Nectin-like protein 2 defines a subset of T-cell zone dendritic cells and is a ligand for class-I-restricted T-cell-associated molecule.

Dendritic cells (DCs) are a phenotypically and functionally heterogenous population of leukocytes with distinct subsets serving a different set of specialized immune functions. Here we applied an in vitro whole cell panning approach using antibody phage display technology to identify cell-surface epitopes specifically expressed on human blood BDCA3(+) DCs. A single-chain antibody fragment (anti-1F12 scFv) was isolated that recognizes a conserved surface antigen expressed on both human BDCA3(+) DCs and mouse CD8alpha(+) DCs. We demonstrate that anti-1F12 scFv binds Nectin-like protein 2 (Necl2, Tslc1, SynCaM, SgIGSF, or Igsf4), an adhesion molecule involved in tumor suppression, synapse formation, and spermatogenesis. Thus, Necl2 defines a specialized subset of DCs in both mouse and human. We further show that Necl2 binds Class-I-restricted T-cell-associated molecule (CRTAM), a receptor primarily expressed on activated cytotoxic lymphocytes. When present on antigen presenting cells, Necl2 regulates IL-22 expression by activated CD8(+) T-cells. We propose that Necl2/CRTAM molecular pair could regulate a large panel of cell/cell interactions both within and outside of the immune system.

many as the sentinels of the immune system (1,2). Indeed, DCs can transmigrate through the endothelium, intercalate between mucosal epithelial cells (3) and through direct cell/cell interaction, trigger or modulate a large panel of immune functions, including CD4 or CD8 T-cell response, NK cytotoxicity (4), and B-cell differentiation into plasma cells (5). Over the past decade, it has become increasingly clear that DCs are not a homogeneous population of cells. Rather, DCs can be divided into distinct subsets each serving a restricted set of specialized functions.
With the exception of PDCs, how mouse, rat, and human DC subsets relate to each other has remained elusive. Consequently, it has been difficult to compare functional DC subset data obtained from different animal models. Our understanding of DC biology in mammals would greatly benefit from a common cross-species phenotypic definition.
Currently, DC subset definition relies heavily upon isolation of antibodies obtained from rodents immunized with allogeneic DCs. We reasoned that, if highly conserved DC subset markers do exist, they would probably not be easily targetable through in vivo immunization. Indeed, since DCs play a critical role in the initiation and the maintenance of immune responses, B-cells, which target conserved epitopes present on host DCs, might be put at a disadvantage and not perceive signals necessary for their survival or differentiation into antibody producing cells.
Whole cell panning with filamentous phage offers an attractive alternative to immunization because it can be conducted in vitro in a competitive setting, thus allowing selection and counter-selection with a tightly controlled stringency and without the limitations linked to interspecies compatibility or to targeting molecules essential to the host immune response. Whole cell panning approaches using phage antibody libraries have been successfully used to isolate antibodies specific to DCs (17), tumor cells (18 -21), lymphocytes (22)(23)(24), as well as other cell types (25,26). Molecular targets of phage antibodies can be identified using cell sorting and biochemical/cloning techniques (18,22).
Here, we applied a whole cell panning phage display approach to select a single chain antibody fragment (scFv) that recognizes an epitope expressed on both human BDCA3 ϩ DCs and mouse CD8␣ ϩ DCs. We show that this scFv binds Nectinlike protein 2 (Necl2), a type-I cell surface protein previously shown to be involved in tumor suppression, synapse formation, and spermatogenesis. We further show that Class-I-restricted T-cell-associated molecule (CRTAM), a surface receptor primarily expressed on activated cytotoxic lymphocytes, constitutes a counter-structure for Necl2. Thus, Necl2 defines a specialized subset of DCs across species and participates in the cross-talk between antigen-presenting cells and cytotoxic lymphocytes.
Mice-Female C57BL/6 (B6) (8 -12 weeks of age) and OT-I TCR-Tg mice specific for chicken ovalbumin (OVA) peptide 357-364 (OT-Ip) in the context of H-2Kb (29) were obtained from Taconic Farms (Germantown, NY) and bred at Amgen (Seattle, WA). All mice were housed under specific pathogen-free conditions and according to federal guidelines.
Patients-Healthy volunteers received either 10 or 25 mg/kg/day of Flt3-ligand (FL; Immunex Co.) for 10 consecutive days in conformity with two institutional review board approved protocols (Baylor Institute, Investigational New Drug number 7805; Duke University, Investigational New Drug number 8209). The volunteers had normal blood counts and chemistries and were Ͼ18 years of age. Prior to entering the study, each volunteer signed an informed written consent that fulfills Institutional Review Board guidelines. FL-treated volunteers underwent apheresis at base line and after 10 days of FL injection. Mononuclear cells were then selected by either standard Ficoll/Hypaque gradient method or by using continuous-flow cell processor COBE® 2991 TM (Gambro® BCT TM ). Purified PBMCs were resuspended at 10 7 to 10 8 cells/ml in medium containing 44% RPMI 1640, 44% fetal bovine serum (FBS), and 12% dimethyl sulfoxide and then cryo-preserved.
Phage Library-Three scFv libraries (Cambridge Antibody Technologies) with total diversity of 10 10 (30) were used to select specific binders to human Dendritic Cells. An aliquot containing 4.10 12 cfu phage in the hyperphage format (31) was first blocked in 500 l of 3% nonfat milk in PBS and dialyzed against PBS overnight at 4°C. The blocked phage were incubated with 10 8 PBMC from FL-treated healthy volunteers for 2 h at room temperature with gentle mixing. Cells were then pelleted, washed three times, and then labeled with anti-BDCA3 antibody. Approximately 10 5 CD162 bright BDCA3 ϩ cells were purified by magnetic-activated and subsequently fluorescence-activated cell sorting. Sorted cells were washed once with PBS, and cell-bound phage were eluted with 200 l of 100 nM HCl for 10 min at room temperature and then neutralized with half-volume of 1 M Tris-HCl (pH 7.5). A log phase culture of Escherichia coli strain of TG1 was infected with eluted phage and plated on 2ϫYT agar plates that contain 2% glucose and 50 g/ml of carbencillin. The resulting colonies were propagated and used to prepare phage for the next round of panning. 10 11 cfu phage were used to repeat the panning process. After the second round of panning, eluted phage were used to infect TG1 cells and plated on 2ϫYT plates with 2% glucose and 50 g/ml carbencillin. After 24 h, 400 individual colonies were picked into 96-well plates and grown overnight to make glycerol stocks.
FACS Analysis of PBMC Using Phage-Polyclonal or individual phage were prepared from the glycerol stock from different stages of panning and used to compare their binding specificity to different subsets of human PBMC. Briefly, 10 11 cfu phage were blocked in PBS ϩ 0.5% bovine serum albumin and then used to bind PBMCs for 1 h at room temperature. PBMCs were obtained by standard Ficoll/Hypaque centrifugation of blood obtained from normal healthy volunteers. Phage revelation was performed by HRP-conjugated anti-M13 major coat protein antibody (Amersham Biosciences, Bucks, UK) followed by FITCconjugated anti-HRP antibody (Jackson ImmunoResearch Laboratories, West Grove, PA).
Production of scFv-Fc Fusion Proteins-The DNA fragment encoding scFv region of 1F12 was cloned into a mammalian expression vector pDC409huIgG1 to produce scFv-Fc fusion protein as described previously (32). The 1F12 scFv-Fc construct was used to transfect COS1-PKB cells, and the fusion protein was purified using protein-A affinity chromatography.
Histology-Human thymi, spleens, lymph nodes, and tonsils were obtained from an in-house tissue bank. Murine tissues were collected from adult BALB/c mice. For paraffin tissues, samples were fixed in 10% neutral buffered formalin and processed routinely overnight in a Shandon automated tissue processor and then embedded. Frozen tissue samples were snap-frozen in OCT compound by immersion in isopentane cooled with liquid nitrogen.
Sections were cut at 5 m and stained with the predetermined optimal concentrations of biotinylated anti-1F12 scFv-Fc. An unrelated scFv-Fc protein served as a negative control. At the same time, sections were stained with commercially available anti-mouse CD8␣ antibody or anti-BDCA3 antibody. For confocal usage, biotinylated reagents were subsequently incubated with Alexa Fluor®488-conjugated Streptavidin fluorochrome (Molecular Probes, Eugene, OR). Non-biotinylated primary antibodies were incubated with Molecular Probes anti-IgG Alexa Fluor®555 fluorochrome (Molecular Probes). All fluorescent slides were examined and photographed on a Zeiss LSM 510 confocal microscope.
Immunoprecipitation/Western Blotting-10 8 Bone marrow-derived DC (33) or anti-CD3⑀-activated C57BL6 mouse splenic CD8 ϩ T-cells were surface-biotinylated (EZ-Link TM Sulfo-NHS-biotinylation kit) and lysed in 1 ml of lysis buffer (PBS ϩ 1% Triton X-100 (Pierce) with protease inhibitors (Complete, Roche Applied Science). Cell debris were spun out (3000 rpm, 15 min) and supernatants were precleared with 100 l of packed protein-A-agarose beads (Roche Applied Science) for 16 h at 4°C before addition of anti-1F12 scFv-Fc, Necl2-Fc, or corresponding controls (unrelated scFv-Fc and Necl1-Fc) at 5 g/330 l. After 3-h incubation on ice, 15 l of packed protein-A beads were added and incubated for 2 h at 4°C. Beads were then washed twice in ice-cold lysis buffer and once in ice-cold PBS before SDS-PAGE analysis on 4 -20% Novex® Tris-glycine gel.
Reduction, Alkylation, and Digestion-Following SDS-PAGE, protein was stained using Colloidal Blue (Invitrogen). Gel bands were manually excised, and the gel pieces were processed using an automated protein digestion station (Genomic Solutions, Ann Arbor, MI) that performed the tasks of reduction and alkylation of cysteines, tryptic digestion (Promega, Madison, WI), and extraction of the peptides.
Mass Spectrometry-Mass spectrometric analysis of tryptic peptides was performed on a Micromass QTOF 1 instrument (Micromass, Manchester, UK). Peptides were sequenced by on-line microcapillary liquid chromatography-electrospray ionization-tandem mass spectrometry (MS/MS) operated in a data-dependent mode. The capillary column was made from a piece of 50 m inner diameter fused silica line with an internal frit (New Objective, Cambridge, MA) that was packed with YMC ODS-AQ resin (Waters, Milford, MA). The liquid chromatography gradient (0 -75% solvent B in 60 min, solvent A ϭ 0.1% formic acid, 0.001% heptafluorobutyric acid, and 5% acetonitrile; solvent B ϭ 80% acetonitrile) was developed using an Eldex Micropro syringe pump (Eldex, Napa, CA) operating at 8 l/min. A preinjector splitter cut the flow to ϳ250 nl/min. The effluent of the column was directed into an Upchurch (Oak Harbor, WA) micro-tee containing a platinum electrode and a New Objective uncoated fused silica tip (360 m outer diameter, 20 m inner diameter, pulled to a 10-m opening).
Identification of Protein-Peptides were identified using the Mascot program (34) in conjunction with the de novo sequencing program Lutefisk (35) as well as an additional data base search program written in-house. Search parameters used were peptide mass tolerance of 0.45 Da, fragment ion tolerance of 0.25 Da, oxidized methionine was the only variable modification allowed, and the nr protein sequence data base from the NCBI was searched. A reversed nr data base was also searched to determine appropriate scoring thresholds to reduce false positive rates (Ͻ0.25%) for individual peptides.
Production of vesicular stomatitis virus-G protein pseudotyped packaged lentiviral constructs was performed by transient transfection of 293ME cells (cell line stably expressing the macrophage scavenging receptor and the Epstein-Barr virus nuclear antigen (EBNA). Each semiconfluent 15-cm plate of 293ME cells was cotransfected with 4. Serial dilution of concentrated virus supernatants were used to transduce 10 5 EL4 cells (ATCC, catalog number TIB-41) in 250 l of culture medium containing 10 g/ml DEAE overnight. Cells were then expanded in 5 ml of culture medium. hNecl1 and hNecl2 transduction efficiency was assessed by flow cytometry using biotinylated mouse anti-hNecl1 antibody and anti-1F12 scFv, respectively. Optimal transduction yielded 10% positive cells that were then enriched by MACS® using streptavidin-coated magnetic beads (Miltenyi Biotec Inc., Auburn, CA).
EL4/OT1 Coculture-EL4 cells or EL4-Necl2 stable transfectants were incubated for 2 h with 100 M H2-K(b)-restricted OVA-derived peptide (SIINFEKL, Anaspec Inc., San Jose, CA), washed extensively, and cocultured with CD8 ϩ T purified from the spleen and lymph nodes of two female OT1 mice using a Miltenyi Biotech CD8 ϩ T-cell-negative enrichment kit according to the manufacturer's specifications.
Quantitative PCR Analysis-mRNA was reverse-transcribed using Roche Applied Science/Applied Biosystems TaqMan reverse transcriptase reagents (catalog number N8080234) after random hexamer priming (catalog number N8080127). Cytokine primers/probe sets were obtained from Applied Biosystems' Assay-on-Demand TM . Murine hypoxanthine phosphoribosyltransferase (muHPRT) primer/probe set was used in parallel reactions as a sample control and for determination of relative cytokine expression.

Identification of a Unique Blood Dendritic
Cell Subpopulation-Peripheral blood was collected from healthy volunteers injected daily with FL. Mononuclear cells were then isolated and subjected to flow cytometric analysis. A rare (0.12 Ϯ 0.04% of PBMC, n ϭ 7) population of cells was identified by their uniquely bright expression of surface CD162 (Fig. 1). Those cells appeared myeloid in origin as they expressed CD11c, CD13, and CD33. In contrast, CD162 bright cells did not show any detectable surface expression of CD2, CD11b, CD14, CD15s(slewX), or Fc␥RII-A (Fig. 1). CD162 bright cells with identical phenotypic characteristics were also found in the blood of normal healthy volunteers although in lower abundance (0.05 Ϯ 0.02, n ϭ 8).
Since FL is a known DC growth factor, we hypothesized that CD162 bright cells might represent a previously uncharacterized population of peripheral blood DCs. To test this hypothesis, CD162 bright cells were purified by flow cytometry and cultured in the presence of allogeneic T-lymphocytes. Unlike CD14 ϩ monocytes, CD162 bright cells were capable of inducing proliferation of allogeneic T-lymphocytes, as assessed by tritiated thymidine incorporation (Fig. 1). Thus CD162 bright cells are allocompetent DCs.
In the course of this study, Dzionek et al. described a monoclonal antibody against a dendritic cell surface antigen (BDCA3) (14). Anti-BDCA3 antibody specifically labels a rare population of human blood CD11c ϩ CD11b Ϫ DCs (14). We tested whether anti-BDCA3 antibody would label CD162 bright DCs in the blood of normal and FL-treated healthy volunteers. CD162 bright DCs were the only blood mononuclear cells labeled with anti-BDCA3 antibodies. Thus, we have identified a rare population of human blood DCs that specifically expresses BDCA3 and is expanded by in vivo administration of FL.
1F12, a CD162 bright BDCA3 ϩ DC Cell Surface Marker-We reasoned that a whole cell panning phage display approach could be applied in vitro to target BDCA3 ϩ DCs in the context of whole PBMCs. Scfv-filamentous phage from a highly diverse library (Ͼ10 10 clones) were incubated with PBMCs from FLtreated healthy volunteers. After extensive washes, CD162 bright BDCA3 ϩ cells were purified by magnetic activated and fluorescence cell sorting. Phage bound to the surface of BDCA3 ϩ DC were then eluted by acid treatment and amplified in E. coli (see "Materials and Methods").
Two successive rounds of selection/amplification were conducted using cells from two different FL-treated healthy donors. Progressive enrichment of cell surface binders was monitored by flow cytometry after each round of selection (supplemental Fig. 1).
Individual phage clones were then assayed for their ability to label PBMCs from normal healthy donors. Ninety-two percent (368/400) of those individual phage labeled BDCA3 ϩ DCs (data not shown). Specificity of BDCA3 ϩ DC surface binders was assessed by measuring their ability to label lymphocytes, monocytes, CD1c ϩ DCs, and PDCs. Interestingly, 93% (344/ 368) of all BDCA3 ϩ DC binders could be clustered into five groups according to their pattern of reactivity (Fig. 2). Phage within each reactivity group shared sequence similarities in their complementarity-determining region sequences suggesting that they recognized a single dominant epitope (data not shown). Thus, our selection procedure successfully selected BDCA3 ϩ DC surface binders, most of which likely recognized one of five dominant epitopes.
Phage within Groups I to IV labeled BDCA3 ϩ DCs as well as other cell populations (Fig. 2). In contrast, Phage within Group V appeared to specifically label the surface of BDCA3 ϩ DCs (Fig. 2). Sequencing of these phage scFv revealed that they

Necl2/CRTAM Interaction and DCs
were clonally related. When expressed in the form of an Fc fusion protein, the scFv portion of this phage retained similar cell surface binding specificity. The target of this scFv-Fc fusion protein will be referred to as 1F12.
1F12 Defines CD11c ϩ CD11b Ϫ Dendritic Cells across Species-We have shown that anti-1F12 scFv specifically labels a subset of DCs in human blood that expresses CD11c but not CD11b.
A population of DCs with a similar CD11c ϩ CD11b Ϫ phenotype has also been described in the T-cell area of mouse lymphoid tissues (39). In mouse spleen, those CD11c ϩ CD11b Ϫ DCs coexpress CD8␣. We tested whether DCs within the T-cell area of human spleen would express 1F12. Human spleen cryo-sections were labeled with anti-1F12 scFv. As shown in supplemental Fig. 2, 1F12 ϩ cells could be detected in the T-cell area and around the periarteriolar lymphoid sheath of human spleen. These 1F12 ϩ cells are rare, exhibit small dendrites, and coexpress BDCA3 (Fig. 3) but not CD8␣. BDCA3 ϩ 1F12 ϩ cells were also found in the T-cell area of human lymph nodes, tonsils, as well as in the thymus (data not shown).
Interestingly, anti-1F12 scFv appeared to be cross-species reactive as it labeled a rare population of CD8␣ ϩ cells within and around mouse spleen T-cell areas (Fig. 3).
To further characterize mouse 1F12 ϩ cells, splenocytes were isolated from FL-treated mice by density gradient centrifugation, labeled with anti-1F12 scFv, CD11c, CD11b, and CD8␣, and then analyzed by flow cytometry.
1F12 ϩ cells expressed CD11c and CD8␣ but not CD11b. Indeed, all splenic CD11c ϩ CD11b Ϫ CD8␣ ϩ DCs were labeled with anti-1F12 scFv (Fig. 4). PBMCs from FL-treated mice contained CD11c ϩ 1F12 ϩ DCs at a proportion comparable with that found in FL-treated healthy human volunteers (see supplemental Fig. 3). Thus anti-1F12 scFv labels an FL-expanded population of CD11c ϩ CD11b Ϫ DCs found in the blood and lymphoid organs of mice and humans.
1F12 Is a Conserved Epitope in the Extracellular Portion of Nectin-like 2-To identify the surface molecule targeted by anti-1F12, we carried out immunoprecipitation experiments using extracts of DCs generated in vitro from mouse bone marrow progenitors cultured for 9 days in the presence of FL (33).
Immunoprecipitation and Western blotting experiments with anti-1F12 scFv were carried out from in vitro derived DC. Those experiments identified a protein that migrated on SDS-PAGE with apparent molecular masses of 100 kDa in reduced and non-reduced conditions. An additional higher molecular weight band at about 200 kDa was observed in non-reduced samples (Fig. 5A).
MS/MS of the trypsinized major SDS-PAGE zone of this protein identified five distinct peptides. Protein data base searching revealed that the primary sequence of those peptides perfectly matched portions of mouse Necl2 protein.
To confirm that anti-1F12 scFv targets Necl2, a soluble protein containing the extracellular domain of human Necl2 fused to the Fc portion of human IgG1 was produced and the ability of anti-1F12 scFv to bind Necl2-Fc was assayed. Fig. 5B shows that anti-1F12 scFv bound to Necl2-Fc, whereas anti-BDCA3 antibody did not. Thus, anti-1F12 scFv recognizes an epitope in the extracellular portion of Necl2.
An Alternative Counter-structure for Necl2 Is Transiently Expressed on the Surface of Activated Cytotoxic T-lympho-

FIG. 2. Off-target reactivity of BDCA3 ؉ DC surface binders.
Four-hundred individual BDCA3 ϩ DC binders were selected after the second round of selection and used to label PBMC from normal healthy volunteers in multicolor flow cytometry experiments with antibodies against CD1c, BDCA3, and BDCA4. Phage labeling was performed as described for Fig. 3. Monocytes were defined as CD1c Ϫ cells with myeloid cells physical parameters (forward and side scatter (FSC/SSC gating), whereas lymphocytes were defined on physical parameters only. This table shows the five most common patterns of reactivity obtained after the second round of selection. 93% of all BDCA3 ϩ binders presented a reactivity profile comparable with one of the five profiles shown here.
FIG. 3. Anti-1F12 scFv labels human BDCA3 ؉ and mouse CD8␣ ؉ DCs in human and mouse spleen. Human or mouse spleen cryosections were labeled with biotinylated anti-1F12 scFv in combination with anti-BDCA3 antibody and anti-mouse CD8␣, respectively. Alexa Fluor®488-conjugated streptavidin fluorochrome and anti-IgG Alexa Flu-or®555 fluorochrome were then used to reveal anti-1F12 and BDCA3/CD8 labeling respectively. Arrows highlight doubly labeled cells.
FIG. 4. Anti-1F12 scFv specifically label CD8␣ ؉ CD11c ؉ CD11b ؊ DC in unchallenged mouse spleens. Female C57BL6 mice were injected daily intraperitoneally with 10 g of Flt3-L for 9 consecutive days. Splenocytes were then purified by Ficoll gradient centrifugation and labeled with anti-1F12 scFv in combination with fluorophore-conjugated anti-CD8␣, CD11b, and CD11c antibodies before analysis on a FACSCalibur TM flow cytometer.

Necl2/CRTAM Interaction and DCs
cytes-Necl2 is an adhesion molecule known to form cis-and trans-homodimers as well as trans-heterodimers with Necl1 and Nectin-3, two other members of the Nectin/Nectin-like family of surface molecules expressed on epithelial cells and neurons (40,41).
Expression of Necl2 on a rare subset of DCs present in the splenic T-cell area suggests that Necl2 might also be involved in an interaction between DCs and other cells, such as T-lymphocytes.
To test this hypothesis, mouse splenic CD4 ϩ and CD8 ϩ Tlymphocytes were purified and labeled with Necl2-Fc before or after anti-CD3 activation.
As shown in Fig. 6, Necl2-Fc did not label resting mouse CD4 ϩ or CD8 ϩ T-cells as assessed by flow cytometry. In con-trast, Necl2-Fc transiently labeled most anti-CD3-activated CD8 ϩ T-lymphocytes as well as a small subset of CD4 ϩ Tlymphocytes over a 3-day time course experiment.
Necl2-Fc binding was detectable as early as 6 h, peaked at 12 h, and declined steadily until 36 h. We failed to detect Necl2, Necl1, or Nectin-3 mRNA or Necl2 protein expression in activated CD8 ϩ T-lymphocytes at 6 and 12 h of culture (data not shown), thus suggesting that Necl2-Fc bound an alternative counter-structure on the surface of activated CD8 ϩ T-lymphocytes.
Necl2 Binds CRTAM-To further characterize the Necl2 counter-structure on activated mouse lymphocytes, CD8 ϩ Tcells were purified from BL6 mouse spleens and activated for 12 h with anti-CD3 antibodies.
Cell membranes were then solubilized and incubated with Necl2-Fc before immunoprecipitation with protein-A beads. SDS-PAGE analysis of the precipitate revealed a protein with apparent molecular mass of 90 kDa under reducing conditions (Fig. 7A). MS/MS analysis of the trypsinized major SDS-PAGE zone of this protein identified one peptide (ESEISEQALESYR) identical to the sequence contained in mouse CRTAM, a C1-V Ig domain-containing type-I surface protein, remotely related to the Nectin/Nectin-like family and primarily expressed in ␣␤TCR ϩ DN-thymocytes and by activated CD8 ϩ T-cells, NK-T-cells, and NK cells (42).
The extracellular domains of mouse and human CRTAM cDNA were fused to mouse IgG1 Fc and produced as soluble proteins. hCRTAM-Fc protein was then used to label human PBMCs, mouse splenocytes, as well as EL4 thymoma cells stably transduced with lentiviral vectors expressing either Necl1 or Necl2.
Like anti-1F12 scFv, CRTAM-Fc exclusively labeled BDCA3 ϩ DCs and CD8␣ ϩ DCs in human PBMCs and mouse splenocytes, respectively (data not shown). hCRTAM-Fc as well as anti-1F12 scFv were also able to label Necl2 but not Necl1 transduced EL4 cells (Fig. 7C). When used in an enzyme-linked immunosorbent assay format, CRTAM-Fc protein was capable of binding recombinant Necl2 but not Necl1 (Fig. 7B). Thus, Necl2 binds CRTAM, a surface molecule transiently expressed on the surface of activated CD8 ϩ T-lymphocytes.
We tested whether CRTAM binding to Necl2 could interfere with Necl2/Necl2 or Necl2/Necl1 interaction in a cell aggregation assay using EL4, Necl1 ϩ EL4, and Necl2 ϩ EL4. Those three cell lines were labeled with a tracer dye (CFDA-SE) and mixed with equal quantity of unlabeled EL4, Necl1 ϩ EL4, or Necl2 ϩ EL4 cells. Cell aggregation was then monitored by ob- FIG. 5. Anti-1F12 scFv binds to Necl2 extracellular domain. A, surface biotinylated cells were lysed in Triton X-100-containing buffer. Lysate was then incubated with anti-1F12 scFv-Fc or unrelated scFv-Fc before adding protein-A beads. Precipitates were analyzed after electrophoresis on 4 -20% Novex® Tris-glycine gel and transfer on nitrocellulose membrane. Presence of surface glycoproteins was then revealed after incubation with streptavidin-HRP. B, Necl2-Fc (huIgG1) or unrelated ctrl-hFc were dotted on the nitrocellulose membrane. Biotinylated anti-1F12 scFv, anti-BDCA3 antibody, or anti-huIgG antibodies were dotted as indicated. After membrane saturation, binding was revealed by adding strepatividin-HRP and peroxidase substrate. 6. A counter-structure for Necl2 is transiently expressed on activated CD8 ؉ T-cells. T-lymphocytes were negatively selected from the spleens of C57BL6 mice by MACS® and cultured for the indicated length of time on anti-CD3 coated microtiter plates. Cells were then harvested and labeled with either Necl2-Fc or Necl1-Fc (negative control) together with anti-CD8 FITC and anti-CD4 APC antibodies. PE-conjugated anti-huIgG1 antibody was used to reveal human Fc. CD4 ϩ and CD8 ϩ T-cells were electronically gated after exclusion of dead cells with propidium iodide. The graph shows labeling of CD4 and CD8 cells with Necl1-Fc (dotted line) or Necl2-Fc (full line).

Necl2/CRTAM Interaction and DCs
serving the presence of CFDA-SE ϩ cell or CFDA-SE ϩ /unlabeled cells mixed aggregates before and after addition of CRTAM-Fc.
Thus, CRTAM efficiently disrupts Necl2/Necl2 homotypic and Necl2/Necl1 heterotypic interactions. Those results suggest that the CRTAM binding site on Necl2 might be partially overlapped with that used by Necl2 to engage other Nectin/ Nectin-like family members.
Necl2 on Antigen-presenting Cells Regulates IL-22 Expression by Activated CTL-CRTAM is expressed on all known cytolytic lymphocytes. In addition, Necl2 is restricted to CD8␣ ϩ DCs, a subset characterized by its ability to prime CTL as well as NK immunity to viruses, in vivo (4,10,11). Those results suggested to us that Necl2, on antigen-presenting cells, might modulate CD8 ϩ T-cell effector function.
To assess whether Necl2 expression on antigen-presenting cells could alter CD8 ϩ T-cells acquisition of cytolytic function, EL4 and Necl2 ϩ EL4 cells were loaded with predetermined optimal concentrations of MHC Class-I (H-2Kb) restricted 357-364 chicken OVA peptide and then cocultured with OVA-specific TCR-transgenic CD8 ϩ T-lymphocytes. The cytolytic activity of CD8 ϩ T-cells was then assessed in this coculture system by monitoring viability of EL4 cells in a time course experiment from 12 to 72 h (see "Materials and Methods").
Although most OVA peptide-loaded EL4 or Necl2 ϩ EL4 cells were lysed by activated OT-I cells within 48 -72 h in culture, Necl2 expression by EL4 cells did not alter the kinetic or the extent of cell lysis (data not shown).
Thus, Necl2 does not seem to alter in vitro differentiation of naïve CD8 ϩ T-cell into CTL. However, Necl2 participates in the cross-talk between by an antigen-presenting cell and CD8 ϩ T-cells by regulating IL-22 expression in activated lymphocytes. DISCUSSION In an attempt to identify surface determinants expressed on a rare CD11c ϩ CD11b Ϫ subset of human blood DCs but not on FIG. 7. Necl2 binds to CRTAM. A, SDS-PAGE analysis of Necl2-Fc immunoprecipitate. CD8 ϩ T-lymphocytes were purified from C57BL6 mouse spleen and activated for 12 h on anti-CD3-coated plates. Cells were then harvested, surface-biotinylated, and solubilized with detergent. Necl2-Fc and/or Necl1(ctrl)-Fc used with protein-A-coupled beads were then added. Western blot analysis of Necl2-Fc precipitate revealed a major glycoprotein at 90 kDa B, direct interaction between Necl2 and CRTAM. 10 g/ml Necl1-Fc or Necl2-Fc were coated in PBS on 96-well microtiter plates. CRTAM-mFc, mIgG1 (negative control), or biotinylated anti-1F12 scFv (positive control) were then added. CRTAM and anti-1F12 binding was revealed after addition of anti-mIgG1-HRP and streptavidin-HRP, respectively. C, CRTAM binds to membrane-bound Necl2 but not Necl1. EL4 thymocytes were stably transduced with lentiviral vectors expressing Necl1 or Necl2 before labeling with CRTAM-mFc ϩ anti-mIgG1-PE (thick line), anti-1F12 scFv ϩ anti-huIgG1-PE (thin line), or anti-Necl1 antibody ϩ anti-mIgG1 PE (dashed line).
other blood leukocytes, we have selected a single scFv phage antibody (anti-1F12), which targets an extracellular epitope of Necl2.
Necl2, also known as Tslc1, SynCaM, SgIGSF, or Igsf4, is a highly conserved type-I transmembrane glycoprotein of 442 amino acids. It has an extracellular portion containing three immunoglobulin-like domains and a cytoplasmic portion harboring two protein-protein interaction motifs: a 4.1/ezrin/radixin/moesin (FERM) domain and a PDZ domain. These two motifs act as docking sites for DAL-1, MPP6(PALS2), and MPP3, three adaptor proteins that connect surface proteins to the actin cytoskeleton (40,43,44).
Necl2 has been previously characterized as an adhesion mol-ecule of the epithelium junctional complex that also participates in the assembly of specialized cell/cell junctions such as neuronal synapses (45) or apical ectoplasmic specializations at the Sertoli cell/spermatid interface (46,47). Consistent with its role in junctional complex assembly, down-regulation of Necl2 expression is associated with oncogenic transformation of epithelial cells (see Ref. 48 for review). Anti-1F12 scFv labels a population of DCs present in the T-cell area of both human and mouse lymphoid organs suggesting that tissue distribution of 1F12 ϩ DCs is conserved in mammals. In hindsight, cross-species reactivity of anti-1F12 scFv is not surprising, since Necl2 is highly conserved throughout evolution with mouse and human Necl2 being 97% identical at FIG. 8. A and B, soluble CRTAM disrupts Necl2/Necl2 homotypic and Necl2/ Necl1 heterotypic interactions. EL4, Necl1 ϩ EL4, or Necl2 ϩ EL4 were labeled as indicated under "Material and Methods" and mixed with equal number of unlabeled cells and cultured at 5 ϫ 10 5 cells/ ml. In B, hCRTAM-Fc (1 g/ml) was added at the culture initiation. C, Necl2 on antigen presenting cells regulates IL-22 expression by activated CTL. CD8 ϩ T-cells were purified from female OT-I TCR-Tg mouse spleens by MACS®-negative selection and then cocultured for 3 days with irradiated OVA peptide-pulsed EL4 or Necl2 ϩ EL4 cells as described under "Materials and Methods." 5 g/ml CR-TAM-Fc was added at the culture initiation where indicated. OT1 T-cells were purified by MACS®-positive selection before mRNA extraction. IL-22 and hypoxanthine phosphoribosyltransferase (HPRT) mRNA expression were then assessed by quantitative PCR. Results are representative of three independent experiments. the protein level. In earlier studies, high throughput sequencing or gene array profiling of the major mouse DC subsets had revealed preferential Necl2 mRNA expression in mouse CD11c ϩ CD11b Ϫ (CD8␣ ϩ ) DCs (49). 2 However, attempts to generate antibodies against recombinant Necl2 through immunization had been unsuccessful, presumably because of the high degree of conservation of Necl2, and therefore we could not formally confirm Necl2 protein expression on the surface of DCs.
When compared with other mouse splenic DCs, CD8␣ ϩ DCs display the highest turnover rate (50,51). Their residence in the spleen does not exceed 3 days. It has been proposed that these DC might either arrive in the spleen directly from the blood stream or derive from an endogenous precursor population (50).
We found that: 1) Necl2 ϩ DCs are present in the blood and in bone-marrow FL-cultures; 2) Necl2 ϩ DCs are not only found within the T-cell area but also in association with the periarteriolar lymphoid sheath of lymphoid organs; 3) human Necl2 ϩ DCs, unlike other blood DC subsets, do not produce sialyl-Lewis x, a complex sugar required for adhesion to E-and P-selectin and for the circulation of leukocytes across the vascular endothelium (52).
Assuming that Necl2 can be used as a tracer for a unique DC subset in both mouse and human, our results would be compatible with the idea that, at steady state, mouse splenic CD8␣ ϩ DCs and human splenic BDCA3 ϩ DCs come from a blood DC population that does not transmigrate through the vascular endothelium but has the capacity to cross high endothelium venules to enter the spleen directly from the blood stream.
Although CD8␣ ϩ DCs and human BDCA3 ϩ DCs share a common CD11c ϩ CD11b Ϫ Necl2 ϩ phenotype, are both expanded in vivo by FL injection, and exhibit comparable abundance and tissue distribution, the question remains as to whether these cells serve similar functions.
Many aspects of the biology of DC subsets remain unsettled, including their ontogeny, function, cytokine production potential, and antigen presentation capacity. Accordingly, only few in vitro assays can be applied to characterize functional activities of DCs in vitro (discussed in Ref. 49). A consensus is emerging concerning mouse splenic CD8␣ ϩ DCs in that they seem to be able to engulf apoptotic bodies in vitro more efficiently than CD8␣ Ϫ DCs (53,54). Although it would be interesting to compare the ability of mouse and human splenic Necl2 ϩ DCs to engulf and cross-present antigens to CTL, establishing functional equivalence between those two subsets would certainly require a larger panel of in vitro assays.
When present on antigen-presenting cells, Necl2 stimulates IL-22 expression in activated CD8 ϩ T-cells. IL-22 is a recently discovered cytokine mostly expressed by activated NK and T-cells (55,56) that promotes the innate immunity of tissues (57). Although the significance of Necl2-mediated regulation of IL-22 expression remains unknown, this result demonstrates that Necl2 serves immunoregulatory functions. We are now assessing whether Necl2 expression correlates with the ability of DCs to trigger in vitro IL-22 production by CD8 ϩ T-cells.
We show here that Necl2 extracellular domain binds CR-TAM, an Ig superfamily member transiently expressed on the surface of activated CTLs. Although CRTAM structure (and notably its N-terminal Ig domain) shows some similarity to that of Necl2 (58), the CRTAM cytoplasmic tail does not contain any of the protein interaction motifs commonly found within Nectin-or Nectin-like cytoplasmic domains. Instead, the CR-TAM cytoplasmic tail contains three conserved phosphorylation consensus sites: two potentially modified by protein kinase C and one by casein kinase II (42). Thus, CRTAM structural features are more akin to those of signaling surface receptors that recruit serine/threonine kinases than they are to surface adhesion molecules.
We suggest here that Necl2 and CRTAM form a specialized cognate pair of surface molecules likely involved in the crosstalk between Necl2 ϩ DC-and CRTAM-expressing cells. In mouse thymus, the Necl2/CRTAM pair could participate in the cross-talk between CD8␣ ϩ DCs and thymocytes involved in the selection of effector cells (59). Necl2/CRTAM interaction could also influence priming of mature CD8 ϩ T-cells, NK T-cells, or NK cells by DCs within and outside lymphoid organs.
Because CRTAM appears as a signaling receptor whose expression within mature lymphocytes includes all known cytolytic cells, we hypothesized that Necl2/CRTAM interaction might regulate cytotoxic activity. Consistent with this hypothesis, genetic linkage analysis in C3H/HeJ mice has revealed that crtam is linked to a minor susceptibility locus for Alopecia areata, a cell-mediated autoimmune disease that targets actively growing (anagen) hair follicles (60). However, in our hands, MHC Class-I OVA peptide-loaded Necl2 ϩ and Necl2 Ϫ EL4 cells appeared equally susceptible to in vitro lysis by naïve OVA-TCR transgenic OT-I T-cells suggesting that CRTAM/ Necl2 interaction does not directly regulate CD8 ϩ T-cells cytotoxic function in vitro.
Necl2 is believed to mediate tumor suppression by inhibiting tumor growth, reducing motility (61), and promoting apoptotic cell death (62). When introduced into lung cancer cells, necl2 suppresses tumorigenicity in nude mice (43,62,63). Inversely, suppression of Necl2 expression through promoter hypermethylation and loss of heterozygosity is associated with oncogenic transformation (64) and correlates with the biologic aggressiveness of tumors (61,65).
Our results raise the intriguing possibility that disruption of Necl2 expression at the tumor site, by interfering with tissue immunity, could also provide tumor cells with a survival advantage. Indeed, suppression of Necl2 expression could affect local IL-22 production by resident CD8 ϩ T-lymphocytes or NK cells, thus compromising the local innate immune response and favoring immune evasion.
Experiments using NK-depleted animals are warranted to formally address whether an immune system component might contribute to Necl2 tumor suppression activity in vivo.
Finally, cell aggregation assays have revealed that CRTAM competes with nectin-like proteins for binding to Necl2. Thus, by displacing interactions between Necl2 and other Nectin/nectinlike family members, CRTAM might regulate cell/cell contact between two neighboring Nectin/Nectin-like expressing cells.
CRTAM-mediated regulation of Necl2/Necl2 or Necl2/Necl1 interaction could have implications beyond the DC/cytotoxiclymphocytes cross-talk. It could, for example, influence the ability of activated (CRTAM ϩ ) lymphocytes to infiltrate epithelial layers. Outside of the immune system, autonomous CR-TAM expression by pachytene 2 spermatocytes (66) could influence Necl2-dependent interactions with Sertoli cells or other spermatogenic cells.
By combining whole cell panning phage display and mass spectrometry, we have identified Necl2 as a unique, evolutionary conserved, DC subset-specific surface marker. Through interaction with CRTAM, Necl2 might participate in the crosstalk between DCs and activated cytolytic lymphocytes and could regulate innate immunity of tissues. Necl2/CRTAM pair could also participate in a large panel of cell/cell interactions both within and outside of the immune system.