Activated protein C protects from GvHD via PAR2/PAR3 signalling in regulatory T-cells

Graft-vs.-host disease (GvHD) is a major complication of allogenic hematopoietic stem-cell(HSC) transplantation. GvHD is associated with loss of endothelial thrombomodulin, but the relevance of this for the adaptive immune response to transplanted HSCs remains unknown. Here we show that the protease-activated protein C (aPC), which is generated by thrombomodulin, ameliorates GvHD aPC restricts allogenic T-cell activation via the protease activated receptor (PAR)2/PAR3 heterodimer on regulatory T-cells (Tregs, CD4+FOXP3+). Preincubation of pan T-cells with aPC prior to transplantation increases the frequency of Tregs and protects from GvHD. Preincubation of human T-cells (HLA-DR4−CD4+) with aPC prior to transplantation into humanized (NSG-AB°DR4) mice ameliorates graft-vs.-host disease. The protective effect of aPC on GvHD does not compromise the graft vs. leukaemia effect in two independent tumor cell models. Ex vivo preincubation of T-cells with aPC, aPC-based therapies, or targeting PAR2/PAR3 on T-cells may provide a safe and effective approach to mitigate GvHD. Graft-vs.-host disease is a complication of allogenic hematopoietic stem cell transplantation, and is associated with endothelial dysfunction. Here the authors show that activated protein C signals via PAR2/PAR3 to expand Treg cells, mitigating the disease in mice.

Analyses of steroid-resistant GvHD revealed involvement of endothelial dysfunction, e.g. increased serum levels of soluble thrombomodulin (TM) [10][11][12][13] , which reflect loss of endothelial TM function 14 . Targeting TM-dependent effects may hence constitute a new therapeutic approach to mitigate GvHD. Indeed, pre-clinical studies in mice suggested that soluble TM ameliorates GvHD, but the underlying mechanism remained unknown 15 . TM is required for efficient activation of the anticoagulant and cytoprotective signaling-competent protease-activated protein C (aPC) 14,16 . aPC signals predominately via G-protein coupled protease activated receptors (PARs) in a cell-and context-specific manner [17][18][19] . The role of aPC in innate immunity is firmly established 17 , whereas its role in adaptive immunity and in particular on T-cells remains largely unknown. In a series of elegant reports Hancock et al. 20 studied the effect of aPC in solid organ transplantation, focusing, however, on innate immune mechanisms. In addition, previous work showed that aPC dampens activation of effector T-cells and increases the frequency of T regs in a model of type 1 diabetes mellitus, but the underlying mechanism, e.g. which immune cell type is targeted by aPC and the receptors involved, remained unknown 21 . Considering the loss of TM in GvHD, the known cytoprotective effects of aPC, and the development of new and safer aPC-based drugs we investigated aPC's role in acute GvHD. Using a combination of in vivo and in vitro approaches we show that aPC signaling in

Preincubation of T-cells with aPC limits T-cell reactivity. As
GvHD is primarily a T-cell driven disease we next explored aPC's effect on T-cell activation using in vitro mixed lymphocyte reactions (MLRs) 23 . Co-culture of C57BL/6 wt pan T-cells with irradiated allogenic antigen-presenting cells (AgPC, BALB/c) for 96 h induced T-cell reactivity, which was markedly blunted in the presence of aPC (20 nM, every 12 h, Fig. 2a), establishing that aPC inhibits MLR-driven allogenic T-cell reactivity. Preincubation of T-cells with aPC (20 nM, once, 1 h, 37°C, followed by a washing step to remove remaining aPC; T(aPC) + AgPC) was sufficient to inhibit allogenic T-cell reactivity in comparison to non aPC preincubated T-cells (T + AgPC) or the positive control T reg -depleted T-cells (T-Tr + AgPC, Fig. 2b). In addition, preincubation of eFluor450 labelled T-cells with aPC reduced T-cell proliferation (Fig. 2c). Of note, preincubation of allogenic AgPC with aPC (T + AgPC(aPC)) following the same protocol had no effect on the MLR (Fig. 2b), suggesting that aPC directly acts on T-cells.
aPC preincubation restricts allo-reactivity of human T-cells. The effect of aPC on allogenic activation of human T-cells was next assessed 23 . Conducting the MLR using human T-cells and AgPC in the presence of aPC (20 nM, every 12 h) reduced T-cell reactivity to 67% as compared to the MLR without aPC (Fig. 4a). Furthermore, preincubation of human pan T-cells with aPC (T(aPC) + AgPC) reduced T-cell reactivity in the MLR, both in comparison to T reg -depleted T-cells (53 vs. 100% 3 H incorporation in (T-Tr) + AgPC, Fig. 4b) and in comparison to pan T-cells without aPC-preincubation (53 vs. 77% 3 H incorporation in T + AgPC, Fig. 4b). An aPC concentration of 5 nM was sufficient for aPC's inhibitory effect ( Supplementary Fig. 4). Preincubation of human AgPC cells with aPC (T + AgPC(aPC)) had no effect (Fig. 4b). Of note, aPC preincubation of human T reg -depleted pan T-cells ((T-Tr)(aPC) + AgPC) partially reduced T-cell reactivity (83 vs. 100% 3 H incorporation in (T-Tr) + AgPC, Fig. 4b). The latter indicates that aPC's inhibitory effect in the MLR is only partially dependent on pre-existing T regs . Congruently with aPC's inhibitory effect on 3 H incorporation, proliferation of aPC preincubated T-cells was reduced (eFluor450-labeling, 54.9% in aPC-preincubated T-cells vs. 67.7% in T-cells without aPC-preincubation, Fig. 4c). Assessment of annexin V and propidium iodide (PI) staining of T-cells demonstrated that preincubation of T-cells with aPC did not change T-cell apoptosis in the MLR ( Supplementary Fig. 5). Preincubation of human pan T-cells with aPC reduced the frequency of Th1 (CD4 + T-bet + ) and Th17 (CD4 + ROR-γt + ) cells, while increasing the abundance of T regs (CD4 + FOXP3 + ) following allogenic stimulation ( Fig. 4d; Supplementary Fig. 6). Concomitantly, expression of IFN-γ, TNF-α, and IL-17A in T-cells was reduced, whereas expression of IL-10 in T-cells and TGFβ1 level in the supernatant were increased (Fig. 4e,f; Supplementary  Fig. 7). Thus, as in mice with GvHD, preincubation of human T-cells with aPC prior to MLR generates a cytokine profile favouring T reg expansion and increases the frequency of T regs .

Preincubation of T-cells with aPC induces and expands T regs .
To determine whether aPC induces T reg differentiation from T reg -depleted T-cells or expands pre-existing T regs , we isolated human T regs (T reg ), T reg -depleted T-cells (T) and AgPC cells. Cells were combined without any aPC-preincubation (T + T reg + AgPC), using T reg -depleted T-cells with aPC-preincubation together with T regs (T(aPC) + T reg + AgPC), or using T regs with aPC-preincubation together with T reg -depleted T-cells (T + T reg (aPC) + AgPC). In all three settings, 0.8 × 10 5 T reg -depleted T-cells, 0.2 × 10 5 T regs , and 3 × 10 5 AgPC were used. Following preincubation of pre-existing T regs with aPC (T + T reg (aPC) + AgPC), a strong increase of T reg frequency (17.6% compared to T + T reg + AgPC, 5.5, and T(aPC) + T reg + AgPC, 10.5, respectively) was observed, suggesting that aPC-preincubation expands pre-existing T regs . However, preincubation of T reg -depleted T-cells with aPC, likewise, increased the frequency of CD4 + FOXP3 + T regs (Fig. 5b), albeit to a lesser degree, reflecting induction of T regs from CD4 + T-cells (iT regs ). The induction of T regs from T reg -depleted T-cells is in agreement with the aPC-mediated induction of IL-10 and TGFβ1 and suppression of IL-6 observed in vivo (Figs. 3f,g), induction of IL-10 and TGFβ1 in vitro (Figs. 4e,f), and the partial suppressive effect observed following preincubation of T reg -depleted T-cells with aPC in the MLR (Fig. 4b).
The predominant increase of T regs (CD4 + FOXP3 + ) following aPC-preincubation of pre-existing T regs suggests that aPC reduces T-cell reactivity primarily by expanding pre-existing T regs . However, given the induction of T regs from T reg -depleted T-cells aPC may in addition reduce T-cell reactivity by inducing T regs from CD4 + CD25 − effector T-cells. To specifically determine the functional relevance of these possibilities, we separately isolated effector T-cells (T eff ; CD4 + CD25 − ) and T regs (CD4 + CD25 + ) by MACS. MLR-experiments were conducted using effector T-cells (T eff (aPC) + AgPC) or T regs (T eff + T reg (aPC) + AgPC) preincubated with aPC, and the effect was compared to that observed in the absence of T regs (T eff + AgPC) or in the absence of aPC-preincubation of T eff cells (T eff + T reg + AgPC). In these experiments, 0.8 × 10 5 T eff , 0.2 × 10 5 T regs , and 3 × 10 5 AgPC were combined. Preincubation of T regs with aPC resulted in a strong suppression of T-cell reactivity compared to control (T eff + AgPC, 100% vs. T eff + T reg (aPC) + AgPC, 42%, Fig. 5c). This effect was significantly stronger than that observed when using T regs not preincubated with aPC (T eff + T reg + AgPC, 69%, vs. T eff + T reg (aPC) + AgPC, 42%). Of note, preincubation of T-effector cells with aPC (T eff (aPC) + AgPC) likewise reduced T-cell reactivity to a comparable level as observed when combining effector T-cells with T regs without aPC-preincubation (T eff (aPC) + AgPC 77% vs. T eff + T reg + AgPC, 69%, Fig. 5c). The suppressive effect observed following aPC-preincubation of effector T-cells was significantly less than that observed following aPCpreincubation of T regs (T eff (aPC) + AgPC 77% vs. T eff + T reg (aPC) + AgPC, 42%). Preincubation of both T eff and T reg separately with aPC (T eff (aPC) + T reg (aPC) + AgPC) suppressed T-cell reactivity more than preincubation of either T eff or T reg only, but this effect was not significantly different from that observed following aPC-preincubation of T regs only (42 vs. 27%; Fig. 5c).
To ascertain increased proliferation of pre-existing T regs following aPC-preincubation and stimulation, we isolated human T regs by FACS. T regs without (T reg ) or with (T reg (aPC)) aPC-preincubation were stimulated with plate-bound αCD3 (10 µg/ml) and αCD28 (8 µg/ml) antibodies. The proliferation of stimulated aPC-preincubated T regs (T reg (aPC)) was almost twice as high as that of control T regs (65.0% in T reg (aPC) vs. 36.5% in T regs , Fig. 5d).
As aPC-preincubation of effector T-cells reduced T-cell proliferation, we next ascertained whether aPC-preincubation induced T regs from effector T-cells. To this end, we determined the frequency of CD4 + CD25 + cells following co-incubation and T-cells (0.4 × 10 6 ) obtained from C57BL/6 mice and T regs (0.1 × 10 6 ) from C57BL/6 (transgene negative DEREG littermate mice) without (B6T + B6T reg + BM + DT) or with (B6T + B6T reg (aPC) + BM + DT) aPC-preincubation or from DEREG mice without (B6T + DTR-T reg + BM + DT) or with aPC-preincubation (B6T + DTR-T reg (aPC) + BM + DT). Diphtheria toxin (DT, 20 ng/g body weight) was injected on day 1 and 2 post transplantation in all groups. b aPC-preincubation of pre-existing human T regs with aPC (T + Tr(aPC) + AgPC) increases T reg frequency to a larger extent than aPC-preincubation of T reg -depleted T-cells (T(aPC) + Tr + AgPC), T reg frequency was measured by flow cytometry after 48 h of MLR. c Preincubation of human T regs with aPC (T eff + T reg (aPC) + AgPC) markedly reduces T-cell reactivity, as determined by thymidine incorporation in the MLR. In the absence of T regs aPCpreincubation of effector T-cells (T eff (aPC) + AgPC) is less efficient, but the effect is comparable to that observed when combing effector T-cells with T regs without aPC-preincubation (T eff + T reg + AgPC). Results are compared with T-cell reactivity of effector T-cells in the absence of T regs (T eff + AgPC). d aPC induces proliferation of human T regs : T regs without (T reg ) or with aPC-preincubation (T reg (aPC)) were labelled with eFluor450 (4 µM) and stimulated with plate bound αCD3 (10 µg/ml) and αCD28 (8 µg/ml) antibodies. Proliferation was measured on day 1 and 6 by FACS. e aPC induces human CD4 + CD25 + cells: effector T-cells (CD4 + CD25 -) were left untreated or were preincubated with aPC (20 nM, 1 h, 37°C). Following allogenic simulation with AgPC the frequency of CD4 + CD25 + cells, determined at day 4, was increased following aPC-preincubation; results from three repeat experiments each with three biological disjunct donors. of effector T-cells with AgPC without (T eff + AgPC) or with (T eff (aPC) + AgPC) aPC-preincubation. After 4 days, the frequency of CD4 + CD25 + cells was increased following preincubation of effector T-cells with aPC (T eff + AgPC, 4.1% vs. T eff (aPC) + AgPC, 7.6%, Fig. 5e; Supplementary Fig. 10). In parallel, we observed a strong induction of nuclear FOXP3 in T effs following preincubation with aPC ( Fig. 5f). Altogether, these data suggest that aPC expanses T regs by both expanding preexisting T regs and inducing T reg differentiation.

T-cell inhibition by aPC depends on PAR2-PAR3 cofactoring.
The above observations demonstrate that aPC promotes T reg expansion and inhibits T-cell reactivity, but the receptors involved remain unknown. Jurkat T-cells express PARs, which are the pivotal receptors for aPC-dependent signaling 17,28,29 . Similarly, primary human and mouse pan T-cells and mouse T regs express all four PARs as well as EPCR (endothelial protein C receptor), whereas human T regs express EPCR and all PARs except PAR4 (Fig. 6a) 28 . Intriguingly, in human T regs PAR3 expression appears to be higher as compared to CD8 + or CD4 + T-cells (Fig. 6b). To ascertain which PAR is required for aPC's effect in the MLR, we first used inhibitory antibodies. These antibodies inhibit proteolytic cleavage of the N-terminal receptor sequence and thus the generation of the corresponding tethered ligand. Inhibition of PAR3 cleavage abolished aPC's effect, whereas N-terminal blocking antibodies to PAR1, PAR2, or PAR4 had no effect (Fig. 6c). The requirement of PAR3 for aPC's inhibitory effect on T-cell activation was confirmed using T-cells isolated from PAR3-deficient mice (Fig. 6d).
Burnier and Mosnier 30 previously established generation of a human PAR3-dependent tethered ligand by aPC. We previously demonstrated that aPC cleaves the N-terminal end of mouse PAR3, but the cleavage site remained unknown 31 . To analyse aPC cleavage of mouse PAR3, we followed established protocols 30 . Using the human PAR3-derived N-terminal peptide, we first validated the approach by replicating previous results 30 ( Supplementary Fig. 11). Cleavage analyses of the mouse PAR3derived N-terminal peptide revealed that both human thrombin and human aPC efficiently cleave at Lys37 within 2 h (Fig. 6e). This result is congruent with our previous mutagenesis studies 31 and establishes the aPC cleavage site within the mouse PAR3derived N terminus.
As PAR3 is not signaling competent itself this finding indicates the involvement of a co-receptor 18 . EPCR is typically, but not in all cell-types or under all conditions, required for aPC signaling 17,31,32 . An antibody blocking aPC binding to EPCR did not abolish aPC's inhibitory effect on T-cell activation (Fig. 6c), suggesting a signaling mechanism of aPC on T-cells independent of EPCR. Of note, the tethered ligand of PAR3 is capable to activate PAR1 or PAR2 ("cofactoring"), presumably by binding to the second extracellular loop of these receptors 18,19 . Inhibitory peptides against PAR1 and PAR4 (blocking binding of a ligand to the second extracellular loop) failed to abolish aPC's effect in the  Fig. 12). Considering the lack of specific PAR2-inhibitory peptides 33 , we knocked down PAR2 expression in primary human T-cells. Efficient shRNA-mediated knockdown of PAR2 was achieved in pan T-cells using transient transfection with lentiviral particles, whereas a scrambled control shRNA had no effect ( Supplementary Fig. 13). PAR2 knockdown in human primary T-cells abolished the inhibitory effect of aPC following stimulation with αCD3 and αCD28 (Fig. 7a) antibodies. Furthermore, the requirement of PAR2 for aPC's inhibitory effect on T-cell activation was confirmed using T-cells isolated from PAR2-deficient mice (Fig. 7b).
Signaling of aPC via PARs involves the MAPK pathway, which is also known to modulate T-cell function and T reg differentiation 17,19,34 . Analyses of T-cells preincubated with aPC and isolated by MACS after 96 h of MLR revealed reduced Erk1/2 and p38 phosphorylation (Fig. 7c), which is consistent with aPC signaling via PAR2 in T-cells. Co-immunoprecipitation of PAR2 or PAR3 confirmed the interaction of PAR2/PAR3 heterodimers on human T regs (Fig. 7d). Taken together, these data indicate the requirement of a PAR2/ PAR3 heterodimer for aPC's inhibitory effect in allogenic activation of human and mouse T-cells.
aPC limits GvHD in a humanized model of GvHD in mice. The above data suggest that ex vivo preincubation of pan T-cells  7 aPC signaling requires PAR2 in addition to PAR3 on T-cells. a Following knockdown of PAR2 in human primary pan T-cells (T) using two different shRNAs (shRNA1 and shRNA2) the inhibitory effect of aPC on T-cells after stimulation with αCD3 and αCD28 antibodies is lost. Scrambled shRNA (shRNAsc) had no effect. Bar graph summarizing results from five repeat experiments using five biological disjunct donors, each in triplicates.
b Preincubation of T-cells from C57BL/6 wild type (B6T(aPC)) mice, but not of C57BL/6 PAR2-deficient (PAR2 −/− (aPC)) mice, with aPC and subsequent stimulation with plate-bound αCD3 and αCD28 antibodies, abrogates aPC's inhibitory effect on T-cell reactivity. c Preincubation of human pan T-cells with aPC reduces ERK1/2 and p38 activity after 96 h MLR as reflected by reduced phosphorylation of these proteins (normalized to the corresponding total protein levels) as determined by immunoblot; bar graph reflecting mean results ± SEM (top) and representative immunoblots (bottom) are shown. d Exemplary immunoblot showing interaction of PAR2 and PAR3 in human T regs analysed using immunoprecipitation for PAR2 followed by immunoblotting of PAR3 and vice versa. Immunoprecipitation using IgG was used as negative control. Immunoblot of PAR2 and PAR3 was done as a input control. Mean value ± SEM (a, b), results of at least three repeat experiments each with three biological disjunct donors (a, c) or from five different mice (b). *P < 0.05, **P < 0.01, NS non significant (a, b ANOVA; c t-test) or T regs prior to transplantation may be an easy, efficient, and safe new therapeutic strategy to mitigate GvHD. To corroborate the translational relevance, we transplanted NSG-Ab°DR4 mice, which lack expression of the murine Prkdc gene, the X-linked Il2rg gene, and MHC class II while expressing the human leukocyte antigen DR4 gene 35 , with 4 × 10 6 human CD4 + T-cells (HLA-DR4 − ) without (hCD4 + ) or with (hCD4 + (aPC)) aPCpreincubation. Survival, physical appearance, and histological damage of NSG-Ab°DR4 mice transplanted with aPCpreincubated human CD4 + T-cells (hCD4 + (aPC)) was markedly improved compared to control (hCD4 + ) mice (Fig. 9).
aPC ameliorates GvHD without impeding the GvL effect. Mice receiving BM and T-cells only (BM + B6T) displayed increased lethality (Fig. 10a) and developed typical hallmarks of GvHD; and GvHD was confirmed by histology (e.g., increased frequency of cryptic apoptosis in the gastrointestinal tract). Contrary, mice receiving only BM and MA9 cells (BM+MA9) lacked morphological and histological signs of GvHD, but these mice nevertheless died as early as BM + B6T mice (Fig. 10a), presumably due to leukaemia progression as indicated by a slightly increased leukaemic burden after 1 week (Fig. 10b). The tumour load, determined 1 week post transplantation, did not differ between mice receiving BM, MA9 cells, and T-cells without (B6T + BM + MA9) or with (B6T(aPC) + BM + MA9) aPC-preincubation (Fig. 10b), indicating comparable tumour engraftment in these mice. Yet, peripheral leukaemic load determined 4 weeks post transplantation was markedly reduced in B6T(aPC) + BM + MA9 mice, but not in B6T + BM + MA9 mice, (Fig. 10c; Supplementary Fig. 14), indicating a sustained GvL effect in B6T(aPC) + BM + MA9 mice. Importantly, mice receiving T-cells preincubated with aPC (B6T(aPC) + BM + MA9) did not display signs of GvHD (e.g., increased frequency of cryptic apoptosis in the gastrointestinal tract). Accordingly, survival of mice receiving T-cells preincubated with aPC was significantly improved compared to all other groups (Fig. 10a).
To corroborate that the GvL effect is sustained following preincubation of T-cells with aPC, we generated an independent leukaemia model using the clinically relevant cooperating oncogenes AML1-ETO and KRAS (AE9aKRAS). BM of primary recipient mice was harvested upon onset of clinical symptoms and a total number of 2 × 10 4 GFP + cells were injected into lethally irradiated C57BL/6 mice. Leukaemic cells were injected along with 5 × 10 6 supporter BM cells and 1 × 10 6 T-cells (with and without aPC preincubation) derived from C3H/HeNRj mice. Again, aPC preincubation of T-cells ameliorated GvHD while maintaining a GvL effect (Supplementary Fig. 15). Hence, ex vivo preincubation of T-cells with aPC ameliorates GvHD without compromising the GvL effect.

Discussion
Here we uncover a new pathway targeting T regs and ameliorating GvHD. Although a crucial role of T regs to protect against GvHD is established, methods to easily and efficiently enrich T regs have been lacking so far 8,9 . The current data demonstrate that aPC-preincubation of T-cells increases T reg frequency through expansion of pre-existing T regs and induction of T regs . The latter observation is supported by a cytokine profile promoting T reg induction (increase TGFβ1 and IL-10, but reduced IL-6 levels). Although expansion of T regs following ex vivo preincubation of pan T-cells with aPC protects from GvHD, it does not compromise the GvL effect when using two independent tumor cell models. A sustained GvL effect despite suppression of GvHD has been previously reported following the adoptive transfer of T regs in animal and clinical studies 6,7 . Hence, our observation is congruent with these earlier reports, but by proposing an easy, safe, and efficient way to expand donor-derived T regs , the current study identifies an approach of potential translational relevance. The sustained GvL effect following preincubation of T-cells with aPC, and subsequent expansion of T regs is congruent with previous studies demonstrating that T regs do not only prevent GvHD, but simultaneously convey an efficient GvL effect 6,7 . Several mechanisms have been proposed for the dual effect of T regs in regard to GvHD and the GvL effect, e.g., inhibition of JAK1/JAK2 signaling, expression of NKG2D by CD8 + cells, IL-21 signaling, or the differential expression of granzyme B by CD4 + CD25 + and CD8 + cells [37][38][39][40] . Whether these mechanisms are regulated by aPC remain unknown. Likewise, the relevance of other T-cell populations, such as Th1, Th17, or CD8 + T-cells for aPC's effects as observed in the current study remains to be evaluated.
The expansion of T regs by aPC-PAR2/PAR3 signaling and the efficacy of aPC to improve GvHD following ex vivo preincubation of T-cells identify a new function of aPC-signaling and PAR-signaling in T-cells. This finding adds to the previous reports suggesting a role of PARs in AgPC 41 . An effect of aPC on T reg expansion was previously reported by Xue et al. 21 in NOD mice in the context of pancreatic islet inflammation. However, the underlying mechanisms, e.g., the immune cells targeted by aPC and the receptors required for aPC's effect, remained unknown. Likewise, expression of PARs on T-cells has been reported before, but these studies evaluated Jurkat (immortalized human T lymphocyte cells) cells and hence the physiological relevance of PARs in adaptive immunity remained unknown 17,28 . To decipher the interaction of PAR2 and PAR3 on T-cells, we first used inhibitory antibodies, which prevent proteolytic cleavage of the N-terminal end of PARs. Only the antibody targeting PAR3 abolished aPC's effect on allogenic T-cell stimulation. Proteolytic cleavage of human and mouse PAR3 by aPC has been previously reported by us and others 30,31 , and we now identify the aPC cleavage site within the mouse PAR3 N-terminal peptide.
PAR3 is generally not considered to be signaling competent itself and requires a co-receptor 19 . Analyses of murine PAR2 −/− T-cells and human PAR2 knockdown T-cells revealed that-in addition to PAR3-PAR2 is required for aPC's inhibitory effect on allogeneic T-cell activation. Furthermore, the functional relevance of PAR2 and PAR3 on T regs was confirmed in vivo, as aPC's protective effect was lost following preincubation of T regs lacking either PAR3 or PAR2. Signaling through heterodimeric PAR-complexes has not been reported in the context of adaptive immunity hitherto 19 . We acknowledge that the exact mechanism of action through which PAR2 and PAR3 interact remains currently unresolved. However, cofactoring of PARs is established and activation of other PARs, including PAR2, by the PAR3-derived tethered ligand has already been proposed 19,29 . Intriguingly, although aPC signals via species-specific PAR-heterodimers in human and mouse podocytes (human: PAR3/PAR2; mouse: PAR3/PAR1), the same PAR-heterodimer (PAR3/PAR2) is required for aPC's effect on mouse and human T-cells and mouse T regs . This observation emphasizes the plasticity of PAR-signaling, both in regard to different cell-types and species.
The current results together with our previous studies 31 establish that aPC cleaves the N-terminal end of PAR3, thus generating a tethered ligand, which may interact with other PARs. As we used human aPC throughout our study, the finding that human aPC cleaves mouse PAR3 is relevant in the context of the current study. Future studies need to address whether likewise mouse aPC cleaves mouse PAR3. Intriguingly, thrombin and aPC cleaved mouse PAR3 at the same residue site (Lys37), contrasting the observations made with the human PAR3-derived peptide 30 (and current study). Signaling specificity of thrombin and aPC via mouse PAR3 can hence not be explained by different tethered ligands generated by either protease. Biased signaling independent of the cleavage site, but rather secondary to involvement of specific receptor and signaling complexes, has been recently proposed for aPC and thrombin-dependent β-arestin-2 signaling via PAR1 42 . Accordingly, we propose that specificity of PAR3 signaling is not only dependent on specific PAR3 cleavage sites, but also on a specific PAR3-containing protein complex.
In addition to PARs, other receptors for aPC have been reported 17 , and we can currently not exclude the involvement of other components within the aPC-receptorsome on T-cells. Of note, the current data suggest that EPCR is not required for aPC-mediated inhibition of allogenic T-cell activation. Signaling of aPC independent of EPCR has been reported in endothelial cells and in particular in non-endothelial cells 31,32 . Intriguingly, aPC increases Akt phosphorylation in human leukaemic monoblast U937 cells independent of EPCR and modulates geneexpression in dendritic cells partially independent of EPCR 43 , supporting the notion that aPC signaling in immune cells does not strictly depend on EPCR.
Previous studies demonstrated that the efficacy of aPC in MLR is enhanced in the presence of protein S 20 . Protein S functions as a non-proteolytic cofactor of protein C in the context of coagulation inhibition. In addition, protein S is expressed by T-cells and engages TAM receptors (Axl and Mertk) on dendritic cells to restrict the immune response 44 . Whether protein C and protein S co-ordinately modulate T-cell activation remains to be shown.
The protective role of T regs in GvHD is well documented in animal models and has led to several clinical studies 9,45 . For example, ex vivo umbilical cord blood-derived T regs reduced the risk of acute and chronic GvHD compared to identically treated historical controls 45 . Both adoptive transfer of T regs and induction of T regs provides protection from GvHD 9, 46 . Efficient and safe methods to expand T regs , as proposed in the current study, may facilitate the use of T regs to combat GvHD. Although the current study provides strong evidence for a protective role of aPC in acute GvHD through expansion of T regs , we cannot exclude that aPC may convey additional beneficial effects during other phases of the GvHD 1 . Thus, aPC's cytoprotective and anti-apoptotic effects are well established 17 , and aPC has been shown to protect from radiation injury 47 , suggesting that aPC may protect from the initial tissue damage during the transplant conditioning regimen and/or from damage ensuing during the allo-response 1 . Protection using aPC during these stages would, however, require systemic therapies with the potential caveat of side effects, such as haemorrhage. Using aPC mutants conveying cytoprotective signaling but lacking anticoagulant function 48 , may avoid the latter. Further studies are needed to evaluate the potential role of aPC in other GvHD phases and the effect of aPC mutants on T regs and T-cell reactivity.
Endothelial dysfunction has been repeatedly demonstrated following whole body irradiation or chemotherapy, therapies typically used in transplant conditioning regimen 14 and in the setting of GvHD [10][11][12][13]47 . On the basis of these findings, endothelial protective therapies have been proposed to convey beneficial effects in GvHD, but their translation was only partial successful 49 . Endothelial dysfunction is intimately associated with a loss of endothelial TM and lower plasma levels of aPC 22,50,51 . Reconstitution of aPC's effect ex vivo may be a safe yet efficient approach to compensate for the inevitable impairment of The efficacy of aPC to protect from GvHD despite ex vivo preincubation and washing of T-cells prior to their transfer into mice is an intriguing observation of potential translational relevance. The mechanism underlying the sustained effect of aPC on T-cells and T regs remains to be fully explored. We demonstrate that aPC-preincubation increases the frequency of T regs by expanding pre-existing T regs and inducing T regs from CD4 + T-cells. T regs mainly exert their suppressive function within the first 2 days after transplantation 8,52 . The presence of T regs specifically during the initial days is sufficient for a long lasting protective effect in GvHD. Accordingly, we speculate that even a transient increase of T regs following aPC ex vivo preincubation might be sufficient to increase the frequency of T regs during this initial phase. Alternatively, we recently demonstrated that aPC epigenetically controls gene-expression in the context of diabetic nephropathy 51 . The importance of epigenetically controlled gene-expression for T reg differentiation is well established 53 . Hence, it appears possible, but remains to be shown, that aPC induces T regs through epigenetic mechanisms.
The findings of the current work suggest a novel mechanistic link between the adaptive immunoreaction, T reg function, and aPC, and thus between endothelial function and the adaptive immune response. Recently, Gur-Cohen et al. established a novel function of TM-dependent PC-activation for retention of hematopoietic stem cell recruitment by limiting NO-production via aPC-EPCR-PAR signaling 54 . These and the current findings provide novel insights into the regulation of leucocyte homeostasis and their function through mechanisms depending on the coagulation protease aPC. Intriguingly, the receptors targeted by aPC on hematopoietic stem cells and T-cells are partially disjunct, which may allow targeting the underlying mechanisms through distinct pharmacological approaches. By identifying a mechanism allowing T reg expansion, the current study proposes a new therapeutic approach to mitigate GvHD while maintaining the GvL-effect.
Cell Proliferation Dye eFluor450 (eBioscience) was used to label the cells and analyse cell proliferation. FITC-conjugated AnnexinV (eBioscience) together with propidium iodide (PI) was used for apoptosis detection. Streptavidin Particles Plus-DM (Becton Dickinson, Franklin Lakes, NJ, USA) and a Dynal magnet (Life Technologies, USA) were used for lineage depletion.
LEGENDplex™ Mouse Th Cytokine Panel (13-plex, BioLegend) was used for cytokine measurement in mouse plasma. Human/mouse TGFβ1 ELISA Ready-SET-Go! (2nd generation, eBioscience) was used to measure TGFβ1 in mouse plasma and in supernatant of primary human T-cell MLR. The details of antibodies are provided in Supplementary Table 1.
APC high mice (female, 8-10 weeks old), which express a transgene resulting in expression of a human protein C variant (D167F/D172K) in the liver, which can be efficiently activated in the absence of thrombomodulin, resulting in high plasma concentrations of aPC, have been previously described and have been backcrossed onto the C57BL/6 (B6;H2 b ) background for more than 10 generation 22  Preparation of aPC. aPC was generated as previously described with slight modifications 55 . Briefly, prothrombin complex (Prothromplex NF600), containing all vitamin K-dependent coagulation factors, was reconstituted with sterile water and supplemented with CaCl 2 (final concentration: 10 mM). The calcium dependent monoclonal antibody to PC (HPC4) covalently linked to a column filled with Affigel-10 resin was used for PC purification. This column was equilibrated at 4°C with 1 column volume (CV) of washing buffer (0.1 M NaCl, 20 mM Tris, pH7.5, 5 mM benzamidine HCl, 2 mM Ca 2+ , 0.02% sodium azide) at a flow rate of 0.5 ml/min, which was used during the whole purification step. After PC-binding the column was washed first with 1 CV of washing buffer, followed by 1 CV with a buffer containing high salt concentration (0.5 M NaCl, 20 mM Tris, pH 7.5, 5 mM benzamidine HCl, 2 mM Ca 2+ , 0.02% sodium azide). Benzamidine was washed off the column using a buffer of 0.1 M NaCl, 20 mM Tris, pH 7.5, 2 mM Ca 2+ , and 0.02% sodium azide using again 1 CV. To elute PC 1 CV of elution buffer (0.1 M NaCl, 20 mM Tris, pH 7.5, 5 mM EDTA, 0.02% sodium azide, pH 7.5) was used. The eluate was fractionated by collecting 5 ml portions. Peak fractions, identified by measuring absorbance at 280 nm, were pooled. Human plasma thrombin (5% w/w, incubated for 3 h at 37°C) was used to activated the recovered PC. Residual thrombin was inhibited by hirudin (1 h at 37°C). To obtain purified aPC thrombin was separated using an ion exchange chromatography with FPLC (ÄKTAFPLC®, GE Healthcare Life Sciences). To this end, a Mono Q anion exchange column (GE Healthcare Life Sciences) was equilibrated with 5 CV of 20 mM Tris pH 7.5, 100 nM NaCl. After applying the solution containing aPC and a washing step with 5 CV of the solutions used for equilibration a 10-100% gradient using 5 CV of a 20 Mm Tris, pH 7.5, 1 M NaCl buffer was applied to the column to elute aPC. aPC eluted at 0.46 M NaCl (~36 mS/cm conductivity). Fractions of 0.5 ml were collected during the peak and pooled. Proteolytic activity, integrity, and purity of aPC was ascertained with the chromogenic substrate SPECTROZYME® PCa and by Coomasie-staining of the purified protein on 10% SDS-PAGE gel.
GvHD models. Pan T-cells were isolated from whole spleen by magnetic bead depletion of non-T-cells using mouse Pan T-cell isolation Kit II (Miltenyi Biotec) following the manufacturer's recommendations. Purity of T-cells was ascertained by FACS and ranged from 95 to 98%.
T regs were isolated from whole spleen using the CD4 + CD25 + Regulatory T-cell Isolation Kit (Miltenyi Biotec) and purity was ascertained by FACS analysis and ranged from 95 to 98%. Bone marrow (BM) was prepared by flushing BM from isolated tibia and femur bones using RPMI complete medium. RBCs in BM were lysed using RBC lysis buffer (Buffer EL; Qiagen).
To Control T-cells were exposed to the same medium without aPC. In further experiments, T regs were isolated from DEREG or transgene-negative littermate (wt-control) mice using the CD4 + CD25 + Regulatory T-cell Isolation Kit (Miltenyi Biotec) and purity was ascertained by FACS analysis and ranged from 95 to 98%. T regs or T-cells were separately preincubated with aPC (20 nM, 1 h, 37°C) prior to BM transplantation. In the experiments in which mice were transplanted with T regs from DEREG or C57BL/6 mice, recipient mice in both groups were injected intraperitoneally with diphtheria toxin (20 ng/g body weight) on day 1 and 2. To induce allogenic GvHD with human T-cells in mice we used the "humanized" NSG-Ab°DR4 mice (N = 12) (see above). For preconditioning these mice received TBI of 2 Gy. After 4 h these mice were intravenously injected with 4 × 10 6 of human CD4 + T-cells (HLA-DR4 − ) without or with aPC-preincubation (20 nM, 1 h, 37°C). In all experiments, cells were washed with PBS following aPC-preincubation and before transplantation.
HLA-DRB1-04 genotyping. To isolate HLA-DRB1-04 (HLA-DR4)-negative Tcells blood samples were obtained from 9 volunteers after getting their written informed consent. Expression of DR4 was assessed by real time polymerase chain reaction (PCR) using a Light Cycler (CFX Connect, Real Time System, BioRad). Briefly, genomic DNA was extracted from donor's PBMC (1 × 10 6 ) using the phenol chloroform extraction method. The HLA-DRB1-04 gene was amplified using the following forward primer: 5′ GTTTCTTGGAGCAGGTTAAACA-3′ and two reverse primers in the same reaction: 5′-CTGCACTGTGAAGCTCTCAC-3′, 5′-CTGCACTGTGAAGCTCTCCA-3′ 56 . The following cycling parameters were used: initially 2 min at 95°C, followed by 40 cycles of 95°C, 10 s; 68°C, 10 s; 72°C, 23 s with a single acquisition per cycle at 72°C. All temperature transitions were 20°C/s. Samples were then subjected to a melting curve analysis with the following conditions: 95°C, 0 s (slope 20°/s); 65°C, 10 s (slope 20°/s) and then heated to 95°C with a slope of 0.3°/s using step acquisition. Positive and negative samples were distinguished by the presence or absence of fluorescence signal during the PCR reaction and the presence of a melting peak (~91°C). Six individuals were identified as being DR4 − . Assessment of GvHD. Individual weights of transplanted mice were obtained and recorded on day l and weekly thereafter until the time of analysis. Survival was checked once daily. The clinical score of GvHD was assessed by a scoring system described in Table 1 that incorporates five physical parameters: weight loss, posture (hunching), activity, fur texture, and skin integrity. Every week mice were evaluated and graded from 0 to 2 for each criterion. A clinical index was subsequently generated by summation of the five criteria scores (maximum index = 10, Table 1) 57 .
Representative samples of GvHD target organs (gut, liver, and skin) were excised from recipients 14 days post-BM transplantation. Formalin-fixed tissues were paraffin embedded and sectioned (5-μm-thick sections). Sections were stained with hematoxylin and eosin for histologic examinations. Pathological scoring was conducted by an experienced pathologist (TK) blinded to the groups. Intestinal GvHD was scored based on the frequency of crypt apoptosis (0, rare to none; 1, occasional apoptotic bodies per 10 crypts; 2, few apoptotic bodies per 10 crypts; 3, the majority of crypts contain an apoptotic body; 4, the majority of crypts contain >1 apoptotic body) 58 . The severity of skin GvHD was assessed by a scoring system that incorporates following parameters: epidermic atrophy, hair follicle loss, increased collagen density in dermis, and inflammation. The slides were graded from 0 to 2 for each parameter 58 . The severity of liver GvHD was assessed by a scoring system that incorporates following parameters: bile ducts infiltrated by lymphocytes (0: not present, 1: one or more lymphocytes in one bile duct, 2: lymphocytes in more than 1 bile duct, 3: lymphocytes in all bile ducts) and portal inflammation (0: not present, 1: inflammatory cells in some portal tracts, 2: inflammatory cells in most portal tracts, 3: packing of inflammatory cells with or without spill-over into adjacent parenchyma in some or most portal tracts) 59 .
Cell isolation and mixed lymphocyte culture. For ex vivo assessment of allogenic T-cell activation, mixed lymphocyte reaction (MLR) were conducted by incubating pan T-cells or effector T-cells (T eff ) with non-T-cells, containing AgPC. For isolation of human T-cells or T eff first PBMCs were obtained from peripheral blood using Ficoll-Paque (GE Healthcare) gradient and then pan T-cells were isolated by magnetic bead depletion of non-T-cells using the human pan T-cell Isolation Kit (Miltenyi Biotec) following the manufacturer's recommendations. In a subset of experiments, regulatory T-cells were isolated using the human CD4 + CD25 + regulatory T-cell Isolation Kit II (Miltenyi Biotec) and untouched CD4 + CD25 − cells obtained during regulatory T-cells isolation were used as effector T-cells. Purity of cells was ascertained by FACS and ranged from 95 to 98%. Non-T-cells were irradiated (30 Gy) and used as AgPC. Pan T-cells and non-T-cells from two genetically distinct (non-related) individuals were co-cultured to trigger allogenic T-cell reactivity. For ex vivo assessment of murine allogenic responses, BALB/c and C57BL/6 splenocytes were isolated by disrupting the spleen with a 100 μm cell strainer. Pan T-cells were isolated from splenocytes by magnetic bead depletion of non-T-cells using mouse pan T-cell Isolation Kit II (Miltenyi Biotec) following the manufacturer's recommendations. Mismatched non-T-cells, used as AgPC, were irradiated with 30 Gy and co-cultured with pan T-cells to trigger the MLR response.
Both human and mouse T-cells were cultured in AIM V serum-free medium (Life Technologies) for 2 h (37°C, 5% CO 2 ) before performing MLR. MLRs with human or mouse cells were either conducted in the presence of aPC or following preincubation of T-cells with aPC. When conducting MLR in the presence of aPC human or mouse 1 × 10 5 pan T-cells were incubated with 3 × 10 5 irradiated allogenic non-T-cells (ratio 1:3) for 96 h and aPC was added every 12 h (20 nM final concentration). For experiments with aPC-preincubation human or mouse, 1 × 10 5 pan T-cells were preincubated with aPC (20 nM) or an equal volume of PBS (control) in AIM V serum free medium (1 h, 37°C), washed with PBS to remove any aPC, and then co-cultured with 3 × 10 5 irradiated allogenic non-T-cells (ratio 1:3) for 96 h. Reactivity of human and mouse T-cells was assessed by measuring [ 3 H] thymidine incorporation during the last 16 h of the incubation time and [ 3 H] thymidine was added at 0.2 Ci per well. At the end of the incubation period, cells were harvested and radioisotope incorporation was measured as an index of lymphocyte proliferation by betaplate liquid scintillation counter (MicroBeta, Wallac, Finland) 60 .
Cell Proliferation assay. Human and mouse pan T-cells were isolated as mentioned above. Preincubation of pan T-cells with aPC (20 nM) or PBS (control) was conducted as described above and cells were washed with PBS. After washing control (without aPC-preincubation) and aPC preincubated T-cells were labelled with eFluor450 (5 µM) for 5 min at 37°C. Cells were then washed with RPMI complete medium three times followed by a final washing step with PBS and again re-suspended in AIM V medium. Human and mouse pan T-cells without (PBS control) or with aPC-preincubation were co-cultured with 3 × 10 5 irradiated allogenic non-T-cells and proliferation of T-cells were measured on day 1 and 4 by FACS.
For isolation of human T regs pan T-cells were isolated as described above, were labelled with FITC-conjugated anti-CD4 and APC-conjugated anti-CD25 antibodies, and were FACS sorted using BD FACS Aria III (BD Biosceince, Germany). Subsets of T regs were preincubated with aPC (20 nM, 1 h, 37°C) or PBS (control) in AIM V medium, followed by 1 washing step with PBS. After washing both control and aPC treated T regs were labelled with eFluor450 as described above. T regs without (control) or with aPC-preincubation were seeded into 96 well plates (25,000 cells per well), coated with αCD3 (10 µg/ml) and αCD28 (8 µg/ml) antibodies and proliferation of T reg cells was measured on day 1 and 6 by FACS Canto II (BD Bioscience, Germany). Data was analysed using FlowJo software.
Apoptosis detection. MLR was set up with T-cells and allogenic AgPC and T-cell apoptosis was detected on day 3 and 6 using anti-CD3, annexin V surface staining, and propidium iodide (PI) DNA staining by flow cytometry. Data was analysed using FlowJo software.
Immunoblotting. To determine receptor-expression on T-cells cell and T-cell subsets (CD4 + , CD8 + , T regs ) lysates of purified human T-cells were prepared using RIPA buffer containing 50 mM Tris (pH7.4), 1% NP-40, 0.25% sodiumdeoxycholate, 150 mM NaCl, 1 mM EDTA, 1 mM Na 3 VO 4 , 1 mM NaF supplemented with protease inhibitor cocktail. Lysates were centrifuged (13,000×g for 10 min at 4°C) and pelleted debris was discarded. BCA assay was used to quantify protein concentration in supernatants. Equal amounts of protein were electrophoretically separated on 10% SDS polyacrylamide gel, and transferred to PVDF membranes. Membranes were incubated with primary antibodies against PAR1 ( horseradish peroxidase-conjugated antibodies for 1 h at room temperature. The enhanced chemiluminescence system was used to develop blots. To compare and quantify levels of proteins, the density of each band was measured using Image J software. GAPDH was used to control for equal loading (full western blot images are shown in Supplementary Figs. 24 and 25).
To analyse the effect of aPC on FOXP3 expression in T eff cells, protein lysates were obtained from T eff cells following MLR without or with aPC-preincubation. Following MLR, T eff cells were re-sorted by MACS and total lysates were prepared using RIPA lysis buffer as above. Immunoblotting was conducted as described in the previous paragraph using a primary antibody against FOXP3 (1:1000) and a secondary anti-mouse IgG (1:2000) horseradish peroxidase-conjugated antibody Lamin was used to control for equal loading (full western blot images are shown in Supplementary Fig. 23).
For analyses of aPC signaling in T-cells, protein lysates were obtained from T-cells following MLR with or without aPC-preincubation. Following MLR T-cells were re-sorted by MACS and total lysates were prepared using RIPA lysis buffer as above. After protein estimation equal amounts of protein were electrophoretically separated on 10% SDS polyacrylamide gel, and transferred to PVDF membranes. Membranes were incubated with primary antibodies against p38 (1:1000), phospho-p38 (1:1000), ERK (1:1000), or phospho-ERK (1:1000). After overnight incubation at 4°C membranes were washed with TBST and incubated with anti-mouse IgG (1:2000) or anti-rabbit IgG (1:2000) horseradish peroxidise-conjugated antibodies for 1 h at room temperature. The enhanced chemiluminescence system was used to develop blots. To quantify and compare protein levels, the density of each band was measured using Image J software. GAPDH was used to control for equal loading (full western blot images are shown in Supplementary Fig. 26) Immunoprecipitation. For immunoprecipitation total cellular proteins were isolated from human primary T regs with RIPA lysis buffer containing Protease/ Phosphatase Inhibitor Cocktail 1X (Cell Signalling Technology, Germany). Lysates (200 μg) were combined with 2 μg of PAR2 (SAM11) or PAR3 (H103) antibody and incubated overnight at 4°C. Immunoprecipitates were purified with protein A/G agarose beads and washed with PBS. Immunoprecipitates were fractionated by SDS-PAGE (10%), transferred to membranes, and subjected to immunoblotting with PAR3 (H103) or PAR2 (SAM11) antibody, respectively, and secondary antibodies as described above (full western blot images are shown in Supplementary Fig. 27).
EPCR blocking assay. To determine the role of EPCR in aPC signaling in T-cells, primary human T-cells were incubated with a rat monoclonal antibody to human endothelial cell protein C receptor (EPCR, 2 µg/ml, 1 h, 37°C). Following incubation, cells were washed with PBS, incubated with aPC (20 nM, 1 h, 37°C), and again washed with PBS. For control the EPCR blocking antibody was omitted, but cells were otherwise treated the same. T-cells with EPCR inhibition and aPC-preincubation, with only aPC-preincubation, or without EPCR inhibition and aPC-preincubation were co-cultured with allogenic AgPC for 96 h for MLR.
Virus production and transduction. To ascertain the role of human PAR2, we generated primary PAR2 knockdown pan T-cells. To this end, HEK-293T cells (ATCC) were routinely maintained in DMEM high glucose medium (Sigma-Aldrich) supplemented with 10% heat-inactivated FBS. Cells were transfected using the calcium-phosphate method and a third-generation lentiviral vector system 24 h after seeding 11 × 10 6 cells per 15 cm dish. In total, 39.5 µg pMD2.G (encoding the VSV G envelope protein), 73 µg psPAX2 (encoding the packaging proteins), and 112.5 µg transfer lentiviral plasmids (each at 1 mg/ml in TE buffer, pH 8) were used for five 15 cm dishes per shRNA construct 67 . Viral supernatants were harvested 24, 48, and 72 h post-transfection and filtered using 0.45 µm filters. Lentiviral particles were concentrated using the PEG 6000-based method 68 . Concentrated viral particles were snap frozen on crushed dry ice and stored at −80°C until usage. For RNA interference we used pLKO.1 lentiviral vectors expressing shRNA for hPAR2 (TRCN0000006769, TRCN0000006770).
Pan T-cells were isolated from PBMCs of healthy donors as described before. In brief, 2 × 10 6 cells were collected in 15 ml polypropylene Falcon tubes, re-suspended in a given volume of concentrated lentiviral vector supernatant (50-100 µl) in the presence of Polybrene (5 µg/ml). Tubes were centrifuged at 1500×g for 1.5 h at 32°C. Cells (1 × 10 6 ) were then seeded in 6-well plates and after 48 h cells were washed and incubated with aPC (20 nM, for 1 h at 37°C in AIM V serum-free medium) and seeded on 96-well plates coated with anti-human CD3 (1 µg/ml) and anti-human CD28 (0.5 µg/ml) at a density of 5 × 10 4 per well for proliferation assay.