Early myeloid-derived suppressor cells (HLA-DR⁻/^lowCD33⁺CD16⁻) expanded by granulocyte colony-stimulating factor prevent acute graft-versus-host disease (GVHD) in humanized mouse and might contribute to lower GVHD in patients post allo-HSCT

NOD-SCID-IL-2Rg−/− mice (NSG mice) were purchased from Vitalstar. All mice were bred and housed in a specific pathogen-free facility in microisolator cages and used at 8 to 12 weeks of age in protocols approved by the local Ethical Committee. Each experimental group included six to eight mice, and the experiment is repeated twice.

This study was approved by the Ethics Committee of Peking University People’s Hospital, Beijing, in accordance with the Declaration of Helsinki. Healthy donors were enrolled from the related donors of patients who had undergone either haplo-identical or HLA-matched allogeneic blood and marrow transplantation at the Peking University People’s Hospital between April 2017 and May 2018. Peripheral blood and BM samples (10 ml in EDTA tube) from recombinant human G-CSF (rhG-CSF)-non-treated or recombinant human G-CSF-treated (filgrastim, 5 μg/kg/day, five consecutive days) healthy allogeneic donors (randomly selected at Peking University People’s Hospital in Beijing) were collected after informed consent was obtained.

Different MDSC types in donor PB and BM samples were analyzed using flow cytometry (FACS Canto II, BD Biosciences) with the following antibodies: CD45, CD3, CD19, CD56, HLA-DR, CD11b, CD33, CD14, CD15, and CD16 (eBioscience or BD Bioscience) (Table 1).

HLA-DR^−/lowCD33⁺CD16⁻ cells were isolated from GPBSC using fluorescence-activated cell sorting (FACS)—sorted (FACS Aria II, BD Biosciences) with antibodies of HLA-DR, CD33, and CD16 (eBioscience or BD Bioscience) according to the manufacturer’s instructions, which included one step of positive selection for CD33⁺ cells and two steps of negative selection for HLA-DR⁻ and CD16⁻ cells.

HLA-DR^−/lowCD33⁺CD16⁻ morphology was studied on cytospins after May-Grünwald-Giemsa staining. Photographs were taken using × 1000 magnification with a smart V350Df digital camera mounted on an OLYMPUS CX31 microscope.

T cells from GPBSC were purified by positive selection using CD3 cell isolation via flow cytometry. Purified CD3⁺ T lymphocytes were labeled with 5 μM 5,6-carboxy-fluorescein diacetate succinimidyl ester (CFSE) (eBioscience) and thereafter were co-cultured with isolated HLA-DR^−/lowCD33⁺CD16⁻ at a HLA-DR^−/lowCD33⁺CD16⁻ to CD3⁺ T cell ratio of 0.25:1 to 2:1 in the presence of anti-CD3/CD28 beads (Thermo) in RPMI 1640 (Biological Industries) supplemented with 10% FBS, penicillin/streptomycin, and 2 mM l-glutamine. 10⁵ T cells were seeded in 96-well U-bottom plates. After 4 days, cells were harvested for flow cytometry analysis. To neutralize IL-10 and TGF-β, blocking antibodies were used at 10 to 20 μg/ml concentrations (blocking antibody for IL-10 and TGF-β from R&D Systems). To inhibit arginase activity, nor-NOHA (Calbiochem) was used at a concentration of 300 μM. For the inhibition of iNOS, l-NMMA (Sigma) was used at a concentration of 300 μM. To inhibit IDO, indoximod (NLG8189) (Selleck) was used at a concentration of 500 μM. To inhibit Cox2, NS398 (Selleck) was used at a concentration of 10 μM. All the in vitro experiment was repeated at least three times.

T cells from G-PBSC were sorted via flow cytometry. Purified CD3⁺ T lymphocytes were co-cultured with isolated HLA-DR^−/lowCD33⁺CD16⁻ at a HLA-DR^−/lowCD33⁺CD16⁻ to CD3⁺ T cell ratio of 1:1 in the presence of anti-CD3/CD28 beads (Thermo) in RPMI 1640 (Biological Industries) supplemented with 10% FBS, penicillin/streptomycin, and 2 mM l-glutamine. Four days later, the percentages of Th1 (CD4⁺IFNγ⁺), Th2 (CD4⁺IL-4⁺), and Th17 (CD4⁺IL-17A⁺) were detected via flow cytometry, then the fold changes of Th2/Th1 and Th2/(Th1+Th17) were calculated. Human Th1/Th2/Th17 Phenotyping Kit (BD Pharmingen) was used.

T cells from G-PBSC were sorted via flow cytometry. Purified CD3⁺ T lymphocytes were co-cultured with isolated HLA-DR^−/lowCD33⁺CD16⁻ at a ratio of 1:1 in RPMI 1640 (Biological Industries) supplemented with 10% FBS, penicillin/streptomycin, and 2 mM l-glutamine. Four days later, the percentage of Treg (CD4⁺CD25⁺FoxP3⁺) was detected via flow cytometry.

NSG mice were sublethally irradiated using 180-cGy total body irradiation by X-ray on day − 1, followed by the intravenous infusion in the caudal vein of 5 × 10⁶ GPBSC and different dose of HLA-DR^−/lowCD33⁺CD16⁻ on day 0. Control groups were transplanted with GPBSC alone. Engraftment of human white blood cell was detected from peripheral blood samples of mice at 7 days, 14 days, and 21 days after transplantation. Engraftment of human eMDSC was detected in the peripheral blood, spleen, and bone marrow of humanized mice at 14 days and 21 days after transplantation. Human Treg (CD4⁺CD25⁺Foxp3⁺), Th1 (CD4⁺T-bet⁺), Th2 (CD4⁺GATA3⁺), Th17 (CD4⁺RORγt⁺), and the cytokine levels were detected from peripheral blood samples of mice at 21 days after transplantation. Levels of cytokines were assessed with pre-designed sets of Luminex immunoassays from eBioscience (for all cytokines measured in supernatants [Human ProcartaPlex Panel #EPX180-12165-901] according to the recommended protocols, using Magpix detection platform (Luminex Corp., Austin, TX) and xPonent software (Luminex). Mice were monitored for survival, weight, and acute GVHD score three times a week. The clinical scoring system was based on six parameters: weight loss, posture, activity, fur texture, skin integrity, and diarrhea. A severity scale of 0 to 1 was used for each parameter, with a maximum score of 6. Mice that had died or had been sacrificed at 1.5 months after HSCT and the tissues were kept in 4% paraformaldehyde. Tissues from GVHD target organs (liver, intestine, and skin) were embedded in paraffin, sectioned, and stained with hematoxylin, eosin, and Safran. Photographs were taken using × 200 or × 400 magnifications with Basler Microscopy ace 2.3 MP Mono camera mounted on an OLYMPUS BX43 microscope.

The conditioning therapy for patients receiving allo-HSCT from matched sibling donors (MSDs) for acute myeloid leukemia (AML), acute lymphoid leukemia (ALL), and myelodysplastic syndromes (MDS) is as follows: intravenous cytarabine (2 g/m²/day) on day − 9, intravenous busulfan (3.2 mg/kg/day) on days − 8 to − 6, intravenous cyclophosphamide (1.8 g/m²/day) on days − 5 to − 4, and oral methyl chloride hexamethylene urea nitrate (Me-CCNU; 250 mg/m²/day) on day − 3. The GVHD prophylaxis regimen consisted post-transplant cyclosporine A, mycophenolate mofetil, and short-term methotrexate [30‐32].

The conditioning therapy for patients receiving allo-HSCT from HLA haplo-identical donors for AML, ALL, and MDS is as follows: intravenous cytarabine (4 g/m²/day) on days − 10 to − 9 and intravenous ATG (2.5 mg/kg/day; Sang Stat, Lyon, France) on days − 5 to − 2, while the application of busulfan, cyclophosphamide, Me-CCNU, and post-transplant GVHD prophylaxis was similar to allo-HSCT from MSDs [30‐32].

The conditioning therapy for patients receiving allo-HSCT from MSDs for severe aplastic anemia (SAA) is as follows: intravenous ATG (2.5 mg/kg/day; Sang Stat, Lyon, France) on days − 5 to − 2 and intravenous cyclophosphamide (50 mg/kg/day) on days − 5 to − 2. Intravenous busulfan (3.2 mg/kg/day) on days − 7 to − 6 was added for haplo-HSCT recipients of SAA. Post-transplant GVHD prophylaxis was similar to allo-HSCT for acute leukemia [33, 34].

The risk stratification for recipients of hematological malignancies is followed by Disease Risk Index (DRI) [35]. The survival functions were estimated by the Kaplan-Meier method using the log-rank test with asymmetric 95% confidence intervals, and the cumulative incidences of aGVHD or cGvHD and TRM/Relapse were calculated using competing risks with 95% confidence intervals. Diagnosis of GVHD was also evaluated with classic and the National Institutes of Health (NIH) consensus criteria [36, 37]. Univariate probabilities of survival were calculated using a left-truncated approach. A bivariable analysis was applied to check with two-way interactions between variables (Spearman correlation for non-normally distributed values; Pearson correlation for normally distributed values). All variables included in the multivariable analysis were tested with time-dependent covariates and constructed in a proportional hazards (Cox) model if PH ≥ 0.05 or a Cox w/Time-Dep Cov model if PH < 0.05, which included patient age, sex, donor-patient sex mismatch, graft components, etc. Variables included in the multivariate analysis were selected by backward elimination process with a criterion of p < 0.10 for retention.

To investigate the types of MDSCs or other myeloid cells expanded in the peripheral blood after G-CSF treatment in donors, the frequency and cell number of different MDSCs were analyzed. Total MDSCs (HLA-DR^−/lowCD11b⁺CD33⁺), M-MDSCs (HLA-DR^−/lowCD11b⁺CD33⁺CD14⁺CD15⁻), G-MDSCs (HLA-DR^−/lowCD11b⁺CD33⁺CD14⁻CD15⁺), CD11b⁺CD14⁻CD15⁺ cells (alternative definition of G-MSCSs), HLA-DR^−/lowCD14⁺ cells (alternative definition of M-MSCSs), and HLA-DR^−/lowCD33⁺CD16⁻ cells in the peripheral blood from ten healthy donors before (homeostasis peripheral blood, H-PB) and after G-CSF mobilization (G-CSF-mobilized peripheral blood, G-PB), as well as G-PBSC, were analyzed. The cell numbers and percentage of all MDSCs were significantly increased in G-PB and G-PBSC. M-MDSC population increased in both frequency and absolute number in G-PBSC. HLA-DR^−/lowCD33⁺CD16⁻ population increased in both frequency and absolute number in G-PB and G-PBSC (Fig. 1a, b, d).

May-Grünwald-Giemsa cytospin results showed that the morphological features of HLA-DR^−/lowCD33⁺CD16⁻ cells were similar to those of immature monocyte-like cells (Fig. 1e). The in vitro immune-suppressive activity of the HLA-DR^−/lowCD33⁺CD16⁻ population identified among the G-PBSC was tested. HLA-DR^−/lowCD33⁺CD16⁻ and autologous CD3⁺ T cells were sorted from the G-PBSC of healthy donors using FACS. HLA-DR^−/lowCD33⁺CD16⁻ cells were co-cultured for 4 days with autologous T cells at different ratios (HLA-DR^−/lowCD33⁺CD16⁻: T = 0:1 to 1:2) in the presence of T cell stimulators (anti-CD3/CD28 beads). Flow cytometry analysis demonstrated that the HLA-DR^−/lowCD33⁺CD16⁻ population had a strong immunosuppressive effect on T cell proliferation in a cell dose-dependent manner (Fig. 1f, g). Moreover, HLA-DR^−/lowCD33⁺CD16⁻ population demonstrated superior immune-suppressive activity compared with CD33⁻ and HLA-DR^−/lowCD33⁺CD16⁺ fraction (Additional file 1: Figure S1). Meanwhile, the immune-suppressive activity of HLA-DR^−/lowCD33⁺CD16⁻ cells was inferior to M-MDSCs and G-MDSCs population (Additional file 1: Figure S2).

To investigate the mechanisms underlying the inhibition of T cell proliferation, the IL-10- and TGF-β-suppressive cytokines expressed by G-CSF-mobilized HLA-DR^−/lowCD33⁺ CD16⁻ cells were detected. The CD33⁻ fraction and non-mobilized HLA-DR^−/lowCD33⁺CD16⁻ cells served as controls. The median percentage of IL-10 or TGF-β positive cells among HLA-DR^−/lowCD33⁺CD16⁻ population in the G-PBSC was significantly higher than those of H-PB (IL-10, 10.1% vs. 0.62%, p = 0.002; TGF-β, 60.24% vs. 10.39%, p = 0.0003). Notably, the TGF-β⁺ cells among the HLA-DR^−/lowCD33⁺CD16⁻ cells significantly increased compared to the CD33⁻ fraction or non-mobilized HLA-DR^−/lowCD33⁺CD16⁻ cells (Fig. 2a, b).

Subsequently, the inhibition of CD3⁺ T cell proliferation by HLA-DR^−/lowCD33⁺CD16⁻ cells was markedly reduced in the presence of an anti-TGF-β antibody but not an anti-IL-10 antibody, indicating that the immunosuppressive activity relied more on the presence of TGF-β (Fig. 2c, d).

Other immunosuppressive mechanisms (arginase, iNOS, IDO, and COX2) were excluded by adding selective inhibitors of arginase (nor-NOHA (N(w)-hydroxy-nor-l-arginine)), iNOS (L-NMMA), IDO (NLG8189) to the co-cultures. However, no reverse effect on T cell proliferation was observed (Additional file 1: Figure S3).

The present study demonstrated that CD4⁺CD25⁺Foxp3⁺Treg cells increased in the HLA-DR^−/lowCD33⁺CD16⁻ cell co-culture group (Fig. 3a, c). Th cell subsets were analyzed after co-culture with or without HLA-DR^−/lowCD33⁺CD16⁻ cells. Though the expression of IFN-γ and IL-17A increased, co-culture with HLA-DR^−/lowCD33⁺CD16⁻ cells led to a significantly higher ratio of Th2/Th1 and Th2/Th1+Th17 cells than the control group (Fig. 3b, d, e).

Because G-CSF expanded HLA-DR^−/lowCD33⁺CD16⁻ cells into the peripheral blood and exerted a strong regulatory effect on T cell immune response, we next investigated if such population prevents aGVHD in vivo. To mimic aGVHD after allo-HSCT in humans, a humanized xenogeneic aGVHD model was established by injection of 5 × 10⁶ donor G-PBSC into irradiated non-obese diabetic (NOD)-severe combined immune deficient (SCID)-IL-2Rg−/− mice (NSG mice) [38, 39] (Fig. 4a). Human white blood cell engraftment was detected by CD45⁺ cells at 7, 14, and 21 days after transplantation, and there was no significant difference of human cells engraftment intragroup (14 days, p = 0.486; 21 days, p = 0.4012) (Additional file 1: Figure S4). We failed to detect the engraftment of HLA-DR^−/lowCD33⁺CD16⁻ MDSCs in PB as these cells could be detected in the spleen and bone marrow at 14 days after transplantation; however, they decreased quickly and almost could not be detected at 21 days after co-transplantation (Additional file 1: Figure S5). Acute GVHD occurred in the humanized mice receiving donor G-PBSC during 2 to 4 weeks after transplantation as evidenced by weight loss, disease score (hunching, activity, ruffling, and diarrhea), and death. Similar to humans [40, 41], humanized mice with aGVHD showed severe inflammation, leukocyte infiltration, necrosis, and tissue damage in target organs, such as the skin, liver, and gut in the G-PBSC group (Fig. 4e). On the base of this model, a prevention model was established by one-time co-infusion of 2.5 × 10⁶, 1 × 10⁶, and 5 × 10⁵ human G-CSF-mobilized HLA-DR^−/lowCD33⁺CD16⁻cells sorted from the same donor with graft at the same day. All of these cell doses reduced weight loss (Fig. 4b), improved overall survival (Fig. 4c), improved pathological score (Fig. 4d), and lessened tissue damage in target organs in medium and high doses of HLA-DR^−/lowCD33⁺ CD16⁻cells compared to mice grafted without or with low-dose human HLA-DR^−/lowCD33⁺ CD16⁻ cells (Fig. 4e). For human Treg detected at 21 days after transplantation, a significant difference was found between the high-dose HLA-DR^−/lowCD33⁺CD16⁻ cells and GPBSC control (p = 0.0498); there was no disparity between medium dose vs. control (p = 0.6825) and low dose vs. control (p > 0.9) (Fig. 4a, Additional file 1: Figure S6); there were no disparity of Th2/(Th1+Th17) ratio between the different groups (Fig. 4b, Additional file 1: Figure S6). For cytokines detected at 7, 14, and 21 days after transplantation, there were significantly less INF-γ and IL-6 at days 7, 14, and 21 and IL-17A and TNF-α at days 14 and 21 (high dose of HLA-DR^−/lowCD33⁺CD16⁻ cells vs. control) and less INF-γ and TNF-α at days 7 and 21, IL-17A at days 7 and 14, and IL-6 at day 21 (medium dose of HLA-DR^−/lowCD33⁺CD16⁻ cells vs. control) (Additional file 1: Figure S7). IL-4/INF-γ and IL-4/IL-17A ratios were higher in the high-dose HLA-DR^−/lowCD33⁺CD16⁻ cells group vs. control at 7, 14, and 21 days after transplantation; by comparison with medium-dose HLA-DR^−/lowCD33⁺ CD16⁻ cells group vs. control, IL-4/INF-γ ratios were higher at days 7 and 14, while IL-4/IL-17A ratio was higher at day 21 (Fig. 5c, d). These results confirmed that immune-suppressive activity of HLA-DR^−/lowCD33⁺CD16⁻ cells which might contribute to alleviating aGVHD in NSG model.

In a prospective study (registered at www.​chictr.​org as ChiCTR-OCH-10000940), 100 allo-HSCT recipients were enrolled to assess the role of HLA-DR^−/lowCD33⁺CD16⁻ in graft on the occurrence of aGVHD. Of these patients, 76 patients received a combination of G-BM and G-PBSC grafts from haplo-donors, and 24 patients received grafts from MSDs. The indications for allo-HSCT were as follows: AML (n = 39), ALL (n = 29), MDS (n = 8), SAA (n = 14), and lymphoma (n = 10). Ninety-nine (99%) patients reached full donor chimerism before day 30 after transplantation as assessed by the analysis of a variable number of tandem repeats on the peripheral blood. Platelet recovery (count > 20,000/ml) was achieved at a median of day 32 post-transplantation, whereas neutrophils reached values greater than 500/ml on day 12. Only 1 patient died due to transplant-related mortality at day 11 post-HSCT. The median follow-up period of the enrolled patients was 11 months. The cell components in the graft were listed in Additional file 1: Table S1. Patients were divided into 2 groups according to the median absolute counts of HLA-DR^−/lowCD33⁺CD16⁻ MDSCs in grafts (1.88 × 10⁷/kg) as follows: 53 patients were in the high MDSCs group (≥ 1.88 × 10⁷/kg), and 47 patients were in the low MDSCs group (Table 2).

The cumulative incidences for different grades of aGVHD at 100 days after transplantation for the total cohort were as follows: 50% of patients developed grade I–IV aGVHD; 28% of patients developed grade I aGVHD (61.8% for haplo-HSCT and 12.5% for MSD-HSCT); 17% of patients had grade II aGVHD (25% for haplo-HSCT and 12.5% for MSD-HSCT); and 5% of patients developed grade III–IV aGVHD (5.3% for haplo-HSCT and 4.2% for MSD-HSCT). Patients who received a high number of MDSCs exhibited lower incidence of grade II–IV aGVHD compared to the low MDSC groups in allo-HSCT (11.3% vs. 31.9%, p = 0.0287) and comparable of grade III–IV aGVHD in allo-HSCT (1.9% vs. 8.5%, p = 0.127) (Fig. 6a, b). In the bivariable analysis, high MDSC dose and CD34⁺ cells in the graft were interacted; for consideration of collinearity in multiple variable analysis (MVA), backward elimination process was applied to choose one factor (high MDSC dose) which was taken into the final MVA model. In the multivariate analysis, absolute counts of MDSCs in allografts emerged as the only independent factor that reduced the occurrence of grades II–IV (HR 0.388, 95% CI 0.158–0.954, p = 0.039). Age, patient gender, HLA disparity, ABO disparity, patient-donor relationship, and CD3/CD4/CD8/CD14/CD34 cells in grafts were not correlated to grades II–IV in the analysis.

For patients with leukemia-free survival (LFS) above 100 days post-transplantation, the estimated 1-year cumulative incidences for chronic GVHD (cGVHD) of the total cohort were as follows: 25.5% patients developed cGVHD, 8.2% patients with extensive cGVHD, and 15.3% patients with moderate to severe cGvHD with NIH criteria. There were trends that patients who received a high number of MDSCs exhibited a lower incidence of cGVHD (17.6% vs. 36.9%, p = 0.067) and NIH moderate to severe cGvHD (10.3% vs. 24.5%, p = 0.045) (Fig. 6c, d). In the univariate analysis, high- vs. low-dose MDSCs are identified as a potential factor that reduced the occurrence of cGVHD (HR 0.481, 95% CI 0.213–1.089, p = 0.072). In the multivariate analysis, no independent risk factor was found.

The estimated 1-year LFS (82.2% vs. 87.8%, p = 0.508) and 1-year cumulative incidence of relapse (CIR, 11.1% vs. 9.8%, p = 0.871) were comparable in patients with hematological malignancies (n = 86) who received high or low dose of MDSCs (Fig. 6e, g). The estimated 1-year overall survival (OS, 86.0% vs. 93.4%, p = 0.254) and 1-year treatment-related mortality (TRM, 7.6% vs. 4.3%, p = 0.488) were comparable in all allo-HSCT recipients (n = 100) with high or low dose of MDSCs (Fig. 6f, h). In the multivariate analysis, DRI is the only potential independent risk factor identified for CIR (HR 2.788, HR 0.905–5.892, p = 0.044); no independent factor was found for LFS, OS, and TRM.