Engagement of distinct epitopes on CD43 induces different co‐stimulatory pathways in human T cells

Summary Co‐receptors, being either co‐stimulatory or co‐inhibitory, play a pivotal role in T‐cell immunity. Several studies have indicated that CD43, one of the abundant T‐cell surface glycoproteins, acts not only as a potent co‐receptor but also as a negative regulator for T‐cell activation. Here we demonstrate that co‐stimulation of human peripheral blood (PB) T cells through two distinct CD43 epitopes recognized by monoclonal antibodies (mAb) CD43‐6E5 (T6E5‐act) and CD43‐10G7 (T10G7‐act) potently induced T‐cell proliferation. However, T‐cell co‐stimulation through two CD43 epitopes differentially regulated activation of nuclear factor of activated T cells (NFAT) and nuclear factor‐κB (NF‐κB) transcription factors, T‐cell cytokine production and effector function. T6E5‐act produced high levels of interleukin‐22 (IL‐22) and interferon‐γ (IFN‐γ) similar to T cells activated via CD28 (TCD 28‐act), whereas T10G7‐act produced low levels of inflammatory cytokines but higher levels of regulatory cytokines transforming growth factor‐β (TGF‐β) and interleukin‐35 (IL‐35). Compared with T6E5‐act or to TCD 28‐act, T10G7‐act performed poorly in response to re‐stimulation and further acquired a T‐cell suppressive function. T10G7‐act did not directly inhibit proliferation of responder T cells, but formed stable heterotypic clusters with dendritic cells (DC) via CD2 to constrain activation of responder T cells. Together, our data demonstrate that CD43 is a unique and polarizing regulator of T‐cell function.


Introduction
CD43 (sialophorin, leukosialin) is a conserved, transmembrane sialoglycoprotein expressed on most haematopoietic cells except resting B cells and erythrocytes. 1 It extends % 45 nm from the cell surface and is one of the most abundant molecules expressed on leucocytes. 2,3 Several studies have addressed the function of CD43 during the last 30 years, but its physiological role is still unclear and is particularly controversial in T cells.
In human and murine T cells, CD43 has been shown to synergize with T-cell receptor (TCR) signalling to induce T-cell activation and proliferation independent of CD28 co-stimulation. 4,5 Signal transduction through CD43 induces Ca 2+ mobilization. When cross-linked with monoclonal antibodies (mAbs), T-cell stimulation via CD43 leads to activation of the mitogen-activated protein kinase pathway and further induces the DNA binding activity of nuclear factor-jB (NF-jB), nuclear factor of activated T cells (NFAT) and activator protein 1 (AP-1) transcription factors. [6][7][8] Downstream of T-cell co-stimulation, CD43 triggers various target genes that may exhibit some overlap with CD28 co-stimulation. 9 T-cell costimulation via CD43 in the presence of TCR signalling has been shown to not only promote interferon-c (IFN-c) production by CD4 + as well as CD8 + T cells but also to negatively regulate T helper type 2 differentiation. [10][11][12] Contrary to the co-stimulatory role of CD43 reported in these studies, CD43 has been suggested to negatively regulate T-cell activation. 13 T cells from CD43-deficient mice are hyper-responsive to various mitogenic stimuli in vitro as well as in vivo. 13,14 Physical properties, such as large size and negatively charged surface, that in turn create a steric barrier for cell-cell contact, are mainly thought to be responsible for negative regulation of T-cell adhesion, T-cell-antigen-presenting cell (APC) interaction and therefore of T-cell activation by CD43. Likewise, TCR signalling has been reported to induce selective exclusion of CD43 from the immunological synapse to a distal polar complex. 15 On the other hand, CD43 along with MHC-I molecule is involved in spontaneous T-cell conjugate formation, an initial step in T-cell activation. 16 Furthermore, expression of only cytoplasmic domain of CD43 in CD43 À/À T cells could reverse the hyper-proliferative effect of CD43 deficiency. The ectodomain of CD43 did not seem to interfere with the T-cell-APC interaction. 17 These observations suggest that negative regulation of T-cell activation via CD43 is mainly facilitated by an intracellular mechanism and is not merely a phenomenon of a physical barrier function. Additionally, CD43 À/À mice showed increased numbers of antigenspecific CD8 + T cells compared with wild-type mice during the course of viral response after the initial peak of expansion, indicating an important role of CD43 during the contraction of an immune response. 18 Hence, CD43 seems capable of acting as both a positive and a negative regulator of T-cell responses.
To elucidate the co-stimulatory role of CD43 in T-cell activation, we took advantage of two well-defined CD43 mAbs 6E5 and 10G7 that bind to different, non-overlapping epitopes on human CD43. 19,20 More importantly, previous studies have demonstrated that targeting CD43 with these two mAbs has different functional effects on T-cell conjugate formation with APC. 16 We demonstrate in this study that engagement of CD43 on human peripheral blood (PB) T cells via two distinct epitopes induces proliferation of T cells, which occurs in large cellular clusters. Yet, targeting of the two epitopes on T cells exerts polarizing effects such as differential activation of transcription factors, cytokine production and also effector functions. T cells co-stimulated via the CD43-6E5defined epitope produced high levels of IFN-c and interleukin-22 (IL-22) similar to CD28 co-stimulation, but only low amounts of IL-4 and IL-17. In contrast, stimulation of PB T cells with mAb CD43-10G7 resulted in poor production of all analysed cytokines except for inhibitory cytokines transforming growth factor-b (TGF-b) and IL-35. Indeed, T 10G7-act showed a suppressive function, which was not critically dependent on these soluble factors but was mediated by inhibiting the T-cell stimulatory function of APC. The suppressive T 10G7-act cells formed stable heterotypic clusters with co-cultured dendritic cells (DC) primarily via CD2/CD58, to further hinder the activation of responder T cells by DC. Taken together, our data demonstrate that CD43 is a unique co-receptor that can exert differential polarization of T-cell function, through its different epitopes.
University of Vienna (both, Vienna, Austria). To isolate peripheral blood mononuclear cells (PBMC), heparinized buffy coats were further separated by standard density gradient centrifugation (450 g for 30 min at room temperature) with Ficoll-Paque TM Plus (GE Healthcare, Chalfont St Giles, UK). Subsequently, total T (CD3 + ) cells were obtained via depletion of CD11b + , CD14 + , CD16 + , CD19 + , CD33 + and MHC class II + cells from total PBMC. CD4 + and CD8 + T cells were also obtained by negative selection and monocytes were separated by positive selection using the MACS technique (Miltenyi Biotec, Bergisch Gladbach, Germany) as described previously. 21 For isolation of CD4 + CD25 + regulatory T cells, CD4 + T cells were further incubated with CD25 antibody and were separated by positive selection using MACS. Naive T cells were isolated from umbilical cord blood (CB). CB samples from healthy donors were collected during fullterm deliveries. Ethical approval was obtained from the Medical University of Vienna, institutional review board. Informed consent was provided in accordance with the Declaration of Helsinki. Briefly, T cells were isolated from CD34-depleted mononuclear cells obtained from CB, using the same protocol as described above. Purity of total T cells (PB T plus CB T cells), CD4 + and CD8 + T cells was checked routinely. Purity of each cell population was found to be ≥ 97%. Monocyte-derived DC were generated by culturing purified monocytes for 7 days with a combination of granulocyte-macrophage colony-stimulating factor (50 ng/ml) and IL-4 (35 ng/ml). 21 T-cell proliferation assay MAXISORP Nunc-Immuno plates (Thermo Scientific, Waltham, MA) were coated overnight at 4°with either CD3 mAb (OKT3) alone or in combination with CD28 mAb (10F3) or one of the CD43 mAbs (6E5 or 10G7). All mAbs were used at 5 lg/ml. The plates were then washed to remove unbound mAbs and purified T cells (2 9 10 5 /well) were added to the respective wells. T-cell proliferation was monitored, measuring [methyl-3 H]thymidine (PerkinElmer, Inc. Waltham, MA) incorporation at day 3. Cells were harvested 18 hr after adding [methyl-3 H]thymidine (0Á05 mCi/well) and incorporated thymidine was detected on a microplate scintillation counter (Topcount; Packard, Meriden, CT) as counts per minute. Assays were performed in triplicates.

Mixed leucocyte reaction
For mixed leucocyte reaction (MLR) purified T cells (2 9 10 5 cells/well) were stimulated with allogeneic DC (5 9 10 4 cells/well). Experiments were performed in 96well round-bottom cell culture plates in the presence of RPMI-1640 medium (Mock) or indicated cell supernatants, as described previously. 22 T-cell proliferation was monitored, measuring [methyl-3 H]thymidine incorporation at day 5. Assays were performed in triplicates.

Flow cytometry analysis
For membrane staining, cells (2 9 10 5 ) were incubated with either unconjugated or conjugated mAbs for 30 min at 4°. For unconjugated mAbs, Oregon Green â 488-conjugated goat anti-mouse IgG antibody (Life Technologies, Carlsbad, CA) and for biotinylated mAbs, PE-conjugated streptavidin was used as the second-step reagents.
Intracellular cytokine production was determined by pre-treating the activated PB T cells, for 12 hr with 5 lM monensin (Sigma-Aldrich) and then by fixing cells in FIX-solution for 20 min at room temperature before incubating with the respective mAbs along with PERM-Solution (both, AN DER GRUB Bio Research GmbH, Kaumberg, Austria) for 20 min at room temperature. Flow cytometry analyses were performed using FACScalibur (Becton Dickinson, Franklin Lakes, NJ).
Before FOXP3 staining, cell surface antigens (CD45RA) were stained as described above. Foxp3/Transcription factor staining buffer set (eBioscience Inc., San Diego, CA) was used for intracellular FOXP3 staining. Briefly, The cells were fixed with fixation buffer in the dark at room temperature for 20 min. Cells were then incubated with AF647 anti-FOXP3 mAb or isotype control mAb in permeabilization buffer in the dark at room temperature for 30 min. Flow cytometry analyses were performed using LSRFortessa (Becton Dickinson).

Analysis of duration of CD43 mAb binding
Peripheral blood T cells were incubated with biotinylated CD43-6E5 or CD43-10G7 mAb at 4°for 1 hr. An initial binding of CD43 mAbs at 0 hr was immediately analysed by flow cytometry. Part of the labelled cells were maintained at 4°. For analysis at 37°, the labelled T cells were incubated with plate-bound CD3 mAb, to ensure survival of T cells throughout 3 days of culture. At the indicated time-point, cells were labelled with PE-conjugated streptavidin as a second-step reagent and were analysed by flow cytometry.

Re-stimulation of T cells
T CD28-act , T 6E5-act and T 10G7-act were harvested at day 3 and were then further cultured for another 4 days in fresh media without stimulation. T cells were then re-stimulated (2 9 10 5 cells/well) via plate-bound CD3/CD28, CD3/CD43-6E5 or CD3/CD43-10G7 in the presence or absence of exogenous IL-2 (20 U/ml). T-cell proliferation was analysed at day 3 by [methyl-3 H]thymidine incorporation. Assays were performed in triplicates.

T-cell suppression assay
T 6E5-act , T 10G7-act and T CD28-act were harvested at day 3 and rested in fresh medium without any stimulus for 4 days, as described above. T cells were then either irradiated (30 Gy, 137Cs source) or pre-treated with 1% formaldehyde. Pre-activated T cells, at various cell numbers, were then co-cultured either with responder PB T cells (1 9 10 5 cells/well) stimulated with immobilized CD3/CD28 mAb or in an allogeneic MLR with DC (5 9 10 4 cells/well) and responder PB T cells (1 9 10 5 cells/well). Where indicated, freshly isolated regulatory T cells were added to allogeneic MLR. Per cent suppression was calculated as described previously. 23 For T cells activated via plate bound mAbs, proliferation was measured at day 3 and for an MLR at day 5 by [methyl-3 H]thymidine incorporation.

Real-time PCR
Total cellular RNA was isolated using peqGOLD TriFast (Peqlab, Erlangen, Germany) with chloroform extraction, followed by isopropanol precipitation according to the manufacturer's protocol. The cDNA was generated using the Revert Aid MuLV-RT kit (Fermentas, Burlington, Canada) using Oligo-dT(18 mer) primers according to the manufacturer's protocol and was stored at À20°until further use. Quantitative real-time PCR was performed with a CFX96 Real-Time PCR Detection System (Bio-Rad, Hercules, CA) using SYBR Green qPCR master mix (Quanta Biosciences, Gaithersburg, MD) for detection. CD3E was used as an endogenous reference gene. 24 Specific primers for human IFNG, IL4, IL22, EBI3, p35, FOXP3, CD3E and p28 were designed using the software PRIMER 3 PLUS 25 and were synthesized at Sigma-Aldrich (see Supplementary material, Table S1). Data analysis was performed using CFX MANAGER software (Bio-Rad).

Cell aggregation assay
Dendritic cells were labelled with 1 lg/ml CellTrace TM Oregon Green â 488 (carboxy-DFFDA-SE) (Invitrogen) in PBS per 1 9 10 7 cells, as per the manufacturer's protocol. The DC were then resuspended in complete RPMI medium. Labelled DC (5 9 10 4 cells/well) were added to irradiated pre-activated T cells (1 9 10 5 cells/well). Where indicated, T 6E5-act , T 10G7-act or T CD28-act cells were pre-treated with either blocking anti-LFA-1 (CD11a/ CD18) or CD2 mAbs (10 lg/ml), before irradiation. Images were taken using Nikon DS-Fi1c under an inverted Nikon Eclipse Ti-S fluorescence microscope (Tokyo, Japan) with 10 9 magnification. Images were captured at three different locations in each well per assay.
Association of five or more cells (DC and T cells) was designated as a heterotypic cluster. The area of a heterotypic cluster was calculated using the image analysis program FIJI. 26 Briefly, the boundary of an individual heterotypic cluster was defined manually. The scale of an image was calibrated according to the scale bar and the area of each heterotypic cluster was then analysed using FIJI. 26 Number of DC per heterotypic cluster was determined by using DEFINIENS analysis software (Definiens AG, Munich, Germany). Oregon Green â 488-positive DC were detected by stain intensity and were further segmented on the basis of preset thresholds for morphology and area of an individual cell.
For analysis of DC-T-cell clustering by flow cytometry, DC were labelled with CellTrace TM Oregon Green â 488 as described above and pre-activated T cells (either T 6E5-act , T 10G7-act or T CD28-act ) were labelled with 5 lM CellTracker TM Red CMTPX dye (Invitrogen) in serum-free medium per 1 9 10 7 cells, as per the manufacturer's protocol. Both DC and pre-activated T cells were then resuspended in complete RPMI medium. Where indicated, pre-activated T cells were pre-treated with respective blocking mAbs. DC (5 9 10 4 cells/well) were added to irradiated pre-activated T cells (1 9 10 5 cells/well). The cells were analysed by flow cytometry after 30 hr. Double-positive cells were counted as T cells and DC clusters.

Multi-channel reporter cell line assay
To analyse the activation of downstream signalling pathways, a multi-channel reporter cell line Jurkat E6 expressing reporter gene under the control of NF-jB, NFAT and AP-1 promoter element was used as described previously. 27 Briefly, a reporter cell line was generated by introducing constructs encoding NF-jB-CFP, NFAT-eGFP and AP-1-mCherry into Jurkat E6 cells. A cell clone that was negative for fluorescent proteins in an unstimulated state and strongly up-regulated CFP, eGFP and mCherry expression upon PMA/ionomycin treatment was selected for further use. 28 The reporter cells were activated via plate-bound CD28, CD43-6E5 or CD43-10G7 mAbs along with CD3 mAb. To assess the activation of the respective transcription factors, cells were harvested after 12 hr and expression of eGFP, CFP and mCherry were measured by flow cytometry using LSRFortessa (Becton Dickinson).

Statistical analysis
Statistical analysis was performed using GRAPHPAD PRISM software (GraphPad, La Jolla, CA). Unpaired, two-tailed Student's t-test followed by Holm-S ıd ak test for multiple comparisons was performed and P-values < 0Á05 were considered significant. Significant values are represented as *P < 0Á05, **P < 0Á01, ***P < 0Á001.

CD43 mAb 6E5 but not CD43-10G7 down-regulates cell surface expression of CD43
Previous studies have shown that both CD43 mAbs bind to different, non-overlapping epitopes on human CD43. 19 As a result of O-linked glycosylation CD43 exists in two isoforms of size 115 000 and 135 000 MW. 29 The 115 000 MW isoform is expressed on naive cells, whereas the 135 000 MW isoform of CD43 is associated with Tcell activation. 30 Results presented in Fig. 1(a(i)) demonstrate that both CD43 mAbs react with unstimulated PB T cells and CB T cells with similar intensity. Both CD43 mAbs also bound efficiently to PB T as well as CB T cells upon stimulation with PMA/ionomycin ( Fig. 1a(ii)). Furthermore, PB T cells stimulated with PMA/ionomycin uniformly expressed the two defined epitopes, similar to unstimulated PB T cells (see Supplementary material, Fig. S1a). The 135 000 MW isoform of CD43 is constitutively expressed more on resting CD8 + T cells than CD4 + T cells. 31 CD43-6E5 and CD43-10G7 mAbs showed comparable reactivity to PB CD4 + and CD8 + T cells (see Supplementary material, Fig. S1b). Binding studies showed that two mAbs react with CD43 on PB T cells with similar affinity (see Supplementary material, Fig. S1c). The two CD43 mAbs also showed similar duration of binding as analysed over the period of 3 days (see Supplementary material, Fig. S1d). Additionally, both CD43 mAbs showed reactivity with Bw5417 cells retrovirally transduced to express human CD43 (Bw-CD43); 32 but not with the parental Bw5417 cells (Bw) (see Supplementary material, Fig. S1e). Bw5417 cell line lacks C2GnT glycosyltransferase that initiates core 2 O-glycan branching and can express only 115 000 MW isoform of CD43. 33 Moreover, two CD43 mAbs bind to parental CD43 + CEM lymphoid T-cell line but not to a CD43deficient CEM cell line. 20 Together data suggest that CD43 mAbs 6E5 and 10G7 are specific for human CD43 and bind to different epitopes present on both isoforms of CD43.
Crosslinking of CD43 with mAbs induces T-cell costimulation and also modulates the cell surface expression of CD43 on leucocytes. 4,6-10,34 Therefore, we next analysed whether cell surface expression of CD43 is modulated by our CD43 mAbs. Results presented in Fig. 1(b) demonstrate that CD43 is strongly down-regulated from cell surface upon activation with plate-bound CD3/CD43-6E5 but not with CD3/CD43-10G7 or CD3/CD28. Downmodulation of CD43 surface expression by CD43-6E5 was fast and efficient and T cells were almost CD43-negative after 6 hr of stimulation with CD3/CD43-6E5 (Fig. 1b). Down-regulation of CD43 surface expression by CD43-6E5 could only be observed in the presence of TCR signalling. T-cell stimulation with plate-bound CD43-6E5 alone did not modulate CD43 expression on T cells (see Supplementary material, Fig. S2). CD43 expression was restored to basic levels after 3 days of culture (data not shown).
Hence, both CD43 mAbs CD43-6E5 and CD43-10G7 show similar affinity and comparable expression profile on various T-cell subsets, but differ in their ability to modulate CD43 cell surface expression (Fig. 1, and see Supplementary material, Fig. S1).

Co-stimulation upon engagement of CD43 with mAb 6E5 or 10G7 induces T-cell proliferation
We next assessed the functional ability of mAbs CD43-6E5 and CD43-10G7, to induce T-cell co-stimulation and T-cell proliferation. Ligation of CD43 mAbs alone did not induce T-cell proliferation (see Supplementary material, Fig. S3a). Both CD43 mAbs along with TCR signalling induced proliferation of PB T and CB T cells, but at marginally lower levels, compared with CD28 costimulation (Fig. 2a, and see Supplementary material, Fig. S3b). Co-stimulation via distinct CD43 epitopes could efficiently activate PB CD4 + as well as CD8 + T-cell subsets (see Supplementary material, Fig. S3c). In line with the proliferation data, T 6E5-act and T 10G7-act also expressed various T-cell activation markers including CD69, CD97 and HLA-DR at comparable levels to T CD28-act (Fig. 2b, and see Supplementary material, Fig. S3d). However, some of the other T-cell activation markers including CD25 and CD71 were differentially regulated. T 6E5-act and T 10G7-act expressed lower levels of CD71 compared with T CD28-act . The T 6E5-act expressed comparable levels of CD25 to T CD28-act , whereas the expression of CD25 was significantly lower on T 10G7-act (Fig. 2b, and see Supplementary material, Fig. S3d).
CD43 has been reported to induce homotypic aggregation in leucocytes, including T cells. [35][36][37][38] T-cell homotypic clustering is considered as a hallmark for efficient T-cell activation in vitro. Likewise, PB T-cell activation using immobilized CD43 mAbs along with TCR signalling induced a homotypic clustering response that was clearly visible after 30 hr (see Supplementary material, Fig. S3e).

Co-stimulation via two CD43 epitopes differentially regulate activation of transcription factors
Previous studies have shown that T-cell co-stimulation via CD43 induces DNA binding activity of NF-jB, NFAT and AP-1 transcription factors. 6 In assays using Jurkat E6 multi-channel reporter cells that express both CD43 epitopes, co-stimulation via CD43-6E5 induced activation of NF-jB (Fig. 2c, and see Supplementary material, Fig. S1f). However, NF-jB activation was weaker compared with CD28 co-stimulation (Fig. 2c). Co-stimulation via CD43-10G7 did not further enhance NF-jB promoter activity compared with CD3 (Fig. 2c). NFAT reporter activity was comparable upon co-stimulation via CD28 or via CD43-6E5 but was lower upon activation via CD43-10G7 (Fig. 2c). Compared with CD3 alone, AP-1 promoter activity was further enhanced only upon CD28 co-stimulation (Fig. 2c).

Differential regulation of helper T-cell cytokines via CD43
T-cell co-stimulation via CD43 uses overlapping as well as distinct signalling pathways from CD28 co-stimulation that in turn may differentially regulate target gene expression in primary human T cells. 9 To further investigate the effect of CD43 co-stimulation on subsequent T helper cell function, the production and secretion of various T-cell cytokines in the supernatants of activated PB T cells was analysed. Similar to T CD28-act , T 6E5-act secreted high levels of IL-22 and IFN-c in the cell supernatant. However, compared with T CD28-act , the measured levels of IL-17 were significantly low in T 6E5-act cell supernatants (Fig. 3a). In contrast to T 6E5-act , supernatants of T 10G7-act contained significantly low amounts of IL-22, IFN-c and also IL-17 (Fig. 3a). Compared with T CD28-act , supernatants of T 6E5-act as well as T 10G7-act contained very low levels of IL-2 and T helper type 2 cytokines including IL-4 and IL-13 (Fig. 3a). These results were further confirmed by intracellular cytokine staining (Fig. 3b, and see Supplementary material, Table S2). The analysis of cytokine production at the protein level correlated with induction of mRNA. Results presented in Fig. 3(c) demonstrate that T 10G7-act expressed low levels of IFNG, IL4 and IL22 mRNA. On the other hand, IFNG and IL22 mRNA levels were high in T 6E5-act similar to T CD28-act . However, T 6E5-act expressed only low levels of IL4 mRNA. The data suggest that T 6E5-act show substantial overlap with T CD28-act , except for the induction of IL-4, IL-13 and IL-2, whereas induction of various T-cell signature cytokines is differentially regulated between the two epitopes on CD43 analysed in this study.

Regulation of anti-inflammatory cytokines by CD43 co-stimulation
Along with inflammatory cytokines, regulatory cytokines play a crucial role in shaping an effective immune response. Therefore, apart from pro-inflammatory T-cell cytokines, the induction of regulatory T-cell cytokines such as IL-10 and TGF-b was also analysed. Production of IL-10 was strongly induced by CD28 co-stimulation, whereas T 6E5-act secreted moderate levels of IL-10 (Fig. 4a). The TGF-b was similarly regulated by all the three co-stimulations tested (Fig. 4a). In addition, the expression of IL-35 subunits, EBI3 and p35 was also    analysed. T-cell co-stimulation via CD43-10G7 induced significantly higher expression of EBI3 mRNA compared with T cell co-stimulation either via CD43-6E5 or CD28 (Fig. 4b). The higher expression of EBI3 protein in T 10G7-act was further confirmed by intracellular staining (Fig. 4c). The expression of p35 was also slightly upregulated in T 10G7-act compared with either T 6E5-act or T CD28-act (Fig. 4b). The IL-35 subunit EBI3 can dimerize with p28 to form IL-27. 39 The expression levels of p28 were in fact lower upon co-stimulation with all three stimuli tested here, compared with stimulation via CD3 alone (Fig. 4b).
Interleukin-35 is regarded as an inhibitory cytokine; therefore supernatants of activated PB T cells were tested for inhibitory effect in an allogeneic MLR. However, we could not observe a soluble factor mediated inhibition of T-cell proliferation (Fig. 4d).

T-cell co-stimulation via CD43-10G7 induces a hypoproliferative state
In the next set of experiments, we analysed the ability of T cells activated in the presence of CD43 co-stimulatory signals to respond to re-stimulation. In contrast to T 6E5-act or T CD28-act that responded efficiently to re-stimulation, T 10G7-act showed reduced proliferative responses irrespective of whether CD3/CD43-6E5, CD3/CD43-10G7 or CD3/CD28 was used as the secondary stimulus (Fig. 5a). As opposed to T 6E5-act and T CD28-act , the addition of exogenous IL-2 had no effect on re-stimulation of T 10G7-act with CD3 alone (Fig. 5a,b). Exogenous IL-2 could restore the proliferation of T 10G7-act in the presence of a co-stimulatory signal during re-stimulation; however, not as effectively as in the case of T6E5-act and T CD28-act (Fig. 5b). The reduced proliferative capacity of T 10G7-act seemed not to be due to increased T-cell death, as indicated by Annexin V and propidium iodide staining (Fig. 5c). Furthermore, the reduced proliferation response was neither due to down-modulation of co-receptors during the first round of activation. We analysed the expression of CD43 epitopes recognized by mAbs CD43-6E5, CD43-10G7 and of CD28 on T 6E5-act , T 10G7-act and T CD28-act (Fig. 5d). Though all three mAbs showed slightly reduced binding to T 10G7-act , this did not correlate with the extent of poor proliferative response induced upon re-stimulation (Fig. 5a,b, d).
Hence, in contrast to the T-cell co-stimulation via the CD43-6E5-defined epitope or via CD28, T-cell co-stimulation via the CD43-10G7-defined epitope induces a deep hypo-proliferative state in T cells.

T 10G7-act acquire an inhibitory function
So far, T 10G7-act exhibited properties similar to many different subsets of inhibitory T cells, such as lower levels of IFN-c, IL-2, IL-4 and IL-22 but up-regulation of IL-35 cytokine subunits, EBI3 and p35 and higher production of TGF-b. [40][41][42] To further analyse the functional properties of these cells, T-cell suppression assays were performed. For this, irradiated T 6E5-act , T 10G7-act and T CD28-act were added to an allogeneic MLR. T 10G7-act inhibited DC-induced T-cell proliferation in an MLR in a dose-dependent manner (Fig. 6a). T 10G7-act were as efficient as regulatory T cells in their suppressive capacity (see Supplementary material, Fig. S4a). Compared with T CD28-act , high numbers of T 6E5-act seem to exert an inhibitory effect. However, with T 10G7-act such inhibition could be achieved when almost 10 times fewer cells were used (Fig. 6a). Prior fixation of T 10G7-act cells with formaldehyde reversed their inhibitory function (Fig. 6b).
Furthermore, T 10G7-act could not inhibit T-cell proliferation when added to allogeneic T cells activated via platebound CD3/CD28 mAb in the absence of APC (Fig. 6c). Hence, the data suggest a crucial role of DC in the inhibition of responder T cells. The suppressive effect of T 10G7-act could also be abrogated when DC were irradiated before being added to an allogeneic MLR (Fig. 6d). This inhibitory effect was not restricted to the presence of DC, as T 10G7-act also exhibited an inhibitory effect when monocytes or monocyte-derived macrophages were used as APC in an allogeneic MLR (data not shown). FOXP3 is a forkhead family transcription factor important for the development and function of natural regulatory T cells. 43 Therefore, expression of FOXP3 was analysed by quantitative PCR as well as by intracellular staining. FOXP3 expression did not differ between different CD43 co-stimulations, suggesting that the suppressive function of T 10G7-act is not directly related to the expression of FOXP3 (Fig. 6e, and see Supplementary material, Fig. S4b). Activation-induced transient expression of FOXP3 in T effector cells has been reported before and has not necessarily been associated with suppressive function of T cells in humans. 22,44,45 Expression of cell surface molecules such as PD-1 (CD279) and CTLA-4 (CD152) has been previously linked with regulatory function in T cells. [46][47][48] However, as analysed by flow cytometry, expression of PD-1 (CD279) as well as CTLA-4 (CD152) was in fact lower on T 10G7-act compared with T 6E5-act or T CD28-act (Fig. 6f).

The inhibitory T 10G7-act cells do not alter the accessory molecule repertoire on DC
The finding that the inhibitory effect of T 10G7-act was observed only in the presence of APC, prompted us to analyse whether a soluble factor secreted by DC co-cultured with T 10G7-act might be responsible for the inhibitory effect. However, results shown in Fig. 7(a) demonstrate that supernatant from an MLR with preactivated PB T cells had no inhibitory effect on the proliferation of responder PB T cells.
Next, the expression of various cell surface receptors on DC co-cultured in an allogeneic MLR with T 6E5-act , T 10G7-act or T CD28-act was analysed. DC co-cultured with T 10G7-act showed comparable expression of MHC-I, HLA-DR to DC co-cultured with either T 6E5-act or T CD28-act . The same is true for DC maturation markers such as CD83 or for co-stimulatory molecules such as CD86 and CD40 (Fig. 7b). Interestingly, the inhibitory receptor B7-H1 (CD274) was not altered on DC cocultured with T 10G7-act (Fig. 7b).
T 10G7-act induce stable heterotypic clustering with DC to constrain activation of responder T cells Dendritic cells co-cultured with irradiated T 10G7-act showed characteristic cluster formations, compared with DC co-cultured with irradiated T 6E5-act or T CD28-act as shown by the area of each cluster and the number of DC per cluster (Fig. 8a,b). Heterotypic aggregate formation was found to be an active process. Neither irradiated T cells co-stimulated via plate-bound CD3/CD43-10G7 mAbs (see Supplementary material, Fig. S5a), nor irradiated DC co-cultured with irradiated T 10G7-act (see Supplementary material, Fig. S5b) were able to form aggregates. Likewise, regulatory T cells have been reported to physically hinder the interaction of DC with responder T cells by forming large aggregates. [49][50][51] Hence, T 10G7-act -induced stable heterotypic clustering may contribute to the inhibitory effect of T 10G7-act on the accessory function of DC in co-culture experiments.
As an underlying mechanism, T 10G7-act -induced heterotypic clustering was found to be mainly CD2 dependent, whereas interaction of T 6E5-act or T CD28-act with cocultured DC was LFA-1 (CD11a/CD18) dependent ( Fig. 8a-c). Heterotypic interaction of T CD28-act with cocultured DC was weaker compared with T 6E5-act and T 10G7-act , as shown by a higher percentage of single-positive DC in T CD28-act (56Á2%) control treated cells compared with T 6E5-act (0Á16%) and T 10G7-act (0Á24%) (Fig. 8c).
Further, pre-treatment of T 10G7-act with a CD2 blocking mAb could abolish the inhibitory effect of T 10G7-act and restore proliferation of responder T cells. The pre-treatment of T 10G7-act with anti LFA-1 blocking mAb could slightly restore proliferation compared with isotype control but not as effectively as pre-treatment with CD2 (Fig. 8d).

Discussion
Signalling via accessory cell surface receptors plays an essential role in the induction, tuning and regulation of T-cell activation and function. 52 A plethora of such costimulatory receptors have been identified, which conventionally provide either positive or negative signals to T cells. CD43 is one of the most abundant cell surface receptors on human T cells. Yet, the functional role of CD43 on T cells is still controversial, with several studies reporting opposing roles of CD43 in T-cell function. [4][5][6]8,9,13,14,17,18 Here, we demonstrate that targeting of distinct epitopes on CD43 can decide the subsequent fate of T-cell function.
This observation was made with two well-defined CD43 mAbs (6E5, 10G7) directed against two non-overlapping binding sites, expressed on both isoforms of CD43 (Fig. 1a, and see Supplementary material, Fig. S1a, b,e). 19,20 CD43-6E5 mAb shares a similar epitope to CD43 mAb MEM-59. 19 Co-stimulation via MEM-59 along with TCR signalling could efficiently induce T-cell activation. 6,53 Previous studies have demonstrated that targeting of CD43 with mAbs CD43-6E5 and CD43-10G7 induces aggregation and oxidative burst formation in neutrophils. 37 Nonetheless, epitope recognized by mAb CD43-6E5 but not CD43-10G7 was found to be involved in T-cell conjugate formation with APC. 16 We observed that both mAbs could potently activate T cells in the presence of TCR signalling, inducing a strong proliferative response in T cells including CB T cells and CD4 + as well as CD8 + T-cell subsets (Fig. 2a, and see Supplementary material, Fig. S3a-c). T-cell activation via CD43 was accompanied by the expression of classical T-cell activation markers such as CD69 and induction of homotypic clusteringa hallmark of T-cell activation in vitro (Fig. 2b, and see Supplementary material, Fig. S3d,e). Mattioli et al., have previously suggested that CD28 and CD43 may use different as well as overlapping signalling pathways. As a result, T-cell co-stimulation via CD43 may trigger expression of similar as well as different target genes compared with CD28 co-stimulation, e.g. the expression of the important T-cell cytokine gene IL2. 9 Similarly, we observed that T 6E5-act and T 10G7-act produced low levels of IL-2 compared with T CD28-act . Various studies have reported that cytokine production and T-cell proliferation are autonomously regulated upon T-cell activation. 54,55 Yet, such low amounts of IL-2 produced upon co-stimulation via CD43 could be sufficient to promote T-cell proliferation (Figs 2a, 3a). 56 Compared with T CD28-act , T 6E5-act produced similarly high levels of IFN-c and IL-22 but only low amounts of IL-2, IL-4, IL-13 and IL-17. Yet, the CD43 mAbs 6E5 and 10G7 exerted polarizing effects on T cells (Fig. 3). We found that T 10G7-act synthesized low amounts of all analysed cytokines except of the inhibitory cytokines TGF-b and IL-35 subunit EBI3 (Figs 3 and 4a-c). Polarizing effects observed upon ligation of our two CD43 mAbs, seem not to be dependent on recognition of specific CD43 isoforms. Downstream of T-cell co-stimulation via different CD43 epitopes, CD4 + and CD8 + T cells showed similar functional responses like bulk T cells. Both CD4 + and CD8 + T cells, activated via CD43-6E5 expressed higher levels of IFNG but lower levels of EBI3 mRNA, compared with their counterparts activated via CD43-10G7 (see Supplementary material, Fig. S7).
Downstream of co-stimulation, signal transduction through CD43 has been reported to induce DNA binding activity of NF-jB, NFAT and AP-1 transcription factors when CD43 is cross-linked with mAbs. [6][7][8] In our test system co-stimulation via CD43 did not induce activation of AP-1 of transcription factor (Fig. 2c). T-cell stimulation via CD43-6E5 mAb induced activation of NFAT and NF-jB. However, co-stimulation via mAb CD43-10G7 could only induce activation of NFAT (Fig. 2c). Following T-cell stimulation, activation of NFAT in the absence of NF-jB and AP-1 activation leads to anergy. 57 This mechanism could explain the hypo-proliferative phenotype of T 10G7-act as opposed to T 6E5-act seen upon re-stimulation (Figs 2c,  5a,b). In T cells, this anergic state has also been observed for low-strength T-cell activation in the presence of co-stimulation. 58 The mAb CD43-10G7 showed similar binding affinity and kinetics as CD43-6E5 (see Supplementary material, Fig. S1c,d). T-cell co-stimulation via CD43-10G7 could efficiently induce T-cell proliferation similar to co-stimulation via CD43-6E5, at levels higher than CD3 alone (Fig. 2a, and see Supplementary material, Fig. S3a-c). Compared with T cells activated via CD3 alone T 10G7-act also showed higher induction of IFN-c, IL-22 and IL-10, but at levels lower than other co-stimuli tested (Figs 3 and 4a, and see Supplementary material, Fig. S7). It will be interesting to elucidate in future studies, whether a distinct signalling pathway or a putative low strength of T-cell activation upon CD43 cross-linking with CD43-10G7 compared to CD43-6E5 is responsible for the observed co-stimulatory functions of CD43-10G7 mAb.
Interestingly, T 10G7-act acquired a prominent inhibitory function in contrast to T 6E5-act or T CD28-act . The suppressive function of T 10G7-act was not mediated via a soluble factor but was dependent on cell-cell contacts. We could also show that T 10G7-act did not directly act on responder T cells. Instead T 10G7-act exhibited their suppressor function via APC such as DC (Fig. 6a,c,d). CD4 + CD25 + FOXP3 + natural regulatory T cells have also been previously reported to route their suppressor function via APC. 59 One of the mechanisms includes modulating DC to produce immunosuppressive factors such as indolamine 2,3-dioxygenase. 60,61 However, in our experiments, the supernatants from DC co-cultured with T 10G7-act were not inhibitory (Fig. 7a). Another reported mechanism is via CTLA-4 that results in the suppression of CD80 or CD86 expression. 59, 62 We did not observe such reduced expression of CD80 or CD86 on DC co-cultured with T 10G7-act (Fig. 7b). Indeed, T 10G7-act expressed lower levels of inhibitory receptors including CTLA-4 (CD152) and PD-1 (CD279) than T 6E5-act or T CD28-act (Fig. 6f). The third mechanism of inhibition is by promoting a long stable interaction of regulatory T cells with DC mediated via various molecules including LFA-1/ICAM-1, CD2/CD58, neuropillin-1 and thereby limiting the activation of responder T cells. [49][50][51]59 Additionally, a cross-talk between CD2 and other co-stimulatory receptors like CD43 and CD28 that facilitates T-cell conjugate formation and T-cell activation, respectively, has been reported. 16,63 Indeed, we have also observed a characteristic large aggregate formation when DC were co-cultured with T 10G7-act that is primarily mediated via CD2. On the other hand, interaction of T 6E5-act with DC was mediated via LFA-1 (CD11a/CD18) (Fig. 8a-c). Blocking LFA-1 (CD11a/CD18) on T 10G7-act showed marginal effect on proliferation, whereas blocking CD2 on T 10G7-act abolished the suppressive effect and restored the proliferation of responder T cells (Fig. 8d).
CD43 is a conserved, sialylated glycoprotein with an elongated extracellular domain. 2,3 Multiple ligands have been described for CD43: ICAM-1 (CD54), MHC-I, Siglec-1 (CD169), galectin-1, E-selectin and albumin. 16,[64][65][66][67][68] The mAb CD43-6E5-defined epitope is involved in mediating the binding of MHC-I molecules, an event that further strengthens the interaction of APC with T cells. 16,19 For the epitope defined by mAb CD43-10G7, no potential ligand has been described so far. One candidate for this binding site might be Siglec-1. We have recently reported that Siglec-1 on DC contributes to the induction of EBI3 expression and IL-35 production in co-cultured T cells. 22 In this study we have observed that T 10G7-act have a preferential bias to produce IL-35, so it is tempting to speculate that Siglec-1 might be a ligand for the CD43-10G7-defined epitope. The ability of the natural ligand to modulate CD43 expression might be an important feature of CD43 function. Depending on the test system and cell type used, cross-linking of CD43 with mAb or phorbol activation has been reported to induce CD43 internalization or proteolytic cleavage in leucocytes. 34,69,70 Such activation-induced proteolytic cleavage of CD43 when observed in T cells, is associated with T-cell stimulation and homeostasis. 69 Hence, results presented in Fig. 1(b) offer a potential mechanism that cross-linking of CD43-6E5, in contrast to CD43-10G7, allows the downmodulation of CD43 from the cell surface and that in the absence of CD43 processing T cells acquire a regulatory function. However, these possibilities need to be analysed in more detail in future studies.
Taken together, our data suggests a unique role of CD43 in polarization of T-cell immunity. These findings could provide an explanation for various contradictory roles of CD43 in T-cell functions that have been previously reported.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1. Reactivity profile of CD43 monoclonal antibodies CD43-6E5 and CD43-10G7 on human T cells. Figure S2. Down-modulation of CD43 surface expression by CD43-6E5 requires T-cell receptor signalling. Figure S3. CD4 + , CD8 + peripheral blood T and cord blood T-cell stimulation upon engagement of CD43 monoclonal antibodies. Figure S4. T 10G7-act acquire FOXP3-independent suppressive function. Figure S5. Heterotypic interaction of dendritic cells with pre-activated T cells is an active process. Figure S6. Gating strategy for flow cytometry analysis of cluster formation between dendritic cells and T cells. Figure S7. The downstream effect of CD43 co-stimulation is similar in CD4 + and CD8 + T cells. Table S1. Real time PCR primer sequences. Table S2. Cytokine profile of T6E5-act , T 10G7-act and T CD28-act .