Introduction

Acute myeloid leukemia (AML) is a heterogeneous disease characterized by uncontrolled proliferation of immature myeloid leukemia cells in the bone marrow with consequent failure of normal hematopoiesis [1]. Although AML cases can be stratified into favorable, intermediate and adverse risk groups based on their genetic profile [2], prognosis within these categories varies widely. Therefore, more prognostic markers are urgently needed to improve patient risk stratification.

The tumor microenvironment (TME) is a highly complex ecosystem containing cancer cells and diverse non-cancerous cells, such as immune cells, cancer-associated fibroblasts, endothelial cells and other cell types. Bone marrow, which resembles the TME in solid tumors to some degree, is the nest of the leukemia progenitors generated and output in AML and has a critical effect on AML occurrence, recurrence and progression [3]. Among the immune cells in the bone marrow, T cells are the prerequisite for immune recognition and elimination of AML blast cells [4]. However, these T cells are often in exhaustion status, that is, they fail to effectively eliminate leukemia cells because of a broad spectrum of immunosuppressive mechanisms such as the checkpoint inhibitory molecules on T cells surface [5]. These immune checkpoints can block immune responses targeting malignant self-cells and lead to AML blast cells’ immune evasion. T cells expressed checkpoints have been illustrated to have prognostic value in a variety of oncological settings including AML. As an example, the expression of cytotoxic T lymphocyte antigen-4 (CTLA-4), the first recognized immune checkpoint, has been demonstrated to be independently associated with AML patients’ prognosis [6].

T cell immunoreceptor with immunoglobulin and ITIM domain (TIGIT), a recently-identified checkpoint inhibitory receptor, is one of the most promising new targets for immunotherapy. TIGIT is expressed on lymphocytes, including cytotoxic T cells, regulatory T cells and natural killer (NK) cells [7]. It was found that TIGIT expression in CD8 + T cells in peripheral blood was elevated in AML patients at diagnosis compared with healthy donors (HDs) and high-TIGIT expression correlated with primary refractory disease [8]. As for the bone marrow specimen, only one study just analyzed its expression pattern as well as its co-expression with PD19 and research on its prognostic significance remained absent to date.

In the present study, we comprehensively compared the single and dual expression of TIGIT and PD1 in bone marrow T cells from newly-diagnosed AML patients and HDs using multiparameter flow cytometry (MFC) and assessed their impacts on patients’ outcomes. We also checked the effect of TIGIT and PD1 expression on the cytokine production of T cells. Moreover, we performed single cell RNA-sequencing (scRNA-seq) in order to understand TIGIT and PD1 expression related pathogenic features.

Materials and methods

Patients and healthy donors

In total, the present study included 71 adult patients with non-acute promyelocytic leukemia (APL) AML who were diagnosed from September 2020 to April 2022 in our institute. The patients were diagnosed according to bone marrow morphology, immunophenotyping, karyotyping and molecular biology. The median age of 71 patients (36 males and 35 females) at diagnosis was 42 (range: 17–62) years. RUNX1::RUNX1T1, CBFB::MYH11, BCR::ABL1, MLL rearrangements and others AML-related fusion transcripts and FLT3-ITD as well as NPM1 mutation were examined on all patients [10,11,12,13]. Moreover, 41 patients’ samples received targeted next-generation sequencing (NGS). TP53 and CEBPA mutation were tested by Sanger sequencing in 43 and 59 patients, respectively. The definition of the European LeukemiaNet (ELN) risk classification by genetics was based on the 2022 ELN guidelines [14]. The patients’ clinical information is summarized in Table 1. It should be noted that the majority of patients enrolled from September 2020 to December 2021 were those with RUNX1::RUNX1T1 fusion gene and all non-APL AML patients were enrolled from December 2021 to April 2022. The frequency of RUNX1::RUNX1T1 fusion were 64.9% (24/37) and 20.6% (7/34), respectively. In addition, 31 HDs were also included in this study. The median age of HDs at enrollment was 47 (range: 31–65) years and 19 (61.3%) were male.

Table 1 The basic clinical information of all patients enrolled at diagnosis (n = 71)
Full size table

As we previously reported, these patients received various regimens as induction including AA (aclarubicin and cytarabine), IA (idarubicin and cytarabine), HAA (homoharringtonine, aclarubicin, and cytarabine), CAG (cytarabine, aclarubicin, and G-CSF) and other targeted therapy (dasatinib, sorafenib or venetoclax) [15, 16]. After achieving complete remission (CR), patients received either cytarabine chemotherapy alone or chemotherapy followed by allogeneic hematopoietic stem cell transplantation (allo-HSCT) as consolidation. Our previous studies have described the implications of allo-HSCT [17]. The last follow-up date was January 2024.

This study was approved by the Ethics Committee of the Peking University People’s Hospital and was in accordance with the Declaration of Helsinki.

Flow cytometry analysis of TIGIT and PD1 expression

MFC was performed on fresh bone marrow specimens collected from 71 AML patients at diagnosis and 31 HDs. Briefly, the samples were incubated with fluorochrome-conjugated antibodies for 15 min at room temperature in the dark. Red blood cells were then lysed using FACS™ lysing solution (BD Biosciences, San Jose, CA, USA). Data were acquired with the FACSCanto™ ‖ (BD Biosciences, San Jose, CA, USA) and analyzed by Kaluza 2.0 (Beckman Coulter, Brea, CA, USA). The monoclonal antibodies used were anti-human CD45-V500 (BD Biosciences, Clone HI30, USA), CD3-APC-H7 (BD Biosciences, Clone SK7, USA), CD4-BV605 (Biolegend, Clone OKT4, USA), CD8-PE-Cy7 (BD Biosciences, Clone SK1, USA), PD1-PE (Biolegend, Clone EH12.2H7, USA) and TIGIT-PerCP-Cy5.5 (Biolegend, Clone A15153G, USA).

The gating strategy is shown in Figure S1. After excluding cellular doublets in the FSC-H/SSC-A plot and removing cellular debris and nonviable cells in the FSC-A/SSC-A plot, cells shown in the CD45/SSC-A plot were defined as nucleated cells (NCs). Lymphocytes were characterized with CD45-high/SSC-A-low in NCs (Figure S1a). T cells were the cluster of CD3-positive cells in the lymphocyte (Figure S1b). CD4 + T cells and CD8 + T cells were CD4-positive CD8-negative and CD4-negative CD8-positive cell subsets, respectively (Figure S1c). Expression of TIGIT and PD1 and co-expression of TIGIT and PD1 were assessed in CD4 + T cells and CD8 + T cells, respectively (Figure S1d–e). All 71 patients were tested TIGIT expression and 64 were simultaneously tested PD1 expression in CD4 + T cells and CD8 + T cells.

T cell activation and function assays

Fresh bone marrow samples collected from 15 AML patients at diagnosis and 6 HDs were obtained bone marrow mononuclear cells (BMMNCs). For T cell activation, BMMNCs were cultured in RPMI Medium 1640 (Gibco, Invitrogen, Paisley, UK) supplemented with 10% fetal bovine serum (FBS) (Gibco, Invitrogen, Paisley, UK), and 1000 IU/ml IL-2 (Shuanglu, Beijing, China) at 37 ℃ for 24 h, then were stimulated with anti-CD3 (Biolegend, Clone HIT3a) in combination with anti-CD28 (Biolegend, Clone CD28.2) for 6 h.

Thereafter these cells were collected for cell surface antigen staining for CD45, CD3, CD4 and CD8 and fixed and permeabilized using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Invitrogen, Waltham, MA, USA) and performed intracellular staining for tumor necrosis factor-α (TNF-α) and interferon-γ (INF-γ), respectively. The stained cells were tested on FACSCantoTM ‖. The monoclonal antibodies used were TNF-α-PerCP-Cy5.5 (Biolegend, Clone MAb11, USA) and INF-γ-BV421 (Biolegend, Clone B27, USA).

Expression of TNF-α- and INF-γ was examined on TIGIT + cells, TIGIT- cells, PD1 + cells, PD1- cells and TIGIT + PD1 + cells in CD4 + T and CD8 + T cells, respectively (Figure S1f-g).

Single cell RNA-sequencing and analysis

Isolated BMMNCs from three newly-diagnosed adult AML patients were used for scRNA-seq. scRNA-seq libraries were prepared using the Chromium Single Cell 5’ Library and Gel Bead Kit from 10 × Genomics, according to the manufacture’s instruction. The output gene expression matrices were analyzed by R software using the Seurat package [18]. Low-quality cells were filtered if they met any of the following criteria: (1) < 500 unique molecular identifiers (UMI); (2) < 250 genes or > 5000 genes; (3) cell complexity < 80%; or (4) > 10% UMIs derived from the mitochondrial genome. After the removal of the low-quality cells and genes, a total of 20,831 cells and 24,892 genes were kept for further analyses. The UMI counts was normalized using Seurat functions “NormalizedData”, “FindVariableFeatures” and “ScaleData”. The RunPCA function was used with default parameters to reduce dimensionality. Batch effect across samples was eliminated by the R package Harmony with the default lambda value [19]. The data after correction of batch effect were used for clustering and Uniform Manifold Approximation and Projection for dimension reduction (UMAP) based on the first 40 principal components (PCs). The main cell types were annotated according to the expression of established marker genes (CD33/LTF/LYZ for myeloid cells, CD3D/CD3E/CD3G for T cells, NCAM1/NCR1/KLRD1 for NK cells, CD19/CD79A/CD79B for B cells and GATA1/HBB/ENG for erythrocytes) (Fig. 5a, b) [20]. AddModuleScore function in Seurat package was used to calculate the gene set scores. T cell feature gene sets from previous researches, including exhaustion and cytotoxic gene signatures were calculated [21, 22].

Definitions and statistical analysis

CR referred to the morphological CR featured with bone marrow blasts less than 5%, no extramedullary disease and the recovery in peripheral blood counts (neutrophils > 1 × 109/L and platelets > 100 × 109/L). Relapse-free survival (RFS) was measured from CR to relapse or the final date of bone marrow morphological evaluation.

For non-normally distributed continuous variables, Mann–Whitney U test and Wilcox matched-pairs rank test were used for the comparison of independent or paired variables, respectively. Categorical variables were compared using the Fisher’s exact test. To determine the optimal cutoff values for each CD4 + T cell and CD8 + T cell subset, receiver operating characteristic (ROC) curves were constructed depending on relapse. RFS was estimated using the Kaplan–Meier curves and compared with the Log-rank test. Covariables selected for the Cox multivariate analysis were those without multicollinearity and related to p < 0.10 in the univariate analysis. The Spearman's rank correlation coefficient was used to measure the statistical dependence between two variables. Significance was determined at p-values < 0.05. The SPSS 26.0 software package (SPSS Inc., Chicago, USA) and GraphPad Prism 9 (GraphPad Software Inc., La Jolla, CA, USA) were used for analyzing the data.

Results

Patients’ outcomes

56 of 71 (78.9%) enrolled patients in the cohort underwent at least one course of induction treatment and were followed up at our hospital. Fifty-four (98.2%) patients achieved CR, among whom 44 (78.6%), 7 (12.5%) and 3 (5.4%) patients individually achieved CR after 1, 2 and 3–5 courses of induction chemotherapy. After achieving CR, 28 (51.9%) patients received chemotherapy alone for consolidation, of whom 9 (32.1%) relapsed subsequently; the remaining 26 (48.1%) patients received chemotherapy followed by allo-HSCT (matched sibling donor, n = 5; haploidentical related donor, n = 21) and 2 relapsed. The median follow-up time for the entire cohort was 21.6 (range: 2.4–39.3) months. Forty-nine (87.5%) patients were still alive at the time of the last follow-up, and the median follow-up time of the survivors was 22.5 (range: 4.4–39.3) months. The 2-year RFS and OS rates were 75.9% (95% confidence interval [CI]: 61.3–85.6%) and 85.8% (95%CI: 72.0–93.1%).

The frequencies of TIGIT + and PD1 + in T cells in the bone marrow of AML patients at diagnosis and healthy donors

As shown in Fig. 1, in the CD4 + T cells, the percentage of TIGIT + cells was significantly higher in AML patients than that in HDs (median (range): 24.6% (5.4–68.7%) vs 15.9% (10.4–37.7%), p = 0.0006) and the percentage of PD1 + and TIGIT + PD1 + cells tended to significantly increase in patients compared with HDs (32.1% (8.6–75.3%) vs 27.1% (12.1–48.5%), p = 0.091; 9.4% (3.7–24.9%) vs 8.8% (4.7–15.3%), p = 0.064) (Fig. 1a). In the CD8 + T cells, the frequencies of TIGIT + , PD1 + and TIGIT + PD1 + cells were comparable between patients and HDs (45.4% (9.5–76.5%) vs 45.9% (27.3–56.3%), p = 0.77; 40.2% (15.7–87.5%) vs 37.8% (17.0–79.6%), p = 0.33; 19.3% (6.3–60.5%) vs 21.4% (6.8–41.2%), p = 0.32) (Fig. 1b).

Fig. 1
Fig. 1
Full size image

The expression pattern of TIGIT and PD1 in CD4 + T and CD8 + T cells of AML patients at diagnosis (n = 71) and healthy donors (n = 31). TIGIT + , PD1 + and TIGIT + PD1 + in CD4 + T cells (a) and CD8 + T cells (b). Numbers in this figure refer to p value

What’s more, in both patients and HDs, the percentage of TIGIT + , PD1 + and TIGIT + PD1 + in CD8 + T cells was all significantly higher than that in CD4 + T cells (AML: p < 0.0001, = 0.0015 and < 0.0001; HD: p < 0.0001, = 0.0002 and < 0.0001), respectively (Fig. S2).

The impact of the percentage of TIGIT + , PD1 + and TIGIT + PD1 + in T cells on patients’ relapse

We performed ROC curve analyses in regard to relapse to determine individual optimal cutoff values of the frequencies of TIGIT + , PD1 + and TIGIT + PD1 + in T cells corresponding to the maximal Youden index (Fig. S3). As a result, 26.0%, 34.0% and 12.2% were the cutoff values for the percentage of TIGIT + , PD1 + and TIGIT + PD1 + in CD4 + T cells and those were 45.0%, 44.3% and 18.5% in CD8 + T cells to categorize patients into high percentage (_H) and low percentage (_L) group, respectively.

In the entire cohort, patients with high percentage of TIGIT + PD1 + in CD8 + T cells at diagnosis had a significantly lower RFS rate than those with low percentage (2-year RFS rate: 59.0% [95%CI: 35.5%-76.4%] vs 91.8% [95%CI: 71.1–97.9%], p = 0.015) (Fig. 2f), while other groupings showed no correlation with RFS rates (all p > 0.05) (Fig. 2a–e). When censoring at the time of allo-HSCT, all 6 groupings had no significant impact on RFS (all p > 0.05) (Fig. S4). We also assessed patients receiving allo-HSCT (n = 26). Patients with high frequency of TIGIT + PD1 + in CD8 + T cells at diagnosis had significantly shorter RFS compared to those with low frequency (2-year RFS rate: 64.8% [95%CI: 25.3–87.2%] vs 100.0%, p = 0.023) (Fig. 3f) and patients with high percentage of TIGIT + in CD8 + T cells tended to have a significant lower RFS rate compared with those with low percentage (2-year RFS rate: 76.2% [95%CI: 42.7–91.7%] vs 100.0%, p = 0.083) (Fig. 3d). The other 4 groupings showed no correlation with RFS rates (all p > 0.05) (Fig. 3a–c, e).

Fig. 2
Fig. 2
Full size image

RFS curves of patients grouped by the percentage of TIGIT + in CD4 + T cells (a), PD1 + in CD4 + T cells (b), TIGIT + PD1 + in CD4 + T cells (c), TIGIT + in CD8 + T cells (d), PD1 + in CD8 + T cells (e) and TIGIT + PD1 + in CD8 + T cells (f) in the entire cohort

Fig. 3
Fig. 3
Full size image

RFS curves of patients receiving allo-HSCT grouped by the percentage of TIGIT + in CD4 + T cells (a), PD1 + in CD4 + T cells (b), TIGIT + PD1 + in CD4 + T cells (c), TIGIT + in CD8 + T cells (d), PD1 + in CD8 + T cells (e) and TIGIT + PD1 + in CD8 + T cells (f)

Univariate and multivariate analysis of RFS in the entire cohort

As shown in Table 2, in addition to the frequencies of TIGIT + PD1 + in CD8 + T cells, patients with high percentage of bone marrow blasts had significantly poor RFS than those with low percentage (2-year RFS rate: 64.0% [95%CI: 42.1–79.4%] vs 87.1% [95%CI: 65.0–95.7%], p = 0.035) and patients who received chemotherapy alone as consolidation regimen tended to have lower RFS rate than patients receiving allo-HSCT (2-year RFS rate: 64.2% [95%CI: 42.2–79.6%] vs 87.5% [95%CI: 66.0–95.8%], p = 0.053). Patients with RUNX1::RUNX1T1 fusion tended to have higher RFS rate than those without RUNX1::RUNX1T1 fusion (2-year RFS rate: 90.0% [95%CI: 65.6–97.4%] vs 66.3% [95%CI: 46.2–80.4%], p = 0.075). Multivariate analysis showed that high percentage of TIGIT + PD1 + in CD8 + T cells was the sole independent poor predictor for RFS (HR: 5.6 [95%CI: 1.2–25.8], p = 0.029).

Table 2 Univariate and multivariate analysis of variables on RFS in the entire cohort (n = 55)
Full size table

Relationship between TIGIT and PD1 expression in T cells and AML patients’ characteristics and molecular abnormalities

As shown in Table S1, the percentage of TIGIT + , PD1 + and TIGIT + PD1 + cells in T cells all had no correlation with patients’ ELN genetic risk classifications and fusion genes (all p > 0.05). Furthermore, they had no correlation with patients’ gender, WBC counts, hemoglobin level, platelet counts and the percentage of bone marrow blasts except the age of patients. In CD4 + T cells, the percentage of TIGIT + , PD1 + and TIGIT + PD1 + cells showed no correlation with age in both patients and HDs (AML: r = − 0.098, p = 0.41; r = − 0.036, p = 0.78; r = − 0.0033, p = 0.98; HDs: r = − 0.19, p = 0.31; r = 0.014, p = 0.94; r = − 0.069, p = 0.72. Fig. S5a–c); whereas in CD8 + T cells, the frequencies of TIGIT + , PD1 + and TIGIT + PD1 + cells were positively correlated with the age of patients (r = 0.25, p = 0.034; r = 0.28, p = 0.024; r = 0.32, p = 0.0093. Fig. S5d–f) while no correlation existed in HDs (r = 0.077, p = 0.68; r = 0.13, p = 0.49; r = 0.12, p = 0.53. Figure S5d–f). What’s more, the age distribution between patients and HDs had no significant difference (median (range): 42 (17–62) vs 47 (31–65), p = 0.091).

Comparison of the cytokine production in TIGIT + and PD1 + T cells in AML and healthy donors

The cytokine production (TNF-α and INF-γ) were represented by their intracellular expression tested by MFC and were made comparison between patients and HDs. We simultaneously compared the levels of TNF-α and INF-γ shown by their positive rates in TIGIT + , TIGIT-, PD1 + and PD1- cells and their patterns shown by TIGIT + to – and PD1 + to – pair-comparison of each case (Fig. 4). TIGIT + cells had or tended to have a significantly lower TNF-α + and INF-γ + rates than the corresponding TIGIT- cells in both CD4 + and CD8 + T cells in the AML individuals (TNF-α + : p = 0.088 and 0.0012; INF-γ + : p = 0.0038 and 0.047), whereas no such difference existed in HDs (p = 0.92, 0.17, 0.75 and 0.75) (Fig. 4c, f). When the levels were considered, TIGIT + CD4 + T cells in AML patients had significantly lower INF-γ + rate than those in HDs (p = 0.016) (Fig. 4b). AML patients and HDs had similar cytokine production patterns in PD1, that is, PD1 + cells had significantly higher INF-γ + rates than the corresponding PD1- cells in both CD4 + and CD8 + T cells (AML: p = 0.007 and 0.015; HD: p = 0.028 and 0.046) and had significantly higher TNF-α + rates than the corresponding PD1- cells just in CD4 + T cells in the individuals (p = 0.0007 and 0.028) (Fig. 4c, f). Furthermore, TNF-α and INF-γ levels of all groups were similar between AML patients and HDs (all p > 0.05) (Fig. 4a, b, d, e). TIGIT + PD1 + cells were evaluated (Fig. 4a, b, d, e). TNF-α + rate in TIGIT + PD1 + CD8 + T cells and INF-γ + rate in TIGIT + PD1 + CD4 + T cells were significantly and tended to be significantly lower in AML patients compared with those in HDs, respectively (p = 0.024 and 0.052) (Fig. 4d, b).

Fig. 4
Fig. 4
Full size image

Impact of TIGIT and PD1 expression on the cytokine production of CD4 + T and CD8 + T cells in AML patients and HDs. TNF-α + % in CD4 + T cells (a), INF-γ + % in CD4 + T cells (b), TNF-α + % and INF-γ + % in pairing of TIGIT + to TIGIT- and PD1 + to PD1- cells in CD4 + T cells of individual cases (c), TNF-α + % in CD8 + T cells (d), INF-γ + % in CD8 + T cells (e), pairing in CD8 + T cells (f). Numbers in this figure refer to p value

scRNA-seq data analysis

TIGIT + , PDCD1(encoding PD1) + and TIGIT + PDCD1 + CD8 + T cells were individually annotated. Exhaustion score and cytotoxic score were calculated using gene sets previous reported [21, 22]. TIGIT + PDCD1 + CD8 + T cells exhibit a higher exhaustion score compared to both TIGIT + and PDCD1 + CD8 + T cells (p = 0.0016 and 0.012, Fig. 5e). Figure 5f shows T cell exhaustion related genes that were higher expressed in TIGIT + PDCD1 + CD8 + T cells compared to PDCD1 + CD8 + T cells and TIGIT + CD8 + T cells. Furthermore, TIGIT + PDCD1 + CD8 + T cells had significantly lower cytotoxic score compared to TIGIT + CD8 + T cells while were similar to PDCD1 + CD8 + T cells (p = 0.0014 and 0.68, Fig. 5c). Profiling of marker genes confirmed that TIGIT + PDCD1 + CD8 + T cells had lower expression of cytotoxicity related genes, including GZMB (encoding granzyme B) and PRF (encoding perforin) (Fig. 5d). These results demonstrated that TIGIT + PDCD1 + CD8 + T cells was a population under severe dysfunction.

Fig. 5
Fig. 5
Full size image

scRNA-seq analysis of bone marrow samples from 3 newly diagnosed AML patients. Uniform manifold approximation and projection (UMAP) of post-QC cells (a). Scaled expression of marker genes for each cell type cluster (b). Violin plot showing the cytotoxic score of TIGIT + CD8 + T cells, PDCD1 + CD8 + T cells and TIGIT + PDCD1 + CD8 + T cells (c). Heatmap displaying average expression of cytotoxic related genes for TIGIT + CD8 + T cells, PDCD1 + CD8 + T cells, and TIGIT + PDCD1 + CD8 + T cells (d). Violin plot showing the exhaustion score of TIGIT + CD8 + T cells, PDCD1 + CD8 + T cells, and TIGIT + PDCD1 + CD8 + T cells (e). Heatmap displaying average expression of exhaustion-related genes for TIGIT + CD8 + T cells, PDCD1 + CD8 + T cells and TIGIT + PDCD1 + CD8 + T cells (f). Color scale in heatmaps (d and f) corresponds to z-scored, log-transformed mean gene-expression counts for each subset

Discussion

T cells are the primary mediators of antitumor immunity and play an important role in the response to immunotherapy [23]. However, T cells in the bone marrow of AML often enter a dysfunctional state, which is widely known as T cell exhaustion. Exhausted T cells often showed the elevated and sustained expression of immune checkpoints, which may fundamentally influence treatment responses and clinical outcomes of AML patients [24]. TIGIT, as a recently-identified immune checkpoint, also plays a crucial role in regulating antitumor immunity of T cells. Reports on the effect of its expression in the bone marrow on prognostic significance in AML remained absent. In the present study, we detected a significant increase of the frequency of TIGIT + cells in CD4 + T cells in the diagnostic bone marrow of AML patients than HDs but that in CD8 + T cells were similar. What’s more, high percentage of TIGIT + PD1 + in CD8 + T cells at diagnosis independently predicted poor RFS. TIGIT + cells in T cells of AML patients were dysfunctional characterized by lower level of intracellular TNF-α and INF-γ compared with their TIGIT- counterpart, which is differing from HDs. Furthermore, TIGIT + PD1 + CD8 + T cells of patients exhibited significantly lower level of intracellular TNF-α compared with that of HDs. scRNA-seq data showed that TIGIT + PDCD1 + CD8 + T cells had significantly higher exhaustion score than both TIGIT + and PD1 + CD8 + T cells and lower cytotoxic score than TIGIT + CD8 + T cells.

The expression pattern of TIGIT in T cells in peripheral blood of AML patients was controversial. Xu et al. reported that the percentage of TIGIT + in both CD4 + T and CD8 + T cells had no significant difference between newly-diagnosed AML patients and HDs9. Other two studies found that the frequency of TIGIT + CD8 + T cell (but not TIGIT + CD4 + T cells) in peripheral blood from newly-diagnosed AML patients was significantly higher than those from healthy controls [8, 25]. The expression discrepancy may be caused by the dynamically changing components of peripheral blood which was affected by bacterial or viral infections or patients’ comorbidities such as autoimmune diseases and allergic diseases. As for TIGIT expression in the bone marrow T cells, Xu et al. demonstrated that the frequencies of TIGIT + in CD8 + T cells but not CD4 + T cells in 32 newly-diagnosed AML patients were significantly higher than that of 14 HDs9. In contrast, we found that TIGIT + in CD4 + T cells was significantly increased but that in CD8 + T cells were similar. The similarity was that we both found the percentage of TIGIT + and PD1 + in CD8 + T cells was higher than that in CD4 + T cells for both patients and HDs. We suspected that difference in the cases enrolled might be one reason for this inconsistency because of the association of TIGIT expression in T cells with age. The range of age of our patients was 17–62 years and that was 14–80 in the study of Xu et al. [9]. Both of us found the positive correlation between the frequencies of TIGIT + and PD1 + in CD8 + T cells (but not in CD4 + T cells) with the age of patients, whereas no such correlation existed in HDs. It indicated that elderly patients had a more immunosuppressive tumor microenvironment at initial diagnosis which may be participated in their poor prognosis [9].

A number of studies documented that overexpression of TIGIT in tumor-infiltrating T cells was associated with poor prognosis in solid tumors, including colorectal cancer [26], muscle-invasive bladder cancer [27] and cutaneous melanoma [28]. Several studies have demonstrated that the expression of TIGIT in T cells in peripheral blood from newly-diagnosed AML patients was associated with the clinical outcomes. Kong et al. found that the high-TIGIT in CD8 + T cells was significantly related to higher rate of primary refractory disease [8]. Wang et al. uncovered that AML patients who achieved the remission after chemotherapy had lower frequencies of TIGIT + PD1 + CD8 + T cells compared with the non-remission group25. To date, no studies ever evaluated the diagnostic bone marrow specimens and performed survival analysis to evaluate its prognostic significance. In this study, we comprehensively analyzed the impact of TIGIT + , PD1 + and TIGIT + PD1 + in both CD4 + T cells and CD8 + T cells on RFS. We observed that only the high frequency of TIGIT + PD1 + in CD8 + T cells at diagnosis was associated with a significantly shorter RFS and it was an independent poor predictor. Coincidentally, Kong er al. found that PD1hiTIM-3+ T cells in peripheral blood were significantly associated with leukemia relapse after allo-HSCT in AML patients [29]. These results suggested that T cells with double-positive immune checkpoints may have more exhausted phenotype and functionality compared with single-positive or negative T cells which more likely led to leukemic cells’ immune evasion and poor clinical outcomes.

In addition, subgroup analysis revealed that high expression of TIGIT + PD1 + in CD8 + T cells was associated with poor RFS in patients receiving allo-HSCT. The percentage of TIGIT + in CD8 + T cells showed same trend. Elevated expression of TIGIT in T cells from post-transplantation blood samples were reported to be correlated with subsequent AML relapse [30, 31]. The above results demonstrated that besides the elevated TIGIT expression in T cells post-transplantation, high expression of TIGIT in T cells at initial diagnosis could also predict leukemic relapse after allo-HSCT, which would make prophylaxis intervention as early as possible. The anti-leukemia effectiveness of allo-HSCT relies on the graft-versus-leukemia (GVL) effect. Relapse after allo-HSCT commonly results from the failure of GVL effect to eradicate minimal residual disease. These results indicated that TIGIT on T cell surface may be involved in GVL failure by mediating immune escape of leukemia cells.

We compared the effect of TIGIT and PD1 expression on T cell function between AML patients and HDs by activating T cells in vitro. AML patients and HDs had similar positive relationship between PD1 expression and TNF-α and INF-γ production. Differing from PD1, TIGIT expression was related to lower TNF-α and INF-γ production in T cell of AML, whereas no such correlation existed in HDs. Furthermore, AML patients as a whole had lower INF-γ production in TIGIT + CD4 + T than HDs. It implied that TIGIT seems to be an immune-inhibitor in AML but not in HDs through inhibiting INF-γ production, and PD1 had similar promoting effect on TNF-α and INF-γ production in both AML and HDs. Combined with TIGIT expression in T cells, we showed that AML patients had a double abnormality compare to HDs, the elevated frequency of TIGIT + in CD4 + T cells and the decreased level of INF-γ of that cell cluster meanwhile. Similarly, Kong et al. also found that TIGIT + CD8 + T cells which was elevated in peripheral blood of AML patients had significantly lower intracellular TNF-α and INF-γ compared with TIGIT-CD8 + T cells8. That is to say, the TIGIT on T cell surface may mediate T cell exhaustion by reducing its cytokine production ability. In addition, we found that patients had significantly lower levels of TNF-α in TIGIT + PD1 + CD8 + T cells compared with HDs.

To further explore the reason why high TIGIT + PD1 + CD8 + T cells led to inferior prognosis of AML patients, we performed detailed transcriptomic characterization of newly-diagnosed AML patients’ immune environments at single-cell resolution. A previous study has shown that PDCD1highTIGIThighLAG3highCD8 + T cells were developed from tissue residual memory T cells (TRMs) with high level expression of PDCD1 and TCF7, which were at the early stage of exhaustion [32]. This result suggested that TIGIT + PDCD1 + CD8 + T cells may enter a terminal exhausted state. In accordance with this, we found that TIGIT + PDCD1 + CD8 + T cells had significantly lower exhaustion score than both TIGIT + and PDCD1 + CD8 + T cells. Furthermore, it has been confirmed that the increasing loss of cytotoxic potential and expression of inhibitory receptor indicate the severity of T cell exhaustion [33]. We found that TIGIT + PDCD1 + CD8 + T cells had significantly lower cytotoxic score than TIGIT + CD8 + T cells. Thus, TIGIT + PDCD1 + CD8 + T cells was a sever exhausted population of T cells. As the exhaustion gene set used to calculate exhaustion score had significant overlap with exhaustion-related genes identified in hepatocellular carcinoma [34], melanoma [35], and in lymphocytic choriomeningitis virus infected mouse models [36], our results suggested that TIGIT + PDCD1 + CD8 + T cells in AML shared similar exhaustion signature with exhausted T cells in solid tumors and chronic viral infections.

There were several limitations in the current study. First, this was a retrospective study, and the treatment regimens were not fully uniform. Second, the study mainly enrolled patients with RUNX1-RUNX1T1 fusion gene during the early stage, which resulted in a higher-than-true proportion of patients with RUNX1::RUNX1T1. In order to clarify the effect of the higher-than-true proportion of patients with RUNX1::RUNX1T1 on RFS, the RUNX1::RUNX1T1 fusion was included into the univariate and multivariate analysis.

In conclusion, the present study demonstrated that the diagnostic bone marrows of AML patients had elevated frequency of TIGIT + in CD4 + T cells compared to HDs. Differing from HDs, AML patients had lower ability of TNF-α and INF-γ production in TIGIT + cells than the TIGIT- counterparts in T cells. In addition, TIGIT + PD1 + CD8 + T cells had significantly lower level of TNF-α in patients compared with HDs. TIGIT + PD1 + CD8 + T cells exhibited significantly higher degree of exhaustion than both TIGIT + and PD1 + CD8 + T cells and more dysfunctional cytotoxicity than TIGIT + CD8 + T cells. Furthermore, high percentage of TIGIT + PD1 + in CD8 + T cells independently predicted relapse in AML. The current study provided evidence that combination of TIGIT and PD1 in CD8 + T cells at diagnosis might be useful prognostic molecules in AML. As single anti-PD-1 monoclonal antibody infusion has been shown rather modest clinical efficacy, the findings of this study offered evidence for PD1 and TIGIT combinational therapy [37]. Moreover, more studies are warranted to confirm the current results.