Acute Myeloid Leukemia (AML) is characterized by a poor prognosis, even in patients who achieve a complete remission to initial therapy. Indeed, leukemia cells exploit a variety of mechanisms to evade T-cell-mediated immunity, leading to disease progression and relapse [1–4].

Among them, activation of immune checkpoint pathways in AML may interfere with effective T-cell antitumor immunity. PD-1/PD-L1 interactions are associated with immune evasion in pre-clinical leukemia models, as we and others have previously demonstrated [5, 6]. It was also reported that overexpression of PD-1 on stroma/non-blast compartment and its ligands (PD-L1 and PD-L2) on CD34⁺ leukemia cells is associated with more aggressive leukemia and progression from Myelodysplastic syndromes (MDS) to AML or AML relapse [7, 8]. Importantly patients resistant to epigenetic therapy had relatively higher increments in expression of these genes compared with patients who achieved response [8]. Clinical studies of PD-1 blockade are currently ongoing in patients with AML and MDS [9].

Although dominant in mediating T cell dysfunction in cancer, it is now well accepted that interruption of PD-1/PD-L1 axes alone does not completely restore T cell function in some patients, indicating the involvement of additional negative regulatory pathways, such as TIM-3/Gal-9, in promoting T cell exhaustion [10, 11]. T cell immunoglobulin and mucin domain 3 (TIM-3) is expressed on Th1, Th17, CD8+ T cells–cells of myeloid lineages [12] in mice. The inhibitory role of TIM-3 in T cell-mediated immune responses is produced by the binding of Galectin-9, one of its ligands, to the carbohydrate motif on TIM-3 inducing Th1 and Th17 cell death [13, 14].

Combinational targeting of these pathways has recently been proposed in hematological malignancies. Administration of TIM-3 and PD-1 mAbs synergistically control tumor growth [10].

Recently, we conducted a prospective phase I clinical trial of Selinexor, a Selective Inhibitor of Nuclear Export (SINE), combined with High-Dose Cytarabine Mitoxantrone (NCT02573363). HiDAC + Mito is an effective induction regimen frequently utilized for patients with high-risk AML, either de novo or relapsed/refractory. The clinical outcomes of patients treated in this study have been recently published [15].

Selinexor (KPT-330) is an exportin 1 (XPO1) inhibitor. XPO1 is a nuclear export receptor involved in a cytoplasmatic translocation of most major tumor suppressor proteins (TSP) and growth regulatory proteins (GRP), including p53, p21, p73, FOXO1, β-catenin and NPM1 [16]. Kojima et al. have demonstrated that increased expression of XPO1 have been independently associated with a worse prognosis in adults with AML. Overexpressed levels of XPO1 lead to enhanced transport of TSP/GRP to the cytoplasm thus, forcing nuclear retention of these proteins is a rational therapeutic strategy of selinexor use in AML [17, 18].

In this short report, we first aimed to characterize the expression of such immune checkpoint molecules, both on CD34⁺ AML cells and on CD34⁻ bone marrow cells and on blood and bone marrow (BM) resident T cells during the treatment course. The rationale and sustainability of incorporating checkpoint blockade once the patients achieve remission as a means of providing immune-mediated protection from relapse is a challenging question of recent years.

Furthermore, in order to better understand the possible mechanism in immune response of the resistance to chemotherapy, patients enrolled in this study were divided in two groups - those in complete remission (CR) and those that experienced resistant disease, relapse, or death prior to, or as the result of treatment failure (TF). The comparison between them was employed at diagnosis, end of induction and at the point of primary induction failure.

The characterization and comparison of immune checkpoint ligands and receptors in the bone marrow at time of diagnosis and end of induction, allowed us to monitor the changes and to identify predictive or prognostic biomarkers to guide future immunotherapy in AML.

In this analysis, we aimed to characterize dynamic changes in expression of immune checkpoint pathways on AML cells and T cells resident in bone marrow environment and peripheral blood prior to and after induction chemotherapy.

Specimens from patients with high-risk AML enrolled in a prospective clinical trial that combined Selinexor with HiDAC+Mito (NCT02573363) were employed to address the question of the incorporating checkpoint blockade in combination with chemotherapy as a means of immune-mediated protection, even for those patients that achieve remission.

To monitor changes in expression profiles of immune checkpoint receptors and ligands, multi-parameter flow cytometry was performed on bone marrow (BM) aspirates and peripheral blood from 26 patients with AML at the time of diagnosis and at the end of induction chemotherapy. Patients were divided into 2 cohorts - those who achieved CR (n = 16), and those who experienced TF (n = 10) (Table 1). Additional file 1: Figure S1A shows the diagram of the strategy of our study and sample collection. Kaplan-Meier plot (Additional file 1: Figure S1B) depicts patient survival from the time of diagnosis to CR or TF populations. The median of days elapsed from diagnosis was 346 and 176 for CR and TF respectively, and Hazard Ratio (Mantel-Haenszel) TF/CR was 1.7; Mantel-Cox test was not significant. The shadow in the chart indicates the timewise of samples collection and analysis.

An anti-CD34 antibody was used to analyze frequencies of CD34⁺ AML cells and the remaining CD34⁻ cell populations and to evaluate expression of costimulatory and coinhibitory ligands on the respective cell populations (Fig. 1a).Fig1Fig. 1

The association between Gal9 and TIM-3 as prognostic marker for Selinexor + HiDAC Mito regimen. At time of diagnosis, multi-parameter flow-cytometry was performed on bone marrow (BM) aspirates from 26 patients. A FITC conjugated anti-CD34 antibody was used to analyze frequencies of CD34+ AML cells and the remaining CD34− cell populations. (a) Patients were divided into 2 cohorts - those who achieved CR (n = 16), and those who experienced TF (n = 10). The comparison of frequency of CD4+ PD-1+ T cells and CD34−Gal-9+ in these two groups are shown. (b-c) Bars represent medians. Percentage of cumulative frequencies are displayed in (e) and (f). We calculated the Spearman correlation coefficients to describe associations between CD4+PD-1+ T cells and CD34+PDL-1+ AML cells in TF patients (d); same for CD34-Gal-9+ between inhibitory and activator markers TIM-3, ICOS, Lag3 on CD8+ cells, at time of diagnosis (g). Lastly, between CD8+TIM-3+ and ICOS and Lag-3. (h) A linear regression according to the Deming procedure and deviation for linearity (Runs Test) was additionally computed. Gal-9 and TIM-3 vs HLA-DR was used as negative control.(i)

The current debate to novel therapeutic approaches that can challenge development of resistance to the treatment or relapse experienced by AML patients is direct toward the evidence of the BM microenvironment as a niche for AML [25]. In this context, despite the limit of our small group of patients, our findings suggest that the Gal9/TIM3 pathway may play a role in patients in remission by subverting ongoing immune surveillance, and suggests that T cells in AML patients, even those who achieve CR to therapy, are likely exhausted or dysfunctional.

In conclusion, the high expression of Gal-9 at diagnosis and the increased expression of TIM-3 at remission in TF patients, provides a rationale for incorporating antibodies against the Gal9/TIM3 pathway during and/or following remission induction therapy for AML. A larger cohort analysis and more mechanistic study will be needed to expand and confirm these results.