Cytotoxic T Lymphocyte-associated Molecule-4, a High Avidity Receptor for CD80 and CD86, Contains an Intracellular Localization Motif in Its Cytoplasmic Tail*

CD28 and CTLA-4, T cell receptors for B7–1 (CD80) and B7–2 (CD86) molecules on antigen-presenting cells, transmit costimulatory signals important for optimal T cell activation. Despite sharing sequence homology and common ligands, these receptors have distinct binding properties and patterns of expression. The function of CTLA-4 during T cell activation is not well understood, although an important role is suggested by complete amino acid sequence conservation of its cytoplasmic tail in all species studied to date. We report here a role of the cytoplasmic tail of CTLA-4 in regulating its subcellular localization and cell surface expression. In activated human peripheral blood T cells, or in several trans- fected or transduced cell types, CTLA-4 is not primarily a cell surface protein, but rather is localized intracellu- larly in a region which overlaps the Golgi apparatus. Transfer of 11 residues, from the CTLA-4 cytoplasmic tail to the homologous position in CD28 was sufficient to confer intracellular localization. Mutation of the tyrosine residue (Tyr 165 ) in this motif to phenylalanine resulted in increased sur- face expression of CTLA-4. Thus, the subcellular localization of CTLA-4 is controlled by a tyrosine-containing motif within its cytoplasmic domain. Contained within this motif is a binding site for SH2 domains of the p85 subunit of phosphatidylinositol 3-kinase. Full activation of T cells requires engagement of the TCR-CD3 complex and ligation of costimulatory receptor(s) (1). T cells activated in the absence of costimulatory

CD28 and CTLA-4, T cell receptors for B7-1 (CD80) and B7-2 (CD86) molecules on antigen-presenting cells, transmit costimulatory signals important for optimal T cell activation. Despite sharing sequence homology and common ligands, these receptors have distinct binding properties and patterns of expression. The function of CTLA-4 during T cell activation is not well understood, although an important role is suggested by complete amino acid sequence conservation of its cytoplasmic tail in all species studied to date. We report here a role of the cytoplasmic tail of CTLA-4 in regulating its subcellular localization and cell surface expression. In activated human peripheral blood T cells, or in several transfected or transduced cell types, CTLA-4 is not primarily a cell surface protein, but rather is localized intracellularly in a region which overlaps the Golgi apparatus. Transfer of 11 cytoplasmic residues, 161 TTGVYVKMPPT, from the CTLA-4 cytoplasmic tail to the homologous position in CD28 was sufficient to confer intracellular localization. Mutation of the tyrosine residue (Tyr 165 ) in this motif to phenylalanine resulted in increased surface expression of CTLA-4. Thus, the subcellular localization of CTLA-4 is controlled by a tyrosine-containing motif within its cytoplasmic domain. Contained within this motif is a binding site for SH2 domains of the p85 subunit of phosphatidylinositol 3-kinase.
Full activation of T cells requires engagement of the TCR-CD3 complex and ligation of costimulatory receptor(s) (1). T cells activated in the absence of costimulatory signals may enter a state of hyporesponsiveness or anergy (2). Recent studies have shown that B7-1 (CD80) and B7-2 (CD86) molecules on antigen-presenting cells (APC) 1 provide a major T cell costimulatory signal in vitro and in vivo (3,4). T lymphocyte receptors for CD80 and CD86 are CD28 and CTLA-4, homologous members of the immunoglobulin superfamily (5) which are encoded by closely linked genes on human chromosome 2 (6). CD28 is a low avidity receptor for both CD80 and CD86, whereas CTLA-4 is a high avidity receptor (7). A highly conserved motif (MYP-PPY) in the CDR3-like region of both CD28 and CTLA-4 is involved in binding of these receptors to CD80 (8). CD28 and CTLA-4 have different patterns of expression during an immune response, with the former being expressed in both resting and activated cells and the latter, only in activated cells (9,10).
A major function of CD28 costimulation is to increase production of interleukin-2 and other T cell cytokines (11,12). The role of CTLA-4 is less clear, although there is complete sequence conservation in the cytoplasmic domains of mouse (13), human (5), and rabbit 2 CTLA-4, which suggests conserved function(s). Studies with mAbs against CTLA-4 have led to two different and distinct models for the role of CTLA-4 in T cell activation (14), either to cooperatively up-regulate T cell activation (9,15) or to antagonize CD28 and down-regulate T cell activation (16).
Studies on CTLA-4 function have been hampered by its low level of expression. Very low levels of CTLA-4 were detected on the cell surface of activated T cells, although RNA transcripts were readily detectable (9,15). We have also found very low levels of CTLA-4 expressed on the cell surface following its transfection or transduction into numerous cell types. 3 In contrast, cell surface expression of CD28 is easily observed in peripheral blood T cells or in transfected or transduced cells. These observations led us to hypothesize that expression of CTLA-4 was subject to more complex regulation than CD28.
Here we show that one reason for the low level of CTLA-4 cell surface expression is that it is localized primarily on intracellular membranes. Furthermore, we show that the cytoplasmic domain of CTLA-4 contains an intracellular localization motif which targets the protein to a perinuclear Golgi or post-Golgi compartment in transfected cells, consequently reducing its cell surface expression. This motif may regulate CTLA-4 surface expression and function during T cell activation.
Recombinant Constructs-CD28 and CTLA-4 expression constructs, OM-CD28 and OM-CTLA-4, have been described previously (18). These and all subsequent constructs contain the signal peptide from oncostatin M. Stable lines expressing CD28 and CTLA-4 were created using retroviral vectors. The full-length cDNAs were amplified using polymerase chain reaction and cloned into the HindIII and ClaI sites of the retroviral vector pLNCX (19) to create pLNC/CD28 and pLNC/CTLA-4. Chimeric constructs C28 -4 (extracellular and transmembrane domains of CD28 fused to the cytoplasmic tail of CTLA-4) and C4 -28 (extracellular and transmembrane domains of CTLA-4 fused to the cytoplasmic tail of CD28) were assembled from polymerase chain reaction fragments amplified from pLNC/CD28 and pLNC/CTLA-4. C28-4 contains CD28 sequences up to R162 (all numbering refers to the mature protein) and CTLA-4 sequences from Lys 155 to Asn 187 (cytoplasmic tail). C4-28 contains CTLA-4 sequences up to Leu 154 and CD28 sequences from Ser 163 to Ser 202 . Retroviral constructs were introduced into the packaging cell line PA317 by electroporation and recombinant viruses were amplified in 2 packaging cells (20).
For transient COS cell transfection, cDNAs were excised from the retroviral constructs with HindIII and ClaI, blunt-ended with Klenow DNA polymerase, and ligated into the EcoRV site of the expression vector pcDNA1 (Invitrogen, San Diego, CA). Truncation, homologue swap and point mutations were constructed using polymerase chain reaction (8) and were cloned into pcDNA1. The amino acid sequences surrounding the replacement regions of R1-R5 are shown in Table I. All mutations were confirmed by DNA sequencing.
Cell Culture and Transfection-Anti-CD3-and phytohemaglutininactivated T cell blasts were prepared essentially as described (9). Peripheral blood mononuclear cells were isolated from heparinized blood and cultured on anti-CD3 coated flasks or with phytohemaglutinin (1 g/ml) for 3 days at 37°C in RPMI 1640 containing 10% FBS, 2 mM glutamine, and pen-strep (100 units/ml penicillin, 75 units/ml streptomycin). In some experiments, culture flasks were coated with F(abЈ)2 fragments (10 g/ml) of anti-CD3 mAb G19 -4 instead of intact mAb. Murine T cell hybridomas C8.A3 (21) and 3B4 were obtained from T. Watts (University of Toronto) and P. Dubois (BMSPRI), respectively, and were maintained in RPMI 1640 with 10% FBS, 2 mM glutamine, 50 M 2-mercaptoethanol, and pen-strep. NIH 3T3 fibroblasts and COS cells were maintained in Dulbecco's modified Eagle's medium with 10% FBS, glutamine, and pen-strep. For retroviral infection, C8.A3, 3B4, or 3T3 cells were incubated for 2 h with recombinant retroviruses in the presence of 4 g/ml polybrene (Aldrich). The following day, transduced cells were selected for neomycin resistance by culturing with G418 at 1 mg/ml. COS cells were transfected as described previously (22) and were harvested 48 h after transfection.
Confocal Microscopy-Intact or permeabilized T cells were stained with mAbs at 10 g/ml, followed by FITC-GAM (Tago Immunochemicals). 3T3 cells were seeded on eight-chamber cover glasses (Nunc Inc., Naperville, IL). After overnight incubation, cells were fixed and permeabilized with 100% methanol at Ϫ20°C for 5 min before staining. In double staining experiments, biotinylated Lens culinaris lectin (LcL, Vector Laboratories, Burlingame, CA) was added at 10 g/ml during incubation with the first step antibody. Texas Red-conjugated avidin (Vector Laboratories, Burlingame, CA) was added with FITC-GAM for detection. Samples were analyzed using a confocal microscope (15).
Radiolabeling and Immunoprecipitation-Intact cells and membrane fractions of anti-CD3-activated T cell blasts were radioiodinated with 125 I (17). Labeled extracts were first precleared by incubation with anti-CD28 mAb 9.3, and the immune complexes were precipitated with immobilized protein A (r-protein A-Sepharose, Repligen, Cambridge, MA). CTLA-4 was immunoprecipitated either with anti-CTLA-4 mAb 11D4 or with CD80Ig. Transiently transfected COS cells were harvested by trypsinization, washed, and incubated in methionine-deficient MEM with 10% dialyzed FBS for 2 h. Metabolic labeling was initiated by adding Tran 35 S-label (a mixture of [ 35 S]methionine and cysteine, ICN Radiochemicals, Costa Mesa, CA) to 500 Ci/ml for 10 -15 min. Preferential metabolic labeling of methionine with this mixture was ensured by the inclusion of Ͼ500-fold excess of unlabeled cystine in the labeling medium. To end labeling, cells were collected by sedimentation and resuspended in complete culture medium containing normal amounts of cysteine and twice the normal concentration of unlabeled methionine.
For immunoprecipitation of cell surface CTLA-4 or CD28 molecules (surface immunoprecipitation), cells were resuspended in 0.5 ml of phosphate-buffered saline containing 10 g of CD80Ig or anti-CD28 mAb 9.3 and incubated on ice for 30 min. After two washes with ice-cold phosphate-buffered saline, cells were dissolved in 0.5 ml of lysis buffer (17) containing 5 g of CTLA-4t or CD28t to saturate unoccupied antibody binding sites. Immune complexes were immediately precipitated with immobilized protein A. For immunoprecipitation of total CTLA-4 or CD28 molecules (totalintraperitoneal), samples were directly lysed and immunoprecipitation was performed with CD80Ig or anti-CD28 mAb 9.3, followed by immobilized protein A. Immunoprecipitates were treated with endoglycosidase H (endo H, Boehringer Mannheim) as suggested by the manufacturer. Immunoprecipitates were analyzed by SDS-PAGE.
RNA Blot Analysis-Total RNA was isolated from anti-CD3-stimulated T cell blasts (9). In vitro RNA transcripts were synthesized using the RiboMax Kit (Promega, Madison, WI), as suggested by the manufacturer. The indicated amounts of RNA were fractionated by electrophoresis on a 1.2% agarose, 0.33 M formaldehyde gel, blotted to a nylon membrane, and probed as described previously (9). Fig. 1. Both CD28 and CTLA-4 cDNA probes gave similar intensities on graded amounts of RNA; the probes had similar specific activities, as judged by hybridization to known amounts of in vitro transcripts for CD28 and CTLA-4. Thus, these cells contained similar levels of CD28 and CTLA-4 transcripts. However, when the same cells were analyzed by surface staining, anti-CD28 mAb 9.3 gave a mean fluorescence intensity (MFI) of ϳ290 (after correction for background) versus ϳ40 for anti-CTLA-4 mAb 11D4. Staining with A, activated T cells contain equivalent amounts of mRNA for CD28 and CTLA-4. Peripheral blood T cells were activated by anti-CD3 mAb stimulation for 3 days. Cells were then harvested, total RNA was extracted, and the indicated amounts were analyzed by blot hybridization analysis using 32 P-labeled probes for CD28 and CTLA-4. The blot was hybridized first with the CD28 probe, stripped, and then hybridized with the CTLA-4 probe. B, cell surface expression of CD28 is much higher than cell surface expression of CTLA-4 on activated T cells. T cells activated as in A were analyzed for cell surface expression of CD28 or CTLA-4 by staining with biotinylated anti-CD28 mAb 9.3 or anti-CTLA-4 mAb 11D4, respectively, followed by PE-SA. Stained cells were then analyzed by flow cytometry. Solid lines, histograms obtained with mAbs 9.3 or 11D4; dotted lines, histograms obtained with isotype control mAbs.

Evidence for Posttranscriptional Regulation of CTLA-4 Expression-A quantitative comparison of levels of CD28 and CTLA-4 transcripts and cell surface expression in activated T cell blasts is shown in
anti-CD28 mAb was always much greater than staining with anti-CTLA-4 mAb, no matter which anti-CTLA-4 mAb, what antibody concentration, or which method of detection were used. Therefore, there was a greater difference in CD28 and CTLA-4 cell surface expression than there was in levels of their RNA transcripts. This suggested a difference in posttranscriptional regulation of these molecules CTLA-4 Is Primarily an Intracellular Protein in Activated T Cell Blasts-There are several mechanisms which might account for the difference in cell surface expression of CD28 and CTLA-4 on activated T cells. We compared the subcellular distributions of CD28 and CTLA-4 by confocal microscopy in Fig. 2. Anti-CD3-activated peripheral blood T cells were stained with isotype control, anti-CD28, or anti-CTLA-4 mAbs before or after treatment with saponin. CD28 staining was detected primarily at the cell surface of intact cells; this staining pattern was not significantly altered by saponin treatment. Intact cells gave little CTLA-4 staining on the cell surface, but significant intracellular staining following saponin treatment. CTLA-4 staining in saponin-treated cells gave a perinuclear vesicular pattern beneath the cell surface. Thus, CD28 is expressed primarily at the cell surface, whereas CTLA-4 expression is primarily intracellular.
The increase in CTLA-4 staining following saponin treatment was quantitated by flow cytometry. In five experiments with T cell blasts activated by anti-CD3 treatment for 3 days, CTLA-4 expression was consistently increased by treatment with saponin prior to staining (mean increase of 5.4 Ϯ 1.7-fold). Excess unlabeled CTLA4Ig blocked the intracellular staining, indicating that it was specific (data not shown). The increase in CTLA-4 staining intensity following saponin treatment was also observed when T cells were activated for 3 days with phytohemaglutinin. In comparison, the intensity of CD28 staining was not consistently affected by saponin treatment (mean increase 1.5 Ϯ 1.1-fold). Staining intensities obtained with anti-CD2, anti-CD4, and anti-CD7 mAbs decreased or remained constant following saponin treatment (data not shown). Thus, CTLA-4 staining was increased ϳ5-fold by mild detergent treatment of activated T cells, whereas staining of other T cell surface antigens was much less affected.
We also examined the subcellular distribution of CD28 and CTLA-4 by comparing their accessibilities with lactoperoxidase-mediated iodination. Similar levels of 125 I-labeled CD28 were immunoprecipitated following iodination of intact cells or isolated membranes, but ϳ7-fold more 125 I-labeled CTLA-4 was immunoprecipitated from isolated membranes than from intact cells. Therefore, CTLA-4 was more accessible to lactoperoxidase-catalyzed iodination when the cell surface membrane was disrupted. This difference could not be explained by exposure of new sites for iodination in the predicted intracellular portion of CTLA-4, since only two of the nine tyrosine residues are contained in this region. Identical results were obtained whether CTLA-4 was immunoprecipitated with mAb 11D4 or CD80Ig, indicating that CTLA-4 exposed by membrane disruption could bind to a physiological ligand(s). Taken together, the data suggest that most CTLA-4 in activated T cells is associated with intracellular membranes, rather than the cell surface.
The CTLA-4 Cytoplasmic Tail Directs Expression to an Intracellular Compartment in Transfected or Transduced Cells-In parallel experiments, we constructed recombinant retroviruses carrying cDNAs of CD28, CTLA-4, and chimeric receptors whose cytoplasmic tails were replaced by the counterpart of the other receptor (C28-4, containing the extracellular and transmembrane domains of CD28 fused to the intracellular domain of CTLA-4; and C4 -28, containing the extracellular and transmembrane domains of CTLA-4 fused to the intracellular domain of CD28). These constructs were introduced into the murine T cell hybridoma C8.A3, and bulk G418-resistant populations were selected for analysis by flow cytometry (Fig. 3, top panels). While CD28 was expressed at readily detectable levels, surface expression of C28 -4 was significantly reduced. Conversely, expression of full-length CTLA-4 was undetectable, but substitution of its cytoplasmic tail with that of CD28 (C4-28) increased surface expression. Therefore, molecules containing the CTLA-4 cytoplasmic tail were poorly expressed on the surface of C8.A3 cells. These findings were repeated in another murine T cell hybridoma, 3B4 (data not shown). NIH 3T3 fibroblasts infected with the same viral constructs also displayed similar pattern of surface expression, but all constructs were expressed at higher level than in C8.A3 (Fig. 3, middle panels). Similar results were also obtained in transiently transfected COS cells (Fig. 3, bottom  panels). RNA blot analysis showed that comparable levels of RNA transcripts were produced for different CD28 or CTLA-4 constructs in C8.A3 and 3T3 transduced cells (data not shown).
In other experiments, we tested by flow cytometry whether the reduced surface expression of constructs containing the intracellular region of CTLA-4 was associated with intracellular accumulation of these proteins. 3T3 cells transduced with CD28 showed similar levels of CD28 expression before or after saponin treatment, but expression of CTLA-4 was markedly increased following saponin treatment. Specificity of intracellular staining following saponin treatment was indicated by the ability of CD28Ig or CTLA4Ig to inhibit staining with PE-conjugated specific mAbs. Similar results were obtained when CTLA-4 was expressed in C8.A3 and COS cells. In all cases, saponin treatment increased staining with anti-CTLA-4 mAb 11D4, but not staining with anti-CD28 mAb 9.3. The cytoplasmic tail of CTLA-4 therefore is responsible for its retention in an intracellular compartment in both lymphoid and non-lymphoid cells.
We further characterized the intracellular compartment in which CTLA-4 was retained in 3T3 cells using confocal microscopy (Fig. 4). 3T3-transduced cells were fixed and permeabilized with methanol and stained with anti-CTLA-4 mAb 11D4. The same cells were also stained with LcL, a marker for the Golgi apparatus (23). This lectin recognizes fucosylated Nlinked carbohydrate side chains that terminate with galactose, N-acetylglucosamine or sialic acid (24), structures found in the Golgi apparatus, but not the endoplasmic reticulum (25). With CTLA-4-transduced cells, the staining pattern obtained with anti-CTLA-4 mAb 11D4 overlapped with staining obtained with LcL. In both cases, intracellular perinuclear vesicular staining was seen. With C4 -28-transduced cells, CTLA-4 staining was primarily cell surface-associated, although intracellular perinuclear vesicular staining was also seen. These findings demonstrate that the CTLA-4 cytoplasmic tail directs its localization to an intracellular compartment which overlaps with the Golgi. Control experiments established that the coincidence of the two staining patterns was not caused by "spillover" of the emission of either chromaphore into the detection system for the other.
Identification of Sequence in the CTLA-4 Cytoplasmic Tail Which Regulates Cell Surface Expression-To identify the sequence(s) responsible for the intracellular localization of CTLA-4, we performed mutational analysis on its cytoplasmic tail (Fig. 5). Since CTLA-4 expression was largely intracellular in several diverse cell types, we chose to perform these exper- iments in transiently transfected COS cells. Mutant proteins (Fig. 5A and Table I) were expressed in COS cells, their cell surface expression was monitored by flow cytometry (Fig. 5B), and subcellular localization was examined by immunofluorescence microscopy (Fig. 6).
Analysis of carboxyl-terminal truncations of C28-4 showed that up to 16 terminal residues could be deleted from the cytoplasmic tail of C28-4 without significantly changing surface expression levels (compare D8 and D16). When 20 residues were deleted (D20), a modest increase in surface expression was detected by flow cytometry. When 23 residues were deleted from C28-4 (D23), surface expression increased further. These results suggested that the carboxyl terminus of a localization motif was between Met 168 and Thr 171 . We then tested homologue swap mutants in which small regions of the cytoplasmic tail of CD28 were replaced by the corresponding residues from CTLA-4 (see Table I). As shown in Fig. 5B, COS cells transfected with mutant R1 had the greatest reduction in surface expression (similar to C28-4). Mutants R2-R4 gave partial reductions in surface expression, whereas mutant R5 had surface expression similar to CD28. Immunofluorescence microscopy experiments showed that mutant R1 had primarily intracellular localization (Fig. 6), whereas mutants R2-R5 were primarily cell surface-associated (data not shown). Mutant R1 therefore contains the smallest CTLA-4 region which inhibits surface expression when transferred to CD28.
We next introduced into CTLA-4 mutations at each of the 11 amino acid residues contained in R1. Substitution of alanine for each residue (except for Tyr 165 ) did not increase surface expression of CTLA-4 (data not shown). To test the involvement of Tyr 165 in regulation of cell surface expression, we introduced a tyrosine to phenylalanine mutation into this position (mutant Y165F). This mutation increased surface staining for CTLA-4 (Figs. 5B and Fig. 6), indicating that Tyr 165 regulates cell surface expression.
We also tested by flow cytometry whether mutant R1 accumulated in an intracellular compartment(s). When COS cells transfected with R1 were treated with saponin prior to staining with PE-conjugated anti-CD28 mAb 9.3, the fraction of cells staining was clearly increased. The MFI of the total population (after correction for background) increased noticeably from 3 to 11 following saponin treatment, whereas the MFI of CD28transfected cells decreased slightly from 55 to 38. Staining was specific because it was decreased down to background levels by addition of CD28Ig. These results are in agreement with immunofluorescence experiments (Fig. 6) showing that mutant R1 was primarily localized in an intracellular compartment.
Mutant R1 also was expressed at lower levels than CD28. We reproducibly detected 3-8-fold lower amounts of mutant R1 in transfected COS cells, as judged by specific immunoassays (data not shown). The basis for this reduced expression is not currently known, although all molecules containing the CTLA-4 cytoplasmic tail showed a similar reduction in expression. With C28 -4, protein expression was reduced when com-pared with CD28 observed, even though similar levels of RNA transcripts were detected. This suggested that as yet unidentified posttranscriptional mechanisms may regulate the levels of CTLA-4 expression.
Biochemical Characterization of the Intracellular Localization of CTLA-4 -The kinetics of cell surface transit of CD28 and mutant R1 were analyzed by pulse-chase analysis in Fig. 7. The amount of CD28 detected by surface immunoprecipitation increased steadily over the 90 min chase period, whereas very little R1 was isolated by surface immunoprecipitation Both CD28 and R1 were present in total extracts, although there was less of the latter. To properly compare the fractions of CD28 and R1 expressed at the cells surface, it was necessary to normalize radioactivity in surface immunoprecipitation to the total amount of radioactivity (Fig. 7, bottom). This analysis revealed that CD28 was steadily translocated to the cell surface, whereas the fraction of R1 expressed at the cell surface increased less. Since the fraction of R1 expressed at the cell surface did not greatly increase, the intracellular expression of mutant R1 could not be attributed to rapid internalization of newly synthesized molecules from the cell surface.
We also used pulse-chase analysis to compare the intracellular localization of CTLA-4 and mutant Y165F (Fig. 8). During the 90-min chase period, very little wild type CTLA-4 accumulated at the cell surface, as detected by surface immunoprecipitation, whereas mutant Y165F rapidly appeared at the cell surface, reaching a maximum at ϳ40 min. Very similar amounts of radioactive CTLA-4 and Y165F were obtained by total immunoprecipitation, in agreement with immunoassays showing that these proteins were expressed at similar levels (data not shown). Normalization of the amount of radioactivity in surface immunoprecipitation for the amount in total immunoprecipitation indicated that ϳ4-fold more mutant Y165F accumulated at the cell surface than wild type CTLA-4. Moreover, accumulation of surface CTLA-4 was steady and slow, suggesting that newly synthesized CTLA-4 molecules were blocked during surface translocation, rather than being rapidly internalized after reaching the surface. In other experiments,  6. Immunofluorescence patterns of CD28, mutant R1, CTLA-4, and mutant Y165F. COS cells were transfected with expression plasmids encoding the indicated proteins. Forty-eight hours following transfection, cells were fixed with methanol and stained with biotinylated anti-CD28 mAb 9.3 or anti-CTLA-4 mAb 11D4 followed by PE-SA (red fluorescence). Nuclei were stained with YO-PRO-1 (green fluorescence). later time points were studied, but the fraction of CTLA-4 and Y165F at the cell surface did not increase further beyond ϳ90 min. CD80Ig was used for immunoprecipitation analysis in Fig.  9. Since similar amounts of CTLA-4 and mutant Y165F were purified by total immunoprecipitation, intracellular CTLA-4 was able to bind its physiological ligand, and hence, was biologically active.
To further characterize the intracellular compartment where CTLA-4 was retained, we performed another pulse-chase analysis and tested the sensitivity of CTLA-4 and mutant Y165F to endo H during the chase period (Fig. 9). Both CTLA-4 and mutant Y165F were initially endo H-sensitive but became resistant during the chase period. The kinetics of acquisition of resistance to endo H were similar for both proteins (t1 ⁄2 ϳ30 min). This indicates that their transit times were similar through the endoplasmic reticulum into the middle to trans-Golgi, where ␣-mannosidases responsible for processing of Nlinked carbohydrates are localized (26). DISCUSSION CD28 and CTLA-4 are homologous molecules with features typical of cell surface membrane receptors (5,6). We have shown here that in activated human peripheral blood T cells CTLA-4 was primarily (ϳ80%) localized on intracellular membranes rather than at the cell surface. Intracellular localization of CTLA-4 was also observed in several cell types transduced or transfected with CTLA-4 cDNA, indicating that intracellular localization is an inherent property of the CTLA-4 protein and not the particular cell type in which it is expressed. In contrast, CD28 was primarily expressed at the cell surface, in agreement with numerous studies implicating it as a T cell surface receptor (11,12). Thus, in addition to having different avidities for B7 ligands (7), and different patterns of expression (9, 10), CD28 and CTLA-4 have different subcellular localization. This further argues that these molecules have different functions during T cell activation (14).
Using mutational analysis, we identified an intracellular localization motif in the CTLA-4 cytoplasmic domain which limits its cell surface expression in COS cells and localizes it to an intracellular region. Mutant R1 showed reduced cell surface expression and primarily intracellular localization. In contrast, mutants R2, R3, and R4, which contain only some of the CTLA-4 residues present in R1, showed greater cell surface Mutant Y165F showed greater cell surface expression and less intracellular expression than wild type CTLA-4. This indicates that Tyr 165 is critical for intracellular localization function. Several other tyrosine-containing intracellular internalization/localization motifs have been identified, such as in transferrin receptor, low density lipoprotein receptor, cationindependent mannose 6-phosphate receptor, and the trans-Golgi protein TGN38 (27,28). However, none of these motifs bears obvious homology with the CTLA-4 localization motif. Moreover, in transferrin, low density lipoprotein, and cationindependent mannose 6-phosphate receptors, phenylalalnine can substitute for the tyrosine with no loss of activity (27). Intracellular localization of CTLA-4 could be due to the lack of transport proteins required to target CTLA-4 to the cell surface or to retention of CTLA-4 by proteins interacting with the intracellular localization motif. Since disruption of the CTLA-4 intracellular localization motif increased cell surface expression, we favor the latter possibility. Although intracellular localization motifs have been identified in many proteins, less is known about protein(s) interacting with these motifs (reviewed in Ref. 29).
Intracellular staining of CTLA-4 overlapped with the staining pattern of LcL, a marker for the Golgi apparatus (23). Further evidence as to the site of intracellular localization of CTLA-4 comes from the kinetics of acquisition of resistance to endo H. CTLA-4 became endo H-resistant with identical kinetics as mutant Y165F, which is transported ϳ4-fold more efficiently to the cell surface. Thus, the block in transport of CTLA-4 to the cell surface occurs subsequent to processing of high mannose N-linked carbohydrate chains, which occurs in the middle to trans-Golgi (25,26). Taken together, these data suggest that CTLA-4 is retained in a trans-or post-Golgi compartment.
The CTLA-4 intracellular localization motif contains a consensus binding site (-YVKM-) for the Src homology 2 (SH2) domains of the p85 subunit of phosphatidylinositol 3-kinase (PI3K) (30). Previous studies showed that anti-CTLA-4 mAbs triggered binding of p85 to this site in the cytoplasmic domain of CTLA-4 in an HTLV I-transformed T cell line (31). The CD28 cytoplasmic domain also contains a similar motif (YMNM) which binds p85 following engagement of CD28 with mAbs or its CD80 ligand (32)(33)(34)(35)(36)(37). Thus, the intracellular localization motif of CTLA-4 and the analogous region of CD28 contain sequence(s) involved in signaling through these receptors. However, our data provide evidence that these sites may not be functionally equivalent in CD28 and CTLA-4, since the site in CTLA-4 has intracellular localization function, whereas that in CD28 does not. Another difference in cytoplasmic domains of CTLA-4 and CD28 was reported by Stein et al. (33), who showed that CD8 chimeras containing the CD28 cytoplasmic tail bound PI3K and triggered interleukin-2 production, but that CD8/CTLA-4-cytoplasmic tail chimeras did not. Taken together, these studies suggest that the YXXM-containing motif in CTLA-4 and CD28 may share some functions, but differ in others.
The overlap between the CTLA-4 internal localization motif and the binding site for p85 PI3K suggests the possibility of PI3K involvement in the intracellular localization of CTLA-4. Recent evidence has implicated PI3K in regulation of cellular protein trafficking. A yeast PI3K homologue functions in protein sorting of vacuolar hydrolases (38,39). Other studies have suggested a role for PI3K in internalization of platelet-derived growth factor receptor following ligand binding (40). If PI3K binding to the cytoplasmic domain of CTLA-4 requires phosphorylation of Tyr 165 (31), it seems likely that involvement of PI3K in intracellular localization would also require phosphorylation of this residue. The disruption of intracellular localization of CTLA-4 by mutation of Tyr 165 to phenylalanine would be consistent with such a requirement. However, other roles for Tyr 165 independent of its phosphorylation are also possible. Conclusive demonstration of a role for PI3K in intracellular localization of CTLA-4 will require direct biochemical demonstration of interaction between these proteins.
The intracellular localization of CTLA-4 was unexpected, and the reason(s) for it are currently unclear. Several possible explanations can be envisioned. Intracellular CTLA-4 could function in an autocrine fashion. Activated T cells express B7-1 (41)(42)(43), so it is possible that intracellular interaction between B7-1 and CTLA-4 can occur in activated T cells and that CTLA-4 in fact functions inside the cell. Alternatively, intracellular retention of CTLA-4 may delay surface transport until it is assembled into a multisubunit complex. Parallels of this scenario can be found in the expression of CD3 ⑀, major histocompatability complex class I and class II molecules, where newly synthesized molecules are retained in the endoplasmic reticulum by the chaperonin calnexin (44 -47). Possibly, such a complex might function at a particular stage of T cell development and/or activation and the cells we tested may lack other components of such a complex.
Finally, intracellular localization of CTLA-4 may indicate a mechanism to achieve subtle control of its expression on the T cell surface. Since CTLA-4 has such high avidity for B7 molecules, its ability to bind B7 molecules is disproportionate to its levels of expression (9). It may therefore be necessary to tightly regulate and focus the small number of CTLA-4 molecules on the cell surface following T cell activation. The pattern of intracellular localization of CTLA-4 overlaps that of the Golgi complex, i.e. it is polarized to one side of the cell. Following cell-cell contact, the Golgi apparatus is typically reoriented such that it faces the site of cell contact (48). This results in the directional release of secretory proteins such as cytokines pro- teins toward sites of cell-cell contact (49,50). Since secretory and membrane proteins share common pathway(s) of protein export, directional release of membrane proteins may also occur. Polarization of the Golgi complex and/or intracellular CTLA-4 during activated T cell-APC interactions could lead to preferential organization of intracellular CTLA-4 facing toward sites of APC contact. This could localize CTLA-4 at sites of APC contact and regulate the ability of CTLA-4 to function during an immune response (14). The existence of such a mechanism to control surface expression of CTLA-4 would further argue for the importance of its role during T cell activation.