All human samples were collected from healthy donors (HD) and from newly diagnosed AML patients after informed consent (local Ethical Committee approval code: 147/2013/O/Tess). Patients’ characteristics are reported in Table S1 in Supplementary Material. AML cells were obtained as mononuclear cells isolated by Ficoll-Hypaque centrifugation (Amersham, USA) from patients’ bone marrow or peripheral blood (PB) samples, including at least 70% leukemic cells, as evaluated by morphology and FACS analysis. CD3⁺, CD19⁺, CD14⁺, and CD4⁺CD25⁺CD127^dim/− cells were purified by magnetic separation (Miltenyi Biotec, Germany), according to manufacturer’s instructions from mononuclear cells separated from buffy coats and patients’ PB by Ficoll-Hypaque centrifugation (Amersham). Purity of cell populations was always >90%. Human HL-60 (DMSZ; ACC 3, FAB M2) and murine WEHI-3B (DMSZ; no. ACC 26) AML cell lines were maintained at 37°C and 5% CO₂. HL-60 cells were cultured in RPMI 1640 medium (Lonza, Milan, Italy), supplemented with 10% heat-inactivated fetal bovine serum (Gibco-Invitrogen, USA), 2 mM l-glutamine, 100 U/ml penicillin, and 100 µg/ml streptomycin (MP Biomedicals, Italy) (complete RPMI). WEHI-3B cells were cultured in Iscove modified Dulbecco’s medium (Sigma-Aldrich, USA) supplemented with 10% fetal bovine serum (Euroclone, Italy), 100 U/ml penicillin and 100 mg/ml streptomycin (Euroclone). AML cells (1 × 10⁶/ml) were treated with DNR 500 ng/ml (Sigma-Aldrich) or cytarabine (ARA-C) 25 µg/ml (Sigma-Aldrich) for 4 h and tested for apoptosis by Annexin-V-FLUOS Apoptosis Detection Kit (Roche, Switzerland) and ATP release (see Datasheet S1 in Supplementary Material).

T cells from newly diagnosed AML patients (n = 23) undergoing standard “7 + 3” induction chemotherapy regimen, including continuous infusion of ARA-C at 200 mg/m² for 7 days and DNR at 60 mg/m² for 3 days, were analyzed before drug administration and at different time points (+7, +14, +21, and +28) after the end of administration regimen. Due to paucity of evaluable cells, in few cases (n = 3), day 7 T cells were not analyzed. For IFN-γ intracellular detection, purified CD3⁺ T cells were co-cultured with autologous AML blasts, at 10:1 ratio, in complete RPMI supplemented with 10% autologous serum (autologous RPMI). CD3⁺ T cells either used as such or pre-activated with ionomycin (500 ng/ml; Sigma-Aldrich) and phorbol-12-myristate-13-acetate (PMA, 10 ng/ml; Sigma-Aldrich) and co-cultured with autologous CD19⁺ cells were used as controls. After 4 h, brefeldin A (2 µg/ml; BD Biosciences, USA) was added, followed by overnight incubation before permeabilization and staining.

In selected cases, 5 × 10⁶ CD3⁺ T cells from IFN-γ-responding AML patients were collected after chemotherapy and stimulated overnight with autologous AML blasts (ratio 1:1). Cells were incubated with IFN-γ and TNF-α Catch reagent (Miltenyi) for 5 min on ice. Finally, after incubation with medium for 45 min at 37°C, cells were stained with the following human monoclonal antibodies (mAbs): IFN-γ APC, TNF-α APC (Miltenyi), CD137 PE-Cy7 (4-1BB, 4B4-1; Biolegend, USA), CD3 APC-eFluor780 (SK7; eBioscience, USA), CD8 Pacific Blue (B9.11; BD Pharmingen, USA), CD33 FITC (HIM3-4), or CD34 FITC (561) (both from Biolegend, USA) to discard blasts population from analysis. CD3⁺CD8⁺IFN-γ⁺TNF-α⁺CD137⁺ were sorted at BD FACS Aria cell sorter (BD Biosciences) and in vitro expanded (2–3 weeks) on a irradiated mononuclear cell feeder layer from two HD (1 × 10⁶ cells/ml) in RPMI supplemented with human serum (HS, 8%), phytohemagglutinin (PHA, 1 µg/ml; Sigma-Aldrich) and IL-2 (150 U/ml; Roche), for 2–3 weeks. IL-2 was added to the culture every other day.

Human monocyte-derived DCs were generated by a 5-day culture of CD14⁺ cells in complete RPMI in presence of granulocyte-macrophage colony stimulation factor (50 ng/ml; GM-CSF Endogen, USA) and IL-4 (800 U/ml; Miltenyi), as previously described (21, 22). DC maturation was induced with a cocktail of cytokine made of TNF-α (10 ng/ml; Endogen), IL-6 (10 ng/ml; Endogen), IL-1β (10 ng/ml; Endogen), and 1 µg/ml PGE₂ (Endogen) (23) or with ATP (1 mM; Sigma-Aldrich). For DC pulsing, chemotherapy-treated HL-60 cells were cultured for 20 h with immature DCs (2:1 ratio) in autologous RPMI. After culture, IDO1 protein expression was evaluated by Western blotting (see Datasheet S1 in Supplementary Material).

Leukemia-reactive IFN-γ producing CD3⁺ T cells were evaluated after two rounds of in vitro stimulation (7 + 7 days) with autologous DCs pulsed with chemotherapy-treated HL-60 cells (ratio 10:1) in autologous RPMI. IL-2 (20 U/ml) was added on alternate days. In selected experiments, allogeneic Tregs induced by the same DCs used for T cell priming were added to cell cultures (ratio 1:1) for 5 days to test their suppressive activity. At the end of culture, T cells were tested for IFN-γ production. DCs loaded for 24 h (1:2 ratio) with HL-60 cell lysate, obtained after three cycles of cells freeze-thawing and filtering through an insulin syringe, were used as target.

Antigen-specific IFN-γ-producing CD3⁺ T cells were obtained as described above, except challenging of DCs with HLA-A0201-restricted Wilms’ tumor-derived peptide WT1-A (10 µg/ml; 126–134; PRIMM, Italy) for 4 h. DCs loaded with WT1-A or WT1-B (187–195) were used as targets.

Immature DCs, chemotherapy-treated HL-60 pulsed DCs and DCs matured in presence of ATP or a cytokine cocktail were co-cultured for 5 days in autologous RPMI with allogeneic CD3⁺ T cells, in presence or absence of the IDO1-inhibitor 1-methyl tryptophan-L (1 mM, 1-MT-l; Sigma-Aldrich). IDO1-silenced DCs were also used (see Datasheet S1 Supplementary Material) (10). Tregs were quantified by FACS analysis. In selected experiments, purified and irradiated CD4⁺CD25⁺CD127^dim/− Tregs (10⁴/well) were added to cultures consisting of CFSE-labeled CD3⁺ T cells (10⁵/well) as responders, stimulated by allogeneic immature monocyte-derived DCs (1:10 ratio), for 5 days, to test their suppressive activity.

T cells, Tregs, and DCs were characterized using the following mAbs, according to manufacturer’s instructions: anti-CD8 Pacific blue (B9.11) or PE, anti-CD15S PE-CF594 (CSLEX1), anti-CD25 BV605 (2A3), anti-CD28 APC or APC-eFluor 780 (CD28.2), anti-CD80 PE-Cy 7 (L307.4), anti-CD152 (CTLA-4; clone BNI3), and anti-Ki-67 Alexa Fluor 700 (B56) from BD Bioscience; anti-CD3 PE-Cy7 or APC (UCHT1) or APC-eFluor 780 (SK7), anti-CD4 FITC (RPA-T4), anti-CD8 APC (SK1), anti-CD25 APC (SK1) or APC-eFluor 780 (CD25-4E3), anti-CD38 Alexa 700 (HIT2), anti-CD83 PE (HB15e), anti-CD127 Pe-Cyanine5 (eBioRDR5), anti-CD197 (CCR7) PE-Cy7 or Brillant Violet 421 (G043H7), and anti-HLA-DR FITC (L243) from eBioscience; anti-CD4 FITC (SFCI12T4D11) and anti-CD45RA ECD (2H4LDH11LDB9) (Beckman Coulter, USA); anti-CD39 PE/Cy7 (A1), anti-CD45RA Brilliant Violet 510 (HI100), anti-CD127 PerCP/Cy5.5 (IL-7Rα; clone A019D5), and anti-CD279 (programmed cell death protein 1; PD-1) APC or Brilliant Violet 711 (EH12.2H7) from Biolegend. LIVE/DEAD Fixable Aqua (Thermo Fisher Scientific, USA) was used to gate out dead cells. For intracellular staining (24), after cells fixation and permeabilization with 4% paraformaldehyde (VWR, USA) and 0.1% saponin (Sigma), the following mAbs were used: anti-IFN-γ PE (4S.B3), anti-Foxp3 PE or FITC (236A/E7), and anti-Ki-67 FITC (20Raj1) from eBioscience. Circulating Tregs in PB of AML patients were assessed by using Human Regulatory T Cell Whole Blood Staining Kit (eBioscience) (see Datasheet S1 in Supplementary material).

Specifically, for ex vivo post-chemotherapy patients’ Tregs screening, anti-CD3/CD4/CD25/FOXP3 mAbs were used. In selected cases (n = 6), a wider panel including anti-CD3/CD4/CD25/CD127/CD15s/CD45RA/FOXP3/PD-1/Ki67 mAbs for identification of Tregs subpopulation (Treg1/2/3) were used.

For each sample, isotype-matched irrelevant mAbs staining was used as control. At least 10,000 events were collected from each sample at Gallios Flow Cytometer (Beckman Coulter) or BD Accuri C6 (BD Biosciences). For expanded T cells, data analyses were performed by using FlowJo Single Cell Analysis Software (FlowJo LLC).

To obtain WEHI-3B cell clones stably expressing plasma membrane luciferase, cells were transfected with the PmeLUC probe, as previously described (25). Briefly, 6 × 10⁶ cells were suspended in electroporation buffer (Life Technologies, USA) with 3 µg of plasmid’s DNA and electroporated in a Microporator MP-100 (Digital bio, Thermo Fisher), at 1,250 V for 40 ms. Stably transfected cell clones were obtained by selection with neomycin/G418 sulfate (0.2–0.8 mg/ml; Sigma) followed by limiting dilution as previously described (26). Male 4–6 weeks old Balbc/J wt or P2X7^−/− mice were subcutaneously injected with 2 × 10⁶ WEHI-3B PmeLUC cells. Tumors became palpable approximately 7 days post-inoculum (p.i.) and were measured with a manual caliper. Tumor volume was calculated according to the following equation: π/6 [w1 × (w2)²], where w1 is the major diameter and w2 is the minor diameter. Luminescence emission was measured daily, days 7–12 from p.i. with a total body luminometer for small animals (IVIS Lumina, Caliper; Perkin Elmer, USA), as previously described (27). Mice anesthetized with 2.5% isofluorane were intra peritoneum (i.p.) injected with 150 mg/kg d-luciferin (Promega, USA) and luminescence was captured from dorsal view. Photon emission was quantified using the Living Image^® software (Perkin Elmer) and averaged as photons/second/cm²/steradian (abbreviated as p/s/cm²/s). DNR (3 mg/kg, Sigma), ARA-C (50 mg/kg, Sigma), or sterile PBS vehicle (placebo) were i.p. administered at p.i. days 7 and 9. Blood samples were collected from the submandibular vein under general anesthesia immediately before sacrificing the animal (p.i. day 12). Cytokines levels were evaluated following 1:2 plasma dilution with Ciraplex CK1 mouse multi-cytokine assay kit (AushonBiosystem, distributed by TemaRicerca, Bologna, Italy) as per manufacturer’s instructions. Tumors were excised and processed for flow cytometry (see Datasheet S1 in Supplementary material). All animal procedures were approved by the University of Ferrara (Ferrara, Italy) Ethic committee and the Italian Ministry of Health in compliance with International laws and policies (EU Directive 2010/63/EU and Italian D.Lgs 26/2014; authorization number 821/2015PR to EA).

Data were expressed as mean ± SEM of values obtained in the experiments. Statistical analyses were performed with GraphPad Prism 6 software (GraphPad Software, Inc., La Jolla, CA, USA), using ANOVA or unpaired t-test. p Values <0.05 were considered statistically significant.

We analyzed the induction of tumor-reactive CD8⁺ cytotoxic T cells (CTLs) in a cohort of AML patients (n = 23) undergoing combined DNR and ARA-C chemotherapy. In 15 out of 23 patients, we observed an increase of leukemia-reactive IFN-γ-producing CD8⁺ T cells (Figure S1 in Supplementary Material) mostly belonging to effector memory (EM) and EM expressing CD45RA (effector memory expressing RA) subsets (Figure S2 in Supplementary Material), which highly expressed the activation marker CD38 and downregulated CD28 as compared to naïve (p = 0.03) or central memory (CM) (p = 0.03) T cells.

Along with the detection of leukemia-reactive CD8⁺ T cells, we observed an increase in CD4⁺CD25⁺Foxp3⁺ T cells after DNR plus ARA-C chemotherapy (Figure 1A) with a peak at day 21. We, then, sought to better characterize Tregs subsets at the phenotypic level. Among CD3⁺CD4⁺ T cells collected at day 21 post-chemotherapy, CD25⁺CD127⁻ cells were subdivided into three different subsets (Treg1, Treg2, Treg3), according to the expression of CD15s (Figure S3 in Supplementary Material) (28, 29). As compared to HD, no differences were observed for Treg1 and Treg3 frequencies, whereas the percentage of Treg2 cells was significantly increased in AML patients after chemotherapy (Figure 1B). Treg2 cells from AML patients expressed Foxp3, which correlated with CTLA-4 and CD39 (Figure 1C), indicating a suppressive phenotype (28, 30, 31). Moreover, as compared to Treg1, Treg2 cells showed higher expression of PD-1, which identifies a novel population of Tregs with crucial suppressive activity in the tumor setting (Figure 1D) (32, 33). Intriguingly, after 21 days post-chemotherapy an increase in Ki67 expression was observed in Treg2 over Treg1 and Treg3 cells, suggesting recent activation (Figure 1E). Indeed, a time-course analysis revealed a selective increase of proliferating Ki67⁺ Treg2 over Treg1 and Tregs3 cell subsets at day 14 post-chemotherapy, which progressively reduced at later time points (Figure S4 in Supplementary Material).

Taken together, the analysis of the composition of T cells, emerging in AML patients after combined DNR and ARA-C chemotherapy, reveals an increase of Tregs with a suppressive phenotype. These ex vivo data prompted us to dissect the contribution of DNR versus ARA-C in Tregs induction and to investigate a possible mechanism responsible for chemotherapy-driven induction of a suppressive microenvironment in AML.

Inflammatory stimuli may promote Tregs via the induction of tolerogenic DCs (16). Since chemotherapy treatment is associated with the abundant release of inflammatory signals, we tested the contribution of DNR and ARA-C in driving DCs toward Tregs induction. Chemotherapy-treated AML cells were pulsed into DCs, which were, then, used in co-culture for inducing leukemia-specific T cells and Tregs. DC-loaded with DNR-treated AML cells were more efficient than DC-loaded with ARA-C-treated cells not only, as expected, in inducing leukemia-reactive CD8⁺ T cells (Figure 2A), but also in increasing the frequency of CD4⁺CD25⁺Foxp3⁺ T cells (11.70 ± 4.15 versus 5.97 ± 3.10%, respectively; p < 0.05) (Figure 2B). DNR-induced CD4⁺CD25⁺Foxp3⁺ T cells were shown to act as Tregs through the complete inhibition of leukemia-specific IFN-γ production by both CD4⁺ and CD8⁺ T cells (Figure 2C). Moreover, these DC-induced Tregs also reduced T cell-mediated IFN-γ-production in response to the leukemia-associated antigen, WT1 (Figure 2D).

These results demonstrate that DNR is more efficient than ARA-C in promoting functional Tregs through tolerogenic DCs.

Within tumor microenvironment, chemotherapy-treated tumor cells release a high amount of ATP, which is a major driver of DC activation and function (5, 6). To test the contribution of ATP release from chemotherapy-treated AML cells in DC-mediated Tregs induction, we preliminarily in vitro measured ATP release from dying AML cells, including cell lines and primary cells, after treatment with DNR and ARA-C. DNR and ARA-C were used at the concentrations capable to induce comparable apoptosis level, but only DNR increased ATP release (Figures 3A,B). These data suggest that the increased capacity of DNR over ARA-C to induce Tregs via DCs may be, at least in part, mediated by ATP. Indeed, when treated with ATP, DCs induced a population of CD4⁺CD25⁺CD127^−/dim T cells (Figure 3C), which acted as bone fide Tregs by reducing allogeneic T-cell proliferation (Figure 3D). Interestingly, in comparison to a cocktail of pro-inflammatory cytokines, ATP treatment of DCs resulted in higher expression of PD-1 on Tregs (Figure 3E), whereas Foxp3 and CD39 expression was similar (data not shown). Moreover, Tregs obtained after culture with ATP-treated DCs expressed Ki-67 at comparable level, suggesting recent activation/proliferation (Figure 3F).

Among the different mechanisms used by DCs for promoting Tregs, IDO1 upregulation is crucial (15, 16). Then, we investigated the involvement of ATP in IDO1 upregulation in DCs during chemotherapy. ATP treatment upregulated IDO1 protein along with the maturation markers, CD80 and CD86 (Figures 4A,B), albeit to a lower extent than pro-inflammatory cytokines, used as positive control (21). Of note, in agreement with the different levels of ATP release (Figures 3A,B), DNR was more efficient than ARA-C in upregulating IDO1 protein (Figure 4C), along with the maturation marker CD83, which was significantly increased by DNR, and not by ARA-C treatment (p < 0.05) (Figure 4D). The inhibition of IDO1 in DCs upon DNR treatment by the IDO1-inhibitor, 1-MT-l (Figure 4E, p < 0.01) or by IDO1-specific RNA interference (data not shown) significantly reduced DC-mediated Tregs induction, thus demonstrating that DNR-induced IDO1 upregulation in DCs may drive Tregs. Taken together, these data suggest that ATP release from DNR- and not ARA-C-treated leukemia cells is correlated with Tregs generation and IDO1 upregulation in mature DCs.

We, then, investigated in a mouse model the link between ATP release from chemotherapy-treated cells and the induction of an immune suppressive microenvironment. In line with data obtained with human AML cells (Figures 3A,B), only DNR, and not ARA-C, increased the release of ATP from murine WEHI-3B AML cells in vitro (Figure S5 in Supplementary Material).

This cell line was then transfected with PmeLUC, a luciferase expressed at the plasma membrane, which allows for ATP measure in the extracellular milieu (25). WEHI-3B PmeLUC live clones were injected in syngeneic mice, which were then treated with either DNR or ARA-C. While DNR and ARA-C treatment had comparable effect on tumor mass reduction (Figure 5A), only DNR increased ATP release (36,094.20 ± 7,420.75 p/s/cm²/sr) over control (13,950.9 ± 4,118.65 p/s/cm²/sr; p < 0.05) (Figures 5B,C). These data prompted us to correlate the composition of leukemia immune suppressive microenvironment, i.e., Tregs and tolerogenic DCs, with the different levels of ATP released upon chemotherapy.

Chemotherapy-induced ATP release correlates with an increase in pro-inflammatory mediators (34, 35). Accordingly, administration of DNR but not ARA-C promoted a significant increase of IFN-γ (p < 0.05), IL-1β (p < 0.01), IL-2 (p < 0.001), and IL-12 (p < 0.05) in the serum of leukemia-bearing mice (Figure S6 in Supplementary Material). To investigate the capacity of chemotherapy to elicit a suppressive, along with its activatory, effect, leukemia infiltrate was analyzed after DNR and ARA-C treatment and compared to placebo. Among CD4⁺ T cells, DNR treatment was associated with a significant increase of total CD25⁺Foxp3⁺ T cells as compared to placebo and ARA-C (Figure 6A). Notably, DNR, and not ARA-C, significantly increased the expression of PD-1 on CD4⁺CD25⁺Foxp3⁺ T cells, both as percentage of positive cells (data not shown) and as mean fluorescence intensity (Figure 6B). This finding in DNR-treated mice paralleled with the increase of PD-1-expressing Tregs in AML patients after induction chemotherapy (Figure 1D) and in Tregs obtained in vitro after culture with human ATP-treated DCs (Figure 3E). Moreover, among CD11b⁺MHCII⁺LY6C⁻ myelo-monocytic cells (Figure S7 in Supplementary Material) only DNR, and not ARA-C treatment upregulated DC expression of CD11c (Figure 6C), which correlated with higher expression of IDO1 (Figure 6D) and MHCII (Figure 6E). Interestingly, in DNR-treated mice, CD11c^high DCs had also increased expression of CD39 (Figure 6F), which is the rate-limiting enzyme during ATP catabolism (36, 37).

To demonstrate in vivo the contribution of ATP release in increasing the infiltration by Tregs and tolerogenic DCs after DNR treatment, P2X7R null mice were injected with WEHI-3B PmeLUC cells and treated with DNR. The levels of ATP after DNR treatment were similar to those observed in wild-type (WT) mice (Figure S8 in Supplementary Material). However, differently from WT mice, in P2X7 null mice DNR failed to increase CD4⁺CD25⁺Foxp3⁺ Tregs (Figure 6G) and IDO1-expressing CD11c^high DCs (Figure 6H). Taken together, these in vivo results demonstrate that ATP release from DNR-treated dying leukemia cells has a role in the induction of an immune suppressive microenvironment, which comprises Tregs and IDO1-expressing DCs.

This study was carried out in accordance with the recommendations of Ethics Committee of the University Hospital of Bologna S. Orsola-MALPIGHI with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Ethics Committee of the University Hospital of Bologna S. Orsola-Malpighi, approval code: 147/2013/O/Tess.

This study was carried out in accordance with the recommendations of Ethical Committee of University of Ferrara. The protocol was approved by the Ethical Committee of University of Ferrara and Italian Ministry of Health (EU Directive 2010/63/EU and Italian D.Lgs 26/2014; authorization number 821/2015PR to EA).