Caveolin-1 Triggers T-cell Activation via CD26 in Association with CARMA1*

CD26 is a widely distributed 110-kDa cell surface glycoprotein with an important role in T-cell costimulation. We demonstrated previously that CD26 binds to caveolin-1 in antigen-presenting cells, and following exogenous CD26 stimulation, Tollip and IRAK-1 disengage from caveolin-1 in antigen-presenting cells. IRAK-1 is then subsequently phosphorylated to up-regulate CD86 expression, resulting in subsequent T-cell proliferation. However, it is unclear whether caveolin-1 is a costimulatory ligand for CD26 in T-cells. Using soluble caveolin-1-Fc fusion protein, we now show that caveolin-1 is the costimulatory ligand for CD26, and that ligation of CD26 by caveolin-1 induces T-cell proliferation and NF-κB activation in a T-cell receptor/CD3-dependent manner. We also demonstrated that the cytoplasmic tail of CD26 interacts with CARMA1 in T-cells, resulting in signaling events that lead to NF-κB activation. Ligation of CD26 by caveolin-1 recruits a complex consisting of CD26, CARMA1, Bcl10, and IκB kinase to lipid rafts. Taken together, our findings provide novel insights into the regulation of T-cell costimulation via the CD26 molecule.

extracellular domain (1)(2)(3) and is capable of cleaving N-terminal dipeptides with either L-proline or L-alanine at the penultimate position (2). CD26 activity is dependent on cell type and the microenvironment, factors that can influence its multiple biological roles (reviewed in Refs. 4 -8). Although CD26 expression is enhanced following activation of resting T-cells, CD4 ϩ CD26 high T-cells respond maximally to recall antigens such as tetanus toxoid (9,10). Cross-linking of CD26 and CD3 with solid-phase immobilized monoclonal antibodies (mAbs) can induce T-cell costimulation and IL-2 production by CD26 ϩ T-cells (2,7,10). In addition, anti-CD26 antibody treatment of T-cells enhances tyrosine phosphorylation of signaling molecules such as CD3 and p56 lck (11,12). Moreover, DPPIV activity is required for CD26-mediated T-cell costimulation (13). CD26 may therefore have an important role in T-cell biology and overall immune function. However, the costimulatory ligand of CD26 has not yet been identified, and the proximal signaling events following CD26 engagement in T-cell remain to be determined.
In our previous study (14), we identified caveolin-1 in antigen-presenting cells (APC) as a binding protein for CD26, and we demonstrated that CD26 on activated memory T-cells directly faces caveolin-1 on tetanus toxoid-loaded monocytes in the contact area, which was revealed as the immunological synapse for T-cell-APC interaction. Moreover, we showed that residues 201-211 of CD26 along with the serine catalytic site at residue 630, which constitute a pocket structure of CD26/DP-PIV, contribute to binding to caveolin-1 scaffolding domain (14). More recently, we demonstrated that caveolin-1 binds to Tollip (Toll-interacting protein) and IRAK-1 (interleukin-1 receptor associated serine/threonine kinase 1) in the membrane of tetanus toxoid-loaded monocytes, and following exogenous CD26 stimulation, Tollip and IRAK-1 disengage from caveolin-1, with IRAK-1 being subsequently phosphorylated to up-regulate CD86 expression (15). It is conceivable that the interaction of CD26 with caveolin-1 on antigen-loaded monocytes results in CD86 up-regulation, therefore enhancing the subsequent interaction of CD86 and CD28 on T-cells to induce antigen-specific T-cell proliferation and activation. However, it is unclear whether caveolin-1 itself is the costimulatory ligand for T-cell CD26.
Recent studies have demonstrated that a newly identified membrane-associated guanylate kinase-like (MAGUK) molecule, CARMA1, is required for TCR/CD3-CD28 costimulation-induced NF-B activation and functions downstream of protein kinase C (PKC) (16 -18). CARMA1, which is predominantly expressed in thymus, spleen, and peripheral blood leukocytes, contains an N-terminal caspase-recruitment domain followed by a coiled-coil domain, a PDZ domain, an SH3 domain, and a guanylate kinase (GUK)-like domain (19,20). After TCR/CD3-CD28 costimulation or PMA-CD28 stimulation, CARMA1 is phosphorylated by PKC, followed by association with Bcl10 and MALT1, and recruitment of these complexes into lipid rafts (21)(22)(23). The recruitment of the CARMA1-Bcl10-MALT1 complex activates IB kinase (IKK) through a ubiquitin-dependent pathway, leading to activation of NF-B (24 -27). However, it remains to be determined whether CARMA1 is associated with lipid rafts directly or is recruited to lipid rafts via undetermined lipid raft-interacting proteins in the immunological synapse of T-cells.
In this study, using recombinant immunoglobulin-caveolin-1 fusion proteins, we identify caveolin-1 as the costimulatory ligand for CD26, and we demonstrate that the N-terminal domain of caveolin-1 induces T-cell proliferation and cytokine production via CD26 costimulation. Furthermore, we show that CARMA1 is bound to the cytoplasmic tail of dimeric CD26 on T-cells and that this interaction of CD26 and CARMA1 plays a pivotal role in CD26-mediated T-cell costimulation. Our data hence identify a previously unknown ligand for CD26 as a costimulatory molecule while elucidating the mechanisms involved in CD26-mediated T-cell activation and differentiation.
For expression of Fc fusion proteins, FreeStyle TM 293 expression system was used according to the manufacturer's instruction (Invitrogen). The Fc fusion proteins expressed in the culture supernatant were then purified by affinity chromatography on protein A-Sepharose (Bio-Rad) followed by size-exclusion purification on Microcon centrifugal filter devices (Millipore), and sterilized using inner diameter 0.22-m filter microcentrifugation tube Spin-X (Corning Glass).
Cells and Reagents-HEK293FT human embryonal kidney, Jurkat T-cell line (JKTwt), and Jurkat T-cells stably transfected with human CD26 (J.CD26wt) were grown as described previously (2,13,14). CARMA1-deficient Jurkat T-cell line, JPM50.6, was developed as described elsewhere (18). Human peripheral blood T-cells were purified from peripheral blood mononuclear cells using MACS Pan T-cell isolation kit II (Miltenyi), collected from healthy adult volunteers and incubated according to the methods described previously (30). Informed consent was obtained from healthy adult volunteers. Biotinylation of recombinant proteins or antibody was generated using EZ-Link TM Sulfo-NHS-LC-Biotin reagents according to the manufacturer's instruction (Pierce). Protease inhibitor mixture, phosphatase inhibitor mixture, and poly-L-lysine were from Sigma. Water-soluble digitonin was purchased from Wako Pure Chemicals Industries, Ltd.
Generating Stable Transfectants-The construct of V5tagged full-length human CD26 (pEF6/V5-CD26wt) was made by PCR, using the primers described in the Supplemental Material. The amplified products were cloned into the pEF6/V5-His B vector (Invitrogen) at the BamHI/EcoRI site. The CD26-CD10 chimeric receptor was composed of the N-terminal cytoplasmic region of human CD10 (1-23amino acid position) ligated to the transmembrane and extracellular regions of human CD26 (7-766-amino acid position), which were made by PCR. The construct of V5-tagged monomeric human CD26 (CD26H750E), which has histidine replacing glutamic acid as a point mutation at amino acid position 750, was made by site-directed mutagenesis method using pEF6/V5-CD26w as a template with the primers described in the Supplemental Material. After constructs were confirmed by DNA sequencing, plasmids were transfected to Jurkat T-cells using Nucleofector II device according to the manufacturer's instruction (Amaxa Biosystems). Two days after transfection of indicated plasmids, the cells were selected for blasticidin (1 g/ml) resist-ance for 4 weeks. Single clone cells expressing CD26wt (V5-CD26wt), CD26-CD10 (V5-CD26 ϩ CD10 cyto), and CD26H750E were then selected using standard limiting dilution method.
For stimulation experiments using the expression system, CHO-K1 cells were transfected with GFP-fused full-length caveolin-1 or SCD-deleted caveolin-1 expression plasmids, with the constructs being described previously (14), using Lipo-fectamine2000 reagent (Invitrogen). Two days after transfection of the indicated plasmids, the cells were selected for G418 (500 g/ml) resistance for 4 weeks. Single clone cells expressing GFP and caveolin-1, detected by anti-caveolin-1 pAb (N20) recognizing the N-terminal region of caveolin-1 using flow cytometry (FACSCalibur TM ), were then selected using standard limiting dilution method.
Flow Cytometric Analysis-For assessment of J.CD26wt that binds biotinylated NT-Fc or NT⌬SCD-Fc, 1 ϫ 10 6 cells were washed in ice-cold phosphate-buffered saline and incubated with Fc␥1 and mouse Ig isotypes (1 g/ml) to block nonspecific binding, followed by reaction with biotinylated NT-Fc or NT⌬SCD-Fc (1 g/ml), and subsequently stained with FITCconjugated streptavidin (1:500). For blocking experiments, unlabeled mouse IgG (20 g/ml) or unlabeled anti-CD26 mAb (20 g/ml) was incubated with cells prior to reaction with biotinylated NT-Fc or NT⌬SCD-Fc. Flow cytometric analysis of 10,000 viable cells was conducted on FACSCalibur TM . Each experiment was repeated at least three times, and the results were provided in the form of a histogram or dot plots of a representative experiment.
Small Interfering RNA (siRNA) against Human CARMA1-We selected two target sequences from nucleotides ϩ305 to ϩ325 (ss1) and ϩ792 to ϩ 802 (ss2) downstream of the start codon of human CARMA1 mRNA (sense1 siRNA (ss1-siRNA), 5Ј AAGAGCCCACUCGGAGAUUCUdTdT, and sense2 siRNA (ss2-siRNA), 5Ј AACUGGAGCGGGAGAAUGAAAd-TdT). Moreover, mis-siRNA at four nucleotides was prepared to examine nonspecific effects of siRNA duplexes (mis-siRNA, 5Ј UAGUGGCCACACGGUGATTCdTdT). These selected sequences also were submitted to a BLAST search against the human genome sequence to ensure that only one gene of the human genome was targeted. siRNAs were purchased from Qiagen. Transfection of siRNA into purified T-cells were conducted using HVJ-E vector (GenomeONE TM ; kindly provided by Ihsihara Sangyo Kaisha Ltd.) as described previously (14). After 48 h of transfection, cell were prepared for examination.
T-cell Proliferation and IL-2 Production Assay-For T-cell proliferation assay, 1 ϫ 10 5 purified T-cells were cultured in 96-well flat-bottomed plates (COSTAR) in a volume of 200 l of AIM-V medium (Invitrogen). For solid-phase stimulation, anti-CD3 (OKT3, 0.05 g/ml) and/or anti-CD26 mAb (5 g/ml), anti-CD28 mAb (4B10, 5 g/ml), or Fc fusion proteins (5 g/ml) were bound on the plates. For stimulation with caveolin-1-transfected CHO cells, purified T-cells were cultured in the presence of soluble anti-CD3 (OKT3, 0.05 g/ml) at 1 ϫ 10 5 cells/well with varying amounts (T-cells: CHO ϭ 800, 400, 200, 100, 50, 25:1 or no CHO cells as background control) of CHO cell transfectants. Before coculturing with T-cells, CHO transfectants were fixed with 0.05% glutaraldehyde for 30 s at room temperature, followed by washing three times with phosphate-buffered saline. T-cell proliferation was measured by [ 3 H]TdR (ICN Radiochemicals) uptake. Cells were incubated for 96 h and were pulsed with 1 Ci/well of [ 3 H]TdR, 16 h prior to harvesting onto a glass fiber filter (Wallac), and the incorporated radioactivity was quantified by a liquid scintillation counter (Wallac). For blocking experiments, cells were treated with soluble anti-CD26 mAb (1F7), anti-CD28 (4B10), or control mouse Ig (each at 20 g/ml) before being cultured in plates coated with stimulatory antibodies and/or Fc proteins.
For IL-2 production assay using Jurkat T-cell lines, JPM50.6, or their transfectants, 5 ϫ 10 5 cells/well in 200 l of culture media were incubated at 37°C in the presence of the indicated plate-bound antibodies and/or NT-Fc proteins. Cells were also stimulated with PMA (10 ng/ml) in anti-CD3-coated wells. After 48 h of incubation, culture supernatants were pooled from the triplicate wells and assayed for IL-2 content using Human IL-2 Biotrack Easy ELISA (Amersham Biosciences) according to the manufacturer's instruction.
Nuclear Protein Extraction and DNA-binding Protein Assay-Nuclear extracts were prepared from Jurkat cells or transfectants stimulated as indicated, and ELISA-based DNA-binding protein assays for NF-B p65 were performed using Mercury TransFactor kits (BD Biosciences) as described previously (14).
Statistics-Student's t test was used to determine whether the difference between control and sample was significant (p Ͻ 0.05 being significant).

RESULTS
We prepared soluble caveolin-1 protein consisting of the putative extracellular N-terminal region or the N-terminal region minus the SCD of human caveolin-1, fused with human IgG 1 Fc (NT-Fc or NT⌬SCD-Fc, respectively). The schematic diagrams of the fulllength human caveolin-1 protein, NT-Fc, NT⌬SCD-Fc and Fc␥1 are shown in Fig 4). Because immunoglobulins are glycosylated posttranslationally, the recombinant Fc fusion proteins produced with mammalian cells had higher a molecular weight in SDS-PAGE than as calculated from their amino acid composition (34). In nonreducing conditions, Fc␥1, NT-Fc, or NT⌬SCD-Fc were observed at ϳ60, 100, or 90 kDa, respectively (lanes 6 -8 in Fig. 1B), indicating that they were expressed as a homodimer. Fc␥1, NT-Fc, and NT⌬SCD-Fc were also evaluated by Western blot analysis using anti-human IgG antibody (Fig. 1C).
We examined whether the generated NT-Fc fusion protein binds to CD26. For this purpose, we used the Jurkat T-cell line that was stably transfected with full-length human CD26 (J.CD26wt) as described under "Experimental Procedures." As shown in Fig. 2A, by using lysates of J.CD26wt, CD26 was coimmunoprecipitated with NT-Fc (lane 2) but not with Fc␥1 (lane 1) nor with NT⌬SCD-Fc (lane 3). We next evaluated binding of NT-Fc to cell surface CD26 using flow cytometry. As shown in Fig. 2B, cell surface CD26 of J.CD26wt was stained with anti-CD26-FITC mAb (panel b, peak 2) (whereas unlabeled CD26 mAb, but not control IgG, blocked staining with anti-CD26-FITC mAb (peak 3 and peak 4 of panel b). J.CD26wt was also stained with biotinylated NT-Fc followed by staining with streptavidin-conjugated FITC (peak 2 of panel c in Fig. 2B). Staining with NT-Fc was blocked by unlabeled anti-CD26 mAb (peak 3 of panel c in Fig. 2B) but not control IgG (peak 4 of panel c in Fig. 2B). On the other hand, J.CD26wt was not stained with NT⌬SCD-Fc (panel d in Fig. 2B). Moreover, native Jurkat  4), NT-Fc (lanes 2 and 5), and NT⌬SCD-Fc (lanes 3 and 6) were subjected to SDS-PAGE (5-20% acrylamide gradient gel) under reducing (ϩ2ME, lanes 1-3) or nonreducing (Ϫ2ME, lanes 4 -6) conditions, followed by Western blot analysis, using horseradish peroxidase-conjugated anti-human IgG. Similar results were obtained in three independent experiments. B, J.CD26wt cells were used for binding activity of Fc fusion proteins. Panel a, forward and side scattergrams of the analyzed cells. Solid circle indicates the gated region for analysis. Panel b, cells were stained with FITC-conjugated control mouse IgG (peak 1) or FITC-conjugated anti-CD26 mAb (peak 2). For blocking assay, cells were first reacted with unlabeled anti-CD26 mAb (peak 3) or unlabeled control mouse IgG (peak 4), followed by staining as described in (peak 2). Panel c, cells were stained with biotinylated Fc␥1 as control (peak 1) or biotinylated NT-Fc (peak 2), followed by reaction with FITC-conjugated streptavidin. For blocking assay, cells were first reacted with unlabeled anti-CD26 mAb (peak 3) or unlabeled control IgG (peak 4), followed by staining as described in peak 2. Panel d, cells were stained with biotinylated Fc␥1 as control (peak 1) or biotinylated NT⌬SCD-Fc (peak 2), followed by reaction with FITC-conjugated streptavidin. For blocking assay, cells were first reacted with unlabeled anti-CD26 mAb (peak 3) or unlabeled control IgG (peak 4), followed by staining as described in peak 2. All four histograms in panel d were stacked in the same position. C, measuring the affinity of Fc fusion proteins to rsCD26 by equilibrium binding. Injections of rsCD26 at 25°C started at 50 nM and were followed by five 2-fold dilutions (50, 25, 12.5, 6. T-cells were not stained with anti-CD26 mAb nor with NT-Fc (data not shown). These data suggested that the soluble N-terminal domain of caveolin-1 binds to cell surface CD26 and that the SCD of caveolin-1 is necessary for binding to CD26, as shown in our previous studies (14,15).
To investigate the properties of binding of NT-Fc to CD26, we next examined the binding affinity with the Biacore system by injecting increasing concentrations of recombinant soluble CD26 (rsCD26) over each sensor surface containing recombinant Fc fusion proteins, Fc␥1, NT-Fc, or NT⌬SCD-Fc (Fig. 2C). rsCD26 did not bind to control recombinant Fc␥1 on a Biacore sensor chip (panel a in Fig.  2C). For each concentration of rsCD26 injected, the binding response at equilibrium was calculated by subtracting the response observed in NT-Fc, resulting in a K d value of ϳ2 ϫ 10 Ϫ5 M by equilibrium binding analysis (panel b in Fig. 2C). rsCD26 did not bind to NT⌬SCD-Fc on a Biacore sensor chip (panel c in Fig. 2C). These results clearly indicated that the N-terminal domain of caveolin-1 binds directly to CD26.
We next evaluated whether NT-Fc stimulation had a similar effect as anti-CD26 mAb on CD26mediated T-cell costimulation in J.CD26wt (13). As shown in Fig. 2D, IL-2 production of J.CD26wt induced by plate-bound anti-CD3 plus NT-Fc was observed to be at a similar level as that induced by anti-CD3 plus anti-CD28 or by anti-CD3 plus anti-CD26 (*, **, and *** in the bar graph), whereas IL-2 production was not observed in JKTwt (which is CD26-negative) following stimulation by anti-CD3 plus anti-CD26 nor by anti-CD3 plus NT-Fc (* and *** in Fig. 2D). IL-2 production by J.CD26wt or JKTwt was not observed with the use of control recombinant Fc␥1 nor NT⌬SCD-Fc (# in Fig. 2D). To further investigate whether the T-cell costimulatory activity of NT-Fc was exerted via CD26, blocking experiments were

Caveolin-1 Activates T-cell via CD26
conducted using CD26-specific mAb, which blocked binding of NT-Fc to J.CD26wt. As shown in Fig. 2E, IL-2 production induced by plate-bound anti-CD3 plus anti-CD26 was blocked by soluble anti-CD26 mAb but not by soluble anti-CD28 mAb (** and ##). In this experimental condition, IL-2 production induced by plate-bound anti-CD3 plus NT-Fc was blocked by soluble anti-CD26 mAb but not by soluble anti-CD28 mAb (* and # in Fig. 2E). Control F␥c1 or NT⌬SCD-Fc did not have any T-cell costimulatory activity (Fig. 2E). Taken together, our results clearly indicated that caveolin-1 binds directly to CD26 and induces T-cell costimulation via CD26.
We next evaluated the ability of NT-Fc to reproduce the effects of anti-CD26 mAb on CD26-mediated T-cell costimulation (11,35). As shown in Fig. 3A, T-cell proliferation induced by plate-bound anti-CD3 plus NT-Fc was observed to be at a similar level as that induced by anti-CD3 plus anti-CD28 or by anti-CD3 plus anti-CD26 (* in the bar graph), whereas T-cell proliferation was not observed using control recombinant Fc␥1 nor NT⌬SCD-Fc (** in the bar graph). Moreover, T-cell costimulation induced by NT-Fc was observed in a dose-dependent manner, whereas increasing doses of Fc␥1 or NT⌬SCD-Fc did not induce T-cell proliferation (Fig. 3B). To  MARCH 30, 2007 • VOLUME 282 • NUMBER 13

JOURNAL OF BIOLOGICAL CHEMISTRY 10123
further define the costimulatory activity of caveolin-1, we prepared CHO cells stably expressing human caveolin-1. As shown in Fig. 3C, we then showed that Cav-wt ϩ CHO cells expressed T-cell costimulatory activity in the presence of anti-CD3 mAb, an effect not observed with mock ϩ CHO cells nor Cav-⌬SCD ϩ CHO cells. The costimulatory activity of Cav-wt ϩ CHO cells was further observed in a cell number-dependent manner (* in Fig. 3C).
To further investigate whether the T-cell costimulatory activity of NT-Fc is exerted via CD26, blocking experiments were conducted using CD26-specific mAb which blocked binding of NT-Fc to J.CD26 (Fig. 2B). As shown in Fig. 3D, T-cell proliferation by plate-bound anti-CD3 plus anti-CD26 was blocked by soluble anti-CD26 mAb but not by soluble anti-CD28 mAb (*). On the other hand, T-cell proliferation by platebound anti-CD3 plus anti-CD28 was blocked by soluble anti-CD28 mAb but not by soluble anti-CD26 mAb (** in Fig. 3D). In this experimental condition, T-cell proliferation by platebound anti-CD3 plus NT-Fc was blocked by soluble anti-CD26 mAb but not by soluble anti-CD28 mAb (*** in Fig. 3D). Control F␥c1 or NT⌬SCD-Fc did not have any T-cell costimulatory activity (Fig. 3D). Moreover, the blocking effect of CD26-specific mAb on NT-Fc costimulation was observed in a dose-dependent manner (* in Fig. 3E), and control IgG or anti-CD28 mAb at concentrations of 0 -50 g/ml did not block NT-Fc costimulation (Fig. 3E). Taken together with data shown in Figs. 1-3, these data suggested that NT-Fc functionally engages CD26 and not nonspecific proteins and that caveolin-1 has a costimulatory effect on T-cell proliferation via the TCR/CD3 pathway.
The proximal signaling molecules of CD26-mediated T-cell costimulation by caveolin-1 were next determined. We first examined whether the cytoplasmic tail of CD26 is responsible for T-cell costimulation by NT-Fc in the presence of anti-CD3 mAb. For this purpose, costimulation experiments were performed on Jurkat T-cells transfected with CD26-CD10 chimeric receptor. Moreover, whereas CD26 is reported to form homodimers on cell surface (36,37), it remains to be determined whether dimeric CD26 is responsible for CD26-mediated T-cell costimulation. Therefore, costimulation experi-ments were conducted using Jurkat T-cells transfected with monomeric CD26 (CD26 H750E). We first verified the Jurkatstable transfectants, as shown in Fig. 4A, and CD26wt, CD26 ϩ CD10 cyto, and CD26 H750E were detected at ϳ100 kDa in reducing conditions (lanes 2-4), bands of CD26wt and CD26 ϩ CD10 cyto migrated at ϳ200 kDa (lanes 6 and 7), and a band of CD26 H750E migrated at 100 kDa (lane 8) in nonreducing conditions, indicating that CD26wt and CD26 ϩ CD10 cyto exist as dimers, and CD26 H750E exists as monomers in the Jurkat transfectants. Moreover, as shown in Fig. 4B, by immunoprecipitation using anti-V5 mAb, V5-tagged CD26 was detected by anti-CD26 mAb in each transfectant (CD26wt, CD26 ϩ CD10 cyto, or CD26 H750E) in reducing SDS-PAGE (lanes 2-4). Furthermore, Fig. 4C showed the cell surface expression of CD3 and CD26 in Jurkat transfectants. CD3 was expressed at similar intensity among transfectants (horizontal axis of panels a-d in Fig. 4C), whereas the intensity of cell surface CD26 expression was different between CD26wt/CD26 ϩ CD10 cyto and CD26 H750E (vertical axis of panels b-d in Fig. 4C), suggesting a difference between dimeric expression and monomeric expression. CD26 was not observed in mock vector-transfected Jurkat (JKT/mock; panel a in Fig. 4C). Using these Jurkat transfectants, costimulation experiments by NT-Fc were conducted. As shown in Fig. 4D, IL-2 production was observed in CD26wttransfected Jurkat T-cells by stimulation with anti-CD3 plus NT-Fc (* in panel a), but not in CD26-CD10 chimera nor in CD26 H750E-transfected Jurkat. Moreover, p65, one of NF-B components, was activated in CD26wt-transfected Jurkat T-cells following stimulation with anti-CD3 plus NT-Fc (* in panel b) but not in CD26-CD10 chimera nor in CD26 H750Etransfected Jurkat. Furthermore, IL-2 production or p65 induction by stimulation with anti-CD3 plus PMA was equally observed in either of CD26wt, CD26-CD10 chimera, or CD26 H750E transfected Jurkat (panels a and b of Fig. 4D). These data strongly suggested that the cytoplasmic tail of dimeric CD26, but not of monomeric CD26, is responsible for T-cell costimulation by NT-Fc in the presence of anti-CD3 mAb.

Caveolin-1 Activates T-cell via CD26
For further confirmation, we next performed IP studies using lysates of J.CD26wt. As shown in Fig. 5D, CD26 was detected in a complex of lysates coprecipitated with anti-CARMA1 pAb (lane 2 of upper panel), and not coprecipitated with control goat IgG (lane 1 of upper panel). Moreover, CARMA1 was detected in a complex of lysates coprecipitated with anti-CD26 mAb (lane 4 of lower panel in Fig. 5D), and not coprecipitated with control mouse IgG (lane 3 of lower panel in Fig. 5D). These data suggested that CARMA1 binds to CD26 in cells. To determine the binding domain between CD26 and CARMA1, coimmunoprecipitation assay was next performed using 293FT-cells cotransfected with Xpress-tagged human full-length CARMA1 (CARMA1wt) and with CD26wt, CD26 ϩ CD10 cyto, or CD26 H750E. As shown in Fig. 5E, CARMA1 was coprecipitated with CD26wt (lane 2) but not with CD26 ϩ CD10 cyto nor CD26 H750E (lanes 3 and 4). These data strongly suggested that the cytoplasmic tail and dimerization of CD26 are necessary to interact with CARMA1. We next explored the binding domain of CARMA1 to CD26. For this purpose, we prepared the C-terminal truncated deletion mutants of CARMA1 (Fig. 5F). As shown in Fig. 5G, CD26 was coprecipitated with CARMA1wt or with CARMA1- (1-742) (lanes 2 and 3) but not with CARMA1- (1-600) (lane 4), suggesting that the PDZ domain in CARMA1 was necessary for binding to CD26.
To explore the role of CARMA1 in CD26-mediated T-cell costimulation, we used CARMA1-deficient Jurkat T-cell lines JPM50.6 to conduct rescue experiments (18). As shown in Fig. 6A, CARMA1 and CD26 were not detected in JPM50.6 (lane 3 of upper and lower panels), whereas CARMA1 was expressed in native Jurkat and J.CD26wt (lanes 1 and 2 of  lower panel), and CD26 was expressed in J.CD26wt (lane 2 of upper panel) but not in native Jurkat (lane 1 of upper panel).
As shown above, costimulation of CD26 is observed to be exerted via interaction of CD26 with CARMA1 in the cytoplasm in Jurkat cells. To confirm this interaction more profoundly, we performed biochemical assays using human T-cells purified from healthy adult peripheral blood mononuclear cells (APB-T-cells). For this purpose, we first conducted IP studies using lysates of APB-T-cells. As shown in Fig. 7A, CD26 was detected in a complex of lysates coprecipitated with anti-CARMA1 pAb (lane 2 of upper panel), and not coprecipitated with control goat IgG (lane 1 of upper panel). Moreover, CARMA1 was detected in a complex of lysates coprecipitated with anti-CD26 mAb (lane 4 of lower panel in Fig. 7A), and not coprecipitated with control mouse IgG (lane 3 of lower panel in Fig. 7A). These data suggested that CARMA1 binds to CD26 in normal T-cells.
We previously showed that nonactivated peripheral blood T-cells treated with the anti-CD26 mAb 1F7 resulted in CD26 recruitment to lipid rafts, concomitant with increased tyrosine phosphorylation of ZAP70, p56 lck , and TCR (33).
Other investigators have reported that in the process of activation of NF-B via CD3 costimulation, CARMA1 was recruited to lipid rafts along with Bcl10 and IKK␤ (18,25,27,38,39). We therefore examined whether CD26 and CARMA1 are recruited to lipid rafts by anti-CD3 plus caveolin-1 costimulation in normal T-cells. For this purpose, using a sucrose gradient separation method, we prepared lipid raft

Caveolin-1 Activates T-cell via CD26
fractions of APB-T-cell lysates in the presence or absence of anti-CD3 plus NT-Fc costimulation. As shown in Fig. 7B, following stimulation with anti-CD3 plus NT-Fc, CD26, CARMA1, Bcl10, and IKK␤ were detected in the lipid raft fractions, whereas CD26, CARMA1, Bcl10, and IKK␤ were not detected in the lipid raft fractions after stimulation with anti-CD3 alone. Moreover, time course analysis revealed that CD26, CARMA1, Bcl10, and IKK␤ were migrated into lipid rafts after stimulation with anti-CD3 plus NT-Fc (Fig. 7C), whereas CD26, CARMA1, Bcl10, and IKK␤ were not detected after anti-CD3 treatment (data not shown). Furthermore, to examine whether CD26 and CARMA1 forms a complex with Bcl10 and IKK␤ in lipid rafts, coprecipitation assay was performed using lipid raft fractions of APB-T-cell lysates from cells costimulated with anti-CD3 plus NT-Fc. As shown in Fig. 7D, CD26, CARMA1, Bcl10, and IKK␤ in lipid rafts were coprecipitated with CD26 (lane 4), although not detected in the lysates of APB-Tcells following stimulation with anti-CD3 alone (lane 2). Taken together, these data indicated that ligation of CD26 by caveolin-1 recruits a complex of CARMA1, Bcl10, and IKK␤ to lipid rafts in normal T-cells.
To examine the role of CARMA1 on CD26-mediated T-cell costimulation more directly, we performed siRNA experiments in freshly isolated APB-T-cells. For this purpose, we prepared two sets of specific siRNA against CARMA1 as described under "Experimental Procedures," and both of these siRNAs decreased CARMA1 expression in APB-T-cells, whereas the expression levels of CD26, TCR-␤, or ␤-actin were not changed in the presence of control siRNA, ss1-siRNA, or ss2-siRNA (inside box of Fig. 7E). After transfection of these siRNAs into APB-T-cells, the proliferation assay was performed in the presence of anti-CD3 plus NT-Fc stimulation. As shown in Fig. 7E, T-cell proliferation stimulated with anti-CD3 plus NT-Fc was decreased in T-cells treated with siRNAs against CARMA1, whereas T-cell proliferation was observed in T-cells treated with control siRNA (* in Fig. 7E). Moreover, T-cell proliferation stimulated with anti-CD3 plus PMA was observed in either of control siRNA, ss1-or ss2-siRNA (** in Fig. 7E). These results suggested that CARMA1 plays an important role in signal transduction following CD26 binding to caveolin-1, leading to T-cell proliferation in normal T-cells.  (panels b, d, and e). D, JPM50.6 transfectants, which were described in B, were stimulated with plate-bound anti-CD3 in the presence or absence of plate-bound NT-Fc as described in Fig. 4D. Panel a, following 48 h of culture, IL-2 concentration of the culture supernatant was measured by ELISA. Values shown are means Ϯ S.E. of determinations from triplicate cultures. * shows points of significant increase (p Ͻ 0.05) compared with control. Panel b, JPM50.6 transfectants were stimulated as described in panel a and harvested for nuclear extract. Each 5 g of nuclear extract was subjected to ELISA-based DNA-binding protein assay. Binding activity to p65 NF-B component was revealed by absorbance value at 450 nm. Data represent mean Ϯ S.E. from triplicate experiments. * shows a point of significant increase (p Ͻ 0.05). Xpress vec. or V5 vec. depicts pcDNA4/HisMax or pEF6/V5 empty vector as a mock, respectively.

DISCUSSION
In this study, we showed that caveolin-1 is the costimulatory ligand for CD26, and that ligation of CD26 by caveolin-1 induces T-cell proliferation and NF-B activation with costimulation of TCR/CD3. Moreover, we showed that the cytoplasmic tail of CD26 in T-cell interacts with CARMA1, resulting in signal transduction leading to NF-B activation and that ligation of CD26 by caveolin-1 recruits a complex of CD26, CARMA1, Bcl10, and IKK␤ to lipid rafts.
Enhancement of CD26 expression in autoimmune diseases may correlate with disease severity (40,41), because patients with autoimmune diseases such as Grave's disease and rheumatoid arthritis have increased levels of CD26 ϩ T-cells in their peripheral blood as well as inflamed tissues, including thyroid and synovial fluids and membranes (9,42). These findings imply that CD26 ϩ T-cells play a role in the inflammation process and subsequent tissue destruction. Originally characterized as a T-cell activation antigen, human CD26 is preferentially expressed on the CD4ϩ memory T-cell subset and is up-regulated after T-cell activation (2,3,10). Along with its enhanced expression on activated T-cells, various lines of evidence have converged to demonstrate that CD26 is functionally associated with T-cell signal transduction processes relating to T-cell activation (2,10,11,43). However, the precise mechanism involved in T-cell activation via CD26 in response to memory antigen such as tetanus toxoid remains to be clearly characterized, including the identification of its costimulatory ligand and the associated proximal signaling molecules. Recently, we demonstrated that CD26 binds to caveolin-1 on APC and that residues 201-211 of CD26 along with the serine catalytic site at residue 630, which constitute a pocket structure of CD26/DPPIV, contribute to binding to the caveolin-1 scaffolding domain (14). This region in CD26 contains a caveolin-binding domain (⌽X⌽XXXX⌽XX⌽; ⌽ and X depict aromatic residue and any amino acid, respectively), specifically WVYEEEVFSAY in CD26 (2, 44). These FIGURE 7. CARMA1 plays an important role in CD26-mediated costimulation by caveolin-1 in human peripheral blood T-cells. A, purified T-cells were lysed, and IP assays were conducted with anti-CARMA1 pAb (goat), anti-CD26 mAb (mouse (ms)), or control Ig (cIgG). IP complexes as well as 10% of input lysates were then separated using SDS-PAGE and immunoblotted with the indicated antibodies. Similar results were obtained in three independent experiments. B, purified T-cells were stimulated for 10 min with anti-CD3 alone (0.05 g/ml) or with anti-CD3 plus NT-Fc (5.0 g/ml), and lysates were prepared by sucrose gradient centrifugation as described under "Experimental Procedures." The distribution of CD26, CARMA1, Bcl10, and IKK␤ was determined by immunoblotting with specific antibodies. Similar results were obtained in three independent experiments. C, purified T-cells were stimulated for 0, 10, and 30 min with anti-CD3 plus NT-Fc, and lysates were prepared by sucrose gradient centrifugation as described under "Experimental Procedures." The distribution of CD26, CARMA1, Bcl10, and IKK␤ was determined by immunoblotting with specific antibodies. Similar results were obtained in three independent experiments. D, purified T-cells were stimulated with anti-CD3 alone (lanes 1 and 2) or with anti-CD3 plus NT-Fc (lanes 3 and 4), and lipid raft fractions were prepared as described in A, and immunoprecipitation of lipid rafts with control IgG (cIgG) (lanes 1 and 3) or anti-CD26 mAb (lanes 2 and 4) was performed as described under "Experimental Procedures." IPs were resolved in SDS-PAGE and immunoblotted with indicated antibodies. Similar results were obtained in three independent experiments. E, purified T-cells were transfected with sense-siRNA (ss1 and ss2) of CARMA1 gene or mismatched siRNA (control) using HVJ-E vector. Cell lysates were resolved by SDS-PAGE and immunoblotted with indicated antibodies, followed by stripping and reprobing with anti-␤-actin antibody (inside box). Purified T-cells treated with siRNA were stimulated and subjected to T-cell proliferation assay as described in Fig. 3A. Values shown are means Ϯ S.E. of determinations from triplicate cultures of five independent donors. * shows points of significant decrease (p Ͻ 0.05), and ** indicates points of no significant change compared with controls.
observations strongly support the notion that DPPIV enzyme activity is necessary to exert T-cell costimulatory activation via CD26 as demonstrated in our previous report using CD26 specific mAbs (13).
To examine the binding of caveolin-1 to CD26 in T-cells, we used soluble Fc fusion proteins containing the N-terminal domain of caveolin-1 (NT-Fc) (Fig. 1), and we found that NT-Fc binds specifically to CD26 to induce T-cell proliferation in the presence of TCR/CD3 costimulation (Figs. 2 and 3). Moreover, the binding affinity between caveolin-1 and CD26 (K d ϳ 2 ϫ 10 Ϫ5 M), as determined by the Biacore system (Fig. 2C), is comparable with that of other costimulatory molecules with important roles in immune responses and their associated ligands, such as CD2-CD5 (K d ϳ 10 Ϫ6 M), CD80-CD28 (K d ϳ 10 Ϫ7 M), and CD86-CD28 (K d ϳ 10 Ϫ6 M) (45)(46)(47). Until now, CD26mediated T-cell costimulation was performed using anti-CD26 mAbs, resulting in various CD26 functions (4,7,48,49). Assuming that the affinity between antigen and antibody is higher (K d ϳ 10 Ϫ9 M) than that of a ligand-receptor system, and that ligand-specific conformations are capable of differentially activating distinct signaling partners (50), ligand-dependent pathways may be predicted to have different signals associated with the antigen-antibody system and ligand-receptor system.
We have demonstrated previously that ligation of CD26 by the anti-CD26 mAb 1F7 induces T-cell costimulation and IL-2 production by CD26-transfected Jurkat T-cell lines, while increasing tyrosine phosphorylation of signaling molecules such as ZAP70, p56 lck , and CD3 was observed (2,7,12). In addition, we have shown that ligation of the CD26 molecules by the anti-CD26 mAb 1F7 increases the recruitment of CD26 molecules with CD45RO to lipid rafts, resulting in increased tyrosine phosphorylation of signaling molecules (33). However, the precise proximal signaling pathway of CD26 has not yet been identified, particularly in view of the fact that the cytoplasmic tail of CD26 contains only 6 amino acid residues without a common signaling motif structure. Moreover, it has been unclear whether the short cytoplasmic tail is responsible for signal transduction associated with CD26-mediated costimulation. In this study, using recombinant CD26-CD10 chimeric receptor, we showed that the cytoplasmic tail of CD26 is indeed responsible for T-cell costimulation induced by anti-CD3 plus caveolin-1 (Fig. 4D). Furthermore, to explore the proximal signaling molecules interacting with the cytoplasmic tail of dimeric CD26, we used proteomic analyses with Fc fusion proteins containing the cytoplasmic amino acid residues of CD26 (Fig. 5, A and B) to identify that CARMA1 binds to the cytoplasmic tail of dimeric CD26 (Fig. 5C). Moreover, we demonstrated here that a PDZ domain in CARMA1 is necessary for binding to CD26 (Fig. 5G). The importance of CARMA1 in CD26-mediated costimulation is also shown by rescue experiments using the CARMA1-deficient Jurkat T-cell line JPM50.6 ( Fig. 6D) and using siRNA against CARMA1 in APB-T-cells (Fig. 7E). CARMA1, containing caspase-recruitment domain and MAGUK domains, plays an essential role in the NF-B activation and IL-2 expression induced by CD3-CD28 or CD28-PMA stimulation (18,22). After being phosphorylated, CARMA1 functions as a signaling intermediate downstream of PKC and upstream of IKK in the TCR signaling transduction pathway leading to NF-B activation (39,51). Because MAGUK domain-containing proteins are generally involved in the organization of multiprotein complexes at the interface of the cytoplasmic membrane (52), it is possible that CARMA1 associates with as yet undefined membrane proteins in the immunological synapse of T-cells. In this regard, our present data suggest a novel mechanism for CAMRA1 function as it complexes with Bcl10 and IKK to transduce CD26-costimulatory signals. Moreover, as shown Fig. 5C, cytoskeletal proteins were also observed in the complex in the pulldown assays with CD26 aa1-10-Fc. Because MAGUK domain-containing proteins are generally involved in the organization of multiprotein complexes in the cytoskeleton (52), the downstream signaling of CD26 may also be associated with cytoskeletal assembly via CARMA1. The association of CD26, CARMA1, and the cytoskeleton will be elucidated in future studies.
CD26/DDPIV is reported to exist as homodimers, a structural organization that allows access of substrates to DPPIV catalytic activity (36,37). Although DPPIV activity is crucial for CD26-mediated T-cell costimulation (13,30), the exact role played by DPPIV in this process is unclear. Our previous study showed that the enzymatic pocket structure of the DPPIV catalytic site is necessary for binding of CD26 to caveolin-1, leading to the up-regulation of CD86 expression on APC (14,15). In this study, we found that monomeric CD26 H750E, which has a 300-fold decrease in catalytic activity (36), does not bind to CARMA1 (Fig. 5E), resulting in loss of CD26-mediated T-cell costimulation by anti-CD3 plus caveolin-1 (Fig. 4D). Therefore, dimerization of CD26 is not only necessary for binding to caveolin-1 but also serves as a scaffolding structure for the cytoplasmic signaling molecule CARMA1. The precise binding position of CARMA1 in the cytoplasmic domain of CD26 remains to be elucidated in future work, because PDZ domains bind primarily to specific C-terminal motifs (X(S/T)X(V/L), where X depicts any amino acid) or internal target motifs as well as other PDZ domains (52).
Based upon this study, we propose the following model to explain the sequence of events leading from CD26-CD3 costimulation to NF-B activation (Fig. 8). In CD3-CD26 costimulation, TCR engagement by peptide-loaded major histocompatibility complex class II presented on APC activates phosphatidylinositol 3-kinase via phosphorylation of immunoreceptor tyrosine-based activation motifs in TCR, leading the recruitment of PKC and IKK complex in lipid rafts (16,18,25,38). Concomitantly, CD26 ligation by caveolin-1 on APC recruits CD26-interacting CARMA1 to lipid rafts, resulting in the formation of a CARMA1-Bcl10-MALT1-IKK complex, and this membrane-associated Bcl10 complex then activates IKK through ubiquitination of the NF-B essential modulator. This study involving Jurkat T-cell lines and human peripheral T-cells represents a different cellular system than those with murine T-cells, where other investigators previously described a role for CD26 in thymic development of murine T-cells (53,54). Our objective with this study was to define a costimulatory ligand for CD26 and proximal signaling molecule of CD26 in human T-cells, with a future aim of analyzing the in vivo role of CD26-mediated T-cell immunity.
In conclusion, we have now demonstrated that CD26 on the T-cell surface binds to caveolin-1, hence identifying the first endogenously expressed CD26 costimulatory ligand in the immune system. Moreover, the caveolin-1-CD26 interaction results in strong T-cell costimulation as a result of the recruitment of a molecular complex consisting of CARMA1-Bcl10-MALT1-IKK in lipid rafts. Our findings will therefore serve as a foundation for future insights into the regulation of T-cell costimulation via the CD26 molecule.