A role for interleukin-12 in the regulation of T cell plasma membrane compartmentation.

The immunological synapse initiates the clustering and stabilization of the T cell receptor by the formation of a large lipid microdomain that accumulates (e.g. CD4/CD8) and segregates (e.g. CD45 and LFA-1) some proteins of the T cell plasma membrane. This work shows that a fraction of transmembrane glycoproteins CD26 and CD45 (the R0 isoform in particular) is present in the rafts of fresh and activated human T lymphocytes. CD26 is proposed as the costimulator of TCR-dependent activation, and CD45 is essential to the T cell activation process because it dephosphorylates at least the inhibitory site of Src kinases. These findings support a more complex model of compartmentation, depending on the stage of T cell maturation and post-transcriptional and post-translational regulation. In addition, interleukin 12 (IL-12; inducer of TH1 responses) drives CD26 and CD45R0 to particular microdomains, thereby involving interleukins in the rules governing raft inclusion or exclusion. The physical association of CD26 and CD45R0 has long been reported. The results presented in this work fit a model in which IL-12 up-regulates a certain type of CD26 expression that interacts on the cell surface with CD45R0, near but outside of the raft core. The use of antisense oligonucleotides for the CD26 mRNAs demonstrated that both events (enhanced by IL-12), CD26-CD45R0 association and membrane compartment redistribution, are related. Thus, CD26 could be part of a shuttling mechanism for CD45 that regulates membrane tyrosine-phosphatase activities, e.g. to control IL-12 receptor-dependent signal transduction.

In the early events of T cell activation, antigen (Ag) 1 presentation results in the clustering of protein-tyrosine kinases, which associate with the CD3 and TCR subunits and the coreceptors CD4 or CD8 (1). The transmembrane tyrosine-phosphatase CD45 is essential in this process because it dephosphorylates at least the inhibitory site of Src family kinases, responsible for the phosphorylation of ITAMs (immunoreceptor tyrosine-based activation motifs) (2). This extremely active phosphatase does not require ligand binding for optimum catalytic activity; later in the process, CD45-dependent dephosphorylation of key substrates, Src, or other protein-tyrosine kinases (e.g. ZAP-70/Syk) and ITAMs must be avoided.
An advance in the understanding of CD45 function was the discovery of specialized membrane domains, called rafts, ganglioside-enriched membranes (GEMs), or detergent-resistant membranes (DRMs). These membranes contain a high density of sphingolipids and cholesterol (2)(3)(4) and serve as attachment sites for a variety of lipid-modified proteins (including GPI) and also integral membrane and cytoplasmic proteins (e.g. Src family kinases Lck and Fyn, CD4, CD8, and LAT (linker for activation of T cells)). A compartmentation model has been proposed in which the immunological synapse initiates the clustering and stabilization of the TCR by the formation of a large lipid microdomain that accumulates (e.g. CD4 and CD8) and segregates (e.g. CD45 and LFA-1) several membrane proteins (1)(2)(3)(4)(5)(6). As the phosphorylation decreases within minutes after the initial response, other phosphatases (not CD45) should be recruited to these rafts later (5).
However, the role of the extracellular domain of CD45 remains elusive despite its high M r and structure, which strongly suggest ligand-receptor interactions (6). The diversity of the structures and sizes of different CD45 isoforms is cell type-dependent and developmentally regulated. Upon T cell activation, naive T cells switch from isoforms containing A, B, or C epitopes, with post-translational (O-and N-linked glycosylation) information, to the lowest M r isoform, CD45R0, which lacks sequences coded by 4/A, 5/B, or 6/C exons (6 -8). Several experiments reported distinct CD45 interactions on naive and memory/effector (CD45R0 ϩ ) cells. CD45 Ϫ T cell lines transfected with cDNAs of different CD45 isoforms or cells from transgenic and knock-out mice had differential responses to Ag. CD45 has also been reported to associate with several surface molecules such as Thy-1, TCR, CD2, CD3, CD4, CD7, CD8, CD26, CD28, LFA-1, B cell receptor, lymphocyte phosphatase-associated protein, endoplasmic reticulum protein glucosidase II, and CD45 itself (6 -9). As the different CD45 isoforms have similar PTP activities, these data suggest that they may interact differentially with other surface molecules that alter PTP accessibility to substrates, modifying in this way the signals received not only through Ag receptors but also through cytokine receptors and integrin-mediated adhesion, to either augment or inhibit T cell activation (9 -12).
The surface CD26 glycoprotein is identical to dipeptidyl-peptidase IV (EC 3.4.14.5). T cells expressing high levels of CD26 constitute a subpopulation of CD45R0 ϩ cells with helper activities in B cell Ig synthesis, proliferative responses to soluble Ags and allogeneic cells, secretion of T H 1-type cytokines, and transendothelial migration capacity (13,14). This study describes the distribution of CD26 and CD45R0 molecules in plasma membrane microdomains of fresh and activated human T cells; it also shows that IL-12 (an inducing T H 1 response cytokine) dramatically changes CD45R0 membrane compartmentation through a CD26-CD45R0 association. The significance of this finding is discussed.
Cell Isolation and Culture-Buffy coats were kindly provided by the Centro de Transfusión de Galicia (Santiago, Spain). Blood was donated by healthy volunteers and human peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll Paque PLUS (Amersham Biosciences) density gradient centrifugation as described elsewhere (20). Cells were cultured (10 6 PBMCs/ml) in RPMI 1640 (Sigma) supplemented with 10% inactivated fetal calf serum (Invitrogen), 100 g/ml streptomycin, and 100 IU/ml penicillin (Sigma) in a humidified atmosphere of 5% CO 2 at 37°C. PBMCs were activated with 1-1.5 g/ml P. vulgaris lectin in the presence or absence of cytokines for the indicated time. The cell lines used were cultured under the same conditions except that Geneticin (100 g/ml, Sigma) was added to the medium of CD26-transfected cell lines.
Inhibition of CD26 Expression with Antisense Morpholino Oligos-Antisense morpholino oligos bound to partially complementary DNA and ethoxylated polyethileneimine (EPEI) were provided by Gene Tools as a Special Delivery Protocol kit. Morpholino subunits are assembled by phosphorodiamidate linkages to obtain an oligo with a modified non-ionic nuclease-resistant backbone. The CD26 antisense sequence, 5Ј-GAACCTTCCACGGTGTCTTCATCGT-3Ј, was designed to bind to the target sequence in the region containing AUG of the post-spliced mRNA for the highest effectivity of translation blocking. The fluoresceinated 5Ј-CCTCTTACCTCAGTTACAATTTATA-3Ј morpholino oligo was used as a standard control for nonspecific effects. EPEI electrostatically binds the anionic morpholino-DNA duplex, generating a cationic complex that approaches the anionic cell surface, leading to endocytosis. Subsequent acidification within the endosome increases EPEI ionization driving to membrane permeabilization and oligo release into the cytosol.
The ratio of EPEI/morpholino oligos/PBMCs was tested for optimal internalization. Briefly, Morpholino-DNA duplex and EPEI (1.6 nM and 0.64 nM, final concentrations) were mixed and preincubated for 20 min in MilliQ (Millipore) water at room temperature, the mixture was added to the cells (4 ϫ 10 6 ) in serum-free RPMI medium, and the samples were incubated for 3 h at 37°C in a humidified atmosphere of 5% CO 2 in air. Cells were then washed with RPMI and cultured as described above.
Immunostaining and Immunofluorescence-Cell surface Ag expression was measured by direct or indirect immunofluorescence as described (21). CD45RA, CD45RB, CD45RC, and sometimes TP1/16 mAbs were used as hybridoma supernatant and revealed with FITC-or PElabeled GAM Abs. The other Abs were applied as primary antibodies. Viable lymphocytes were identified according to their forward and right angle scattering. The percentage of cells positive for the Ag was evaluated by setting negative controls as the omission of primary antibody or the inclusion of isotype controls. Direct immunofluorescent protocol was used for two-color experiments (21). For studies of detergent resistance of proteins associated with rafts, the work of Janes et al. (22) was adapted to flow cytometry. Briefly, cells were treated with 1% Triton X-100 for 5 min on ice, with 10 mM methyl-␤-cyclodextrin (M␤CD; Sigma), which depletes cellular cholesterol, for 15 min at 37°C, or with M␤CD followed by Triton X-100 extraction, before fixation (3% PFA in PBS for 30 min at room temperature) and staining as described above. Samples were processed on a BD Biosciences FACScalibur cytometer, except where indicated. WinMDI software (a kind gift of J. Trotter, Scripps Institute, La Jolla, CA) was used to analyze data.
Protein Concentration Determination-The protein concentration of samples was determined using the Bradford procedure (Sigma). Bovine serum albumin was used as a standard.
Membrane Phosphatase and Peptidase in Vitro Assays-Cells were washed twice in RPMI 1640 and resuspended at 3 ϫ 10 6 cells/ml in hypotonic lysis buffer (HLB) (25 mM Tris-HCl, pH 7.5, 25 mM sucrose, 0.1 mM EDTA, 5 mM MgCl 2 , 5 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin) before sonication. Individual samples were made up in duplicate in the presence and absence of 1 mM Na 3 VO 4 . The membranes were sedimented from post-nuclear supernatants at 100,000 ϫ g for 60 min at 4°C, the resulting pellet was resuspended in 200 l HLB by sonication, and the protein concentration was determined. Plasma membrane (PM) protein (20 g) was incubated as described by Cayota et al. (23) in a reaction mixture of 5 mM p-nitrophenyl phosphate (Sigma 104 ® substrate), 80 mM MES, pH 5.5, 10 mM EDTA, and 10 mM dithiothreitol at 37°C. After a 20-min incubation, reaction was stopped by the addition of 1 ml of 0.2 N NaOH. PTP activity was expressed as absorbance units at 405 nm after removal of values of the samples duplicated in the presence of vanadate. Dipeptidyl-peptidase IV activity was measured as described previously (24) and expressed as absorbance units at 405 nm. Briefly, PM protein (20 -40 g) was incubated for 1 h at 37°C in the presence of 0.5 mg/ml Gly-Pro p-nitroanilide tosylate (Sigma).
Isolation of GPI-enriched Membranes by Equilibrium Density Gradient Centrifugation-The following steps were carried out at 4°C unless indicated otherwise, basically following the works of Ilangumaran et al. (25,26). Cells (50 ϫ 10 6 ) were washed twice in PBS and once in TKM buffer (50 mM Tris-HCl, pH 7.5, 25 mM KCl, 5 mM MgCl 2 , and 1 mM EDTA). Detergent lysates were prepared in TKM containing 0.5% Triton X-100 and the protease inhibitors Pefabloc SC (2 mM), leupeptin (10 g/ml), and aprotinin (5 g/ml) for 20 min on ice. For equilibrium gradient centrifugation, cell extracts were adjusted to 40% sucrose, loaded into SW55Ti tubes (Beckman L8-M), overlaid with 2.7 ml of 36% sucrose, and finally completed with 1.575 ml of 5% sucrose in TKM buffer. After centrifugation at 200,000 ϫ g for 18 h, 450-l fractions were collected and stored at Ϫ20°C. Fraction proteins were evaluated by Western blotting or immunoprecipitation or were detected by dot immunoassay. Serial dilutions in PBS (200 l) of detergent-lysates or sucrose density fractions were applied to the wells and dotted onto nitrocellulose filter sheets using a Bio-Rad dot-blot apparatus (25).
For immunoprecipitation studies, cell surface proteins were sometimes biotinylated following the manufacturer's instructions (Pierce). Briefly, 25 ϫ 10 6 lymphocytes/ml were resuspended in PBS, pH 8.0, containing 0.5 mg/ml sulfo-NHS-biotin for 30 min at room temperature. When required, the depletion of microdomain-linked proteins was carried out according to Cheng et al. (28). Briefly, washed cells (50 ϫ 10 6 ) were resuspended in 1 ml of ice-cold Hanks' balanced salt solution with 1 g/ml HRP-conjugated cholera toxin B subunit (which binds membrane GM1; Sigma) for 30 min at 37°C. Then cells were washed twice and resuspended in 1 ml of 0.5 mg/ml 3-3Ј-diaminobenzidine in Hanks' balanced salt solution with or without H 2 O 2 for 45 min at 4°C. After lysis, polymerized raft proteins were removed as described above.
Immunoprecipitation was performed with anti-CD26 TP1/16 (in contrast to 1F7, anti-CD26 TP1/16 was more effective with Triton than digitonin lysates), anti-CD45 D3/9, or anti-CD45R0 mAbs, previously coupled for 1 h at 4°C to anti-mouse-agarose beads (Sigma; 50 l of 1:1 lysis buffer). Occasionally, precleared lysates were incubated first with antibodies. Precipitated protein was recovered by centrifugation, washed three times in lysis buffer, eluted by boiling in SDS-PAGE sample buffer, and analyzed as described above, except for the filters with biotinylated proteins, which were incubated sequentially with streptavidin, biotinylated alkaline phosphatase, and CDP-Star substrate (New England Biolabs). Developed filters were exposed on X-Omat S film (Eastman Kodak Co.) several times depending on the intensity of the signal. When needed, the blots were stripped as described elsewhere (29).
Dot blots were treated as described above, except when measuring alkaline phosphatase (AP) levels. AP activity was developed directly with bromochloroindolyl phosphate/nitro blue tetrazolium substrate (BCIP/NBT, BioRad). The spots were quantitated by scanning the filters and densitometry (ImageMaster 1D, Amersham Biosciences).
Cell Proliferation and Calcium Assays-Three-day PHA-blasts were cultured in 96-well culture plates at 1 ϫ 10 5 cells/well in a total volume of 100 l of complete medium with/without 20% conditioned medium (medium in which PBMCs were activated previously for 3 days). In the preincubation assays, cells were cultured with the anti-CD45R0 and CD45RA Abs for 30 min before adding the cytokines, whereas in the postincubation assays, the Abs were added 30 min after the cytokines. Controls were cells incubated in the absence of cytokines. After 2 days of culture, 20 l/well CellTiter 96 AQ ueous One Solution Reagent (Promega, Madison, WI) was added to the plate 4 h before the absorbance was recorded at 492 nm in a Labsystems Multiskan MS plate reader. All cultures were performed in triplicate.
anti-CD3 mAb OKT3 (t ϭ 80 s) plus GAM HϩL mAb (t ϭ 280 s). Analyses were performed by flow cytometry and MFI 3.4J2 software, a kind gift of Eric Martz (Scripps Institute), for mean fluorescence intensity data.

IL-12 Enhances PM CD26-CD45R0 Interaction-Previously
we reported a strong IL-12-dependent surface CD26 up-regulation on activated human T cells, including the effector/memory CD45R0 subset (25,30). A weaker staining of these IL-12 blasts with anti-CD45R0 UCHL-1 mAb, previously observed (20), is shown on the same cells (Fig. 1A). Explanations such as loss of sialylation, isoform switching, or CD45 internalization in the presence of IL-12 were ruled out (data not shown). Down-regulation was discarded, because IL-12 enhanced the levels of PM PTP enzymatic activity, which is ϳ90% CD45specific (31) (Fig. 1B).
Because CD26 is known to associate with CD45 (6 -9), we performed immunoprecipitation studies. Fig. 1C shows that, in the presence of IL-12 together with the expected CD26 up-regulation, anti-CD26 TP1/16 mAb also coprecipitates more CD45 (the two low M r isoforms). In consonance with the above result, neither CD45 nor R0 Ags (which were surface biotinylated) were down-regulated by IL-12. The nature of the coprecipitated bands is confirmed later on in this article.
An attractive model to explain these results is that IL-12 is up-regulating a certain type of CD26 expression (already existing; see Fig. 1C) that induces interaction with CD45R0 and/or other molecules and masks the anti-R0 Ab epitope. Thus, as observed in Fig. 1 2. Detection of membrane CD26, CD45, CD4, and AP in sucrose density gradient fractions. Costimulation with IL-12 excludes CD45R0 from rafts. Protein from 50 ϫ 10 6 PBMCs, PHA (basically CD3 ϩ cells), or PHAϩIL-12 lymphoblasts was extracted in TKM buffer containing 1% TX-100, adjusted to 40% sucrose, and subjected to equilibrium density gradient centrifugation in a SW55 Ti rotor. After overnight centrifugation at 200,000 ϫ g, 11 (except 10 fractions in A) 0.45-ml fractions were collected. 2-Fold serial dilutions (starting at a volume of 100 l) of the fractions were dotted and later probed for the indicated Ag (except for AP). To ensure that the signal strength for any dilution throughout the gradient was within the saturation limit, several 2-30-min exposures of the membrane were performed using ECL detection; the 15-min exposures are shown. For AP, the dotted material was developed directly with bromochloroindolyl phosphate/nitro blue tetrazolium substrate, and the reaction was stopped with MilliQ water. The data are representative of three (AP, CD4) or five (CD26, CD45R0) separate experiments. Note that CD26, total CD45, and CD45R0 are present in soluble HDF (fractions 9 -10 or 10 -11) but also in insoluble LDF (fractions 4 -6 and 3-5) (C-E) and that CD26 is present in fraction 3 (in contrast to AP and CD4). Controls from the same gradients show the GPI-anchored AP (present only in LDF) and CD4 (with the well known bipolar LDF and HDF distributions) (A and B). Note the up-regulation of CD26 and CD45 in the PHA-activated cells and the differential redistribution of CD26 to LDF after this activation; under the influence of IL-12, CD26 redistributes to central fractions and CD45R0 migrates spectacularly to HDF. The weaker CD4 staining in IL-12-treated cells can be explained by the IL-12-primed CD8 subset proliferation (19,65), but AP staining was also weaker in these fractions (34 Ϯ 8%, n ϭ 3). F, relative distribution of proteins. Films were scanned and subjected to densitometry. The signal strength (arbitrary values) of each dot is expressed as a percentage of the total for the respective Ag. Means from the experiments performed are shown, S.D. being irrelevant.
lines (25)(26) and porcine lung (32). We considered a possible relationship between the CD26-CD45R0 association and GEMs/PM rafts. The well known GPI-anchored (32) AP was used as a control for microdomain purification by equilibrium density gradient centrifugation. AP was detected in the light density fractions (LDF) (Fig. 2A, lanes 4 -6). Transferrin R (CD71) was used as control for soluble, non-raft protein (data not shown).
The dual distribution of CD26, with ϳ28% of the material detected in the LDF and 51% in the heavy density fractions (HDF) (Fig. 2A, lanes 10 and 11) of the sucrose gradient, is shown in human T cells (because monocytes and B and NK cells from PBMCs are essentially CD26 Ϫ ) (Fig. 2C). In activated T cells, in addition to a higher intensity, CD26 was redistributed to the GEMs (37% in LDF and 30% in HDF, percentages that are in agreement with data from mouse T cell line (25), indi- , and CD71 (always absent from GEMs). GAM-FITC or GAM-PE was used as negative control (gray-filled lines). In internal density gradient control, TX-100 treatment extracted approximately two-thirds of the GPI-linked AP (data not shown). As expected, CD3, CD4, HLA-DR, and CD71 staining was deleted, but a significant percentage of CD8, CD26, and CD45R0 was resistant, supporting a strong GEM anchorage. B, PHA-blasts cultured with (right) or without (left) IL-12 (2 ng/ml) were either fixed directly or treated previously with TX-100 as described in A, with 10 mM M␤CD for 15 min at 37°C, or with a combination of both. The cells were then labeled for CD26 and CD45R0 and analyzed by two-color immunofluorescence on a log scale with identical settings to allow comparison of staining. Numbers in the upper right quadrants represent mean immunofluorescence data for anti-CD45R0-PE UCHL-1 mAb staining. Note that in IL-12 blasts, only TX-100 treatment had an effect on R0 extraction. IL-12-dependent CD26-CD45R0 interaction was lost only after M␤CD and TX-100 treatment.
cating the activated status of T cells). IL-12 enhanced CD26 intensity and enriched the intermediate fractions (33 and 26%, respectively in LDF and HDF, n ϭ 4; Fig. 2, C and F). Note that there is more CD26 than the well described CD4 glycoprotein (which shows a bipolar pattern) in the rafts. In fact, CD26 is present in almost all sucrose gradient fractions, including fraction 3 (AP and CD4 are not detected in this fraction) (Fig. 2B).
As described previously in a murine T cell lymphoma (25) but not in Jurkat or B cells (22,28,33,34), CD45 also associates with GEMs in human lymphocytes (Fig. 2D). The ratio of LDF/ HDF was higher for total CD45 and R0 than for CD26. The differences in intensity between PBMC and PHA-activated lymphocyte fractions again reflect total and CD45R0 up-regulation. In comparison with CD26, PHA hardly affects CD45 redistribution (34% in LDF and 37% in HDF, n ϭ 4; 37 and 32%, respectively, for CD45R0). In contrast to CD26, where the ratio LDF/HDF was constant, the presence of IL-12 in T cell activation reduced the percentages of CD45 in the rafts (or in fractions near the rafts). This redistribution can be attributed to the R0 isoform (Fig. 2E) (23% in LDF and a spectacular 51% in HDF, n ϭ 5).
CD26-CD45R0 Association Is Near but Outside of Rafts-Do CD26 and CD45R0 interact in a particular PM microdomain? A new approach was used, based on the resistance of rafts to solubilization by nonionic detergents and flow cytometry. As stated previously, a fraction of CD4, CD8, and CD26 molecules is present, and CD3, CD71, CD45, and HLA-DR are absent or weakly associated with lipid rafts (3, 5) (TCR cross-linking stabilizes CD3 against detergent) (22). After treating effector T cells with TX-100, only CD8, CD26, and interestingly CD45R0 staining was partially maintained (Fig. 3A). CD4 staining was lost, suggesting that our treatment was harsh enough to deplete CD4 from lipid rafts. The different solubility of CD8 may be explained by its preferential interaction with raft-associated LAT protein (35). Additional controls for AP activity in the extracted fraction and the remaining PM proteins showed very restrictive conditions for this approach (data not shown). As observed in Fig. 3B, the loss of CD45R0 staining in IL-12 cultures was maintained after Triton X-100 treatment. Cells were also treated with M␤CD, which disrupts rafts by depleting cellular cholesterol (22, 34 -36). The treatment alone did not avoid the loss of CD45R0 staining. Even after M␤CD plus TX-100 treatment, CD26 and CD45R0 were detected, suggesting that a fraction of both molecules is tightly attached to the lipid microdomain core. Note that in IL-12 blasts, the most R0 was extracted with TX-100 alone, i.e. it is present in the soluble fraction. These results support those reported in the previous section. In addition, IL-12-dependent loss of CD45 staining occurred only under M␤CD plus Triton X-100 treatment. It is interesting to note that CD26 ϩ cells present weaker CD45R0 staining than CD26 cells even in the absence of IL-12 (Fig. 3B, control PHA), which could represent the CD26/CD45R0 association observed by immunoprecipitation.
The specificity of these results was confirmed by coprecipitation assays following two different experimental approaches. In the first one (Fig. 4A), HDF and LDF pooled samples from PHA or PHAϩIL-12-stimulated blasts containing the same amount of protein were incubated with TP1/16 mAb, and immunoprecipitates were revealed with UCHL-1 mAb. Although coprecipitation could not be detected in pooled insoluble fractions from the density gradient (possibly because of sample manipulation), IL-12 enhanced CD45R0 coprecipitation. In the second approach, the selective elimination of raft-associated proteins by the HRP-3-3Ј-diaminobenzidine reaction (28) did not affect coprecipitation between CD26 and CD45R0 Ags in PHA-blasts and diminished CD45R0 coprecipitation (still higher) in IL-12-treated cells (Fig. 4B). The results altogether demonstrate that IL-12 induces a stronger CD26-CD45R0 association near (or on the outer edge of) PM raft regions. However, if loss of raft-associated CD45R0 after the HRP-3-3Јdiaminobenzidine reaction correlates with the R0 Ag remaining after TX-100 treatment (Fig. 3B, IL-12 dot plots), it is implicit that part of the striking IL-12-dependent CD45R0 microdomain redistribution described in the previous section could be explained by CD26 masking the UCHL-1 epitope (in rafts or next to rafts).
CD26 Is Involved in Pulling Out CD45R0 from Rafts-A related question is whether both events enhanced by IL-12, the association and the membrane compartment redistribution, are connected. With the addition of antisense oligos for the CD26 mRNAs to the above described culture conditions, which resulted in a drop of the cell surface CD26 expression, it was demonstrated that loss of R0 staining in IL-12-cultured cells is associated specifically (oligonucleotide controls did not show this effect; data not shown) with a de novo expression of certain CD26 protein (Fig. 5A, right). This expression can be observed even with PHA stimulation alone (Fig. 5A, left). Experiments with brefeldin A, which avoids surface protein expression, corroborated this finding (data not shown). FIG. 4. Physical association of CD26 and CD45R0 takes place next to rafts. A, PHA or PHAϩIL-12 cells were processed as described in the legend to Fig. 2. Normalized protein samples from both pooled F3-F6 TX-100-insoluble (Insol.) and F10 -F11 TX-100-soluble (Sol.) sucrose gradient fractions were subjected to immunoprecipitation with anti-CD26 mAb or negative isotype control IgG 1 mAb. Immunoprecipitated samples (IP) and a complete lysate from 10 6 PHA-blasts as positive control (Cϩ) were run on a 7.5% SDS-PAGE and transferred to polyvinylidene difluoride membrane, and the presence of CD45R0 Ag (180 kDa) was revealed with UCHL-1 mAb and ECL. CD45R0 was detected only in soluble fractions from PHAϩIL-12 blasts. B, PHA and PHAϩIL-12 lymphoblasts (25 ϫ 10 6 ) were incubated with HRP-conjugated cholera toxin B subunit (C.T.; 1 g/ml, 30 min, 37°C) and 3-3Јdiaminobenzidine (D.A.B.) with (ϩ ϩ) or without H 2 O 2 (ϩ Ϫ) as indicated prior to lysis in 0.5% TX-100 TKM buffer. After elimination of polymerized proteins and immunoprecipitation with anti-CD26 TP1/16 (or IgG 1 isotype control (CϪ lane)) mAb, CD45R0 detection in polyvinylidene difluoride membranes was performed with UCHL-1 mAb, GAM-HRP, and ECL. It can be seen again that anti-CD26 coprecipitates more CD45R0 from IL-12 blasts than from PHA-blasts. Moreover, elimination of raft proteins affected R0 coprecipitation in these cells but not in PHA-blasts.
Inhibition of CD26 biosynthesis in these cells scarcely affected CD26 distribution or R0 detection (Fig. 5B); however, IL-12-dependent redistribution of CD45R0 from GEMs to soluble membrane is avoided in cells treated with antisense, but not control, oligonucleotides. in consonance with flow cytometry results, the presence of more R0 in rafts is observed even in PHA blasts.
Is the IL-12-dependent CD26-CD45R0 Association Functionally Relevant?-To verify whether incubations with soluble UCHL-1 Ab could block CD26-CD45R0 association, proliferation assays of 3-day PHA-blasts cultured for 2 additional days with UCHL-1 Ab or RP1/11 (anti-CD45RA) (data not shown), added to the cultures 30 min before or after the addition of cytokines, were carried out. A small Ab concentration was able to induce proliferation. In IL-12 cultures, the enhancing effect of UCHL-1 Ab was lost at the lowest doses of preincubation (Fig. 6A). The behavior was totally different when the Ab was added after the IL-12. If UCHL-1 effectively blocked the CD26-CD45R0 association, at least that induced by IL-12, these results suggest that the interaction regulates T cell proliferation. However, the results of postincubation imply a process of fast kinetics.
Another approach was to stimulate 5-day PHA or PHAϩIL-12 lymphoblasts (the latter with less CD45R0 in rafts) through the TCR and measure changes in [Ca 2ϩ ]. A reduced calcium response was detected in PHAϩIL-12 blasts (Fig. 6B), although an identical percentage (64%) of responding cells in both blast types was observed. These results were reinforced by those from antiphosphotyrosine-Western blotting (data not shown). DISCUSSION This study reveals that a fraction of the transmembrane proteins CD26 and CD45 is associated to GEMs in fresh and activated human lymphocytes. CD26, unlike CD4, which shows a bipolar pattern, is present in almost all sucrose gradient fractions. Effector T cells have much more raft CD26 than fresh lymphocytes. CD26 has been proposed as a costimulatory molecule of TCR-dependent T cell activation (13,14,(37)(38)(39). Independent of its enzymatic activity (38), which is important in the T cell response through modulation of the activity of several biological factors (chemokines, etc.) (13, 14, 40 -42), co-crosslinking of anti-CD26 and CD3 mAbs enhances phosphorylation of CD3 Tyr residues and increases CD4-associated p56 lck tyrosine kinase activity (16,27). This effect could now be ascribed to the aggregation of lipid rafts that facilitates colocalization of kinases and TCR, thereby triggering phosphorylation, as observed with the GPI-associated proteins (3-5, 22, 26, 43-45).
However, a role in this process for a CD26-CD45 association (37)(38)(39)44) cannot be discarded, because a percentage of the PTP CD45 is associated to GEMs in human fresh T cells (B lymphocytes from PBMCs constitutively lack CD45 in GEMs (28)) and PHA-blasts. This result is not unexpected because CD45 is associated with the GPI-protein Thy-1 (CD90) and with raft-associated TCR in cell lines, probably resembling an effector state (38); nevertheless, models of membrane compartmentation after B-or T-response to antigen have been proposed showing that CD45 may be excluded from the signaling complex raft (1-6, 33). It is interesting to note for later discussion FIG. 5. Relationship of CD26-CD45R0 association with GEMs. A, flow cytometry analyses show that IL-12dependent loss of CD45R0 staining is dependent on cell surface CD26 de novo expression. PBMCs (10 6 cells/ml) treated with CD26 antisense (thick lines) or control oligos (not shown) and mock-treated (thin lines) were stimulated with PHA (1 g/ml) in the absence (left) or presence (right) of recombinant IL-12 (2 ng/ml). On day 5, cultured cells were stained with anti-CD26, anti-CD45R0, or anti-CD3 (not shown) mAbs-FITC. Anti-CD3 staining was never affected. B, dot blots from sucrose gradients of the same cells show that whereas treatment with CD26 antisense oligos hardly affected CD26 membrane distribution, more CD45R0 Ag was observed in LDF even in PHA-blasts. Dots from one experiment representative of several were treated as described in the legend to Fig. 2. that these experiments have used B or Jurkat cells and both do not express CD26.
This CD45, present in rafts of effector/memory cells in proportions similar to CD26, can be ascribed to the R0 isoform. From our data on cells treated with Triton X-100 and M␤CD, it is easily deduced that a fraction of CD26 and CD45R0 is more resistant to the treatment than is CD4 or CD8, suggesting that this fraction is inside of the raft core. A recent report confirms that CD26 and CD45R0 can be found in membrane rafts of CD26-transfected Jurkat cells under certain conditions (44).
These data do not invalidate the model of compartmentation, which implies increased stability of phosphorylation (3,5,33), when considering naive T cells (they do not express the CD45R0 isoform). However, our findings appear to play a role in the far more efficient in responding to stimulation effector/ memory CD45R0 ϩ cells. In these T cells, in which the rafts are bigger than in naive cells where CD45 can be exposed at the edges of smaller rafts (5,46), low M r CD45 PTP activity may be necessary to maintain the activation of Src family kinases in rafts and to prime them to triggered TCR. Early observations of cocapping (46 -48) have shown that CD4, low M r CD45, and the CD3-TCR complex behave as independent entities on the surface of naive T cells, whereas they are stably associated on memory T cells and T H 1 clones. These data, taken together, fit into the idea that low M r isoforms of CD45 could be associated with rafts, whereas the large CD45 isoforms would be excluded (6), and support a more complex model of compartmentation depending on the stage of T cell maturation (49).
The present work shows that IL-12 drives CD26 and CD45R0 to particular PM regions. It has been reported that IL-2 pulls out the ␣ chain from rafts to bind the ␤ and ␥ chain of the IL-2 receptor complex in soluble fractions (50). Hence, interleukins are involved in the rules governing the inclusion and exclusion of proteins into rafts. The finding that IL-12 dramatically changes the distribution of CD45R0 from rafts to soluble fraction may explain the impaired Ca 2ϩ responses observed upon activation via TCR in IL-12-cultured cells. R0 redistributes from rafts perhaps to control IL-12 receptor-dependent signal transduction through suppression of the JAK-STAT (Janus kinase/signal transducers and activators of transcription) pathway as described for other cytokine receptors (12,51).
Physical association between CD26 and CD45 (particularly the R0 form) has long been reported (27). The results presented here show that in cells triggered by IL-12, even more CD26 and CD45R0 proteins are associated. Although some interleukins can regulate CD45 isoforms (52,53), the close CD26-CD45R0 association was seen only with IL-12. Taking these results together with those of in vivo elimination of rafts and CD26 expression inhibition, it can be concluded that the new surface CD26 molecules are directed to the soluble fraction near the lipid raft core and anchor CD45R0. As IL-12-dependent R0 membrane compartment redistribution is avoided in the absence of de novo CD26 expression, CD26 may be part of a shuttling mechanism for CD45. It has been reported recently that raft CD45 exchanges with the larger pool of freely diffusing CD45 after the formation of the immunological synapse (54).
Only a fraction of CD26 molecules associate with CD45R0, and IL-12 did not change the percentage (obtained from coprecipitation experiments). This finding suggests that the extracellular domains of both proteins control the interaction, although a direct binding of CD26 to the cytosolic domain of CD45 in an internalization model cannot be excluded (44). The fact that the extracellular domain of low M r CD45 controls the association with the CD4-TCR complex (46 -48) and the differential homodimerization of isoforms (a way of negative regulation of PTP CD45) (9) supports our results. In this sense, one should be aware that recombinant proteins can mask certain post-translational modifications necessary for the binding and modulation of protein function. CD45R0 itself is an example (55).
PTP activity is also important in the regulation of adhesiontriggered tyrosine kinase cascades (note that several of these studies were done with CD26 Ϫ Jurkat cells (3, 6, 10, 56), and FIG. 6. IL-12-dependent CD26-CD45R0 association may be functionally relevant. A, proliferation assays in the presence or absence of interleukins and soluble anti-CD45R0 UCHL-1 Ab. Three-day blasts were cultured at 10 5 cells/well in 20% conditioned medium (see "Experimental Procedures") for 2 additional days with or without IL-12 (2 ng/ml) or IL-2 (50 units/ml). In the preincubation assays (solid lines), cells were incubated with anti-CD45R0 Ab for 30 min before the addition of cytokines, whereas in the postincubation assays (dotted lines), UCHL-1 was added 30 min after the cytokines. A 1:1 dilution represents 5 g/ml. Data are recorded from absorbances at 492 nm after an additional 4-h incubation with CellTiter 96 AQ ueous One Solution Reagent, and the values are relativized to proliferation data without Ab. Points are the means of three experiments, with S.D. being irrelevant. No difference was found when soluble anti-CD26 Ab was used. B, anti-CD3-induced calcium flux. PHA and IL-12 blasts were washed and loaded for 30-45 min at 37°C with Fluo-3 AM (4 g/ml) in the presence of probenecidin and Pluronic F-127. Cells were stimulated with anti-CD3 mAb OKT3 (at t ϭ 80 s; left arrowhead) plus goat anti-mouse HϩL (at t ϭ 280 s; right arrowhead) and analyzed by flow cytometry. CD26 can be a functional adhesion receptor (14, 39, 56 -59), and thus microdomain redistribution could regulate their presence at integrin-dependent adhesion sites. One can speculate on the importance of this mechanism for normal T cell behavior and, further, on the possibility that it could be blocked by particles or molecules from pathogens. Both CD26 and CD45 molecules are involved in the regulation of lymphocyte death (2,17,60) and in the pathophysiology of AIDS (8,23,39,61). At least HIV-1 gp-120 (62), but perhaps also Tat (63,64) and Gag (34), modulate CD4 lateral interactions with them.