Tocilizumab increases regulatory T cells, reduces natural killer cells and delays graft‐versus‐host disease development in humanized mice treated with post‐transplant cyclophosphamide

Graft‐versus‐host disease (GVHD) is a life‐threatening complication following donor hematopoietic stem cell transplantation, where donor T cells damage host tissues. This study investigated the effect of tocilizumab (TOC) combined with post‐transplant cyclophosphamide (PTCy) on immune cell engraftment and GVHD development in a humanized mouse model. NOD‐scid‐IL2Rγnull (NSG) mice were injected intraperitoneally with 2 × 107 human (h) peripheral blood mononuclear cells and cyclophosphamide (33 mg kg−1) or saline on days 3 and 4, then TOC or control antibody (0.5 mg mouse−1) twice weekly for 28 days. Mice were monitored for clinical signs of GVHD for either 28 or 70 days. Spleens and livers were assessed for human leukocyte subsets, and serum cytokines and tissue histology were analyzed. In the short‐term model (day 28), liver and lung damage were reduced in PTCy + TOC compared with control mice. All groups showed similar splenic hCD45+ leukocyte engraftment (55–60%); however, PTCy + TOC mice demonstrated significantly increased (1.5–2‐fold) splenic regulatory T cells. Serum human interferon gamma was significantly reduced in PTCy + TOC compared with control mice. Long‐term (day 70), prolonged survival was similar in PTCy + TOC (median survival time, > 70 days) and PTCy mice (median survival time, 56 days). GVHD onset was significantly delayed in PTCy + TOC, compared with TOC or control mice. Notably, natural killer cells were reduced (77.5%) in TOC and PTCy + TOC mice. Overall, combining PTCy with TOC increases regulatory T cells and reduces clinical signs of early GVHD, but does not improve long‐term survival compared with PTCy alone.


INTRODUCTION
Donor hematopoietic stem cell transplantation (HSCT) is a curative therapy for hematological malignancies, which are impervious to conventional chemotherapy and radiotherapy. 1 The therapeutic effect of donor HSCT is attributed to the graft-versus-leukemia (GVL) effect, which involves the elimination of residual host malignant cells via alloreactive donor T cells. 2 However, without post-transplant treatments, these alloreactive donor T cells recognize the healthy host tissues as foreign and mount a destructive inflammatory immune response known as graft-versus-host disease (GVHD) in 40-60% of HSCT recipients. 3 GVHD is fatal in 15-30% of cases 3 and results in severe damage primarily to the skin, liver, gastrointestinal tract 4 and lungs. 5 Damage to these tissues results in the release of proinflammatory cytokines, such as interleukin (IL)-6, which leads to the further activation of antigen-presenting cells and T cells, establishing a positive feedback loop that amplifies the inflammatory immune response. 6 Cyclophosphamide is a cytotoxic compound that selectively impairs alloreactive T cell proliferation. 7 Cyclophosphamide is often given at 50 mg kg À1 on days 3 and 4 post-donor HSCT 8 and referred to as posttransplant cyclophosphamide (PTCy). While PTCy reduces GVHD occurrence and severity, 51% of patients still develop the disease. 9 Additional therapies are therefore required to combat the development of GVHD following donor HSCT, which may involve the combination of PTCy with another therapeutic agent.
IL-6 is a pleiotropic cytokine with a central role in T cell differentiation 10 and GVHD development. 11 When IL-6 is present with transforming growth factor beta, naive T cells develop into T helper (Th) 17 cells, which have been shown to exacerbate GVHD. 12 However, in the absence of IL-6, the same naive T cells develop into regulatory T cells (Tregs), which have a protective role against GVHD. 13 As such, driving the T cell differentiation pathway away from inflammatory Th17 cells and toward Treg development is a potential strategy for the prevention of GVHD.
Tocilizumab (TOC) is a humanized monoclonal antibody directed against the IL-6 receptor (IL-6R), which blocks both classical and trans-signaling 14 and is specific for the human IL-6R. 15 TOC is currently used as a treatment for rheumatoid arthritis, 16 coronavirus disease 2019 (COVID- 19) 17 and other inflammatory disorders. The use of TOC as a therapy for GVHD has had limited success clinically, where a single 8 mg kg À1 dose on day-1 partly reduced grade II-IV acute GVHD, but did not improve overall survival. 18 Given the role of IL-6 in GVHD, the effectiveness of IL-6R blockade in preclinical mouse models and trends observed in human clinical studies, the use of TOC to treat GVHD still holds great potential. The current study aimed to examine the effects of PTCy or TOC alone and in combination on human immune cell subsets, cytokines and tissue damage at early and late time points in GVHD development using a preclinical humanized mouse model.

RESULTS
TOC combined with PTCy reduces clinical score, improves survival and reduces histological GVHD early in disease In our humanized mouse model, NOD-scid-IL2Rc null (NSG) mice injected with 2 9 10 7 human peripheral blood mononuclear cells (hPBMCs) develop clinical signs of GVHD from day 21 post-hPBMC injection. 19 To investigate the effect of PTCy, TOC and combined treatments on immune mechanisms occurring early in disease, NSG mice were humanized by injecting 2 9 10 7 hPBMCs (from eight healthy donors) and injected intraperitoneally with either TOC or a control antibody (0.5 mg mouse À1 ) twice weekly for 28 days. On days 3 and 4 post-hPBMC injection, mice were injected with cyclophosphamide (33 mg kg À1 ; PTCy) or phosphatebuffered saline (PBS), resulting in four treatment groups: control, TOC alone, PTCy alone and PTCy + TOC (Figure 1a). GVHD development was assessed by monitoring clinical symptoms, including weight loss, as described. 20 Weight loss was observed in control mice from day 12 post-hPBMC injection (Figure 1b). By contrast, PTCy-or TOC-treated mice gained weight over 28 days and similarly, PTCy + TOC mice continuously gained weight over 28 days, with reduced weight loss at day 28 compared with control mice (P = 0.05). At day 28, the clinical score was reduced by 76% and 65%, respectively, in TOC (P = 0.05) and PTCy mice (P = 0.01) compared with control mice (Figure 1c). Similarly, PTCy + TOC mice showed an 83% reduction in clinical score (P = 0.01) compared with control mice. Clinical scores between PTCy + TOC mice were similar to PTCy (P = 0.88) or TOC mice (P = 0.67). In the short-term model, survival outcomes were significantly longer in mice treated with TOC alone [median survival time (MST), > 28 days; P = 0.007] and PTCy alone (MST, > 28 days; P = 0.001) compared with control mice (MST, 26 days; Figure 1d). Likewise, PTCy + TOC mice showed prolonged survival (MST, > 28 days) compared with control mice (P = 0.001). Survival outcomes of PTCy + TOC mice were similar to PTCy (P > 0.99) and TOC (P = 0.34) mice.
The liver, skin and lung are target tissues of GVHD in humanized NSG mice, 21 but previous studies using PTCy have only examined these target tissues at endpoint. 19 Thus, histological assessments were performed on day 28 to examine the effects of treatment on early GVHD in these organs. Control mice exhibited substantial immune cell infiltration in the liver (Figure 1e). Treatment with TOC or PTCy alone reduced GVHD severity (histological grades 0-2, mean > 1) in the liver compared with control mice (histological grades 1-4, mean > 2), but this was not statistically significant. PTCy + TOC resulted in a significant reduction of GVHD severity (histological grade 0-1, mean < 0.5) compared with control mice (histological grade 1-4, mean > 2; P = 0.002). PTCy + TOC mice also showed a 50% reduction in GVHD severity (histological grade 0-1, mean < 0.5) compared with TOC or PTCy (histological grades 0-2, mean > 1) but this was not statistically significant. Schematic overview of humanized mouse model. NSG mice were injected intraperitoneally (i.p.) with 2 9 10 7 human peripheral blood mononuclear cells (hPBMCs; n = 8 donors) and subsequently injected i.p. with cyclophosphamide (33 mg kg À1 ) or saline on days 3 and 4 post-hPBMC injection (PTCy) and/or TOC or control antibody (Ctrl Ab; 0.5 mg mouse À1 ) twice weekly for 28 days. Mice were monitored for clinical signs of GVHD including (b) weight, (c) clinical score and (d) survival over 28 days. Tissues were sectioned and stained with hematoxylin and eosin. Sections of (e) liver, (f) skin and (g) lung were examined for histological GVHD. (e) Liver and (f) skin were assessed using a standardized grading system. (g) Lung GVHD was assessed as the percent of clear alveoli area of total lung area. Images are representative of up to 10 mice per treatment group. (b, c, e-g) Data are presented as the mean AE standard error of the mean. Data are from two independent experiments with four donors per experiment. Significance was analyzed using (b, c) two-way ANOVA, (d) log-rank (Mantel-Cox), (e) Kruskal-Wallis and (g) one-way ANOVA tests. **P < 0.01, *P < 0.05.
Evidence of GVHD in the skin was similar among all treatment groups (histological grades 0.5-3, mean > 0.5; Figure 1f). Immune cell infiltration was low to moderate, and evidence of epidermal thickening was negligible across all treatment groups. The mean percentage of clear alveoli space (an inverse measure of lung GVHD) was 40.7% in control mice (Figure 1g). The mean percentage of clear alveoli space was similar in TOC mice (41.4%) and was increased in PTCy mice (47.6%) compared with control mice. The alveoli space in the lung was significantly increased (1.3-fold) in PTCy + TOC-treated mice compared with control mice (P = 0.006). Lung pathology in PTCy + TOC mice was similar to PTCy mice (P = 0.39), but the clear alveoli space was significantly increased (1.3-fold) in comparison to TOC mice (P = 0.02). Histological evidence of GVHD was negligible in the ear as well as the duodenum in all groups and similar but minor histological GVHD was evident in the kidney of all groups (Supplementary figure 1).

TOC combined with PTCy increases splenic human Tregs early in disease
PTCy alters the proportion of human leukocytes in humanized mice 19 and TOC was found to increase Tregs in in vitro experiments prior to exploring its efficacy as a GVHD treatment in humanized mice (Supplementary figure 2). Furthermore, human leukocytes engraft in the spleens of humanized NSG mice. 22 Therefore, the effect of treatments on human leukocyte engraftment was assessed by flow cytometry of spleens from humanized mice, 28 days post-hPBMC injection or at ethical endpoint ( Figure 2). Flow cytometric analysis was performed using the gating strategy shown (Figure 2a). TOC or PTCy alone or in combination did not impact engraftment of hCD45 + cells or hCD3 + T cells, with each group showing similar proportions of hCD45 + leukocytes (53.9-55.3%) compared with control mice (P = 0.99; Figure 2b). These cells were predominantly hCD3 + T cells (86.2-88.3%) with similar proportions in all four groups (P = 0.73) (Figure 2c).
The splenic hCD4 + :hCD8 + T cell ratio correlates with clinical GVHD severity in humanized NSG mice. 23 Similar proportions of hCD4 + and hCD8 + T cells were observed in TOC, PTCy + TOC and control mice, but the proportion of hCD4 + T cells was 38% less than hCD8 + T cells in PTCy mice (P = 0.04; Figure 2d). However, there was no significant difference in the hCD4 + :hCD8 + T cell ratio between all four groups (P = 0.41; Figure 2e).
As TOC inhibits IL-6 signaling, which is involved in the development of Tregs and Th17 cells, we examined whether each of these subsets were altered by TOC or PTCy, alone or in combination. Tregs reduce GVHD severity by inhibiting the activation and proliferation of pathogenic T cells. 24 Proportions of splenic hTregs (hCD4 + hCD25 + hCD127 low ) were similar in TOC, PTCy and control mice (6.4-6.9%; Figure 2f). Notably, the proportion of splenic hTregs was increased (approximately 2-fold) in PTCy + TOC mice compared with control mice (P = 0.001), as well as with PTCy (P = 0.01) and TOC mice (P = 0.03). The proportions of highly suppressive hCD39 + hTregs, which can also influence GVHD, 25 were similar in all four groups (63.9-76.7%; P = 0.53; Figure 2g).
Natural killer (NK) cells are the main non-T cell population that engraft NSG mice in this model, 29 with minimal engraftment of B cells, monocytes and dendritic cells. 30,31 Moreover, NK cells can suppress GVHD in mice, 32 but promote GVL responses. 33 Thus, NK cells were also examined in the current study. The proportions of splenic hNK (hCD3 À hCD56 + ) cells were similar in all four groups (8.4-9.8%; P = 0.67; Figure 2k). NK T cells (NKT) can be suppressive and protect against GVHD. 34 The proportions of splenic hNKT (hCD3 + hCD56 + ) cells, which suppress GVHD, 35 were relatively low and also similar in all four groups (0.9-2.2%; P = 0.30; Figure 2l).

TOC combined with PTCy increases hepatic human Tregs early in disease
The liver is one of the primary target organs of GVHD and previous studies have shown that human leukocytes engraft in the livers of humanized NSG mice at early time points. 19 Therefore, the effect of PTCy, TOC and combined treatments on human leukocyte engraftment was assessed by flow cytometry of livers from humanized mice, 28 days post-hPBMC injection, or at ethical endpoint, using the aforementioned gating strategy ( Figure 2a). As was observed in the spleen, TOC or PTCy alone or in combination did not impact the engraftment of hCD45 + cells in the liver, with each group showing similar . Tocilizumab (TOC) combined with post-transplant cyclophosphamide (PTCy) significantly increases the proportion of splenic Tregs in humanized NOD-scid-IL2Rc null (NSG) mice. Spleens from humanized mice were collected at day 28 and human (h) immune cell subsets were analyzed by flow cytometry using (a) a consistent gating strategy. Live cells were gated based on Zombie NIR staining. Single cells were gated based on forward scatter height (FSC-H) and forward scatter area (FSC-A; not shown) before viable human and mouse (m) leukocytes were gated using side scatter area (SSC-A) and FSC-A. The proportion of (b) hCD45 + and mCD45 + leukocytes was then identified before gating, (c) hCD3 + T cells, (d) hCD4 + and hCD8 + T cell subsets, (e) hCD4 + :hCD8 + T cell ratio, (f) hCD4 + hCD25 + hCD127 low regulatory T cells (hTregs), (g) hCD39 + hTregs, (h) hCD4 + hCD161 + hCD39 + T helper 17 (hTh17) cells, (i) hTh17:hTreg ratio, (j) hCD8 + hCD161 high cytotoxic T cells (hTc17) cells, (k) hCD3 À hCD56 + natural killer (hNK) cells and (l) hCD3 + hCD56 + natural killer T (hNKT) cells. Data are presented as the mean AE standard error of the mean. Symbols represent individual mice. Data are from two independent experiments with four donors per experiment. Significance was determined using either (b, f, h-j) one-way ANOVA or (c-e, g, k, l) Kruskal-Wallis tests. **P < 0.01, *P < 0.05. Ctrl Ab, control antibody.
TOC combined with PTCy reduces serum hIFN-c and hepatic hIFN-c + and hIFN-c + hIL-17 + early in disease Interferon gamma (IFN-c) and IL-17 promote GVHD. 36 Therefore, the concentration of hIFN-c in sera was assessed in all groups by ELISA, and the proportion of hIFNc-and hIL-17-producing cells in splenocytes and hepatocytes was assessed by flow cytometry at day 28 post-hPBMC injection. The concentration of serum hIFNc was reduced by 33% in PTCy mice (P = 0.17) and by 18% in TOC mice (P = 0.66) compared with control mice, but this was not significant. Notably, the serum hIFN-c concentration was significantly reduced by 46% in PTCy + TOC compared with control mice (P = 0.04). PTCy + TOC mice also showed reduced serum hIFN-c concentrations when compared with PTCy or TOC mice (20% and 35%, respectively), but this was not significant (Figure 4a). Intracellular staining and flow cytometry ( Figure 4b) revealed that the proportion of splenic hIFN-c + (P = 0.62; Figure 4c) or splenic hIL-17 + (P = 0.44) producing cells was similar in all four groups ( Figure 4d). Likewise, the proportion of splenic doublepositive hIFN-c + hIL-17 + -producing cells was similar between groups (P = 0.90; Figure 4e). In both instances, apparent decreases were largely attributed to one high value in the respective control group. Similar to splenocytes, hIFN-c + cells and hIL-17 + cells in the hepatocytes from all four groups were similar (Figure 4f, g). Notably, the proportions of hepatic hIFN-c + hIL-17 + -producing cells were reduced by 94% in TOC (P = 0.12), 93% in PTCy (P = 0.82) and 95% in PTCy + TOC (P = 0.10) mice compared with control mice. The proportions of these double-positive cells were similar between the three treatment groups (P > 0.99; Figure 4h). Treatment with TOC or PTCy alone or in combination did not affect the relative expression of IFNG and IL17A at day 28 (Supplementary figure 3).
TOC combined with PTCy delays GVHD development, but does not improve long-term outcomes compared with PTCy alone PTCy given on days 3 and 4 post-hPBMC injection has been shown to reduce clinical score and weight loss, and improve survival in humanized mice. 19 To examine further whether the addition of TOC to PTCy treatment could prevent GVHD, NSG mice were injected with 2 9 10 7 hPBMCs, treated as above ( Figure 1a) and monitored for up to 70 days for clinical signs of GVHD (Figure 5a). Weight gain and loss were similar to those observed in the 28-day study (Figure 1b) with weight loss continuing for control mice (Figure 5b). Notably, TOC mice also lost weight from day 28. By contrast, both PTCy and PTCy + TOC mice continued to gain weight, which plateaued at about day 35. The overall difference in weight between treatment groups was not significant (P = 0.21). Clinical scores also paralleled the 28-day study (Figure 1d), with scores increasing in all four groups, but quickest in the control group, followed by the TOC group. Notably, the score in the TOC group was similar to PTCy and PTCy + TOC until day 48 before increasing sharply. Prolonged survival was observed for TOC (MST, 38 days) and PTCy (MST, 56 days) mice and was greatest for PTCy + TOC mice (MST, > 70 days) However, there was no significant difference in overall survival between groups (P = 0.17; Figure 5d).
Given the clinical data above, time to GVHD onset, defined as a clinical score of ≥3, was assessed. Time to GVHD onset was similar between TOC (28 days) and control mice (21 days). By contrast, time to disease onset was significantly increased in PTCy (40 days; (P = 0.04) and PTCy + TOC mice (51 days; P = 0.008). Further, disease onset in PTCy + TOC mice was significantly delayed compared with TOC mice (P = 0.0005) and PTCy mice, although the latter did not reach statistical significance (P = 0.23; Figure 5e). Notably, at endpoint, 55% of PTCy + TOC mice survived, compared with only 25% of PTCy and 20% of TOC and control mice.
Histological assessment of the liver, skin and lung (Figure 5g, h) was performed on samples at endpoint. Substantial immune cell infiltration and tissue damage was observed in the liver of control mice. Treatment with TOC or PTCy did not reduce GVHD-associated tissue damage (histological grades 0.5-4, mean > 1.5) compared with control mice (histological grades 1-4, mean > 2). PTCy + TOC resulted in a 30% reduction in histological GVHD (histological grades 0.5-4, mean > 1) compared with control mice; however, this was not significant (P > 0.99) and was similar to TOC-and PTCy-treated mice (Figure 5f). Moreover, evidence of GVHD in the skin was similar among all treatment groups (histological grades 0.5-3, mean > 1.5), with mice across all groups experiencing epidermal thickening and immune cell infiltration (P = 0.92; Figure 5g). The mean percentage of clear alveoli space in the lungs of control mice was 34.6%. The mean percentage of clear alveoli space was similar in TOC and PTCy alone mice (34.8-38.0%) compared with control mice. In PTCy + TOC mice, the mean percentage of clear alveoli space was greater (1.2fold) (43.0%) than that in control and PTCy mice; however, this was not significant and was similar to TOC mice (P > 0.99; Figure 5h). Histological evidence of GVHD was present in the ear and kidney of mice from all groups but was negligible in the duodenum (Supplementary figure 4). Outlier data points (Figure 5e-g) were more closely examined to determine whether these points occurred as a result of a correlation with Treg percentage. No such correlation was found (data not shown).

TOC combined with PTCy does not impact human cell engraftment, but TOC reduces the proportion of splenic and hepatic NK cells at endpoint
The effect of PTCy, TOC or combined PTCy + TOC therapy on human leukocyte engraftment, hCD4 + and hCD8 + T cell subsets and hNK and hNKT cells was analyzed by flow cytometry in the spleen and liver of humanized NSG mice at endpoint in the long-term model using a consistent gating strategy (Figure 2a).
The proportions of splenic hTh17 cells were similar across all four groups (30-38%) (P = 0.88; Figure 6g). Likewise, the hTh17:hTreg ratio was similar between all four groups (P = 0.99), with the apparent increase in the PTCy group resulting from a single outlier (Figure 6h). Moreover, the proportions of hTc17 cells were similar between all groups (2.0-4.3%; P = 0.36; Figure 6i).
Although hNK cell engraftment was typically quite low in this study (0.6-2.8%), treatment with TOC significantly reduced (77.5% reduction; P = 0.05) and PTCy + TOC partly reduced (P = 0.12) hNK cells, whereas treatment with PTCy had no impact compared with control mice. TOC and PTCy + TOC resulted in similar hNK cell proportions (P > 0.99), which were low compared with PTCy alone (P = 0.80; Figure 6j). Proportions of splenic hNKT cells were similar in all four groups (P = 0.70; Figure 6k).
In the liver, TOC, PTCy or PTCy + TOC did not impact the engraftment of hCD45 + cells compared with control mice (75.3-82.9%; P = 0.73; Figure 7a), similar to day 28. As observed in the spleens of humanized mice at endpoint (Figure 6), the proportions of hCD3 + T cells were greater in TOC and PTCy + TOC mice (P = 0.19) than in PTCy and control mice which were similar (P = 0.94; Figure 7b). Despite this increase in hCD3 + T cells, the proportions of hCD4 + and hCD8 + T cells were similar among all four groups (P = 0.91; Figure 7c), as was the hepatic hCD4 + : hCD8 + T cell ratio (1.1-2.4; P = 0.97; Figure 7d).
Treatment with TOC, PTCy or PTCy + TOC did not affect the proportions of hepatic hTregs compared with control mice (5.3-9.4%; P > 0.99). However, TOC or PTCy + TOC resulted in similar proportions of hTregs, which were slightly greater (approximately 1.6-fold) than PTCy alone, with PTCy + TOC approaching significance (P = 0.17; Figure 7e). Although there was a minor difference observed in the proportions of hTregs, the proportions of hCD39 + hTregs were similar across all treatment groups (P = 0.94; Figure 7f).

DISCUSSION
The current study investigated the effects of TOC and PTCy alone on GVHD development, immune cell engraftment, cytokines and tissue damage at an early time point (28 days), none of which have been compared previously. Further, this study investigated whether the therapeutic benefits of PTCy, which has shown some efficacy clinically and in mouse models of GVHD, 9,19,37,38 could be further improved by the addition of TOC in this preclinical model long-term. The results demonstrated that combined therapy with PTCy and TOC delayed GVHD onset, reduced weight loss and prolonged survival compared with control mice. Furthermore, combination therapy with PTCy and TOC increased splenic hTregs, reduced the concentration of serum hIFN-c and reduced histological GVHD in the liver and lung early in the disease. However, this combination therapy did not significantly improve long-term survival outcomes or the development of clinical GVHD compared with PTCy alone. Notably, the hNK cell population, which plays a pivotal role in GVL responses and post-transplant survival, was reduced at endpoint following treatment with TOC.
In the current study, combination therapy with PTCy and TOC delayed GVHD onset and attenuated GVHD severity. Regarding the use of PTCy alone, this delay in GVHD onset is consistent with findings in mice from our group 19 and others 38 and similar results have been observed in the clinical setting. 9 Regarding the use of TOC alone, findings from the current study are consistent with those in allogeneic 11,13 and humanized 25 mouse models, and with findings from a phase III clinical trial, 18 which found that a single dose of TOC on day À1 of transplantation resulted in a trend toward reduced acute GVHD, but did not improve overall survival.
Tregs are reportedly resistant to depletion by PTCy 39 and previous studies have shown that the presence of Tregs is essential for PTCy to elicit therapeutic effects against GVHD. However, our group has previously found that PTCy reduced GVHD severity despite reducing the population of Tregs. 19 Hence, we sought to improve the efficacy of PTCy by combining it with TOC, targeted at increasing the Treg population. In the current study, PTCy combined with TOC increased the proportion of hTregs early in disease and at endpoint. The attenuation of GVHD severity observed aligns with findings from studies in allogeneic mouse models which employ IL-6R blockade as a therapeutic strategy for GVHD prevention 11,13 However, observations from these studies are conflicting, with one study reporting an increase in Tregs 13 and the other reporting no change. 11 Another study examining the effects of TOC in humanized mice did not examine Treg proportions. 25 The current study did not examine the impact of PTCy and TOC on naive and memory T cells, as we previously found that PTCy had minimal impact on na€ ıve and memory T cell subsets in humanized mice. 19 Moreover, the current study did not investigate the effect of TOC on B cells, because of consistently low B cell engraftment in this model. 40,41 Likewise, the effect of TOC in the absence or presence of PTCy on naive and memory T cell subsets and B cells was not reported in recent clinical studies. 18 Thus, the impact of TOC on these cell types following allogeneic (or xenogeneic) transplantation remains to be explored.
Although combination therapy with PTCy and TOC significantly delayed the onset of GVHD compared with control mice, this delay was not significant compared with PTCy alone. The combination of PTCy and TOC reduced clinical score and improved survival compared with PTCy alone, but this did not reach significance. The dose and timing of both PTCy and TOC may be suboptimal. In this study, PTCy was administered at 33 mg kg À1 on days 3 and 4 post-hPBMC injection, which has shown efficacy in delaying GVHD development in our humanized mouse model of GVHD. 19 A recent study in an allogeneic HSCT model has demonstrated that 25 mg kg À1 PTCy administered on days 3 and 4 post-donor HSCT is optimal for limiting GVHD. 37 Notably, in the current study, cessation of TOC injections occurred after 28 days, at which point, TOC mice began to succumb to GVHD. However, previous studies have utilized a single injection at a higher dose. In several clinical trials 18,42 and one other humanized mouse model, 25 a single 8 mg kg À1 dose of TOC was given. In this previous humanized mouse model, a single injection of TOC resulted in improved survival, but no improvement in weight loss, 25 paralleling the findings of the current study. Given that TOC does not cross-react with the murine IL-6R, future studies using an anti-murine IL-6R antibody may better establish the therapeutic potential of IL-6R blockade in GVHD and help determine the relative contribution of host (murine) IL-6 signaling in this model of GVHD.
PTCy has been shown to reduce histological GVHD in the liver of humanized NSG mice. 19 In the current study, combining PTCy with TOC further reduced histological liver and lung GVHD early in disease, but did not affect the long-term outcomes. These findings align with previous studies which show that IL-6R blockade reduced liver and lung damage 5 weeks after transplant in mice. 13 The lack of difference observed at endpoint could be a result of cessation of treatment with TOC at day 28, which may have impacted Tregs and Th17 cells that differentiate depending on the absence or presence of IL-6 signaling, respectively. Future studies may test longer-term TOC therapy to address this.
Despite efficacy in mouse models, clinical trials using TOC as prophylaxis for GVHD have had limited success. In one study where TOC was given as a treatment for steroid-refractory GVHD, a clinical response in four of six patients (67%) treated with TOC was observed, however, two patients succumbed to gastrointestinal GVHD. 43 More recently in a phase III clinical trial, 18 trends toward reduced incidence of GVHD and improved GVHD-free survival in patients treated with TOC were observed. Although results from these studies may indicate that IL-6R blockade with TOC has limited therapeutic benefit in GVHD, it is important to note that TOC has not been administered in combination with PTCy nor long-term. Nevertheless, combination therapy with PTCy and TOC in the current study resulted in a significant increase in the time to disease onset. This outcome indicates that this combination therapy is of some therapeutic benefit and future studies should seek to determine the optimal dose and treatment regimen for using combination therapy with PTCy and TOC.
IFN-c has a complex and pleiotropic role in GVHD. 44 We have previously shown that reduced IFN-c in the sera of humanized NSG mice is correlated with reduced GVHD, 21,23 yet treatment with PTCy increases this IFN-c concentration. 19 In the present study, combination of PTCy and TOC led to reduced hIFN-c as well as increased hTregs early in disease. Markedly, the transplantation or expansion of Tregs has shown to reduce serum IFNc in both allogeneic and humanized mouse models. 45,46 As such, the reduction of IFN-c observed in the sera of PTCy and TOC mice at day 28 in the current study may be a result of the increase in hTregs. This is further supported by the similarity between groups in hIFN-c concentration and the loss of significance in the increase of hTregs at endpoint. Double-positive hIFN-c + hIL-17 + cells are a subset of CD4 + T cells that are similar to Th17 cells and are significantly elevated in inflamed tissues. 47 Treatment with PTCy, TOC or a combination of both lead to a reduction of hIFN-c + hIL-17 + cells in the liver of NSG mice at day 28, suggesting a reduction of liver inflammation early in disease.
Notably, the proportion of splenic and hepatic hNK cells was reduced in TOC alone and PTCy + TOC mice at endpoint. IL-6 has been shown to improve the proliferation and cytotoxicity, as well as other important functions of NK cells. 48 As such, it is perhaps unsurprising that a loss of the hNK cell population was observed. Previous studies have shown that donor HSCT grafts containing higher numbers of NK cells are associated with less severe GVHD. 49 More recently, the recovery of NK cells following donor HSCT has been identified as a risk marker for overall survival. 50 In this recent study using machine learning analysis, the rate of relapse and nonrelapse mortality was significantly higher in patients with NK cell counts < 50.5 cells lL À1 on day 28 following transplant. This may have detrimental implications for the use of this combination therapy as a treatment for GVHD.
The proportions of splenic hNKT cells were not impacted by PTCy + TOC nor either treatment alone. By contrast, each of these treatments resulted in partial increases in the proportion of these cells in the liver at day 28 but not at day 70. These findings are consistent with a preventative role of NKT (invariant NKT) cells in GVHD. Whether PTCy or TOC have direct effects on NKT cells or act indirectly via effects on T cells, for example, remains to be explored. However, a limitation of the current study is that hNKT cells were identified using only CD3 and CD56, which may have included other T cell subsets. 34 Thus, future studies could use antibodies to the Va24-Ja18 T cell receptor or CD1d tetramer loaded with alphagalactosylceramide to confirm that PTCy and TOC do not impact proportions of invariant NKT cells.
In conclusion, this study has demonstrated that combining PTCy with TOC delays GVHD onset compared with PTCy alone and increases protective Tregs in a humanized mouse model. Although this combination therapy delayed GVHD onset and increased the proportion of Tregs early in disease, a high proportion of mice treated with PTCy and TOC in combination still succumbed to disease. Further investigation of this combination therapy is warranted and should seek to develop an improved treatment regimen to maintain the protective Treg population and reduce potential cyclophosphamideinduced toxicity. Furthermore, the addition of TOC depleted the population of hNK cells which have an important role in the GVL response. Future studies may also need to address whether the GVL response is maintained when PTCy combined with TOC is given as a treatment for GVHD.

Mice
Experiments involving mice were approved by the University of Wollongong (Wollongong, NSW, Australia) Animal Ethics Committee (AE 16/23 and AE 20/01). Female NSG mice 5-7 weeks of age were obtained from the Animal Resources Centre (Canning Vale, WA, Australia) or Australian BioResources (Moss Vale, NSW, Australia) and were acclimatized for 2 weeks prior to commencing experimental work. Mice were housed in individually ventilated cages (Tecniplast, Buguggiate, Italy) and provided with irradiated food and autoclaved water, ad libitum.

Humanized mouse model of GVHD
Experiments involving human blood were approved by the University of Wollongong Human Ethics Committee (approval number HE 12/290). Peripheral blood was collected with consent from healthy human donors (seven males and two females; 23-40 years of age) and hPBMCs were isolated by density gradient centrifugation as described. 20 Isolated hPBMCs were washed with sterile Dulbecco's PBS (Thermo Fisher Scientific, Waltham, MA, USA; 440g, 10 min) and resuspended in PBS at 10 9 10 7 cells mL À1 for injection into NSG mice. NSG mice, not subjected to any pretransplant conditioning, were injected intraperitoneally with 2 9 10 7 hPBMCs and then injected intraperitoneally with either TOC (Chugai Pharmaceutical, Tokyo, Japan) or rat immunoglobulin G (control antibody; Sigma-Aldrich, St Louis, MO, USA; 0.5 mg mouse À1 ) twice weekly for 28 days. Mice were injected intraperitoneally with either cyclophosphamide (33 mg kg À1 ) (Sigma-Aldrich) or PBS on days 3 and 4 post-hPBMC injection. Mice were monitored for clinical signs of GVHD for up to 28 days (short-term model) or 70 days (long-term model), in a blinded fashion, as described. 20 Tissues were collected from mice euthanised by CO 2 inhalation at days 28 and 70 (or at the ethical endpoint).

Immunophenotyping
Splenocytes and hepatocytes from humanized mice were isolated as described 21 prior to live/dead staining with Zombie Near Infrared dye (BioLegend, San Diego, CA, USA) and subsequent incubation with fluorochrome-conjugated monoclonal antibodies (Supplementary table 1) in PBS containing 2% fetal calf serum (Thermo Fisher Scientific). Samples were washed in PBS (300 g for 5 min) and resuspended in PBS. Data were collected using an LSRFortessa X-20 flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA) using the appropriate band pass filters (Supplementary table 1) and FACSDiva software version 8.0 (BD Biosciences). Immune cell subsets were analyzed using FlowJo software version 10.8.1 (BD Biosciences).

Hematoxylin and eosin staining
Tissue samples were collected from euthanised mice and fixed overnight in neutral buffered (10%) formalin (Sigma-Aldrich) before being embedded in paraffin. Tissues were sectioned (3 lm) using an RM2255 microtome (Leica Biosystems, Wetzlar, Germany) and stained with hematoxylin and eosin (POCD, Artarmon, NSW, Australia) as described. 30 Histological damage was assessed using a Leica DM750 inverted light microscope with the 209 objective for all tissues, except with the 109 objective for duodenum. Images were captured using Leica Application Suite Software version 4.7. Histological GVHD in the liver and skin was graded in a blinded fashion using a standardized grading system (using grades from 0 to 4) as described. 46 Histological GVHD in the lung was determined by blinded area measurements of open alveoli space using Fiji 51 and quantified as a percentage of the total lung area measured. 29 ELISA Serum was isolated from mouse blood collected via cardiac puncture as previously described 30 and stored at À80°C. IFN-c was measured using a human IFN-c ELISA kit (Life Technologies, Carlsbad, CA, USA), per the manufacturer's instructions. Absorbances (570 nm and 450 nm) were measured using a SpectraMax Plus 384 plate reader (Molecular Devices, Sunnyvale, CA, USA).

In vitro effects of TOC
Freshly isolated hPBMCs were suspended in RPMI-1640 medium containing 10% fetal calf serum and 2 mM GlutaMAX at 1 9 10 6 cells mL À1 and seeded at 0.5 9 10 6 cells mL À1 in a 24-well plate. Cells were stimulated with 1 lM phytohemagglutinin (Sigma-Aldrich), treated with 10 lg mL À1 of either control antibody or TOC and cultured for 6 days at 37°C with 95% air/5% CO 2 . At day 6, cells were transferred to 5-mL polystyrene tubes, washed twice with PBS and stained for immunophenotyping as described earlier.

RNA isolation and complementary DNA synthesis and gene expression analysis
Tissues were collected from humanized mice and stored in RNAlater (Sigma-Aldrich) at À20°C. RNA was isolated from the spleen and liver using TRIsure (Thermo Fisher Scientific), per the manufacturer's instructions, and stored at À80°C. Complementary DNA was synthesized from RNA using a qScript SuperMix Kit (Quanta Biosciences, Beverly, MA, USA), per the manufacturer's instructions. The quality of complementary DNA was checked as described 20 and stored at À20°C. Relative gene expression was analyzed using TaqMan Universal Master Mix II (Thermo Fisher Scientific), per the manufacturer's instructions. Quantitative real-time PCR was conducted with the following primers (Thermo Fisher Scientific): FAM-labeled-human IFNG (Hs00989291_m1) and human IL17A (Hs00174383_m1) with human T cell receptor beta-chain constant region TCRBC (custom primer) as the house keeping gene. Amplifications were performed in triplicate, as described, 20 using an Applied Biosystems QuantStudio 5 real-time PCR System (Thermo Fisher Scientific) with QuantStudio Design and Analysis Software version 1.4.3. Gene expression was quantified using the DDCt method and was calculated relative to the gene expression of a single control mouse.

LEGENDplex
Serum was isolated from mouse blood collected via cardiac puncture as described 30 and stored at À80°C. The concentration of human IL-2, IL-6, IL-10, IFN-c and TNFa in mouse sera was determined using a 5-plex Th1 panel LEGENDplex kit (BioLegend, San Diego, CA, USA) per the manufacturer's instructions. Data were collected using an Invitrogen Attune NxT acoustic focusing cytometer and Attune NxT software version 3.1 and analyzed using LEGENDplex data analysis software version 8.0.

Data presentation and statistical analysis
All data are presented as the mean AE standard error of the mean. Data were tested for normality using a Shapiro-Wilk test. Statistical differences were determined using a two-tailed Students t-test (parametric) or Mann-Whitney U test (nonparametric) for single comparisons or a one-way ANOVA (parametric) or Kruskal-Wallis (nonparametric) with Tukey's post hoc test for multiple comparisons. Comparisons of mouse weight and clinical score over time were analyzed using a twoway ANOVA. Mouse survival was analyzed using a log-rank (Mantel-Cox) test. All statistical analysis and graphs were generated using GraphPad Prism software version 9.1.1 (GraphPad Software, La Jolla, CA, USA). For all analyses, differences with P < 0.05 were considered statistically significant.