Anti-GARP protein

Abstract

The present invention relates to a protein binding to GARP in the presence of TGF-β and uses thereof.

Claims

1. An anti-human Glycoprotein A repetitions predominant (hGARP) antibody, comprising a heavy chain variable region comprising the following CDRs: TABLE-US-00034 VH-CDR1: (SEQ ID NO: 52) GYGIN;  VH-CDR2: (SEQ ID NO: 3) MIWSDGSTDYNSVLTS; and VH-CDR3: (SEQ ID NO: 4) DRNYYDYDGAMDY. a light chain variable region comprising the following CDRs: TABLE-US-00035 VL-CDR1: (SEQ ID NO: 5) KASDHIKNWLA; VL-CDR2: (SEQ ID NO: 6) GATSLEA; and VL-CDR3: (SEQ ID NO: 7) QQYWSTPWT.

2. An antibody or antigen binding fragment produced by a hybridoma registered under the accession number LMBP 10246CB on May 30, 2013.

3. A hybridoma cell line producing an antibody against GARP registered under the accession number LMBP 10246CB on May 30, 2013.

4. A polynucleotide sequence encoding the antibody molecule of claim 1.

5. An expression vector comprising the polynucleotide sequence of claim 4.

6. An isolated host cell comprising the vector of claim 5.

7. A method of producing an anti-hGARP antibody, comprising culturing the host cell of claim 6 under conditions to express the antibody, and recovering the expressed antibody.

8. The antibody of claim 1, which is labeled with a detectable label, wherein the detectable label comprises at least one element, isotope, or chemical compound attached to the antibody to enable detection.

9. A kit comprising the antibody of claim 1, wherein the kit is used to identify the presence of a complex of hGARP and latent TGF-β1 in a biological sample.

10. A kit comprising the antibody of claim 1, wherein the kit is used to identify activated Tregs in a biological sample.

11. The anti-hGARP antibody of claim 1, wherein the heavy chain variable region comprises a VH-CDR1 having the amino acid sequence of GFSLTGYGIN (SEQ ID NO:2).

12. The anti-hGARP antibody of claim 1, wherein the antibody is a humanized antibody.

13. The antibody of claim 12, which is an IgG1 or IgG4 antibody.

14. The anti-hGARP antibody of claim 11, comprising a heavy chain variable region having the amino acid sequence of SEQ ID NO:50, and a light chain variable region having the amino acid sequence of SEQ ID NO:51.

15. The anti-hGARP antibody of claim 11, comprising a heavy chain variable region having the amino acid sequence of SEQ ID NO:8, and a light chain variable region having the amino acid sequence of SEQ ID NO:9.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1. New monoclonal antibodies that recognize human GARP on the cell surface. Murine BW5147 T cells, transfected or not with human GARP (hGARP) were stained with biotinylated in-house anti-hGARP antibodies (MHGARP1 to 9) and streptavidin-PE (SA-PE, top panels), or with a commercial anti-hGARP antibody (clone Plato-1) and secondary anti-mouse IgG2b coupled to AlexaFluor 488 (AF488, bottom panels).

(2) FIGS. 2A-2B. MHGARP8 inhibits active TGF-β production by a human Treg clone. Clone Treg A1 was stimulated during 24 hours with anti-CD3/CD28 antibodies, 10 alone or in the presence of the indicated anti-hGARP mAbs (20 μg/ml). (FIG. 2A) Cell lysates were analyzed by WB with anti-pSMAD2 and anti-β-ACTIN antibodies. (FIG. 2B) Quantification of ECL signals from WB shown in A.

(3) FIG. 3A-3E. (FIG. 3A) Regions in the hGARP protein required for binding by anti-hGARP antibodies. Murine BW5147 T cells expressing the HA-tagged proteins schematized on the left were stained with anti-hGARP (MHGARP1 to 9, as indicated on top of the figure) or anti-HA antibodies, and analyzed by flow cytometry. Histograms are gated on live cells. Based on the FACS results, regions required for binding by the various MHGARP mAbs were identified and are indicated by horizontal bars above the representations of the HA-tagged chimeras.

(4) FIG. 3B Abundance of the epitope recognized by MHGARP-8 increases upon overexpression of TGF-β1. Parental BW5147 T cells (BW untransfected) or clones stably transfected with hGARP alone (BW+hGARP) or with hTGFB1 (BW+hGARP+hTGF-b1) were stained as in A, or with anti-mLAP-AF647 or anti-hLAP-APC antibodies, and analyzed by flow cytometry.

(5) FIG. 3C MHGARP-1, -2, -3, -4 and -5 recognize free hGARP, but not hGARP bound to TGF-β1. Cell lysates from parental BW5147 T cells or a clone stably transfected with hGARP and hTGFB1 were immunoprecipitated with anti-hGARP mAbs (MHGARP1 to 9, as indicated on top of the figure). Cell lysates (30% input) or IP products were analyzed by Western blot with a commercial anti-hGARP mAb (clone Plato-1, top panels) and with an antibody directed against a C-terminal epitope of TGF-β1, which detects pro-TGF-β1 as a 50 kDa band and mature TGF-β1 as a 13 kDa band (bottom panels). * Aspecific product detected in untransfected cells.

(6) FIG. 3D Overexpression of hTGFB1 in hGARP-transfected 293T cells decreases binding of MHGARP-1, -2, -3, -4, and -5, but increases binding of MHGARP-8. 293T cells were co-transfected with a hGARP-encoding plasmid (0.25 μg), the indicated amounts of a hTGFB1-encoding plasmid, and an empty plasmid to bring the total amount of transfected DNA to 2.5 μg in all conditions. Transfected cells were stained with anti-hGARP mAbs (MHGARP1 to 9, as indicated on top of the figure), and analyzed by flow cytometry.

(7) FIG. 3E Silencing of hTGFB1 in hGARP-transduced JURKAT cells decreases binding of MHGARP-8. JURKAT cells, transduced or not with hGARP, were transfected with siRNA specific for the TGFB1 mRNA (siTGFB1) or a scramble siRNA control. Transfected cells were stained with anti-hGARP mAbs (MHGARP1 to 9, as indicated on top of the figure) or with an anti-hLAP antibody, and analyzed by flow cytometry.

(8) FIGS. 4A-4B. Presentation of hTGF-β1 on the cell surface is not sufficient for binding by MHGARP8. 293T cells were transfected as indicated below, stained with anti-hLAP antibodies or with MGARP8, then analyzed by flow cytometry.

(9) FIG. 4A Transfection with constructs encoding the HA-tagged proteins schematized on the left, without a hTGFB1 construct.

(10) FIG. 4B Co-transfection with constructs encoding the HA-tagged proteins schematized on the left, with a hTGFB1 construct.

(11) FIG. 5. Binding of MHGARP-2, -3 and -8 requires amino-acids 137-138-139 of hGARP. Parental BW5147 T cells (BW untransfected) or clones stably transfected with plasmids encoding HA-tagged forms of hGARP were stained with the indicated anti-hGARP or anti-HA antibodies, and analyzed by flow cytometry. The HA-tagged forms of hGARP tested here comprised aa 20-662 of hGARP (wild type, WT), or aa 20-662 of hGARP in which groups of 3 amino-acids located in region 101-141 were replaced by the amino-acids found in the corresponding region of mGARP (Mut I, Mut II and Mut III). Amino-acid sequences of region 101-141 of hGARP-WT (SEQ ID NO: 53), -Mut I (SEQ ID NO: 54), -Mut II (SEQ ID NO: 55), -Mut III (SEQ ID NO: 56) and mGARP (SEQ ID NO: 57) are indicated on the left. Amino-acids that differ between human and mouse GARP are highlighted by grey vertical boxes, and amino-acids mutated in Mut I, Mut II and Mut III are indicated by black horizontal boxes.

(12) FIGS. 6A-6B. MHGARP8 inhibits Treg function in vivo. On day 0, the indicated groups of NSG mice received i.v. injections of human PBMCs, in combination or not with human Tregs. Mice from groups III and IV were treated with the MHGARP8 antibody, injected i.p. once a week, starting on day −1. Objective signs of GvHD development in the recipient mice were monitored bi-weekly. A GvHD score was established based on weight loss (0: <10%; 1: 10%-20%; 2: >20%; 3: >30%), anemia (0: red or pink tail; 1: white tail), posture (0: normal; 1: hump), general activity (0: normal; 1: limited), hair loss (0: no hair loss; 1: hair loss) and icterus (0: white or red tail; 1: yellow tail). Maximum disease severity or death corresponded to a score of 7. FIG. 6A Experiment 1. Values represent mean scores. (FIG. 6B) Experiment 2. Values represent mean scores+SEM.

(13) FIGS. 7A-7B. New anti-hGARP mAbs. (FIG. 7A) Schematic representation of the experimental strategies used to derive anti-hGARP mAbs. (FIG. 7B) Flow cytometry analyses of clone ThA2 (human CD4+Th cells which do not express hGARP), or ThA2 cells transduced with hGARP, after staining with biotinylated MHG-1 to -14 mAbs and streptavidin coupled to PE (SA-PE), with LHG-1 to -17 mAbs and a secondary anti-hIgG1 antibody coupled to PE, or with a commercially available mouse anti-hGARP mAb (clone Plato-1) and a secondary anti-mIgG2b antibody coupled to AF647.

(14) FIGS. 8A-8B. Immune responses from immunized llamas. (FIG. 8A) shows immune responses from DNA immunized llamas. (FIG. 8B) shows immune responses from llamas immunized with BW cells expressing hGARP/hTGFβ.

(15) FIG. 9. Cross-reactivity to cynomolgus GARP-TGFβ measured on cells by FACS. 293E cells were transfected with human/cyno GARP and human/cyno TGFB. LHG-10-D and the affinity optimized variants are cross-reactive with cynomolgus GARP-TGFB.

(16) FIG. 10. Sequences of LHG-10 antibodies and its shuffle variants. FIG. 10 discloses SEQ ID NOS 34, 35, 35, 37, 39, 36, 38, 35, 37, 39, 36, 38, 13-15 and 19-33, respectively, in order of appearance.

(17) FIGS. 11A-11B. MHGARP8 and LHG-10 inhibit production of active TGF-β by human Tregs. After a short in vitro amplification, human CD4+CD25hiCD127lo cells (Tregs) were re-stimulated with anti-CD3/CD28 coated beads during 24 hours, in the presence or absence of the indicated mAbs (10 μg/ml). Cells lysates were analyzed by Western Blot with antibodies against phosphorylated SMAD2 (pSMAD2), as a read-out for active TGF-β production, or β-ACTIN (loading control).

(18) FIGS. 12A-12B. MHGARP8 and LHG-10 inhibit the suppressive activity of human Tregs in vitro. (FIG. 12A) Freshly isolated human CD4+CD25−CD127hi cells (Th; 2×10.sup.4 per microwell) were seeded alone or with clone Treg A1 (Stockis, J. et al. Eur. J. Immunol. 2009, 39:869-882) at a 1/1 Treg to Th ratio. Cells were stimulated with coated anti-CD3 and soluble anti-CD28, in the presence or absence of the indicated anti-hGARP mAbs (10 μg/ml). 3H-Thymidine (3H-Thy) was added during the last 16 hours of a 4-day culture and incorporation was measured in a scintillation counter as a read-out for proliferation. Bar histograms indicate kcpm (means of triplicates+SD). Clone Treg A1 did not proliferate in the absence of Th cells (Treg alone: 0.5±0.04 kcpm). Suppression of Th proliferation in the presence of Tregs is indicated above each black bar, and is calculated as follows: % suppression=1−(kcpm (Th alone)/kcpm (Th+Treg). (FIG. 12B) Clone ThA2 cells (Th; 1×10.sup.4 per microwell) were seeded with clone Treg A1 at the indicated Treg to Th ratios, in the presence or absence of MHGARP8 (MHG-8), anti-hTGF-11 mAb (clone 1D11) or an isotype control (mIgG1). Stimulation, measure of proliferation and calculation of suppression were performed as in A.

(19) FIGS. 13A-13B. Forms and regions of GARP bound by anti-GARP mAbs. (FIG. 13A) Schematic representations of GARP and GARP/TGF-β complexes. Protein GARP is represented by a thick curved grey line. Numbers indicate amino-acid positions. TGF-β is represented with the Latency Associated Peptide (LAP) as thick black lines, and the mature TGF-β1 peptide as thick straight grey lines. Thin black lines represent inter-chain disulfide bonds. (FIG. 13B) Classification of anti-hGARP mAbs based on their binding requirements.

(20) FIGS. 14A-14B. Three groups of anti-hGARP mAbs bind free GARP only, free GARP and GARP/TGF-β1 complexes, or GARP/TGF-β1 complexes only, 5 respectively. (FIG. 14A) Cell lysates of BW cells transfected with hGARP and hTGFB1 were immunoprecipitated with the indicated anti-hGARP mAbs. Total lysates (BW+hGARP+hTGFB1 or untransfected controls) and IP products were analysed by Western Blot with antibodies against hGARP (clone Plato-1), LAP or the mature TGF-β peptide. (FIG. 14B) Flow cytometry analyses of 293T cells, untransfected or transfected with hGARP, hTGFB1 or both, and stained as indicated with anti-LAP-APC, biotinylated MHG mAbs and streptavidin-PE, clone Plato-1 and anti-mIgG2b-AF647, or LHG mAbs and anti-hIgG1-PE.

(21) FIGS. 15A-15B. Amino-acids of hGARP required for binding by MHG and LHG mAbs. (FIG. 15A) Flow cytometry analyses of 293T cells transfected with plasmids encoding the HA-tagged mGARP/hGARP chimeras schematized on the left (numbers represent amino-acid positions in hGARP). Cells were stained with biotinylated MHG mAbs and strepatvidin-PE, LHG mAbs and anti-hIgG1-PE, or anti-HA and anti-mIgG1-AF647. hTGFB1 was co-transfected with mGARP/hGARP chimeras for the analyses of mAbs that bind hGARP/hTGF-β1 complexes only (LHG-3, MHGARP8 (MHG-8), LHG-10). (FIG. 15B) As above, except that 293T cells were transfected with plasmids encoding mutated forms of full-length HA-tagged hGARP. In each mutant, 3 amino-acids of hGARP were replaced by the 3 amino-acids found in mGARP, as illustrated in the alignment on the left (numbers represent amino-acid positions in hGARP). FIG. 15B discloses SEQ ID NOs 53-57, respectively, in order of appearance.

(22) FIGS. 16A-16C. Inhibition of human Treg function by anti-hGARP in vivo. (FIG. 16A) shows the protocol on day 0, the indicated groups of NSG mice received i.v. injections of human PBMCs, in combination or not with human Tregs. (FIG. 16B) shows the results of 4 independent experiments (I to IV), performed with cells from donors A, B or C, with the indicated numbers of mice per group (n). Disease onset is the day when mean disease score becomes ≥1, and is indicated for 3 experimental groups in which mice were grafted with PBMCs only (group a), PBMCs and Tregs (group b), or PBMCs and Tregs and treated with MHGARP8 (MHG-8) (group c). (FIG. 16C) Detailed results from experiment IV, showing the evolution of mean disease score (left) and survival curves (right) in the indicated groups of mice. Statistical significance of differences between groups b (PBMCs+Tregs) and c (PBMCs+Tregs+MHG-8) were calculated using 2-way Anova analysis for progression of disease scores (p=0.0001), and a Log-rank (Mantel-Cox) test for survival (p=0.0027).

EXAMPLES

(23) The present invention is further illustrated by the following examples.

Example 1: New Monoclonal Antibodies Directed Against Human GARP (Anti-hGARP Monoclonals)

(24) DBA/2 or Balb/c mice were immunized with murine P1HTR cells transfected with human GARP. Sera from immunized mice were tested for the presence of anti-hGARP antibodies, by screening for binding to hGARP-expressing BW cells by FACS. Splenocytes from mice with high titers of anti-hGARP antibodies were fused to SP2/neo cells. Hybridomas were selected in HAT medium and cloned under limiting dilution. Supernatants of +/−1600 hybridoma clones were screened by FACS for the presence of antibodies binding to hGARP-expressing BW cells. We identified 38 clones producing anti-hGARP monoclonals in this screening. Nine clones were selected and amplified for large scale-production and purification of 9 new anti-hGARP monoclonals (MHGARP1 to 9).

(25) As shown in FIG. 1, MHGARP1 to 9 bind to murine BW5147 cells transfected with hGARP, but not to untransfected cells. MHGARP1 to 9 also bind 293T cells transfected with hGARP and two human T cells lines (clone Th A2 and Jurkat) transduced with a hGARP-encoding lentivirus, but not the corresponding parental cells (not shown). This recognition pattern is identical to that of a commercially available anti-hGARP mAb (clone Plato-1) used here as a positive control. These results show that MHGARP1 to 9 recognize hGARP on cell surfaces.

(26) As shown in FIG. 7, 5 additional MHGARP antibodies were produced and purified. MHGARP antibodies (MHG-1 to -14 on the figure) do not bind clone ThA2 (human CD4+T helper cells, which do not express hGARP), but bind ThA2 transduced with hGARP.

Example 2: MHGARP8, but None of 12 Other Anti-hGARP Monoclonals, Inhibits Active TGF-β Production by Human Treg Cells

(27) A human Treg clone (1E+06 cells/ml) was stimulated in serum-free medium with coated anti-CD3 (1 μg/ml) and soluble anti-CD28 (1 μg/ml) antibodies, in the presence or absence of 20 jag/ml of an anti-hGARP monoclonal antibody. Thirteen anti-hGARP monoclonals were tested in this assay: our 9 new monoclonals (MHGARP1 to 9), and commercially available antibody clones Plato-1 (Enzo Life Sciences, catalog No. ALX-804-867), 272G6 (Synaptic Systems, catalog No. 221 111), 50G10 (Synaptic Systems, catalog No. 221 011) and 7B11 (BioLegend, catalog No. 352501). Cells were collected after 24 hours, lysed and submitted to SDS-PAGE under reducing conditions. Gels were blotted on nitrocellulose membranes with the iBlot system (Life Technologies). After blocking, membranes were hybridized with primary antibodies directed against phosphorylated SMAD2 (pSMAD2, Cell Signaling Technologies) or β-ACTIN (SIGMA), then hybridized with secondary HRP-coupled antibodies and revealed with Enhanced ChemiLuminescent (ECL) substrate (ThermoFisher Scientific). The presence of pSMAD2 indicates production of active TGF-β1 by the stimulated Treg clone. ECL signals were quantified by measuring the density of the 55 kDa pSMAD2 and 40 kDa β-ACTIN bands on autoradiographs, using the Image J software.

(28) To examine whether hGARP is required for active TGF-β production by TCR-stimulated Treg cells, we stimulated a human Treg clone through its T cell receptor (TCR), alone or in the presence of anti-hGARP mAbs. Active TGF-β produced by stimulated Tregs triggers an autocrine signal, which leads to the phosphorylation and activation of SMAD2 and SMAD3 transcription factors. We measured the presence of phosphorylated SMAD2 (pSMAD2) by Western Blot (WB), as read-out for active TGF-β production by the stimulated Treg clone. As shown in FIG. 2, pSMAD2 was reduced more than 10 fold in the presence of MHGARP8. This reduction is similar to that observed in the presence of an anti-TGF-β mAb, used here as a positive control. None of the 12 other anti-hGARP mAbs (8 other in-house produced MHGARP and 4 commercially available anti-GARP antibodies) inhibited active TGF-β production by the Treg clone. Altogether, our data demonstrate that GARP is required for active TGF-13 production by human Tregs, as MHGARP8, an antibody directed against hGARP, prevented active TGF-β production.

Example 3: MHGARP8, but not Other Anti-hGARP mAbs, Recognizes a Conformational Epitope that Requires the Presence of TGF-β

(29) Mapping the Regions Recognized by Anti-hGARP Monoclonals

(30) Murine BW5147 T cells were electroporated with plasmids encoding the HA-tagged proteins schematized in FIG. 3A, corresponding to hGARP, mGARP or mGARP/hGARP chimeras. Stable clones selected in neomycin were stained with biotinylated anti-hGARP antibodies (anti-hGARP1 to 9) and streptavidin-PE, with the commercial anti-hGARP antibody (clone Plato-1) and a secondary anti-mIgG2b-AF488, or with an anti-HA antibody and secondary anti-mouse IgG1-AF488. Histograms are gated on live cells. Black histograms show signals on untransfected BW cells, white histograms show signals on BW cells expressing the HA-tagged hGARP, and grey histograms show signals on BW cells expressing HA-tagged mGARP or mGARP/hGARP chimeras.

(31) Parental BW5147 T cells (BW non-transfected) or clones stably transfected with hGARP alone (BW+hGARP) or with hTGFB1 (BW+hGARP+hTGF-β1) were stained with biotinylated anti-hGARP antibodies (anti-hGARP1 to 9) and streptavidin-PE, with the commercial anti-hGARP antibody (clone Plato-1) and a secondary anti-mIgG2b-AF488, or with anti-mLAP-AF647 or anti-hLAP-APC antibodies.

(32) We investigated the mechanism by which MHGARP8, but not other anti-hGARP mAbs, inhibits active TGF-β production by Tregs. We hypothesized that MHGARP8 may recognize an epitope in hGARP that is distinct from the epitopes recognized by the other anti-hGARP mAbs.

(33) With the exception of MHGARP-1, our MHGARP mAbs do not recognize murine GARP (mGARP). We thus constructed plasmids encoding HA-tagged hGARP, mGARP or hGARP/mGARP chimeras to map the hGARP regions recognized by our mAbs. We transfected murine BW cells and derived stable clones expressing the HA-tagged proteins (schematically represented in FIG. 3). All clones expressed similar levels of HA-tagged protein on the surface, as indicated by similar fluorescence intensities after staining with an anti-HA mAb (FIG. 3A). As expected, all the MHGARP mAbs bound to the clone expressing HA-tagged hGARP, whereas none, except MHGARP-1, bound to the clone expressing HA-tagged mGARP. Four groups of mAbs emerged from the analysis of binding to the HA-tagged hGARP/mGARP chimeras (FIG. 3A). Monoclonal antibodies in the first group (MHGARP-6, -7 and -9) bound none of the chimeras, indicating that they recognize an epitope located between aa 20 and 101 of hGARP (region 20-101). mAbs in the second group (MHGARP-2, -3 and -8) bound to only 1 of the 5 chimeras, and thus recognize an epitope in region 101-141. A third group comprises MHGARP-5, which bound to 2 of the chimeras and therefore recognizes region 141-207. This group probably also contains MHGARP-1, which is cross-reactive but bound these 2 chimeras more efficiently than it bound mGARP or the 3 other chimeras. Finally, mAbs in the fourth group (MHGARP-4 and Plato-1) bound 4 of the 5 chimeras, and thus recognize region 265-333.

(34) Based on the above, we grouped the anti-hGARP mAbs into 4 families of antibodies that recognize 4 distinct regions of the hGARP protein. MHGARP-8, which displays neutralizing activity, binds to region 101-141. This region is also recognized by MHGARP-2 and -3, which are not neutralizing. Therefore, the ability to bind region 101-141 is not sufficient to confer neutralizing activity.

(35) To further define the epitopes recognized by MHGARP-2, -3 and -8, we compared the binding of the anti-hGARP antibodies to clones of BW cells expressing hGARP alone (BW+hGARP), or hGARP and hTGF-β1 (BW+hGARP+hTGF-β1). With the notable exception of MHGARP8, all anti-hGARP antibodies stained BW+hGARP+hTGF-β1 with the same intensity as BW+hGARP, indicating that the two clones express the same levels of hGARP on the cell surface. The MHGARP8 antibody in contrast, stained BW+hGARP+hTGF-β1 more intensely than BW+hGARP (FIG. 3B). This indicates that although hGARP levels are similar on the two clones, the epitope recognized by MHGARP8 is more abundant on BW+hGARP+hTGF-β1 than on BW+hGARP cells.

(36) A plausible explanation for this observation is that the epitope recognized by MHGARP8 appears only when hGARP is bound to murine (m) or human (h) TGF-β1. This could be due to one of two mechanisms: either the epitope comprises amino-acids from both hGARP and TGF-β1 (mixed conformational epitope), or it comprises amino-acids from hGARP only, but that adopt a different conformation in the presence of TGF-β1 (binding-induced conformational epitope). BW cells express murine TGF-β1, and murine TGF-β1 binds to hGARP (FIG. 3B). Therefore, binding of MHGARP8 to BW+hGARP (in the absence of transfected hTGF-β1) could be due to recognition of hGARP/mTGF-β1 complexes.

(37) To explore the hypothesis that MHGARP8 recognizes GARP when it is bound to TGF-B1, we performed co-immunoprecipitation experiments. We used the different anti-GARP antibodies to immunoprecipitate GARP from BW+hGARP+hTGF-β1 cells, then checked if TGF-β was co-immunoprecipitated with GARP. As shown in FIG. 3C, all anti-GARP antibodies efficiently immunoprecipitated GARP (FIG. 3C, top panels). Co-immunoprecipitation of TGF-β1 was observed with MHGARP-6, -7, -8, and -9 mAbs, indicating that these antibodies bind GARP bound to TGF-β1. In contrast, MHGARP-1, -2, -3, -4 and -5 immunoprecipitated GARP as efficiently as the other anti-GARP mAbs, but they did not co-immunoprecipitate TGF-β (FIG. 3C, bottom panels). This indicates that MHGARP-1, -2, -3, -4 and -5 recognize free GARP, but not GARP that is bound to TGF-β. It is important to note that MHGARP-2 and -3, which require the GARP.sub.101-141 region for binding, recognize only free GARP, whereas neutralizing MHGARP8, which also requires GARP.sub.101-141, recognizes GARP bound to TGF-β.

(38) To confirm this observation, we used 293T cells, which express low levels of endogenous TGF-β1, to co-transfect hGARP with increasing amounts of hTGFB1 (FIG. 3D). Binding of MHGARP-1, -2, -3, -4 and -5 decreased dose-dependently when hTGFB1 was co-transfected with hGARP. It was completely abrogated at the highest doses of hTGFB1. This confirms that MHGARP-1 to -5 bind only free GARP. Binding of MHGARP-6, -7, and -9 was not modified by co-transfection of hTGFB1, indicating that these mAbs bind hGARP whether or not it is bound to TGF-β1 (i.e. they bind both free GARP and GARP bound to TGF-β1). In striking contrast, binding of MHGARP8 increased dose-dependently when hTGFB1 was co-transfected with hGARP. This suggests again that in contrast to all other antibodies, MHGARP8 does not bind free GARP, but only GARP bound to TGF-β1.

(39) To demonstrate that MHGARP8 binding requires the presence of TGF-β1, we used siRNAs to silence the expression of TGFB1 in Jurkat cells transduced with hGARP (FIG. 3E). The siRNA against TGFB1 mRNA efficiently reduced expression of TGF-β1, as illustrated by the decrease in surface LAP detected on Jurkat+hGARP cells (FIG. 3E, right panel). Reduced expression of TGF-β1 in Jurkat+hGARP decreased the binding of the MHGARP8 antibody, but it did not modify the binding of the other anti-GARP antibodies (FIG. 3E, foreground histograms). This confirms that in contrast to the other anti-GARP antibodies, MHGARP8, does not bind free GARP, but only binds GARP in the presence of TGF-β1.

(40) Finally, we sought to exclude the unlikely hypothesis that presentation of TGF-β on the cell surface, irrespective of hGARP expression, is sufficient for binding by MHGARP8. In other words, we sought to demonstrate that MHGARP8 recognizes a mixed or a binding-induced conformational epitope that requires expression of both hGARP and TGF-β. For this, we transfected 293T cells with constructs encoding hGARP, mGARP or the hGARP/mGARP chimeras described above, with or without a construct encoding hTGF-β1. Transfected cells were analyzed by FACS to measure binding of the MHGARP8 antibody, and presentation of hTGF-β 1 on the cell surface with an anti-hLAP antibody (FIG. 4). By comparison to unstransfected cells, transfection of hGARP, mGARP or hGARP/mGARP constructs alone (no hTGFB1) induced low levels of surface LAP, due to low levels of endogenous hTGFB1 expression (FIG. 4A, left). Surface LAP levels dramatically increased upon transfection of hTGFB1 in cells transfected with hGARP, mGARP, or any hGARP/mGARP construct (FIG. 4B, left histogram). This indicates that hTGF-β1 is presented on the cell surface by hGARP, by mGARP and by all the hGARP/mGARP chimeras. Importantly, MHGARP8 bound only to the surface of cells transfected with hGARP, or with the hGARP/mGARP constructs encoding amino-acids 101 to 141 of hGARP (FIGS. 4A and 4B, right). It did not bind to cells transfected with hTGFB1 and mGARP, nor to cells transfected with hTGFB1 and hGARP/mGARP constructs that do not encode hGARP101-141 (FIG. 4B, right), although these cells presented high levels of LAP on their surface (FIG. 4B, left). This demonstrates that presentation of TGF-β1 on the cell surface (by mGARP or hGARP/mGARP chimeras) is not sufficient for binding by MHGARP8. Binding of MHGARP8 requires the presence of both hGARP (region 101-141) and TGF-β1 on the cell surface.

(41) As indicated above, MHGARP8 does not bind mGARP. Its binding to hGARP requires a region comprising amino-acids 101 to 141. To further define the epitope recognized by MHGARP8, we compared the sequences of region 101-141 in human and murine GARP. In this region, only 13 amino-acids differ between hGARP and mGARP (FIG. 5, amino-acids highlighted by grey boxes). We constructed 3 HA-tagged mutant forms of hGARP. In each mutant (Mut I, Mut II and Mut III), 3 consecutive amino-acids were replaced by the corresponding amino-acids of the mGARP protein (FIG. 5, black boxes). We derived stable clones of BW cells transfected with these HA-tagged forms of wild type (WT) or mutant hGARP. All clones expressed similar levels of HA-tagged protein on the surface, as demonstrated by staining with an anti-HA antibody (FIG. 5, histograms on the right). We then analyzed the clones after staining with MHGARP-2, -3 and -8, i.e. antibodies which require region 101-141 of hGARP for binding. The three antibodies bound to cells expressing WT, Mut I and Mut II forms of hGARP. In contrast, binding was greatly reduced on cells expressing the Mut III form of hGARP, indicating that MHGARP-2, -3 and -8 require amino-acids 137-138-139 of hGARP for binding.

(42) Altogether, our data show that MHGARP8 is the only available anti-GARP antibody that inhibits active TGF-β1 production by human Tregs. This neutralizing activity is linked to binding of MHGARP8 to an epitope that is distinct from those bound by all other anti-GARP antibodies: binding of MHGARP8 requires both region 101-141 of hGARP and the presence of hTGF-3, whereas binding of non-neutralizing antibodies require other regions of hGARP (for MHGARP-1, -4, -5, -6, -7 and -9), or occurs only in the absence of TGF-β1 (for MHGARP-2 and -3). In region hGARP101-141, amino-acids 137 to 139 are required for the binding of MHGARP-2, -3 and -8. Affinity of MHGARP8 antibody to immobilized shGARP-TGFβ was measured by BIACOR analysis: Kd of said antibody is 0.2 nM.

Example 4: MHGARP8 Inhibits Human Treg Cell Function In Vivo

(43) To examine whether MHGARP8 also inhibits human Tregs in vivo, we used a model of xenogeneic GvHD induced by transfer of human PBMCs (Peripheral Blood Mononuclear Cells) into immuno-compromised NOD-Scid-IL2Rg.sup.−/− (NSG) mice. NSG mice lack functional T, B and NK cells. This allows efficient engraftment of human hematopoietic stem cells (HSCs), which proliferate and generate a functional human immune system in recipient mice. When human PBMCs are used instead of HSCs, efficient engraftment of T cells occurs, but is soon accompanied by the development of a xenogeneic Graft-versus-Host Disease (GvHD). In this model, GvHD results from the activity of human donor cytotoxic T lymphocytes that recognize tissues of the recipient NSG mice as foreign (Shultz, et al. Nature 2012, 12:786-798). The severity of the GvHD can be decreased by co-transferring human Treg cells with human PBMCs (Hannon et al. Transfusion 2014).

(44) Human PBMCs were isolated from total blood of a hemochromatosis donor by centrifugation on density gradients (Lymphoprep™), and frozen for later use. Autologous Tregs were generated as follows: CD4+ T cells were isolated from the blood of the same donor using the RosetteSep™ Human CD4+ T Cell Enrichment Cocktail (StemCell Technologies) and stained with anti-CD4, anti-CD25 and anti-CD127 antibodies coupled to fluorochromes. CD4+CD25hiCD127lo cells were sorted by flow cytometry (>99% purity) then stimulated with anti-CD3/CD28 coated beads (Dynabeads® Human T-Activator CD3/CD28 for T-Cell Expansion and Activation, Life Technologies) in the presence of IL-2 (120 IU/ml) during 14 days. These expanded Treg cells were frozen for later use.

(45) NSG mice were irradiated (2.5 Gy) on day −1, then injected in the tail vein with human PBMCs (2.7×106 per mouse) alone, or mixed with expanded human Tregs (1.4×106 per mouse) on day 0. Mice also received weekly i.p. injections of MHGARP8 antibody (400 μg on day −1 (day minus 1), 200 μg at later time points), or control PBS. Mice were monitored bi-weekly for GvHD development as indicated in the text.

(46) We transferred human PBMCs with or without Tregs in NSG mice, and treated the mice with i.p. injections of MHGARP8 antibody or control PBS. The large number of human Treg cells required for the transfers were obtained through short in vitro amplification of CD4+CD25+CD127lo cells sorted from human PBMCs by flow cytometry. Objective signs of GvHD development in the recipient mice were monitored bi-weekly. We performed two independent experiments, which yielded similar results. In experiment 1 (FIG. 6A), signs of GvHD (mean score ≥1) appeared 29 days after injection of human PBMCs (group I; n=2). Disease severity increased quickly, and one of the 2 mice was euthanized for ethical reasons on day 55. In mice injected with PBMCs and Tregs (group II; n=3), the appearance of GvHD was delayed by comparison to PBMCs alone (mean score ≥1 reached after 58 days). This indicates that Tregs, as expected, partially protected NSG mice against GvHD. Importantly, treatment of mice receiving PBMCs and Tregs with the MHGARP8 antibody (group III, n=6) aggravated the disease: signs of GvHD appeared earlier (36 days) than in mice from group II. The effect of MHGARP8 appears to depend on the presence of Tregs, as no difference in disease score was observed between mice receiving PBMCs only (group I) or PBMCs and MHGARP8 (group IV; n=4). We repeated this experiment with a larger number of mice per group (FIG. 6B). Again, co-injection of Tregs with PBMCs delayed the appearance of GvHD by comparison to PBMCs alone (day 46 in group II versus day 28 in group I), and treatment with the MHGARP8 antibody aggravated GvHD in mice receiving PBMCs and Tregs (day 28 in group III) by comparisons to untreated mice (day 46 in group II). Altogether, this shows that MHGARP8 inhibits the immune-suppressive function of human Tregs in vivo.

Example 5: New Anti-hGARP Monoclonal Antibodies (mAbs) Using Immunization of Llamas Approach

(47) Production of Recombinant Soluble GARP-TGFβ1 Complex

(48) Human and murine GARP-TGFβ1 complex was produced as a soluble complex using a truncated GARP expression construct. The human GARP protein sequence was truncated after Leucine 628, followed by a cleavable TEV-3× strep tag (EAAENLYFQGAAWSHPQFEKGAAWS HPQFEKGAAWSHPQFEKGAA*) (SEQ ID NO: 40). Murine GARP protein sequence was truncated after leucine 629, followed by the same cleavable TEV-3× strep tag. The GARP-TGFβ1 complexes were produced by co-expression of the truncated GARP and the TGFβ1 in HEK293E cells, followed by purification via the Strep-Tag.

(49) Immunization of Llamas

(50) Immunizations of llamas and harvesting of peripheral blood lymphocytes (PBLs) as well as the subsequent extraction of RNA and amplification of antibody fragments were performed as described by De Haard and colleagues (De Haard H, et al., J. Bact. 187:4531-4541, 2005). Four llamas were immunized with BW cells over-expressing human GARP and TGF 31 (FIG. 7A) as confirmed by flow cytometry using MHGARP8 (MHG-8) monoclonal antibody described in this patent application. The llamas were immunized with intramuscular injections in the neck once per week for a period of six weeks. Approximately 10.sup.7 cells were injected into the neck muscles and Freund's incomplete adjuvant was injected in a second region located a few centimeters from the injection site of the cells. Another four llamas were immunized with a mix of human GARP cDNA and human TGFβ1 cDNA expression vectors, once per two weeks, with four repetitive injections.

(51) Blood samples of 10 ml were collected pre- and post-immunization to investigate the immune response. Three-to-four days after the last immunization, 400 ml blood was collected for extraction of total RNA from the PBLs prepared using a Ficoll-Paque gradient and the method described by Chomczynski P, et al., Anal. Biochem. 162: 156-159, 1987. On average, RNA yields of 450 μg were achieved, which was used for random cDNA synthesis and PCR amplification of the V-regions of the heavy and the light chains (Vλ and Vκ) for construction of the Fab containing phagemid libraries as described by De Haard H et al., (J Biol Chem. 1999 Jun. 25; 274(26): 18218-30), to obtain diverse libraries of good diversity (1-7×10.sup.8).

(52) The immune response to the GARP-TGF β1 complex was investigated by ELISA on coated recombinant soluble GARP-TGF β1 complex (1 μg/ml). Five-fold serial dilutions of sera, starting from 10% sera were prepared and 100 μl of diluted sera was added onto the coated wells and incubated for 1 hour at RT. After washing with 3×PBS/Tween, the plates were blocked with PBS supplemented with 1% casein (FIG. 8). Binding of conventional llama IgG1 to its target GARP-TGFβ was measured in ELISA using a mouse anti llama IgG1 antibody (clone 27E10, Daley L P, et al. Clin. Diagn Lab Immunol. 12, 2005) and a HRP-conjugated donkey anti-mouse antibody (Jackson) for detection.

(53) Selections and Screenings of GARP-TGFβ1 Specific Fabs

(54) Phage expressing Fabs were produced according to standard protocols and selections performed on immobilized recombinant soluble GARP-TGFβ1 with total elution of the GARP-TGFβ1 binding phage with trypsin according to standard phage display protocols.

(55) Two to three rounds of selections were performed to enrich for human GARP-TGF β1 specific Fabs expressed by the phage. hGARP and hTGF β1 (LAP) counter selections were used to enrich for Fabs binding the hGARP-TGFβ1 complexes. Individual colonies were isolated and periplasmic fractions (peris) in 96-well plates were produced by IPTG induction from all the libraries according to standard protocols.

(56) Screening of the hGARP-TGFβ specific Fabs was performed using ELISA. hGARP-TGF β1 was immobilized on a maxisorb plate. After blocking with 1% casein in PBS for 1 h, Fab from 20l periplasmic extracts were allowed to bind to hGARP-TGF β1.

(57) Characterization of Monoclonal Antibodies

(58) GARP-TGFβ1/GARP specific clones were sequenced in the VH and the VL regions and divided into VH families based on the sequence of the CDR3 in the VH. 17 families were identified. Of each VH family identified we cloned at least one representative clone in to a full human IgG1 (LHG1-LHG17). These monoclonal antibodies were analyzed on Biacore for their binding characteristics to soluble human GARP-TGFβ1 complex. Recombinant soluble human GARP-TGFβ1 was immobilized at approximately 4,000 RU on a CM5 chip (GE Healthcare).

(59) Binding of monoclonal antibodies to the human and cynomolgus GARP-TGFβ1 complex expressed on HEK-293 cells was analyzed by FACS. Cynomolgus GARP and cynomolgus TGFβ1 encoding sequences were cloned from a cDNA sample from cynomolgus peripheral blood lymphocytes (PBMCs). Primers were based on the predicted sequences of cynomolgus GARP (XM_005579140.1; SEQ ID NO: 41) and cynomolgus TGF β1 (XM_005589338.1; SEQ ID NO: 42) by amplification of overlapping parts of the full sequence. For both cynomolgus GARP and cynomolgus TGFβ1 three separate PCR amplicons were DNA sequence analyzed. They fully aligned with the predicted sequences. Cynomolgus GARP and cynomolgus TGFβ1 were cloned into pCDNA3.1 for transient over-expression in HEK293E cells. Binding to cynomolgus GARP-TGFβ1 was compared to binding to human GARP-TGFβ1 on FACS. LHG-10 and the shuffled variants (LHG-10.3 to LHG-10.6) can be considered as cross-reactive with cynomolgus GARP-TGFβ1 (FIG. 9).

(60) Primers Used:

(61) TABLE-US-00031 >cyno TGFB S1: (SEQ ID NO: 43) cgcctc CCCCATGCCG ccctccg >cyno TGFB S2: (SEQ ID NO: 44) acaattcctg gcgatacctc >cyno TGFB AS1: (SEQ ID NO: 45) CTCAACCACTGCCGCACAAC >cyno TGFB AS2: (SEQ ID NO: 46) TCAGCTGCATTTGCAGGAGC

(62) VK Shuffling for Improved Affinity

(63) VK chain shuffling was used to improve the affinity of the mAb LHG-10 (FIG. 10). In this method, the heavy chain of the parental clone (VHCH1 of LHG-10) was reintroduced in the phagemid-light chain library. The heavy chain was extracted from an expression vector, which lacks the bacteriophage derived gene 3 necessary for display, to further avoid contamination of the parental light chain in the selection procedure. The heavy chain was cloned into the phagemid-light chain library and the ligated DNA was electroporated into E. coli TG1 cells to create the light chain shuffled library. The size of libraries was above 10.sup.8.

(64) Affinity selections, combined with off-rate washes, were performed to select for chain shuffled Fabs with an improved affinity for human GARP-TGFβ1. A set-up was chosen where Fab expressing phages were incubated with different concentrations of recombinant soluble human GARP-TGFβ1 directly coated to the microsorb plate.

(65) By adding the recombinant soluble human GARP-TGFβ1 in excess over the coated recombinant soluble human GARP-TGFβ1, the binding of the higher affinity phage was favored. Each round the time of washing was increased (Table 3) to select for phages with a better off-rate by washing away the lower affinity variants. Phages were eluted with trypsin and used for infection of E. coli TG1 cells. In total, 5 rounds of selection were done. In addition the amount of input phage was decreased in subsequent rounds to reduce background on the one hand and on the other hand to lower the mAb concentration thereby increasing the stringency of the selection.

(66) TABLE-US-00032 TABLE 3 Parameters varied for each round of selection for VK shuffling RI RII RIII RIV RV Concentrations 10 μg/ml 10 μg/ml 10 μg/ml 10 μg/ml 10 μg/ml rhGARP-TGFβ 1 μg/ml 1 μg/ml 1 μg/ml 1 μg/ml 1 μg/ml 0.1 μg/ml 0.1 μg/ml 0.1 μg/ml 0.1 μg/ml 0.1 μg/ml Vol. Phage 10 μl 1 μl 1 μl 1 μl 1 μl Time of washing 0 h 2 h O/N O/3N O/6N Conditions — 37° C., 37° C., 37° C., 37° C., 100 μg/ml 100 μg/ml 100 μg/ml 100 μg/ml rhGARP-TGFβ rhGARP-TGFβ rhGARP-TGFβ rhGARP-TGβ in 1% casein in 1% casein in 1% casein in 1% casein

(67) Screenings of at least 24 clones from selection rounds III, IV and V were performed. The clones were grown in deep well plates (1 ml expressions) and periplasmic fractions were prepared. These periplasmic extracts were analyzed on Biacore for improved off-rates. Top four Fab clones with improved off-rates were cloned into hIgG1 (LHG-10 series) and also an effector-dead variant hIgG1 with an N297Q substitution in the Fc region (LHG-10-D series), and the resultant IgGs were analyzed for improved binding characteristics on Biacore (Table 4). In addition, the LHG-10-D IgGs were checked for cross-reactivity on cyno GARP/cyno TGF-β1 in a FACS-based assay using HEK-293E cells transfected with cyno GARP/cyno TGFβ1 or human GARP/human TGFβ1. MHGARP8 was also tested in this cross-reactivity assay. All LHG-10-D and MHG-8 are cross-reactive against cyno GARP/cyno TGFβ1 (FIG. 9).

(68) TABLE-US-00033 TABLE 4 Binding characteristics of shuffled clones association dissociation affinity fold Fold fold ka (1/Ms) improvement kd (1/s) improvement KD improvement mIgG1 MHGARP8 1.25E+05 N/A 3.39E−05 N/A 2.64E−10 N/A hIgG1 - LHG-10-D 1.42E+05 1.0 2.62E−05 1.0 1.85E−10 1.0 N297Q LHG-10.3-D 2.31E+05 0.6 5.18E−06 5.1 2.24E−11 8.3 LHG-10.4-D 3.71E+05 0.4 1.21E−05 2.2 3.27E−11 5.7 LHG-10.5-D 3.83E+05 0.4 1.07E−05 2.4 2.80E−11 6.6 LHG-10.6-D 2.84E+05 0.5 6.15E−06 4.3 2.16E−11 8.6 hIgG1 LHG-10 2.39E+05 1.0 3.12E−05 1.0 1.31E−10 1.0 LHG-10.3 2.87E+05 0.8 6.38E−06 4.9 2.22E−11 5.9 LHG-10.4 4.48E+05 0.5 1.30E−05 2.4 2.91E−11 4.5 LHG-10.5 4.15E+05 0.6 1.37E−05 2.3 3.31E−11 4.0 LHG-10.6 2.76E+05 0.9 4.40E−06 7.1 1.59E−11 8.2

Example 6: Two Anti-hGARP mAbs (MHGARP8 and LHG-10) Inhibit Active TGF-β1 Production by Human Tregs

(69) Stimulated human Tregs produce active TGF-β1 close to their cell surface. Autocrine and paracrine TGF-β1 activity induces SMAD2 phosphorylation in Tregs themselves, and in Th cells co-cultured with Tregs (Stockis, J. et al. Eur. J. Immunol. 2009, 39:869-882). To test if GARP is required for TGF-β1 activation by Tregs, we stimulated human Tregs in the presence or absence of anti-hGARP mAbs, and measured phosphorylation of SMAD2 by Western Blot. As a source of human Tregs we used CD4+CD25.sup.hiCD127.sup.lo cells sorted from PBMCs and amplified in vitro during 12-14 days (Gauthy E et al PLoS One. 2013 Sep. 30; 8(9):e76186). As determined by methyl-specific qPCR, amplified cell populations contain 44 to 82% cells with a demethylated FOXP3i1 allele, indicating that they are still highly enriched in Tregs.

(70) As expected, phosphorylated SMAD2 was detected in the stimulated Tregs, but not in non-stimulated Tregs, nor in Tregs stimulated in the presence of a neutralizing anti-TGF-β1 antibody (FIG. 11). Phosphorylated SMAD2 was greatly reduced in Tregs stimulated in the presence of MHGARP8 (named MHG-8 on FIG. 11A) or LHG-10 (FIG. 11B), indicating that these two anti-hGARP mAbs block active TGF-β production. The 29 other new anti-hGARP mAbs, as well as 4 commercially available anti-hGARP mAbs, did not block TGF-β production by Tregs (FIG. 11).

(71) The inhibitory activity of MHGARP8 and LHG-10 shows that GARP is required for active TGF-β1 production by human Tregs.

Example 7: MHGARP8 and LHG-10 Inhibit the Suppressive Activity of Human Tregs In Vitro

(72) We previously showed that human Tregs suppress other T cells at least in part through production of active TGF-β1 (Stockis, J. et al. Eur. J. Immunol. 2009, 39:869-882). We therefore tested whether MHGARP8 (MHG-8) and LHG-10 also inhibit human Treg function in in vitro suppression assays. We used a Treg clone as a source of Tregs, and freshly isolated CD4.sup.+CD25.sup.−CD127.sup.hi cells or a CD4.sup.+ T cell clone (Th cells) as targets for suppression. Tregs and Th cells were stimulated with anti-CD3 and anti-CD28 in the presence or absence of various additional mAbs. As shown in FIG. 12, clone Treg A1 inhibited the proliferation of CD4.sup.+CD25.sup.−CD127.sup.hi Th cells by 66% in the absence of anti-hGARP mAb. Suppression was reduced to 36% and 32% in the presence of MHG-8 or LHG-10, respectively, but was not reduced in the presence of 6 other anti-hGARP mAbs. We also measured suppression by clone Treg A1 on another Th target (clone Th A2) in the presence of MHGARP8, an anti-hTGF-11 mAb or an isotype control. MHGARP8 (MHG-8) inhibited the in vitro suppressive activity of Treg A1 in a manner similar to that of the anti-TGF-β1 antibody, whereas the isotype control showed no effect (FIG. 12).

Example 8: Epitopes Recognized by Inhibitory Anti-hGARP mAbs

(73) Only a minority (2/35) of anti-hGARP mAbs block active TGF-β production and suppression by Tregs. This could be due to their ability to bind epitope(s) that are distinct from those bound by non-inhibitory mAbs. We therefore mapped the regions required for binding by inhibitory and non-inhibitory mAbs.

(74) GARP associates with pro- or latent TGF-β1 to form disulfide-linked GARP/TGF-β1 complexes (FIG. 13 and Stockis 2009b Eur. J. Immunol. 2009. 39: 3315-3322 and Gauthy E et al). We first sought to determine whether anti-hGARP mAbs also bind GARP/TGF-β1 complexes, using co-immunoprecipitation (IP) experiments in murine BW cells transfected with hGARP and hTGFB1. We tested 32 anti-hGARP mAbs: our 31 new mAbs and the commercially available Plato-1 mAb. All mAbs efficiently immunoprecipitated GARP (top panel of FIG. 14A, showing IPs with 12 representative mAbs). Pro-TGF-β1, as well as LAP and mature TGF-β1 (i.e. latent TGF-β1) were co-immunoprecipitated with 24 mAbs indicating that they bind GARP/TGF-β1 complexes (6 mAbs shown in FIG. 14A, middle and lower panels). In contrast, 8 mAbs (3 shown in FIG. 14A) did not co-immunoprecipitate pro- or latent TGF-β1, suggesting they bind free GARP but not GARP/TGF-β1 complexes.

(75) We confirmed this by FACS analyses of transfected 293T cells (FIG. 14B). Untransfected 293T cells express no GARP and very low levels of endogenous TGF-β1. No latent TGF-β is detected on their surface with an anti-LAP antibody. Transfection of GARP or TGFB1 alone induces no or low surface LAP, respectively, whereas co-transfection of GARP and TGFB1 induces high surface LAP as a result of latent TGF-β1 binding and presentation by GARP (FIG. 14B, left histograms). Three groups of anti-hGARP mAbs emerged from the analysis of transfected 293T cells, and are classified in 3 columns in FIG. 13B. A first group (left column) comprises the 8 mAbs that did not co-immunoprecipitate pro- or latent TGF-β1: they bound 293T cells transfected with hGARP alone, but not with hGARP and hTGFB1. This confirms that these mAbs bind free GARP only, as binding to surface GARP is lost in the presence of TGF-β1 (FIG. 14B shows 3 representative mAbs of this group). A second group comprises most other mAbs (19 mAbs, middle column of FIG. 13B): they bound 293T cells equally well upon transfection with hGARP alone or with hGARP and hTGFB1, indicating that they bind both free GARP and GARP/TGF-β1 complexes (FIG. 14B shows 6 mAbs of this group). Interestingly, a third group of 5 mAbs bound 293T cells transfected with hGARP and hTGFB1, but not cells transfected with hGARP alone (right column of FIG. 13B). These mAbs bind GARP/TGF-β1 complexes but not free GARP, and include inhibitory MHGARP8 (MHG-8) and LHG-10 (FIG. 14B shows 3 mAbs of this group).

(76) From the above, we concluded that most mAbs bind free GARP only (8/32) or free GARP and GARP/TF-β1 complexes (19/32). Only 5 mAbs, including inhibitory MHGARP8 (MHG-8) and LHG-10 but also 3 non-inhibitory mAbs, bind GARP/TGF-β1 complexes, but not free GARP. This pattern of recognition does not explain why only MHGARP8 and LHG-10 are inhibitory.

(77) We next sought to define the regions of hGARP required for binding by the various mAbs. The vast majority of the anti-hGARP mAbs do not cross-react on mouse GARP (mGARP). We thus constructed plasmids encoding HA-tagged mGARP/hGARP chimeras (FIG. 15A, left panel) and transfected them in 293T cells, with or without hTFGB1 depending on the binding requirements determined above. All chimeras were expressed at similar levels on the surface of 293T cells, as evidenced by staining with an anti-HA mAb (FIG. 15A, histograms on the right). Binding patterns to mGARP/hGARP chimeras (FIG. 15A, 10 representative mAbs) allowed to identify the region of hGARP required for binding by each anti-hGARP mAb. This is summarized in FIG. 15B, where mAbs are distributed in rows corresponding to various regions of hGARP: mAbs in the first row require a region comprising amino-acids 20 to 101 (hGARP.sub.20-101), mAbs in the second row require hGARP.sub.101-141, those in the third require hGARP.sub.141-207, the fourth, hGARP.sub.265-332, and finally, a fifth group requires hGARP.sub.332-628. However, even when considering the regions required for binding, the epitope recognized by inhibitory MHGARP8 (named MHG-8 on the figure) and LHG-10 could not be distinguished from that of non-inhibitory mAbs: MHGARP8 and LHG-10, like LHG-3, -12 and -13, bind GARP/TGF-β complexes that contain hGARP.sub.101-141.

(78) Sequences of mouse and human GARP.sub.101-141 differ at 14 amino-acid (aa) positions, comprising 3 clusters of 3 contiguous positions (FIG. 15B, left panel). We constructed 3 mutated versions of hGARP. In each mutant, a series of 3 contiguous aa from region 101-141 were replaced by the aa found in mGARP. We transfected 293T cells with the HA-tagged mutants, alone or with hTGFB1 depending on the binding requirement of the mAbs tested. Binding patterns to mutants revealed 3 types of mAbs (FIG. 15B, right panel), which required amino-acids hGARP.sub.111-113, hGARP.sub.126-127, or hGARP.sub.137-139 for binding, respectively. Six mAbs, including MHGARP8 (named MHG-8 on the figure) and LHG-10, required hGARP.sub.137-139 (FIG. 13B). Whereas 4 of 6 can bind free hGARP, MHG-8 and LHG-10 are the only mAbs that require hGARP.sub.137-139 in the context of GARP/TGF-β1 complexes.

(79) From the above, we concluded that inhibition of TGF-β production by MHGARP8 and LHG-10 is associated with the ability to bind an epitope that is distinct from those recognized by all other, non-inhibitory, anti-hGARP mAbs.

Example 9: Inhibition of Human Tregs Function by Anti-hGARP In Vivo

(80) We next sought to evaluate whether inhibitory anti-hGARP mAbs could inhibit human Treg function in vivo. We used a model of xenogeneic graft-versus-host disease (GVHD) induced by transfer of human peripheral blood mononuclear cells (PBMCs) into immuno-compromised NOD/Scid/IL2Rg.sup.−/− (NSG) mice. NSG mice have defective cytokine signaling and lack functional T, B and NK cells, allowing very efficient engraftment of human T cells upon i.v. injection of PBMCs. Thirty to forty days after PBMC transfer, recipient mice develop xenogeneic GVHD, due to the activity of human cytotoxic T lymphocytes against murine tissues Shultz, Nat Rev Immunol. 2012 November; 12(11):786-98. In this model, co-transfer of human Tregs with human PBMCs attenuates GVHD (Hannon et al. Transfusion. 2014 February; 54(2):353-63), providing a model to test the inhibitory activity of anti-hGARP mAbs on human Tregs in vivo.

(81) We transferred human PBMCs (3×10.sup.6/mouse) with or without autologous Tregs (1.5×10.sup.6/mouse) in NSG mice (FIG. 16A). As a source of human Tregs, we used blood CD4.sup.+CD25.sup.hiCD127.sup.lo cells that had been shortly amplified in vitro, as described above. In addition, mice were injected with MHGARP8 (named MHG-8 on the figure), anti-TGF-β1, an isotype control or PBS, one day before the graft and weekly thereafter. Objective signs of GVHD were monitored bi-weekly, to establish a disease score based on weight loss, reduced mobility, anemia or icterus, and hair loss. We performed four independent experiments (FIG. 16B), and detailed results are shown for one (FIG. 16C). Depending on the experiment, onset of disease (mean GVHD score ≥1) was observed 28 to 41 days after PBMC transfer in groups of mice that received no mAb or an isotype control. Co-transfer of Tregs delayed disease, which occurred 46 to 72 days after transfer, indicating that human Tregs were able to suppress human T cell responses against xenogeneic antigens. Administration of MHGARP8 to mice transferred with PBMCs and Tregs abrogated the protective effect of Tregs: disease occurred as early as in mice receiving PBMCs only (28 to 44 days after transfer). Inhibition of Treg suppressive function by MHGARP8 was similar to that observed with a neutralizing anti-TGF-β1 antibody. An isotype control had no effect.

(82) Altogether, this shows that MHGARP8 inhibits the immune-suppressive function of human Tregs in vivo.