the location of the KK motif in a short cytoplasmic tail or near an amphipathic helical region, which prevented efficient binding of these proteins to the coatomer complex.
Recently it has been reported that specific tyrosine residues in the cytoplasmic tail of the envelope protein of HIV-1 are responsible for endocytosis of the protein (45). Mutations or deletion of these tyrosine residues resulted in a dramatic in-crease in cell surface expression, and this phenomenon was attributed to the reduced rate of endocytosis (45). This phenomenon is not likely to account for the differences that we observed in the expression of proteins at the cell surface because SIVenv103RS, which was expressed at the cell surface, and SIVenv164RS, which was retained intracellularly, contain both of the tyrosine motifs thought to be responsible for endocytosis. Furthermore, other mutants that are expressed at the cell surface (SIVenv6RS and SIVenv13RS) as well as those that are not expressed at the cell surface (SIVenv18RS and SIVenv37RS) all contain a tyrosine residue that is conserved in many SIV and HIV isolates (21).
In summary, we have provided evidence that the ability of lysine residues at Ϫ3 and Ϫ4 positions to function as an efficient retrieval signal depends on the length and structure of the cytoplasmic domain. The mutants in which the KK motif was functional failed to undergo proteolytic processing, were not detected on the cell surface, and were defective in inducing cell fusion. Thus, our studies provide evidence that there is a requirement for lysines to be placed at a minimum distance from the membrane spanning domain.