SIRPγ-CD47 Interaction Positively Regulates the Activation of Human T Cells in Situation of Chronic Stimulation

A numerous number of positive and negative signals via various molecules modulate T-cell activation. Within the various transmembrane proteins, SIRPγ is of interest since it is not expressed in rodents. SIRPγ interaction with CD47 is reevaluated in this study. Indeed, we show that the anti-SIRPγ mAb clone LSB2.20 previously used by others has not been appropriately characterized. We reveal that the anti-SIRPα clone KWAR23 is a Pan anti-SIRP mAb which efficiently blocks SIRPα and SIRPγ interactions with CD47. We show that SIRPγ expression on T cells varies with their differentiation and while being expressed on Tregs, is not implicated in their suppressive functions. SIRPγ spatial reorganization at the immune synapse is independent of its interaction with CD47. In vitro SIRPα-γ/CD47 blockade with KWAR23 impairs IFN-γ secretion by chronically activated T cells. In vivo in a xeno-GvHD model in NSG mice, the SIRPγ/CD47 blockade with the KWAR23 significantly delays the onset of the xeno-GvHD and deeply impairs human chimerism. In conclusion, we have shown that T-cell interaction with CD47 is of importance notably in chronic stimulation.


INTRODUCTION
T cells are adaptive immune cells that are educated in the thymus to defend the organism and eliminate pathogens efficiently by recognizing the MHC-antigen complex via their TCR. Mature T cells express a variety of receptors which engagement will stabilize, enhance, or decrease their activation upon antigen recognition. In this study, we focused on SIRPg, a T-cell restricted surface molecule from the signal-regulatory protein (SIRP) family. SIRP family comprise transmembrane glycoproteins expressed on immune cells and central nervous system (1)(2)(3). Three main members have been described: SIRPa, SIRPb, and SIRPg with homologous extracellular immunoglobulin (Ig)-like domains but distinct transmembrane and intracytoplasmic domains (4)(5)(6)(7)(8). Briefly, SIRPa (CD172a) is expressed on hematopoietic progenitors, myeloid cells, dendritic cells, NK cells, and neurons (4,9). SIRPa cytoplasmic tail contains immunoreceptor tyrosine-based inhibition motifs (ITIMs) (10) resulting in inhibitory signals. SIRPa ligates CD47, an ubiquitously expressed molecule, and this interaction is responsible of the "don't eat me signal" which leads to the inhibition of phagocytosis by myeloid cells (11). SIRPb (CD172b) associates with DAP12 (12,13) which contains an immunoreceptor tyrosine-based activation motif (ITAM). Yet, ligand of SIRPb is still unknown. Finally, SIRPg (CD172g) has been described recently (14) probably because the gene of SIRPg is absent in rodents and found only in humans and primates (15). SIRPg has a very short intracytoplasmic tail of 4amino acids (aa) uncapable of transducing signals on its own. SIRPg also interacts with CD47 as SIRPa (14)(15)(16), but the SIRPa/CD47 affinity (Kd 2 µM) is 10 times stronger than the one of SIRPg/CD47 (Kd 23 µM) (17).
It has been shown that the interaction between SIRPg and CD47 plays a key role in T-cell transendothelial migration under shear flow conditions (17) in cell-cell adhesion and promotes antigen-specific T-cell proliferation and costimulation (16). However, these observations rely on the use of the anti-SIRPg clone LSB2.20 which we report in this study as not being able to efficiently block the SIRPg/CD47 interaction at the concentration tested. Thus, we reevaluate the impact of SIRPg in human T-cell biology using a different anti-SIRP monoclonal antibody (clone KWAR23), which we characterize here as capable to efficiently block SIRPg and SIRPa interactions with CD47. We show that SIRPg expression varies on CD4 and CD8 T-cell subpopulations and, while being expressed by regulatory T cells (Tregs), we demonstrate that SIRPg/CD47 interaction is not implicated in their suppressive function. Upon antigen presenting cell (APC) recognition, SIRPg clusters at the immune synapse independently of its ligation to CD47 on the APC. IFN-g secretion by chronically activated T cells is impaired by SIRPg/ CD47 interaction blockade in vitro. Finally, when used in vivo in a T-cell-dependent xeno-GvHD model mediated by the injection of human PBMC in NSG-recipient mice, the anti-SIRPa-g mAb KWAR23 impairs human cell engraftment, leading to a delayed GvHD onset. Importantly, all CD4 and CD8 T-cell compartments were compromised in KWAR23-treated mice. In conclusion, we have shown that T-cell interaction with CD47 is of importance notably in chronic stimulation.

Reagents
All the reagents used in the study are listed in the Table 1.
Jurkat SIRPg-KO cells were then transduced with lentivirus encoding OST-SIRPg. SIRPg expression on transduced cells was higher than in their WT counterpart. SIRPg high cells were FACS sorted and amplified for ulterior use.

ELISA-Binding Assay
For human SIRPa-binding assay, recombinant human SIRPa was immobilized on plastic at 0.5 µg/ml (SIRPa) in carbonate buffer (pH 9.2). After saturation, purified antibodies were added in range (from initial concentration at 10 µg/ml) to measure binding. After incubation and washing, peroxidase-labeled donkey antihuman IgG was added and revealed by conventional methods.

Coculture of Tregs and Effector T Cells
Purified T cells were stained with anti-CD4, anti-CD25, and anti-CD127 for 30 min at 4°C in the dark, cells were washed, and DAPI was added to discriminate dead cells. Cells were then sorted using the ARIA II (BD Biosciences). Tregs and effector T cells were further stained with two different cell proliferation dyes (CPD) for 10 min at 37°C and cultured either alone (control conditions) or together, in U-96-well costar plates previously coated or not with anti-CD3 (clone OKT3) at 2 µg/ml. When indicated, soluble anti-SIRPa/g (KWAR23) antibody was added at 10 µg. After 5 days in culture, T-cell proliferation was assessed by flow cytometry using the LSRII (BD Biosciences).

T-Cell Activation
T cells were stained with CPD prior to be activated in plates coated with 0.5 µg/ml of anti-CD3 (OKT3) and 2-3 µg/ml of soluble anti-CD28.2. Cells were incubated 3 or 5 days at 37°C, 5% CO 2 , then assessed by flow cytometry. In some experiments, T cells were stimulated with 0.5-1 µg/ml of coated OKT3 with or without coated CD47-Ig (10 µg/ml).
Chronically Activated T Cells PBMC at 7.5 × 10 6 were stimulated in complete medium three times in six-well plates coated with 3 µg/ml of anti-CD3 (OKT3) and 3 µg/ml of soluble anti-CD28.2. Each stimulation is consist of a 48-72-h culture in six-well plates coated with 3 µg/ml of anti-CD3 (OKT3) and 3 µg/ml of soluble anti-CD28.2. Each time, cells were washed and counted prior to be restimulated. Chronically stimulated PBMC were then cultured for 48 h in plates coated with 0.5 µg/ml of anti-CD3 (OKT3) and 10 µg/ml of human CD47-Fc recombinant protein in the presence or not of mAbs targeting the SIRP/CD47 interactions. Supernatants were collected after 48 h of stimulation with OKT3 and CD47-Fc to measure IFN-g secretion, and cell fitness was analyzed by Alamar Blue detection at 530 nm after 5 days stimulation.

Confocal Microscopy
Jurkat, resting, or activated T cells were stained for 30 min at 4°C in the dark with an anti-SIRPg (LSB2.20) and an anti-CD47 mAb. Cells were fixed with PFA at 4% for 10 min at room temperature (RT), permeabilized with PBS + 0.1% Triton for 5 min at RT, and further stained with DAPI for 5 min at RT in the dark. ProLong was added to the cells and the solution carefully place between a glass slide and a 12-µM round coverslip and left for at least 48 h polymerization before being analyzed with confocal microscopy (Nikon Super Resolution SIM Confocal A1) using the NIS software. The images obtained were analyzed using ImageJ software. SIRPg and CD47 colocalization was analyzed using ImageJ software. Briefly, a region of interest (ROI) was drawn on the whole cell surface and analyzed with coloc2 plugin.

Time-Lapse Microscopy
IBIDI (Gräfelfing, Germany) 8-well slides were coated with poly-L-lysine at 0.001%. Raji cells were primed with 10 ng/ml of SEE for 30 min at 37°C, and WT or SIRPg high Jurkat cells were stained with a cell tracker green for 15 min at 37°C, washed thoroughly then stained with an anti-SIRPg mAb (clone LSB2.20) for 30 min at 4°C. Cells were then washed, and Raji and Jurkat cells were deposited in the IBIDI slide. Once the cells are stabilized, the time lapse video starts. To assess the role of SIRPg in the synapse formation, SIRPg/CD47 interaction was blocked on SIRPg high cells for 15 min at 37°C with the KWAR23 mAb at 10 µg/ml, then mixed with Raji for 15 min prior to be plated in IBIDI wells. SIRPg polarization into the synapse was analyzed using ImageJ software. Briefly, a ROI was drawn at the cell/cell contact between Jurkat and Raji and the intensity of pixels for SIRPg channel was analyzed. The same ROI was then moved at the opposite side of the synapse to measure the intensity of pixels for SIRPg channel. For control condition, cells were divided randomly in half and the intensity of pixels for SIRPg channel was measured on the two parts of the cells. The size of the ROI does not influence the intensity of pixels measured.

Animals
We used a xeno-GvHD model of immunodeficient mice to assess the role of SIRPg in GvHD occurrence. NSG mice were purchased from the Jackson Laboratory (Bar Harbor, ME) and bred by the LabEx IGO humanized rodent platform. Mice (males and females) were used between 8 and 12 weeks old. All animals were housed under specific pathogen-free conditions, according to institutional guidelines. NSG mice were irradiated at 1.5 Gy on Day 0 prior to an i.v. injection of 10 7 human PBMC from healthy volunteers on Day 0. Mice were treated 2 times/week with PBS or anti-SIRPa-g (KWAR23 mAb) at 5 mg/kg from D1 to D42. Human cell engraftment and weight lost were monitored during the experiment. If losing more than 20% of its maximum body weight, the mice were sacrificed. Human chimerism analyzed by flow cytometry is calculated as follows: (% of hCD45)/(% of hCD45 + % of mCD45) * 100. The relative number of CD4/CD8 T cells in the spleen is calculated as follows: (% human chimerism in spleen) * (% human CD4/CD8 in spleen). This study was carried out according to authorization from the French Ministry of Research, APAFIS # 4678. Intraperitoneal residual macrophages were collected as previously described in {Ray:2010cp}. Fc block was added to the cells following manufacture recommendation prior the incubation with the antibodies.

RESULTS
The KWAR23 Is a Pan Anti-SIRP mAb, Which Efficiently Inhibits SIRPg-CD47 Interaction KWAR23 mAb is a recombinant antibody which has been selected after immunization with a SIRPa recombinant molecule by Ring and collaborators (18). Given the homology of 79% when comparing the extracellular portions of the SIRPa, SIRPg, and SIRPb (15), we hypothesized that KWAR23 could also recognize other SIRP protein family members. We directly addressed this question by labelling human cells which are known to either uniquely express SIRPa, i.e., monocytes or described to express SIRPg, i.e., T lymphocytes. KWAR23 efficiently binds 100% of monocytes when the concentration used is sufficient as does the anti-SIRPa-specific mAb ( Figures 1A, B, left panels); similar results were obtained when U937 (SIRPa + cell line) were stained (data not shown). Interestingly, increased concentration of KWAR23 resulted in T-cell staining which reached a plateau when 60% of human T cells were stained. No binding was seen with the anti-SIRPa-specific mAb (clone 18D5), suggesting that KWAR23 recognizes a molecule other than SIRPa, presumably SIRPg on T cells ( Figures 1A, B, right panels). Furthermore, Jurkat WT and Jurkat SIRPg cells were stained with KWAR23 and LSB2.20, a specific anti-SIRPg, developed by Piccio and colleagues (16). While both Abs similarly stained Jurkat WT cells, no staining was observed on Jurkat SIRPgKO cells with either Ab, identifying SIRPg as the sole KWAR23 epitope on Jurkat cells ( Figure 1C). To confirm these results, we setup two ELISA in which recombinant SIRP molecules were used. The anti-SIRPa-specific mAb recognized SIRPa but not SIRPg-recombinant protein, while the KWAR23 recognized SIRPa, SIRPg, and even SIRPb ( Figure 1D and data not shown for SIRPb). Thus, KWAR23 appears as a pan anti-SIRP antibody. KWAR23 has been selected by Ring and collaborators for its ability to specifically block the SIRPa/CD47 interaction as does the anti-SIRPa Ab (19) (Figure 1E, left panel). LSB2.20 inhibits the SIRPg/CD47 (16). Thus, we evaluated LSB2.20, KWAR23, and the anti-SIRPa for their ability to interfere with SIRPg/CD47 binding. Surprisingly, the LSB2.20 did not efficiently, block the SIRPg/CD47 interaction at the concentration tested, neither did the anti-SIRPa. On the contrary, KWAR23 was found to efficiently abrogate SIRPg/CD47 interaction ( Figure 1E, right panel).
Thus, we redefined the KWAR23 as a Pan anti-SIRP antibody able to specifically block SIRPa-g/CD47 interactions.

SIRPg Expression Varies on T-Cell Populations and While Being Highly Expressed on Treg, Is Not Implicated in Their Suppressive Functions
SIRPg is expressed on T cells (16), and this expression can be altered by the presence of an eQTL (rs2281808) (20). So far, SIRPg expression on different T-cell subsets has not been described. Thus, we analyzed SIRPg expression on PBMC by FACS analysis and showed that naïve (CD3 + CD45RA + CD28 + ) and central memory (CM, CD3 + CD45RA − CD28 + ) T cells highly express SIRPg whereas effector memory (EM, CD3 + CD45RA − CD28 − ) T cells express to a lesser extent SIRPg on their surface. Interestingly, TEMRA CD8 T cells (TEMRA, CD3 + CD8 + CD45RA + CD28 − ) express SIRPg at a lower level than EM T cells (Figures 2A, B). Since SIRPg ligates CD47, a ubiquitously expressed molecule, we investigated CD47 expression on T-cell subsets. We show that it follows the same expression pattern as SIRPg (Figures 2A, C). We then evaluated SIRPg expression on T cells upon polyclonal stimulation. Proliferating T cells rapidly and significantly upregulates SIRPg at their surface and reaches a plateau once T cells have divided more than twice ( Figures 2D, E). This upregulation of SIRPg is increasing from Days 3 to 5 of activation. Conversely, the expression of CD47 on proliferating T cells decreases upon T-cell divisions with a higher impact at Day 5 as compared with Day 3 ( Figures 2D, F).
We then investigated the expression of SIRPg and CD47 on Tregs by gating on CD3 + CD4 + CD25 + CD127 low cells and demonstrated that both molecules are highly expressed on naïve (CD45RA + ) and memory/activated (CD45RA − ) Tregs while CD47 is significantly less expressed by memory/activated Tregs ( Figures 2G-I). Prior to questioning the role of SIRPg on Treg functions, we evaluated the consequences of SIRPg blockade on T-cell proliferation upon CD3 and CD28 stimulations. As shown in Figure 2K, the specific blockade of SIRPg by KWAR23 slightly reduces T-cell proliferation. We have previously shown that CD28 stimulation can inhibit Treg function (21), thus in an experimental set-up where T effector cells were only stimulated with coated OKT3, we analyzed the impact of SIRPg blockade on Tregs suppressive function. After 5 days of culture in anti-CD3coated wells, effector T cells (Teff) proliferated, while Tregs cultured in the same conditions did not ( Figure 2J). When cocultured at a 1:1 Teff : Treg ratio, Tregs were able to inhibit Teff proliferation regardless of the presence or not of KWAR23 ( Figure 2L), suggesting that SIRPg is not involved in Tregs  Figures S1D, E). SIRPg was not expressed on B cells neither on NK cells (data not shown) as already reported by Piccio (16).
Altogether, our results show that SIRPg expression varies upon T-cell differentiation and T-cell stimulation, suggesting that this molecule could impact T-cell functions.   Figure 3A); SIRPg and CD47 therefore do not seem to colocalized on Jurkat cells. In the absence of SIRPg, i.e., on SIRPg KO Jurkat cells, CD47 expression is not altered and seems equivalent to that observed on WT Jurkat cells ( Figure 3A). Surprisingly, the knockout of CD47 modifies the patterns of SIRPg. SIRPg is expressed on the whole cell surface instead of being in a big punctuate pattern at a specific localization of the cell surface as in WT Jurkat cells ( Figure 3A). This might suggest that CD47 would impact SIRPg distribution on the plasma membrane but not reciprocally. We also evaluated the expression of SIRPg and CD47 on freshly isolated T cells; we identified a similar expression as on WT without any colocalization ( Figures 3B, C).
Given that SIRPg and CD47 expression varies upon T-cell stimulation ( Figures 2D-F), we analyzed their distribution on resting and on activated T cells. Upon T-cell stimulation, SIRPg is upregulated and expressed on the whole T-cell surface as compared with resting T cells (i.e., freshly isolated T cells), in which SIRPg is localized in big punctuate patterns ( Figure 3B). However, CD47 distribution does not change upon T-cell activation ( Figure 3B). The Pearson value clearly showed that there is no colocalization of SIRPg and CD47 neither at steady state nor upon T-cell activation since the value obtained were not close to 1 ( Figure 3C).

SIRPg Polarizes at the Immune Synapse
SIRPg/CD47 interaction has been shown to enhance superantigen-mediated T-cell costimulation (16); we therefore hypothesized that SIRPg might be localized at the immune synapse. Interestingly, we observed a polarization of SIRPg by time-lapse microscopy when Jurkat and SEE-primed Raji cells are in close contact. Jurkat cells emit pseudopods when they are closed to Raji cells and SIRPg clearly polarized at the cell-cell contact ( Figures 4A, B; Supplementary Video S1). To ensure that this cell-cell contact was a synapse, in one experiment, cells were fixed, permeabilized, and stained with phalloidin. The staining showed a polarization of actine at the cell-cell contact indicating the synapse formation ( Figure 4C). Interestingly, CD47 (stained only on Raji cells) was localized at the immune synapse ( Figure 4C). We obtained similar results when using T cells (stained with an anti-SIRPg Ab) from healthy volunteer (HV) rather than Jurkat ( Figure 4D). We further quantified SIRPg expression intensity at the synapse and on the opposite site of the cell as explained in Figure 4E in all acquired time-lapse microscopy videos and showed a clear SIRPg accumulation at the immune synapse ( Figure 4F). Interestingly, blocking the SIRPag/CD47 interactions with KWAR23 mAb did not affect SIRPg polarization at the synapse ( Figure 4F), suggesting that SIRPg spatial reorganization of the synapse does not depend on its interaction with CD47 on the Raji cells.

Blocking SIRPg/CD47 Interaction Impairs IFN-g Secretion in Chronically Activated T Cells
To further decipher the importance of the SIRPg/CD47 interaction on T-cell function, we stimulated T cells with low doses of OKT3 (0.5 or 1 µg/ml) and CD47-Fc, both coated, to provide a costimulatory signal as shown by Piccio and colleagues (16). With an experiment setup different than the one used by Piccio, we showed that the addition of coated CD47-Ig during Tcell stimulation did not increase their proliferation ( Figure 5A);  however, coated CD47-Ig increases T-cell metabolism independently of its interaction with SIRPg since the addition of KWAR23 has no impact ( Figure 5B). Importantly, coated CD47-Fc significantly enhances IFN-g secretion of OKT3stimulated T cells; once again the addition of KWAR23 in the culture has no effect ( Figure 5C). Our results show that T cells freshly stimulated with coated OKT3 and coated CD47-Fc benefited from CD47 in a SIRPg-independent pathway. CD47 is upregulated by different cells upon chronic infection, notably viral infection (22), and tumor cells. We therefore evaluated the impact of CD47-Fc stimulation on T cells chronically stimulated with three rounds of anti-CD3 anti-CD28 stimulations. Chronically activated T cells further stimulated with CD47-Fc kept their fitness as shown in the Figure 5D. However, their culture on a OKT3-CD47 coating increased their IFN-g secretion ( Figure 5E). To evaluate the role of the SIRPg/CD47 interaction in this observation, the blocking anti-SIRPg antibody KWAR23 or the blocking anti-CD47 antibody B6H12 were respectively added. As shown in Figure 5F, the addition of KWAR23 did not impair cell fitness, while B6H12 seriously impaired T-cell viability when used at 10 µg/ml ( Figure 5F). This mAb has already been shown to induce T-cell apoptosis (23). KWAR23 significantly decreased IFN-g secretion while a tendency was observed with the blocking anti-CD47 antibody B6H12 ( Figure 5G). This observation suggests a role of the SIRPg/CD47 interaction in chronically activated T cells.

Blocking SIRPa-g/CD47 Interactions In Vivo Impairs Human T-Cell Chimerism and Delays the Xeno-GvHD Onset
Since SIRPg is absent in rodents, we took advantage of a graftversus-host disease (GvHD) model, in which human PBMC injected in NSG-recipient mice would be chronically stimulated by the xenogeneic environment. The role of SIRPg in this experimental condition was evaluated by treating or not the irradiated recipient mice with 5 mg/kg of KWAR23 twice a week. In this situation, the xenogenic GvHD is human T cells  driven due to their cross-reactivity with host MHCs (24). We first observed that mice survival was significantly extended when treated with KWAR23 ( Figure 6A). At Day 6, mice treated with the KWAR23 showed a tendency of a lower human chimerism, although the frequency of human cells within the blood was quite low 1-week posthumanization. This observation was clearly significant at Day 14 ( Figure 6B), suggesting that SIRPg blockade impairs T-cell survival, activation, proliferation, or localization within the host. This chimerism mostly comprised human CD3 + cells in both groups as shown by blood phenotypic analysis at Day 14 ( Figure 6C). Human CD8 T cells played a primary role in xeno-GvHD pathogenicity (24); however, the CD4/CD8 ratio analyzed at D14 did not show an accumulation of CD8 T cells in the blood ( Figure 6D). At Day 14, lower absolute numbers of human CD4 and CD8 circulating T cells were detected in the KWAR23-treated group ( Figure 6E) which could explain the delay of the observed xeno-GvHD onset. Mice with a NOD background such as NSG mice, are superior to other strains for human immune cell engraftment thanks to their SIRPa whose affinity for human CD47 is even higher than the one for their natural ligand (mCD47), ensuring a functional "don't eat me" signal (25). NSG mice are known to have functional residual macrophages. Given the weak chimerism observed in the KWAR23-treated group and the specificity of the KWAR23 to block human SIRPa-g/hCD47 interactions, we wondered whether the KWAR23 cross-reacted with mouse SIRPa leading to the disruption of mSIRPa/hCD47 don't eat me signal and eventually the phagocytosis of human cells. Intraperitoneal macrophages collected from NSG mouse are all CD11b + SIRPa + and are not stained by the KWAR23 (Supplementary Figure S2A), ruling out our previously depicted hypothesis.
We evaluated the consequences of the KWAR23 treatment on T-cell activation. We analyzed PD1 expression on human circulating cells. All human T cells were PD1 + regardless of the treatment ( Figure 6F). Therefore, the delay of the GvHD onset in the KWAR23-treated group was not due to a defect of human Tcell activation. In vitro, we have shown that chronically activated T cells secreted less IFN-g when SIRPg/CD47 interaction was blocked ( Figure 5G). Thus, we measured IFN-g concentration in mice's sera collected at Day14. As expected, KWAR23-treated mice secreted significantly less IFN-g than control mice ( Figure 6G). This can be explained either by the lower amount of human T cells within the KWAR23-treated group, or by a direct effect of the treatment on chronically activated T cells (as shown in vitro in Figure 5G). Because human T cells are fully activated in the xenogeneic environment, we wondered if the KWAR23 treatment would impact the CD47's downmodulation observed in vitro following T-cell stimulation ( Figure 2F). Indeed, the downmodulation of human CD47 might reduce the "don't eat me" signal and increase their susceptibility to residual mouse macrophages. KWAR23 did not stress CD47 down-regulation (Supplementary Figure S2B) indicating that the treatment would not increase T-cell susceptibility to mouse macrophages.
We then evaluated SIRPg on human T cells in both groups by using the anti-SIRPg mAb clone LSB2.20. We first validated that the KWAR23 did not impair the staining using LSB2.20 (Supplementary Figure S2C). The frequency of human CD4 and CD8 SIRPg + T cells in the KWAR23-treated group was not different to that in the PBS-treated mice (Figures 6H, I). However, SIRPg geometric mean was significantly lower in the KWAR23-treated group (Figures 6H-J).
Thus, blocking SIRPg/CD47 interaction in vivo impaired human T-cell chimerism by specifically affecting human T-cell numbers leading to the delay of xeno-GvHD.
Blocking SIRPa-g/CD47 Interactions In Vivo Impair Human T-Cell Differentiation Because SIRPg expression varies upon T-cell differentiation in the blood from HV (Figures 2A-C), we evaluated the impact of the treatment on T-cell differentiation in vivo. Three out of four HV whose PBMC were i.v. injected were phenotyped prior to their injection in NSG mice; naïve CD4, central memory CD4, and central memory CD8 T cells were the most abundant subpopulations (Supplementary Figures S3A, B). After i.v. hPBMC injection in untreated recipient mice, naïve T-cell compartment shrank in favor to central memory and effector memory cells regardless of the treatment. Whether or not this observation reflects naïve T-cell differentiation into central memory and effector memory cells, together with a memory Tcell compartment activation is difficult to evaluate. However, the inhibition of SIRPa-g/CD47 interactions in vivo did not bias neither CD4 nor CD8 T-cell subpopulation frequencies ( Figures 7A, B) while it clearly affected the absolute number of CD4 and CD8 T in all subpopulations except the TERMA CD8 T cells (Figures 7C, D). To further document the effect of KWAR23-treatment, some mice were sacked and their splenocytes were analyzed at Day 14. We confirmed that human cell engraftment was impaired in the spleen also upon KWAR23 treatment ( Figure 8A). The CD4/CD8 ratio was significantly lower in the KWAR23-treated group ( Figure 8B). A relative number of CD4 was significantly lower in the KWAR23-treated group ( Figure 8C) without affecting the frequency of CD4 or CD8 T-cell subpopulations ( Figures 8D, E) nor the relative number of those subpopulations (Figures 8F, G).
Altogether, our results suggest that KWAR23 treatment affects survival of all human T-cell subpopulations independently of mice macrophage implication.

DISCUSSION
SIRPg protein is only expressed in humans and nonhuman primates (26) and binds CD47, an ubiquitous protein with an affinity 10-fold lower than that of SIRPa for CD47 (15). So far, reports on the potential role of the SIRPg were performed with an anti-SIRPg antibody (clone LSB2.20) which has not been appropriately characterized limiting the interpretation of previous reports on the SIRPg biology. Thus, we decided to reevaluate the impact of SIRPg in human T-cell biology using a different anti-SIRP monoclonal antibody (clone KWAR23) which we characterized here as capable to efficiently block SIRPg and SIRPa interactions with CD47. Addressing this question by blocking at the CD47 protein as previously reported (16) is challenging since CD47 is a multifacetted target engaging with various ligands in cis and in transinteraction (27,28), and some anti-CD47 mAb known to block SIRP/CD47 interaction (clone B6H12) can also trigger T-cell apoptosis (23).
Here, we report that SIRPg is expressed by all naïve and central memory T cells and is upregulated after activation as previously reported (15), confirming that T cells might benefit (J) SIRPg geometric mean on human CD4 and CD8 T cells. Data were analyzed with a one-way ANOVA test followed by Kruskal-Wallis. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001; ns, not significant.
Dehmani et al. SIRPg/CD47 Implication on Chronically Activated T Cells from this expression. While a previous study (16) reported that anti-SIRPg (LSB2.20 clone) or anti-CD47 (clone B6H12) negatively impact human T-cell activation, we were first surprised that the specific blockade of SIRPg/CD47 interaction using appropriate tools did not seem to directly impact T-cell activation upon one round of stimulation at least on T-cell functions addressed in our experiments, i.e., proliferation, metabolism, and IFN-g secretion which are one of the many aspects of T-cell activation. The benefits of CD47-Ig coating on OKT3-stimulated T cells might therefore be dependent on a trans CD47/CD47 interaction as mentioned by others (27,28). SIRPg extracellular portion consists of three domains (D1, D2, and D2), and isoforms containing D1/D2/D3, D1/D2, or D1 only, have been described (14,16). SIRPg/CD47 interaction was suggested to occur the SIRPg D1 portion (29,30). The LSB2.20 was raised against a SIRPg-D1/D2-Ig fusion protein. Thus, one could consider that LSB2.20 epitope is not on the same extracellular domain than CD47-binding site. This would explain that targeting SIRPg with either CD47-Ig or LSB2.20 on T cells would differently act on their effector functions as presented here and in a former study by Piccio for CD47-Ig or LSB2.20 stimulation, respectively (16). Recently, a patent by Ruozhen and colleagues in which epitopes of different anti-SIRPg are mentioned, indicates that LBS2.20-binding domain is on SIRPg D1 (31), so on the same domain than the CD47binding site. Furthermore, they also showed LSB2.20 inefficient to increase T-cell proliferation upon CD3 stimulation (31).
Given the broad distribution of CD47, one could consider that its implication in T-cell biology through SIRPg ligation should be finely tuned. Thus, we hypothesized that an upregulation of CD47 could impact SIRPg ligation and lead to a biological effect. Indeed, CD47 is upregulated in a number of cancer cells, notably by MYC oncogene, a transcription factor, which binds promotor of cd47 and Pdl-1 genes (32). CD47 upregulation has been very recently shown in infected cells (22,33). Here, we describe that SIRPg clusters at the immune synapse and CD47 on the target cells also clusters at the cell/cell contact in a juxtacrine fashion. These observations confirm that SIRPg expression and reorganization should also benefit human T cells. Unlike CD28 (34) on T cells or SIRPa on macrophages (35), SIRPg localization at the immune synapse seem independent of its ligation to its known ligand CD47. This may reflect that SIRPg/CD47 interaction needs to exceed a threshold for its effector functions, explaining SIRPg localization to the immune synapse independently of CD47 engagement as a sensor searching for sufficient CD47 expression on the target cells.
It is worth noting that we could not reveal CD47 on WT Jurkat cells when using SIRPg-Fc by FACS, while the SIRPa-Fc could. This may be due to the 10-fold lower affinity of SIRPg, to the level of CD47 expressed by Jurkat cells or to the conformation of either CD47 or SIRPg. pH may influence protein/protein interactions as acidic pH has just unrevealed VISTA as a new ligand of PSGL1 (36), given that CD47 is  upregulated by tumor cells in a usually acidic environment, pH variation might also modify CD47/SIRPg affinity. In this study, all our experimental readouts focused on T-cell biology considering SIRPg as a ligand. However, the 4 aa long SIRPg intracellular tail is unexpected to transduce a signal by its own. Unlike SIRPb, SIRPg is not associated with DAP12 (4,15,16). Recently, the polymorphism of SIRPa (i.e., SIRPa-V1 and SIRPa-V2) has been shown to stimulate CD47 when donor differed from the recipient in one or both Sirpa alleles resulting in innate immune responses (37). Although these experiments were addressed in mice and not yet confirmed in humans, one can consider that upregulation of SIRPg could contribute in a CD47 signaling.
Interestingly, all poxviruses that infect vertebrates (except four) contain predicted vCD47s (38). Even though their amino acid sequence similarity with their host parent gene is comprised between 18% and 29%, all of them conserved the structure with a N-terminal IgV domain, five transmembrane helices, and a short cytoplasmic tail (38). Although smallpox was recently eradicated by vaccination, the virus was present for thousands of years; SIRPg may have interacted with vCD47. Recently, Myers and colleagues have shown that mice chronically infected with Friend retrovirus or LCMV Clone 13 accumulate some SIRPa + CD8 + T cells (39). They also evidenced this population in patients with chronic HCV infections by using a CyTOF approach. The expression of SIRPa was lower in these CD8 T cells than in monocytes. In mice, the SIRPa expression on T cells was associated with effector functions; in humans, SIRPa + CD8 + T cells show higher activation status as compared with their SIRPa − CD8 + counter parts (39). KWAR23, blocking anti-SIRPg antibody, treatment was mostly effective in chronically activated T cells in vitro ( Figure 5) and in vivo ( Figure 6A). We previously reported that SIRPg/CD47 blockade inhibits IFN-g release in human tumor-antigen T-cell cross-priming assays using a CD8 + T-cell clone isolated and expanded from a melanoma patient (40,41). Altogether, this suggests that SIRPg/CD47 interaction might be more important in the activation biology of memory or chronically activated human T cells, than in naïve T-cell costimulation as previously hypothesized. Moreover, SIRPg/CD47 interaction could play other roles, such as positively control the human T-cell transendothelial migration under shear flow conditions (17) or tissue migration (40). Since KWAR23 antibody blocks both SIRPa and SIRPg interactions with CD47, we might consider that the observed effects were not only due to the sole SIRPg blockade but also to SIRPa/CD47 blockade. However, on one hand in vivo, this antibody displays no cross-reactivity for mouse SIRPa and on the other hand in vitro in our chronically activated human T-cell models, only T cells remained after several polyclonal stimulation rounds. Finally, we were surprised that the SIRPg spatial reorganization at the synapse did not depend on its interaction with CD47. As previously mentioned, SIRPg affinity for CD47 is relatively low and CD47 homophilic interactions have already been shown (27) which could explain CD47 clustering at the Raji membrane, although CD47KO target cells (here Raji) would be needed to fully confirm our observation. SIRPg might interact with a yet unknown other receptor (beside CD47) which interaction might not be blocked by the KWAR23 antibody.
In the study, using a blocking anti-SIRPa-g mAb, we report that T-cell interaction with CD47 is of importance not only for T-cell migration but also for their activation particularly in chronic stimulation situations. Whether these observations only involve SIRPg or SIRPg and SIRPa proteins remain to be clarified.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

ETHICS STATEMENT
Blood from healthy individuals was obtained at the Etablissement Francais du Sang (Nantes, France). Written informed consent was provided according to institutional guidelines. The patients/participants provided their written informed consent to participate in this study. This study was carried out according to authorization from the French Ministry of Research, APAFIS # 4678.

ACKNOWLEDGMENTS
This work was realized in the context of the IHU-Cesti project which received French government financial support managed by the National Research Agency via the "Investment Into the Future" program ANR-10-IBHU-005. The IHU-Cesti project is also supported by Nantes Metropole and the Pays de la Loire Region. Supplementary Figure S2 | The KWAR23 treatment does not affect the analysis of human cells in vivo. (A) Intraperitoneal NSG cells were collected and stained for mouse SIRPa and mouse CD11b. The cross reactivity of the KWAR23 (antihuman SIRPa-g) against mSIRPa was evaluated. CD11b and SIRPa + expression on intraperitoneal macrophages is presented as dot plot and the cross reactivity of the KWAR23 (revealed with the antihuman IgG PE) is presented in the red histogram while the control staining with the antihuman IgG PE only is presented in the black histogram (upper panel). In the same experiment, human PBMC were used as control and were either stained with KWAR23 and revealed with the antihuman IgG PE (red histogram) or were only stained with the antihuman IgG PE (black histogram). (B) Geometric mean (± SEM) of CD47 expression on CPD-stained T cells is presented as a function of their division numbers, in the presence (red) or not (black) of KWAR23 during an anti-CD3+ anti-CD28.2 stimulation. n = 4 HV analyzed in one experiment. (C) SIRPg staining with LSB2.20 on human CD3 + PBMC (left dot