A Functional γδTCR/CD3 Complex Distinct from γδT Cells Is Expressed by Human Eosinophils

Background Eosinophils are effector cells during parasitic infections and allergic responses. However, their contribution to innate immunity has been only recently unravelled. Methodology/Principal Findings Here we show that human eosinophils express CD3 and γδ T Cell Receptor (TCR) but not αβ TCR. Surface expression of γδTCR/CD3 is heterogeneous between eosinophil donors and inducible by mycobacterial ligands. Surface immunoprecipitation revealed expression of the full γδTCR/CD3 complex. Real-time PCR amplification for CD3, γ and δ TCR constant regions transcripts showed a significantly lower expression in eosinophils than in γδT cells. Limited TCR rearrangements occur in eosinophils as shown by spectratyping analysis of CDR3 length profiles and in situ hybridization. Release by eosinophils of Reactive Oxygen Species, granule proteins, Eosinophil Peroxidase and Eosinophil-Derived Neurotoxin and cytokines (IFN-γ and TNF-α) was observed following activation by γδTCR-specific agonists or by mycobacteria. These effects were inhibited by anti-γδTCR blocking antibodies and antagonists. Moreover, γδTCR/CD3 was involved in eosinophil cytotoxicity against tumor cells. Conclusions/Significance Our results provide evidence that human eosinophils express a functional γδTCR/CD3 with similar, but not identical, characteristics to γδTCR from γδT cells. We propose that this receptor contributes to eosinophil innate responses against mycobacteria and tumors and may represent an additional link between lymphoid and myeloid lineages.


Introduction
Eosinophils are polymorphonuclear granulocytes mainly found in increased numbers during helminth parasitic infections and allergic reactions [1,2]. They are classically considered as mediator-releasing cells during effector phase of adaptive immunity, under the influence of T cell dependent cytokines or chemokines and antibodies [2], whereas eosinophil-derived chemokines have been recently shown to locally attract Th2 lymphocytes at lung inflammatory sites [3,4]. Nevertheless, their precise function as beneficial or deleterious to the host still remains ambiguous, since highly toxic proteins present in eosinophil granules exert potent cytotoxic effects against non self targets such as parasites [5,6] but also against stressed or necrotic host cells [7] and in asthma [8]. Eosinophils are foremost present in mucosal tissues in contact with the environment such as in gastro-intestinal tract and skin [2] and are characterized by their wide morphological and functional heterogeneity [9].
However, the early appearance of eosinophils in agnathans, predating the appearance of the classical adaptive immune system [20] and the expression by eosinophils of several receptors involved in innate immunity, such as formyl peptide receptor [21], protease-activated receptors [22,23] and TLR [24] further point toward a role for eosinophils in innate immunity. Eosinophils contribute to TLR-mediated immunity against viruses and mycobacteria [25,26]. Indeed, we recently showed that TLR-2-dependent activation of human eosinophils induced a-defensin and ECP release and decreased mycobacteria growth [24]. Furthermore, expulsion of mitochondrial DNA by eosinophils has been shown to contribute to innate immune defences against bacteria [27]. Finally, eosinophil-tumor cell interactions and IL-4dependent tumoricidal activity of eosinophils have been reported [28,29]. Thus eosinophils appear functionally located at the interface between innate and adaptive immunity.
Strikingly, cdT cells are ancestral to other lymphoid populations such as abT cells and B cells, they participate to both innate and adaptive immune responses, have a preferential mucosal localisation and might act as professional antigen-presenting cells [30] recognizing non-peptide antigens found on several pathogens, including mycobacteria, necrotic and tumor cells [31,32].
These surprising similarities between cdT cells and eosinophils prompted us to investigate, whether, in addition to other T cellassociated receptors, human eosinophils expressed a cdT cell Receptor (TCR). We here report that human blood eosinophils express low levels of surface cdTCR/CD3, inducible by mycobacterial ligands. We show that eosinophils release granule proteins and cytokines upon activation by cdTCR agonists, including mycobacteria. Furthermore, we provide evidence that cdTCR contributes to eosinophil cytotoxic potential towards tumors.

Human eosinophils express CD3 and cdTCR but not abTCR
In order to investigate expression by human blood eosinophils of T cell associated receptors, we purified eosinophils by negative selection using magnetic beads. These highly purified eosinophils ( Figure S1A) expressed specific granule proteins such as eosinophil peroxidase (EPO) and major basic protein (MBP) but not myeloperoxidase (MPO) associated to neutrophil and monocyte/ macrophage lineages [33] ( Figure S1B). Lymphocytes expressed neither of these myeloid markers ( Figure S1B).
Following permeabilization, binding of anti-CD3 but not anti-CD8 antibodies was detected in eosinophils ( Figure 1A). In T cells, CD3 associates with either abTCR or cdTCR. We did not detect abTCR on eosinophils but cdTCR expression was evidenced ( Figure 1A). While lymphocytes from PBMC fraction expressed these markers at the expected frequencies ( Figure 1B) and unlike a previous report [34], we were unable to detect CD3 or abTCR expression in neutrophils ( Figure 1C). Likewise neither CD8 nor cdTCR expression was detected in neutrophils ( Figure 1C).
Surface expression of CD3 and cdTCR was detected on a fraction of eosinophils following double staining with antibodies against two specific eosinophil granule proteins EPO (Figure 2A) or MBP ( Figure 2B). Similarly to intracellular staining, neither CD8 nor abTCR were detected at eosinophil surface. By contrast, another cdT cell marker, NKG2D, was expressed in a lower proportion of cells (Figure 2A-B). As CD3 is required for receptor complex surface expression, CD3 + cdTCR + eosinophils were thus identified following triple staining ( Figure 2C). In human blood, cdT cells express either Vc9/Vd2 or Vd1 variable regions. We also evidenced that EPO + or MBP + eosinophils expressed Vc9, Vd1 and Vd2 and unexpectedly co-expressed Vd1 and Vd2 at their surface (Figure 2A-C).
Expression of cdTCR/CD3 complex was investigated in healthy donors and patients with either reactive eosinophilia (allergies, various skin diseases) or hypereosinophilic syndrome (HES). A wide heterogeneity of surface expression of cdTCR/ CD3 and NKG2D was not only found among the 23 donors but also within each group, including healthy donors, however no correlation was observed between receptor expression levels and either eosinophilia, or a specific pathology ( Figure 3A-B).
Following in vitro activation of cdT cells, CD3, CD8, cdTCR as well as Vc9, Vd2 and Vd1 were detected at the expected frequencies ( Figure S2A). Likewise, freshly isolated PBMC from the same eosinophil donors, in which cdT cells represent a very minor population, expressed CD3, abTCR or low levels of cdTCR ( Figure  S2B). By contrast, monocytes did neither express CD3 nor cdTCR ( Figure S2C). Finally, neutrophils neither expressed CD3, cdTCR nor the full abTCR complex at their surface ( Figure S2D). Thus, following overnight culture, a fraction of eosinophils, unambiguously identified by the presence of specific eosinophil markers, express the cdTCR/CD3 complex at their surface.
Besides receptor reexpression at the membrane following a resting period, as demonstrated for NKG2A on cdT cells [35], ligand-dependent induction or upregulation of receptor surface expression has been widely reported, including in the case of FceRI [9]. Thus, we investigated whether a specific cdTCR ligand was able to induce cdTCR/CD3 surface expression on human eosinophils. While CD3 and cdTCR were barely detected at the surface of freshly purified eosinophils ( Figure S3A), upon incubation for 2 h with TUBag1, a natural non peptidic ligand from Mycobacterium tuberculosis [36], about 50% of eosinophils expressed CD3 and cdTCR at their surface ( Figure S3B). By contrast, neither CD8 nor abTCR expression was induced upon incubation with TUBag ( Figure S3B).
Finally, besides mature eosinophils present in peripheral blood, eosinophils can be differentiated in vitro from CD34 + cord blood progenitors. In such EPO + eosinophils, in the absence of possible contamination by cdT cells, cdTCR/CD3 complex was detected at cell surface, after 3 weeks differentiation ( Figure S4). As for blood eosinophils, neither CD8 nor abTCR (not shown) could be detected.
Eosinophils express all the subunits from cdTCR/CD3 complex To evidence the cellular distribution of cdTCR/CD3 complex and to further exclude that results obtained by flow cytometry could be due to contaminating cdT cells, blood eosinophils were analysed at low magnification by immunofluorescence and at high magnification by confocal microscopy. In full agreement with flow cytometry data, in a preparation without contaminating cells, only a fraction of blood eosinophils, identified by their typical binucleated morphology, expressed the four chains of CD3 and cdTCR. No signal was detected with anti-aTCR, anti-bTCR antibodies or control IgG (Fig 4A). We next characterized the structure of the cdTCR/CD3 complex expressed at the surface of eosinophils, using surface biotinylation, followed by lysis with a mild detergent, coimmunoprecipitation with an anti-cdTCR antibody, SDS-PAGE in reducing conditions and detection of biotinylated proteins. Bands corresponding to TCR, CD3d, e, c and f subunits were detected at the expected molecular weights both on cdT cells and eosinophils ( Figure 4B). This result indicates that all the necessary chains are present to allow for surface expression of a complete cdTCR/CD3. However, the amount of material used for eosinophils was 10 times higher than for cdT cells, results consistent with flow cytometry data showing much lower surface expression in eosinophils than in cdT cells ( Figure 2 and Figure S2). cdTCR/CD3 is less abundant and more restricted in eosinophils than in cdT cells Transcripts specific for the four chains (e, f, c and d) of CD3 complex, cTCR and dTCR constant regions and CD8 were   (A-C) CD3, cdTCR, Vc9, Vd2, Vd1, abTCR, CD8 and NKG2D surface expression on EPO + (A) and MBP2 + (B) purified peripheral blood eosinophils analysed after 18 h culture and gated as described in Figure S1. Results in (A) and (B) represent to 2 distinct representative donors. Staining with control isotype antibodies is represented. (C) Presence of CD3 + cdTCR + EPO + and Vd1 + d2 + EPO + eosinophils identified by triple staining. Staining with control isotype matched antibodies as well as corresponding double stainings is represented. doi:10.1371/journal.pone.0005926.g002 amplified and quantified by real-time PCR in eosinophils purified to homogeneity (100%) and cdT lymphocyte subsets or Colo-205 colon carcinoma cells as positive and negative controls respectively. Expression of the various chains was 300 to 4,000 fold lower in eosinophil population than in cdT lymphocyte population, while CD8 expression was absent in eosinophils but present in cdT lymphocytes ( Figure 5A). Amplification of Vd1-Jd3, Vd2-Cd and VcI-JcP rearranged transcripts but not Vd1-Jd4 was also detected Immunofluorescence and confocal microscopy (insets) analysis of cdTCR/CD3 chains expression on cytospin eosinophil preparations after 18 h culture. Staining with anti-aTCR, anti-bTCR antibodies and control goat IgG is represented. Arrowheads indicate some positive cells. (B) Surface of eosinophils and cdT cells was biotin-labelled. Cells were lysed and complexes were immunoprecipitated, using anti-cdTCR or isotype control antibodies. Immunoprecipitated proteins were resolved on a reducing 14% SDS-PAGE and transferred to PVDF membrane. Biotinylated proteins were revealed using ABC-HRP and chemiluminescence. Positions of the TCR and CD3 subunits are marked. Material corresponding to the total and to 1/ 10 th of the material was loaded for eosinophils and cdT cells respectively. Dashed lines indicate that non adjacent lanes of the same gel have been joined on the Figure  in eosinophils. These rearrangements were all found in cdT lymphocytes but neither in abT lymphocytes nor Colo-205 cells ( Figure 5B). Spectratyping analysis of CDR3 length profiles for Vc9-Jc1/2, Vc9-JcP and Vd2-Cd evidenced a limited Vd2-Cd diversity (2 peaks) on eosinophils compared to cdT cells ( Figure 5C), while no signal was detected on eosinophils for the 2 other rearrangements, in contrast to cdT cells (data not shown). Furthermore, to unambiguously demonstrate that TCR rearrangements occur in eosinophils, at single cell level, we performed in situ hybridization using sequences from rearranged TCR genes as RNA probes. Hybridization with anti-sense probes corresponding to Vd2-Cd and VcI-JcP rearrangements, that were detected by RT-PCR, gave a positive signal on nuclei from eosinophils at times clearly bilobed. In keeping with RT-PCR results, a significantly stronger signal was detected on cd T lymphocytes, while corresponding sense probes only gave background amplification on both cell types (Fig. 5D). Vd1-Jd4 anti-sense probe gave positive signal on cd T lymphocytes but not on eosinophils in which this rearrangement was not detected by RT-PCR.
Finally, DNA sequencing of Vd1Jd3 Vd2-Cd regions from eosinophils and cdT cells show that V and J regions were virtually identical to the IMGT sequence, while significant differences between eosinophils and 2 different cloned sequences of cdT cells from the same donor were found in D regions and in junctional regions ( Figure 6). This further confirms that cdTCR/CD3 complex expressed by eosinophils is less expressed and diverse than the corresponding receptor on cd T cells.
cdTCR/CD3-mediated eosinophil activation induces ROS production, degranulation and cytokine release Upon activation, eosinophils very rapidly produce Reactive Oxygen Species (ROS) and release, by selective degranulation, eosinophil-specific granule proteins, including highly cytotoxic cationic proteins such as EPO and eosinophil cationic protein (ECP) [6]. In purified blood eosinophils, cdTCR/CD3 complex activation by immobilized antibodies to CD3, cdTCR or Vd1 led to a similar kinetics of ROS production ( Figure 7A) and to a significant EPO and eosinophil-derived neurotoxin (EDN) release ( Figure 7D and G). Stimulation with a soluble ligand, bromohydrin pyrophosphate (BrHPP), a phosphoantigen agonist selective for human c9d2 + T lymphocytes [37], induced dose-dependent ROS production as well as EPO release ( Figure 7B and E). Furthermore, eosinophil activation by sec-butylamine (SBA) [38], an inducer of endogenous phosphoantigens, yielded to a sustained dose-dependent ROS and EPO production ( Figure 7C and F), comparable to values obtained following activation with IgA/anti-IgA immune complexes, one potent physiological eosinophil activator. Of note, plate-bound antibodies induced stronger ROS release but weaker EPO release than soluble cdTCR activators. Since eosinophils also release several immuno-regulatory and proinflammatory cytokines [10], we further show that activation with anti-CD3, -cdTCR or -Vd1 antibodies efficiently induced IFN-c and TNF-a production by eosinophils ( Figure 7H and I). Thus, cdTCR/CD3-expressing eosinophils respond to selective agonists similarly to human cdT lymphocytes and produce both specific eosinophil granule proteins, myeloid cell mediators as well as cytokines.

Mycobacterium bovis induce cdTCR/CD3-dependent eosinophil activation
Human cdT cells respond to non-peptide antigens including mycobacterial and tumor-derived ligands. Patients with mycobacterial infections may exhibit blood and tissue eosinophilia and mycobacteria-induced eosinophil-associated experimental acute inflammation [39] as well as rapid eosinophil chemotaxis in vivo. Mycobacteria-eosinophil interactions might thus reflect the reactivity conferred by cdTCR/CD3 complex. Accordingly, incubation of blood eosinophils with increasing ratios of Mycobacterium bovis-BCG induced dose-dependent ROS production and EPO release ( Figure 8A-B), that could be inhibited, in a dosedependent manner, by an anti-cdTCR antibody [40] ( Figure 8C-D). Similar eosinophil activation, and inhibition by an anti-cdTCR antibody, was also obtained following incubation with TUBag, a natural mycobacterial component ( Figure 8E-H). Thus mycobacteria appear capable of activating eosinophils, at least partly, in a cdTCR-dependent manner.
cdTCR/CD3-dependent eosinophil cytotoxicity toward tumor cells Eosinophils have been associated with many tumors in vivo. Furthermore, both cd T cells [41] and eosinophils [19] exert potent cytotoxic activity towards many tumor cells. Therefore, we investigated cdTCR-mediated eosinophil cytotoxicity towards Colo-205 carcinoma cell line. Indeed, eosinophils induced timedependent apoptosis of Colo-205 tumor cells in vitro, which was significantly inhibited following addition of neutralizing anti-cdTCR Ab ( Figure 9A). This effect was more prominent at earlier time-points suggesting that cdTCR-mediated eosinophil-tumor interactions are important for the initiation of the cytotoxic reaction ( Figure 9B). Finally preincubation of eosinophils with cdTCR ligands, SBA and TUBag, potentiated eosinophil cytotoxicity towards tumor cells ( Figure 9C-D). Altogether, these data confirm the potential role of eosinophils in anti-tumor cytotoxicity and strongly suggest the involvement of cdTCR/CD3 complex in this mechanism.

Discussion
We here provide evidence that highly purified human blood eosinophils, as well as eosinophils differentiated in vitro from CD34 + cord blood progenitors, gated on the basis of presence of specific eosinophil granule proteins, express a functional cdTCR/ CD3 complex, so far almost exclusively associated to T lymphocytes, but lack abTCR. As previously observed for other receptors such as CD25 [42], CD4 [18], FceRI, CD89 and CD28 [10], cdTCR/CD3 surface expression on eosinophils is highly heterogeneous between individual donors. Such an heterogeneity in cdTCR/CD3 expression has also been observed for human blood cdT lymphocytes [35]. As it is the case for several receptors, including FceRI [9], surface expression of cdTCR/CD3 can be  Table S1). (B) TCRc and d rearrangements in peripheral blood eosinophils analysed by RT-PCR using primers listed in Table S2. CD8 amplification is performed to exclude lymphocyte contamination. Colo-205 are chosen as negative control (C) Spectratyping analysis of CDR3 length profiles for Vc9-Jc1/2 ; Vc9-JcP and Vd2-Cd families in eosinophils and in cd T cells from the same donor. (D) Expression of rearranged c and dTCR at the single cell level in peripheral blood eosinophils and cdT cells detected by in situ hybridization using anti-sense and control sense probes specific for Vd2-Cd, VcI-JcP and for Vd1-Jd4 used as negative control. doi:10.1371/journal.pone.0005926.g005 induced/up-regulated, in particular upon incubation with ligands. Our results show that cdTCR surface expression was induced upon incubation in the presence of a natural mycobacterial ligand as well as following overnight culture. The molecular mechanism underlying this later finding remains unknown. Our results indicate that CD3e expression, both at mRNA and protein levels, was consistently lower than cdTCR. Due to the requirement of this CD3 subunit for surface expression of the receptor complex, it is tempting to speculate that the very low CD3e expression probably accounts for the low surface expression of the whole complex on eosinophils although further experiments would be required for a formal demonstration. This low receptor expression as well as the fragility of eosinophils was a constraint to obtain significant amounts of viable cdTCR + eosinophils following cell sorting.
Eosinophils display a rearranged cdTCR with a more restricted repertoire compared to cdT cells, further excluding the possibility of lymphocyte contaminants in eosinophil preparations. Besides T lymphocytes, dTCR rearrangements have only been reported in malignant B lymphoblasts [43], NK cells [44] and in mixed-type leukaemia retaining both T-lymphoid and myeloid characteristics. Furthermore, presence of double positive Vd1 + Vd2 + and equal numbers of Vd1-and Vd2-expressing eosinophils in most of normal donors suggests incomplete allelic exclusion as found in some instances in cdT cells [45]. Direct cdTCR activation of eosinophils by cdTCR/CD3-specific antibodies and selective agonists induced eosinophil-specific responses. Release of such myeloid-associated mediators and the scarcity of circulating cd T cells further exclude the possibility that our results obtained using highly purified eosinophils would be due to cdTCR expression by contaminating lymphocytes.
Thus, mature eosinophils expressing lymphoid markers might represent hitherto unattended cdTCR-expressing myeloid cells. Our results might be in agreement with observations that myeloidderived plasmacytoid DC are able to express a ''lymphoid program'' (RAG1, rearranged IgH genes) [46]. By contrast, in the recently proposed myeloid-based model of haematopoiesis, Tor B-committed progenitors would each keep a myeloiddifferentiation potential [47]. For instance, genetic reprogramming from differentiated B cells to macrophages has been reported [48]. Further experiments are required to unravel by which mechanisms eosinophils have acquired this lymphoid program. Of note, eosinophils are found in human thymus [8]. Furthermore, similarly to gut intraepithelial cdT cells, eosinophils are able to undergo peripheral differentiation, for instance in allergic lungs [49]. Investigating TCR expression by eosinophils in various organs and tissues would greatly benefit from experiments on animal models. Unfortunately, we were unable to detect cdTCR/ CD3 on mouse blood and tissue eosinophils. This is not extremely surprising since mouse eosinophils are strikingly different from human eosinophils, regarding their granular content [50] and membrane expression of various immune receptors, in particular IgE receptors [51] and IgA receptors [52]. Furthermore, cdT cells also display significant differences between mouse and humans. Indeed, a high proportion of skin lymphocytes express cdTCR in mouse but not in human [53], while only human Vc9Vd2 are responsive to phosphoantigens [54].
Our findings indicate that eosinophils might also contribute, through cdTCR-dependent mechanisms, to immune defences in tissues, where 90% eosinophils are located. One might speculate that cdTCR/CD3 expression is higher in tissue than in blood eosinophils, as previously reported for FceRI in a model of ''humanized'' mice where human FceRI was quantified and increasingly expressed from blood to peritoneal cavity and spleen eosinophils [9]. Tissue eosinophils are associated to granulomatous diseases such schistosomiasis [55], Crohn's disease [56] and mycobacterial infections [39]. Eosinophils can damage cell wall and lyse Mycobacterium tuberculosis in vitro [57]. Vc9d2 and Vd1 expression, rapid response to phosphoantigen ligands, as well as BCG-induced activation so far represented hallmarks of cdT lymphocyte-mediated cytotoxicity towards phosphoantigen-expressing mycobacteria [58]. Since tumors produce phosphoantigens and are targeted by cdT lymphocytes [58], cdTCR-mediated activation might thus contribute to the anti-tumoral activity of eosinophils. Indeed, while cdT lymphocytes and NK cells display granzyme B/perforin-mediated cytotoxicity, eosinophils additionally possess a unique set of cationic proteins with a strong cytotoxic potential [50]. Furthermore, tumor-associated tissue eosinophilia (TATE) is considered by some authors as a marker of good prognosis, including in colon carcinoma [59]. Anti-tumoral activity of eosinophils should be considered in the context of current development of innovative cancer immunotherapies, based on synthetic phosphoantigens [60]. Indeed, due to their high cytotoxic potential and despite their low cdTCR expression in comparison to T cells, eosinophils might thus represent very efficient ''innate'' effector cells in defence against targets bearing cdTCR ligands, such as tumor cells, in particular when eosinophils surround or infiltrate solid tumors.
Our work outlines a previously unexpected link between eosinophils and cdT lymphocytes, which not only share the expression of the cdTCR/CD3 complex but also some major functions in innate immunity. It might also shed light on the

Ethics statement
Peripheral venous blood was obtained from healthy donors or eosinophilic patients after written informed consent. Research protocol was approved by the Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale de Lille (France).

Cell purification and culture
Eosinophils were isolated as described [61] on Percoll followed by negative selection using anti-CD16-coated microbeads. Purity determined after cytospin and RAL555 staining was above 98% and 100% for RNA analyses. Purified eosinophils were cultured overnight at 37uC in complete RPMI without phenol red.
Cord blood was obtained from the maternity unit of Lille University Hospital. Mononuclear cells were isolated by Ficoll-Hypaque density centrifugation. CD34 + cells were purified using CD34 + isolation kit (Miltenyi). Purified CD34 + cells were cultured in complete medium supplemented with 2.5610 25 M b-mercaptoethanol (b-ME) at 0.5610 6 cells/ml. Eosinophil differentiation was induced upon addition of 50 ng/ml SCF, 1 ng/ml each GM-CSF, IL-3 and IL-5 (Peprotech).
Colo-205 colon carcinoma cell line was obtained from ATCC. Cells were maintained in RPMI 1640 medium supplemented with 10% FCS, 10 mM L-Glutamine and 10 mg/ml Gentamycin.

Flow cytometry
Antibodies and isotype controls used for flow cytometry (origin, clone, working dilution or concentration) are listed in Table S1. Anti-EPO (AHE) and anti-MBP2 [62] (a kind gift from Dr. D Plager, Mayo Clinic, Rochester) were biotinylated using biotinsulfosuccinimidyl ester (Molecular Probes) according to manufacturer's instructions and used at a 1:100 dilution. Samples were analysed on a FACSCalibur TM using CellQuest TM software (BD).
Multiple staining experiments are represented as color density plots [63] with linear scales used for both forward (FSC) and side scatter (SSC) and logarithmic scales used for fluorescence parameters (Log density: 50%).
Intracellular staining was performed after fixation of purified eosinophils, neutrophils or PBMC (2610 5 ) with 2% paraformaldehyde and permeabilization with 0.01% saponin in PBS. After blocking of non-specific binding with 5 ml mouse serum for 10 min, cells were incubated with anti-CD3-FITC, anti-abTCR-FITC, anti-cdTCR-FITC or anti-CD8-FITC or corresponding isotype controls for 30 min in the presence of saponin.
Following incubation for 2 h with or without TUBag-1 [36] (40 nM) in RMPI 1640 or after overnight culture, purified blood eosinophils were gated prior to the analysis, on the basis of their FSC and SSC characteristics in order to exclude dead cells ( Figure 1B). Within PBMC fraction, monocytes and lymphocytes were gated on the basis of their FSC and SSC characteristics prior to the analysis. Quadrants were set on samples after double staining with FITC-or APC-and PE-or biotin-conjugated control isotype antibodies. At least 10 4 events were acquired per sample.
For double and triple staining, cultured peripheral eosinophils were preincubated with 5% human serum albumin (HSA) for 30 min at 4uC and labelled for membrane receptor with the relevant fluorochrome-conjugated antibodies. After washing, cells were fixed 10 min with 2% paraformaldehyde at 4uC and permeabilized for 10 min at room temperature (RT) with 0.01% saponin in PBS-1% BSA. Non-specific binding was blocked with 5 ml mouse serum for 10 min. Purified peripheral blood eosinophils were further identified by incubation with biotinylated anti-EPO or anti-MBP or isotype control for 30 min and streptavidin-PE (Molecular Probes) (1:200) for 20 min in the presence of saponin. Neutrophils were identified by intracellular staining with anti-MPO-PE following an identical procedure. Monocytes and T lymphocytes were identified by staining with anti-CD14-PE, anti-abTCR or anti-cdTCR respectively.
For conventional immunofluorescence a Leica DM-RB microscope equipped for epifluorescence with a 100 W mercury lamp was used. Filter and dichroic mirror sets were TX2 for Texas Red and A for Hoechst 33342. Images were acquired with a 10061.32 NA oil immersion objective, a Photometrics Cool SNAP TM camera using RSImage TM software (Roper Scientific). Samples were analysed by confocal microscopy using a DM-IRE2 inverted microscope with SP2-AOBS scan-head (Leica) at the Imaging Facility of Institut Pasteur de Lille. Excitation was performed at 543 nm for Texas Red. Fluorescence emission wavelengths pass bands were selected between 581 and 621 nm according to the emission spectral analysis. Excitation power was between 100 and 400 mW. Acquisitions were performed using a 10061.4 NA oil immersion objective. 3D pre-treatment, analysis and edition were performed with Edit3D (free software, Y. Usson, GDR2588, CNRS, France). Analysis and editing was performed on Photoshop (Adobe).

Surface immunoprecipitation
Surface labelling and co-immunoprecipitation was performed as previously described for cd T cells [64]. Purified human eosinophils or differenciated cdT cells were washed and then resuspended in PBS (pH 8.0) at 3.10 7 cells/ml at room temperature. Surface biotinylation was performed for 30 min at room temperature using 2 mM Sulfo-NHS-LC-Biotin (Pierce Biotechnology). Cells were washed three times with PBS containing 100 nM glycine and once in PBS.
Labelled cells were lysed in a buffer containing 20 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, protease inhibitors and 1% Brij 97 (Sigma-Aldrich). Lysates were cleared by centrifugation for 15 min at 12,000 g and sequentially incubated at 4uC in the presence of Protein-G Sepharose (Gamma bind Plus Sepharose, Pharmacia), Protein G Sepharose-bound normal mouse IgG (Jackson Immunoresearch) and Protein G Sepharose-bound anti-cdTCR. Immunoprecipitated samples were washed three times with lysis buffer then boiled in reducing sample buffer (30 mM Tris, pH 6.8, 5% glycerol, 2% SDS, 2.5% b-mercaptoethanol, 0.1% bromophenol blue). Material was run on 14% polyacrylamide gels. Material corresponding to the total and to 1/10 th of the material was loaded for eosinophils and cdT cells respectively. Following separation, proteins were transferred to PVDF membrane (Biorad). Blots were developed using ABC-HRP (Vector Laboratories) and the enhanced chemiluminescence (ECL) plus Western blot detection system (Amersham).

Real time PCR
Total RNA was prepared from eosinophils by guanidium/CsCl centrifugation method. Briefly, purified eosinophils were pelleted by centrifugation and then lysed in 4 M guanidium isothiocyanate, 1 mM EDTA, 25 mM sodium acetate, 4.9% b-mercaptoethanol, 68 mM N-lauryl sarcosine. Lysate was drawn through a 20G needle. RNA was obtained by ultracentrifugation (28000 rpm, 20 h, and 20uC) through a CsCl gradient. RNA pellet was washed twice and dissolved in water and precipitated overnight at 220uC with ethanol 70% and sodium acetate 0.08 M. After centrifugation (10000 rpm, 45 minutes at 4uC), RNA was washed twice in ethanol 70%, resuspended in water and store at 280uC until use. RNA from cd T cells and Colo205 cells was prepared using RNeasy mini-spin columns (Qiagen, UK) according to the manufacturer's instructions. All samples were quantified by absorbance measurement at 260 nm on a spectrophotometer (Biorad), and RNA quality was checked by running samples on 1.5% agarose gel in RNA loading buffer (Sigma).
Total RNA was first submitted to DNAse I (Invitrogen) treatment (15 min at room temperature), and cDNA was generated using Superscript II reverse transcriptase (Invitrogen) according to the manufacturer instructions. Samples were analyzed by quantitative real-time PCR, performed according to manufacturer's protocol, using SYBR Green PCR Master Mix (Applied Biosystems) and the ABI Prism 7000 Sequence Detection System (Applied Biosystems). Primers were designed using the Primer3 Website and are listed on Table S2.
Samples were run in triplicate in a reaction volume of 25 ml. Amplification was carried out for 40 cycles, with denaturation at 95uC for 10 min during the first cycle and subsequently for 15 seconds, annealing and extension for 1 minute at 60uC. A dissociation temperature gradient was included at the end of the run. Gene expression was normalized according to GAPDH expression. Relative gene expression was calculated with the 2 2DDC T method [65].

TCR repertoire analysis
Qualitative analysis of CDR3 length for Vc9-Jc1/2, Vc9-JcP and Vd2-Cd families was performed in triplicate on eosinophils and in cd T cells from the same donor (TcLand, Nantes, France) [66]. Briefly, cDNA was amplified by PCR using a Cc or Cd reverse primer and Vc9 or Vd2-specific forward primers, respectively. The amplifications were performed in a 384-Well GeneAmpH PCR System 9700 (Applied Biosystems, Foster City, CA). Briefly, each amplification product was used for an elongation reaction using a dye-labeled Jc specific-labelled primer for the rearrangement analysis and Cd labeled primer for the d chain analysis, then heat denatured, loaded at least in duplicate onto a sequencing analyzer (48-capillary 3730 DNA Analyzer -Applied Biosystems). Then, GeneMapper software (Applied Biosystems) was used to display the distribution profiles of CDR3 lengths, in amino acids, of the amplified and elongated products.
cdTCR junctional sequences analysis Vd1-Jd3 and Vd2-Cd amplification products (400 bp) were cloned into a TA vector (pCR-TOPO, Invitrogen) using standard protocols. For each rearrangement, sequencing was performed on 5 randomly picked clones. TCR sequences were obtained from the IMGT website (imgt.cines.fr). Sequence alignments were performed using DNAstar software. GeneBank sequence accession numbers are: FJ890312 and FJ890313 respectively.
In situ hybridisation PCR products, amplified from purified cdTCR lymphocyte RNA, and corresponding to VcI-JcP, Vd2-Cd and Vd1-Jd4 rearrangements were cloned into the pCRII-TOPO vector (Invitrogen). Clones were isolated and sequenced on both strands. Clones corresponding to the published sequence were used to generate hybridization probes. Sense and anti-sense probes were synthesised from linearised plasmids using 350 mM digoxigenin-UTP (Roche) and SP6 or T7 polymerases as described [67]. In situ hybridisation was performed using a Discovery automat and corresponding kits (Ventana Medical Systems). Slides were incubated in EZprep buffer before processing with a standard RiboMap Kit. Slides were pretreated 30 min with CC2 buffer, then 20 min with protease 3, followed by 6 h hybridization with sense or anti-sense probes (100 ng/slide). Slides were then washed twice 10 min in 26 SSC at 60uC, twice 10 min in 0.16SSC at 60uC. Slides were incubated for 30 min with a mouse anti-digoxigenin antibody (Jackson Immu-noResearch) and then for 30 min with a rabbit anti-mouse antibody before treatment with UltraMap kit. Labelling was detected after a 150 min (Vd2-Cd), 110 min (VcI-JcP) or 90 min (Vd1-Jd4) incubation in the presence of NBT/BCIP substrate. Slides were counterstained for 1 h with Nuclear Fast Red and mounted in permanent mounting medium (Vector).

EDN and cytokine release
Eosinophils (2610 6 /ml) were activated with the same stimuli as for ROS production. After 18 h culture, supernatants were collected. EDN levels were measured by specific ELISA (MBL). The lower detection limit was 0.62 ng/ml. IFNc and TNFa cytokines were measured by specific ELISA (Diaclone). Detection limit was 5 pg/ml for IFNc and 8 pg/ml for TNFa.

Tumor cell apoptosis
Colo 205 cells were labelled with 10 mM PKH26 (Sigma) according to manufacturer's instructions. Then, eosinophilmediated cytotoxicity against PKH-26-labelled Colo-205 (10 mM) was measured in complete medium at 1.6610 4 targets/ well into U-bottom plates containing eosinophils (25:1 Effector:-Target ratio) [61]. After different time points, apoptosis was measured by flow cytometry after annexinV-FITC staining (Pharmingen) for 15 min at RT. For inhibition experiments, blocking antibodies against cdTCR or isotype-matched control (10 mg/ml) were added to eosinophils immediately prior to the incubation in the presence of targets. In some experiments, eosinophils were preincubated for 30 min with 50 mM SBA or 20 mM TUBAg 1 before addition to the target cells.

Statistical analysis
Individual data or mean6S.E.M values were presented. All statistical analyses were performed using SPSS software. Normality of data samples was assessed with the Normality test of Shapiro and Wilk. Then parametric paired-samples Student's t-test was used to compare variables and the one-tailed statistical significance level was as represented on Figures.