Mucosal-associated invariant T cells predict increased acute graft-versus-host-disease incidence in patients receiving allogeneic hematopoietic stem cell transplantation

The study prospectively recruited 86 patients who underwent allogeneic HCT in the Blood Diseases Hospital, Chinese Academy of Medical Sciences from May 1, 2018 to June 30, 2019. Patients eligible for this analysis included all consecutive adult and pediatric recipients who achieved donor-derived neutrophil engraftment after allogeneic HCT. One patient who experienced engraftment failure was excluded. Peripheral blood samples were collected from recipients at day 15, 30 (± 5), 60 (± 5), and 90 (± 5) post allogeneic HCT and peripheral blood samples from healthy subjects as the normal control. For cytokine analysis and RNA-sequencing (RNA-seq), peripheral blood samples were collected when patients were diagnosed with aGVHD, and samples were also collected from patients who did not develop aGVHD at the same time point and used as controls. Graft samples transfused into these recipients were collected from the abandoned testing samples. All blood samples were then examined using flow cytometry and RNA-seq using standard protocols. This study was approved by the Ethics Committee of Blood Disease Hospital, Chinese Academy of Medical Sciences (KT2018104-EC-2). Written informed consent was acquired from all patients and healthy subjects.

All the peripheral blood and graft samples were purified by density gradient centrifuge using Ficoll–Paque to achieve mononuclear cells. The MAIT cells and subgroups were examined using the following antibodies: anti-CD3-APC-Cy7, anti-TCRγδ-FITC, anti-CD161-Percp-Cy5.5, anti-TCRVα7.2-APC, anti-CD4-PE, anti-CD8-PE-Cy7, anti-CXCR6-PE, anti-PD1-PE, and anti-CD38-PE-Cy7. Mononuclear cells (1 × 10⁶) were incubated in complete media for intracellular cytokine staining. All samples were acquired on an LSRFortessa flow cytometer. Data analysis was performed using the FlowJo Version 7.6 software.

For peripheral blood MAIT cell sorting, samples obtained from patients and healthy controls were purified by density gradient centrifuge. Mononuclear cells were stained with surface antibodies (anti-CD3-APC-Cy7, anti-TCRγδ-FITC, anti-CD161-Percp-Cy5.5, anti-TCRVα7.2-APC). Then, the cells were sorted immediately using an FACSAria III cell sorter (BD Bioscience, Franklin Lake, NJ, USA), selecting CD3⁺ TCRγδCD161⁺TCR-Va7.2⁺ cells.

Phorbol myristate acetate (PMA), ionomycin (IM) and 1 μL monensin were added to the media at the beginning of incubation. After incubation for 4 h, cells were aspirated and stained with the surface markers, followed by fixation in Cytofix/Cytoperm and permeabilized with Perm/Wash solution according to the manufacturer’s introductions. The cells were then stained with cytokine antibodies.

RNA-seq was performed on MAIT cells from peripheral blood samples as previously described [19]. Each patient collected 1 × 10⁴ MAIT cells for sequencing, and totally, ten samples were sequenced, including four samples from patients with aGVHD, three from patients without aGVHD, and three from healthy controls. RNA concentration was measured using Qubit RNA Assay Kit (Life Technologies, CA, USA), and RNA integrity was assessed using Bioanalyzer 2100 system (Agilent Technologies, CA, USA) with RIN > 6.5.The libraries were generated using the NEBNext UltraTM RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA) following the manufacturer’s recommendations. The clustering of the samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumina), then the library preparations were sequenced on an Illumina Hiseq platform with a target depth of 3 million reads. The data are available at Gene Expression Omnibus (GEO) with the accession number of GSE157959.

aGVHD was diagnosed clinically according to the International Bone Marrow Transplant Registry [20]. Bacterial infections were defined as: positive clinical symptoms of bacterial infection and positive culture results, including blood, sputum, sterile tissue, bronchoalveolar lavage fluid and urinary fluid. Fungal infections were defined as: diagnosis of proven and probable fungal infection following guidelines of Infectious Diseases Society of America [21, 22]. Viral infections were defined as: cytomegalovirus or Epstein–Barr virus > 400 copies/mL in plasma using quantitative polymerase chain reaction. The cumulative incidence of aGVHD was estimated and compared using Gray’s test. The multivariate Cox regression model was used to investigate the association between MAIT cells and the hazard of aGVHD. All factors found to influence aGVHD in univariate analysis with P < 0.20 were included in the Cox proportional hazard model. Correlations were analyzed using the Spearman’s rank correlation test. A significant difference for continuous variables was assessed using Mann–Whitney analysis. P ≤ 0.05 was considered significant. All Statistical analysis was performed using SPSS 12, except that for Gray’s test, which was performed using R 3.6.1.

The characteristics of all 86 patients in the study are shown in Table 1. The median age of recipients at the time of allogeneic HCT was 37 (range 7–63) years. 41 patients (47.6%) received allogeneic HCT for acute myeloid leukemia, 19 (22.1%) for acute lymphoid leukemia, 25 (29.1%) for myelodysplastic syndrome, and one (1.2%) for primary myelofibrosis. All patients received myeloablative conditioning, either a regimen based on busulfan and cyclophosphamide (n = 69) or a regimen based on total body irradiation and cyclophosphamide (n = 17). 45 patients received HLA-matched sibling donor grafts, 37 received haploidentical donor grafts, and only four received HLA-matched unrelated donor grafts. The GVHD prophylaxis regimens included cyclosporine A/methotrexate (n = 26), cyclosporine A/methotrexate/mycophenolate mofetil (n = 33), and tacrolimus/mycophenolate mofetil (n = 27). Anti-thymocyte globulin (ATG) was used as an alternative approach to the post-transplant cyclophosphamide (PT-Cy) for in vivo T cell depletion in haploidentical donor transplantation, HLA matched unrelated donor transplantation and sibling donor transplantation with the recipients’ age older than 45 years. Totally, 64 patients of the enrolled patients used ATG. Of all the donors, 35 cases were female donor, including 21 cases of female to male and 14 cases of female to female. CMV-seropositive donor to CMV-seropositive recipients were the most common donor–recipient CMV serostatus (n = 81). The median number of infused total mononuclear cells and CD34⁺ cells were 9.7 × 10⁸/kg (range 5.2–18.5) and 3.1 × 10⁶/kg (range 1.6–6.7), respectively. None of the patients underwent ex vivo T cell depletion. In our study, 44 cases developed aGVHD in 100 days with a cumulative incidence of 50.6% (95% confidence interval [CI] 39.9–61.2%), including 15 cases of grade A aGVHD, 17 cases of grade B aGVHD, six cases each of grade C aGVHD and grade D aGVHD; Of these patients who developed aGVHD, 21 cases had skin involvement, 14 had liver involvement, and 17 had intestinal involvement; 13 patients had ≥ 2 organ involvement. The median onset time of aGVHD was 31 (range 11–58) days post graft transfusion.

Among the enrolled 86 patients, 69 peripheral blood stem cell graft (PB-graft) samples mobilized by granulocyte-colony-stimulating factors were collected. MAIT cells were identified using flow cytometry and defined as CD3⁺TCRγδVα7.2⁺CD161⁺ cells as previously described [23] (Fig. 1A). According to our result, no significant differences had been found in MAIT cell proportions between PB-grafts (n = 69) and blood samples from healthy donors (n = 48) (P = 0.87, Fig. 1B). According to the expression levels of CD4/8 on the surface of MAIT cells, we then subdivided MAIT cells into four subsets: CD4 positive, CD8 positive, CD4/CD8 double negative (DN), and CD4/CD8 double positive (DP) subsets. Results showed that the percentage of CD8 positive MAIT cells in PB-grafts (n = 66, median percentage: 75.1%) was significantly lower than that in healthy donors (n = 40, median percentage: 81.1%, P = 0.006). Otherwise, the frequency of DN-MAIT cells (median percentage: 12.4%) and DP-MAIT cells (median percentage: 4.7%) from PB-grafts were both significantly higher than those from healthy controls (DN: median percentage 9.6%, P = 0.046; DP: median percentage 2.1%, P = 0.001; Fig. 1C), which indicated that the transfused MAIT cell subpopulations in grafts had transformed after mobilization. In addition, the number of MAIT cells in the graft was significantly higher than that in the healthy control (134.8 × 10⁷/L vs. 4.1 × 10⁷/L, P < 0.001, Fig. 1D); the number of each CD4/8 subset was also significantly higher than that in the control group (CD4 + : 91.3 × 10⁶/L vs. 1.6 × 10⁶/L, P < 0.001; CD8 + : 963.2 × 10⁶/L vs. 36.5 × 10⁶/L, P < 0.001; DN: 178.0 × 10⁶/L vs. 4.8 × 10⁶/L, P < 0.001; DP: 46.1 × 10⁶/L vs. 0.73 × 10⁶/L, P < 0.001; Fig. 1E).

We also analyzed the clinical factors that may affect the reconstitution of MAIT cells post allogeneic HCT, and the results showed that the application of ATG, type of donor, and viral infection (Cytomegalovirus and Epstein–Barr virus) significantly affected the early reconstitution of MAIT cells. Patients receiving matched-sibling donor grafts had a better MAIT reconstitution compared with those receiving alternative donor grafts (Fig. 2A). In addition, application of ATG in conditioning regimens could significantly reduce the proportion and number of MAIT cells in the recipients’ peripheral blood, and this effect could still be observed at day 90, post graft transfusion (Fig. 2B). In order to avoid the effect of donor type on MAIT cell reconstitution, we further investigated the ATG usage on the proportion and number of MAIT cells in recipients receiving HLA-matched sibling donor transplantation, and the effect of ATG on MAIT cells can still be observed (Fig. 2C). For the factor of infection, we found that viral infections were significantly related to the decrease in the proportion and number of MAIT cells post allogeneic HCT, while bacterial infections and fungal infections had no obvious correlation with MAIT cell reconstitution (Fig. 2D).

As it has been reported that MAIT cells in recipients are partly derived from the transfer of MAIT cells in the grafts and are related to the GVHD occurrence post allogeneic HCT [17], we further investigated whether MAIT cell frequency and the absolute number in the grafts could predict the incidence of aGVHD post allogeneic HCT.

The cumulative incidence of grade B-D aGVHD for the entire cohort was 40.8% (95% CI 29.5–53.9) at day 100 post allogeneic HCT (Fig. 3A). The median MAIT cell percentage in PB-grafts was 3.03% (interquartile range 1.63–6.18%). We used this value as a threshold to divide patients into two subgroups, MAIT cell percentage ≥ 3.03% group (interquartile range 3.49–12.6%) and MAIT cell percentage < 3.03% group (interquartile range 0.72–2.23%), to evaluate the impact of MAIT cell frequency on the incidence of grades B‒D aGVHD. Results showed that the cumulative incidence of B‒D aGVHD at day 100 significantly increased in patients receiving grafts with MAIT cell percentage ≥ 3.03% than that with percentage < 3.03% (56.3%, 95% CI 37.1–71.2 versus 23.1%, 95% CI 13.8–46.2; P = 0.038; Fig. 3A). We additionally found that a higher proportion of CD4-CD8 + MAIT cells in the grafts were significantly associated with an increased incidence of B‒D aGVHD post allogeneic HCT (P = 0.035, Fig. 3B, Table 2). Besides, we further used the median absolute number of MAIT cells (3.17 × 10⁶/kg, interquartile range 1.56–7.48 × 10⁶/kg) as a threshold to evaluate the relationship between MAIT cell number in grafts and incidence of B‒D aGVHD. The cumulative incidence aGVHD in patients with MAIT cell number ≥ 3.17 × 10⁶/kg (interquartile range 4.36–14.83 × 10⁶/kg) was 51.1% (95% CI 35.8–68.9) versus 26.6% (95% CI 14.6–47.2) in patients with MAIT cell number < 3.17 × 10⁶/kg (interquartile range: 0.77–2.80 × 10⁶/kg); but no statistical difference was drawn (Fig. 3A, Table 2; P = 0.173), According to the analysis of CD4/8 MAIT cell subsets, it showed that more CD4CD8⁺, CD4⁺CD8, CD4⁺CD8⁺MAIT cells in the grafts indicated an increased incidence of B‒D aGVHD, yet, the CD4CD8MAIT cell number had no impact on aGVHD development (Fig. 3BII–IV, Table 2). We also tried several other cell subsets in PB-grafts, including CD4⁺ T cells, CD8⁺ T cells, CD4⁺CD25⁺T cells, Natural killer cells, and B cells. Using a strategy similar to MAIT cells, the median proportion of each cell population was used as a threshold to divide them into groups with higher proportions and those with lower proportions. No significant impacts were found in these cell subsets on B‒D aGVHD incidence (data are shown in Additional file 1).

We further performed univariate and multivariate analyses to better explore the risk factors affecting the occurrence of GVHD post allogeneic HCT. Therefore, 69 patients, who obtained the MAIT cell frequencies both in the transfused graft samples and PB samples post allogeneic HCT were enrolled. In addition to the above-mentioned MAIT cell proportion and number in grafts that had a significant impact on the incidence of B‒D aGVHD (Fig. 3), univariate analysis also identified that donor type significantly associated with aGVHD development. The incidence of grade B-D aGVHD in patients receiving HLA-matched sibling donor grafts (28.6%, 95% CI 10.9–39.8) was significantly lower than that in patients receiving alternative donor grafts (53.3%, 95% CI 31.5–73.2; P = 0.031, Table 2). Factors influencing aGVHD incidence with P < 0.20 in the univariate analysis were included in a multivariate analysis. Since ATG has clearly been reported to have an impact on the occurrence of GVHD, this study also included ATG usage in the multivariate analysis; however, the P value of the ATG usage was above 0.2. In total, the following factors were included in the multivariate analysis: donor type, MAIT cell percentage, MAIT cell number, GVHD prophylaxis, and ATG usage. We found that MAIT cell frequency in PB-grafts was still an independent factor for aGVHD occurrence. The aGVHD incidence significantly decreased in patients receiving grafts with MAIT cell percentage < 3.03% when compared with MAIT cell percentage ≥ 3.03% (hazard ratio [HR] = 0.36; 95% CI 0.15–0.88; P = 0.025; Table 3). Alternative donor graft was another independent factor associated with increased aGVHD incidence (hazard ratio [HR] = 2.25; 95% CI 1.01–5.05; P = 0.048; Table 3). No impact was observed for ATG usage, MAIT cell number, or GVHD prophylaxis on the occurrence of aGVHD (ATG usage: P = 0.367; MAIT cell number: P = 0.604; GVHD prophylaxis: P = 0.839).

We further investigated the reconstitution of MAIT cells after allogeneic HCT. The frequencies of circulating MAIT cells at days 15 and 30 post allogeneic HCT were similar, whereas it significantly decreased at day 60 compared to that in patients at day 15, without prominent recovery till day 90 (day 60: 1.78% vs. 3.81%, P < 0.0001; day 90: 1.26% vs. 3.81%; P < 0.0001, Fig. 4A). However, the absolute number of MAIT cells showed different trend lines. The absolute number of circulating MAIT cells were at the lowest point at day 15 post allogeneic HCT, followed with a significant recovery at day 30 (7.23 × 10⁶/L vs. 12.98 × 10⁶/L, P < 0.0001; Fig. 4B).

Our results further identified that both frequency and absolute number of MAIT cells at day 15 in patients with B‒D aGVHD were significantly lower than that in patients without aGVHD (3.32% vs. 5.71%, P = 0.327, and 2.18 × 10⁶/L vs. 8.27 × 10⁶/L, P = 0.048; Fig. 4C, D). Both frequency and number of MAIT cells at day 30 in patients with B‒D aGVHD (average percentage: 1.91%, average cell number: 6.31 × 10⁶/L) were still lower than those in patients without aGVHD (average percentage: 4.21%, P = 0.046; average cell number: 14.92 × 10⁶/L, P = 0.024, Fig. 4C, D).

After analyzing the frequencies and numbers of MAIT cells post allogeneic HCT, we further studied the functional role of MAIT cells in aGVHD development. The MAIT cell activation, migration, and immune response capacity was evaluated through the examination of CD38, CXC chemokine receptor-6 (CXCR6) and programmed death 1 (PD-1) expression levels on MAIT cells [9, 24, 25]. Peripheral blood samples were obtained from 26 PB-grafts, 23 patients with aGVHD, and 22 patients without aGVHD (Fig. 5A). Based on the reported study, we used CD38 as a marker of MAIT cell activation [9]. Our data revealed that the proportion of CD38⁺ MAIT cells were significantly increased in recipient peripheral blood post allogeneic HCT when compared to that in grafts (with aGVHD: 72.3% vs. 20.6%, P < 0.001; without aGVHD: 76.2% vs. 20.6%, P < 0.001; respectively, Fig. 5B); however, no difference was observed between patients with and without aGVHD (P = 0.641). CXCR6 is involved in the trafficking of T cells to peripheral organs such as the liver or intestines [26], and MAIT cells show abundant CXCR6 expression levels [3, 24]. In our study, we found that the circulating MAIT cells exhibited higher levels of CXCR6 in patients with aGVHD than in patients without aGVHD (68.3% vs. 54.9%, P = 0.026, Fig. 5B), supporting their migration to the inflammation site and circulating number decrease in aGVHD. In addition, the PD-1 expression was significantly higher in patients with aGVHD (median percentage: 28.0%) than in patients without aGVHD (28.0% vs. 15.2%, P = 0.011) and PB-grafts (28.0% vs. 12.4%, P < 0.001; Fig. 5B).

We also detected the PMA/ionomycin stimulated cytokine production in MAIT cells from 14 patients with aGVHD, 16 patients without aGVHD, and 16 PB-grafts (Fig. 5A). The IFN-γ⁺ and TNF-α⁺ MAIT cell levels were comparable between the patients without aGVHD and PB-grafts (Fig. 5C); however, we observed a markedly elevated production of IFN-γ and TNF-α in patients who developed aGVHD when compared with patients who did not develop aGVHD (IFN-γ: 67.4% vs. 25.2%, P < 0.001; TNF-α: 65.0% vs. 38.1%, P = 0.005), as well as the PB-grafts (IFN-γ: 67.4% vs. 34.7%, P = 0.006; TNF-α: 65.0% vs. 40.1%, P = 0.024; Fig. 5C).

We further used Spearman’s rank test to investigate the correlation between circulating MAIT cell frequencies and their expression levels of IFN-γ, TNF-α, PD-1, CXCR6, and CD38. The frequencies of MAIT cells showed a moderate inverse correlation with PD-1, IFN-γ, and TNF-α expression levels (PD-1: r = ˗0.324, P = 0.029; IFN-γ: r = ˗0.512, P = 0.004; TNF-α: r = 0.419, P = 0.021; Fig. 5D), however, it did not correlate with CXCR6 and CD38 levels. The results demonstrated that MAIT cells from patients with aGVHD had increased inflammatory responses compared to patients without aGVHD and grafts, especially through the IFN-γ and TNF-α activation pathway.

To identify cell-intrinsic regulatory mechanisms of MAIT cells in aGVHD patients, we performed RNA-seq to identify the transcriptional profile of the circulating MAIT cells from patients with aGVHD (n = 4), patients without aGVHD (n = 3), and healthy controls (n = 3). For the four aGVHD patients, peripheral blood samples were collected when patients were diagnosed with grades C‒D aGVHD, including three cases of grade C aGVHD and one case of grade D aGVHD. Samples from patients without aGVHD were collected at a similar time point according to the collection time from aGVHD patients. There were 754 differentially expressed genes in MAIT cells between patients with and without aGVHD, and 698 genes between aGVHD patients and healthy controls at minimum log (2) fold change of ± 1 (Fig. 6A, Additional file 2). GSEA revealed significant enrichment of interferon-alpha response pathway in patients with aGVHD (NES = 1.24, FDR q-value = 0.241) and patients without aGVHD (NES = 1.83, FDR q-value = 0.004) when compared with healthy controls (Figs. 6B i–ii and 5C i–ii). Some of the shared hub genes with core enrichment in this pathway were ISG15, CD74 and IRF2 (more details are shown in Additional file 3). Besides, we also observed an enrichment of Moserle IFN-α response gene set between patients with and without aGVHD at a possible trend toward significance (NES = 1.89, FDR q-value = 0.051; Fig. 6B iii and C iii). Some of the hub genes in this gene set were IFITM1, IFIH1, and IFIT family members such as IFIT5, IFIT3, and IFIT1. Additionally, the primary immunodeficiency gene set was enriched in patients without aGVHD when compared with patients who developed aGVHD (NES = 2.05, FDR q-value = 0.015) and healthy controls (NES = 1.59, FDR q-value = 0.098; Fig. 6D). The shared hub genes were JAK3, RAG1, and TAP1.