A B compared to controls both in percentage and absolute number (Figure 1A-C). Moreover, we found a positive correlation between the number of circulating ILCs and the percentage of circulating leukemic blasts (R2= 0.3477, P= 0.0031)(Figure 1D). Then, we analyzed the relative frequency of the different ILC subsets, characterized by the combination of surface marker expression detailed above and of specific transcription factors: T-bet for ILC1, GATA-3 for ILC2 and Rorγt for ILC3 (Figure 1E). Strikingly, we found a significant enrichment of ILC1 in AML patients and a concomitant profound reduction of ILC3 NCR+ cells. No differences were detectable in the levels of ILC2 and ILC3 NCR–cells (Figure 1F). No difference was observed either in the frequency of total circulating ILCs and in the distribution of the ILC subsets among the three cytogenetic risk groups (data not shown). In addition, since AML is a malignancy originating in the bone marrow (BM), we compared the frequencies of ILCs between paired BM and PB samples from 17 patients. We did not find any difference either in percentage of total ILCs (data not shown), or in the relative proportions of the different ILC subsets (Figure 1G), suggesting a similar composition of ILCs in BM and PB in AML patients at disease onset. To test the functionality of ILCs in terms of their capacity to produce type 1 (IFN-γ and TNF-α), type 2 (IL-5 plus IL-13) and type 3 (IL-17A and IL-22) cytokines, we stimulated MNCs with 1 mg/mL PMA plus 0.5 mg/mL Ionomycin, in the presence of BrefeldinA (Sigma-Aldrich), for 3 h and stained them for intracellular cytokine production. Following this direct ex vivo shortterm activation, ILCs of AML patients were dramatically impaired in their production of IFN-γ, TNF-α as well as type 2 cytokines (eg IL-5/IL-13) as compared to ILCs of healthy subjects, while we found no difference in type 3 cytokines (eg IL-17A and IL-22)(Figure 2A and B). Moreover, in order to assess the role the ILCs might have in vivo, we used fresh PB of 4 newly diagnosed AML patients and, by adapting a method recently described, 11 we isolated AML blasts from MNCs using a monoclonal antibody cocktail against CD14 and CD33 (Miltenyi). Then, we depleted the CD14–CD33–MNCs’ flow through fraction from CD4+ and CD8+ cells so as to obtain MNCs that are both depleted of leukemic cells, and enriched in ILCs. It was seen that ILC-enriched MNCs, of both healthy donors and leukemic patients, incubated with target cells (ie blasts) are able to produce cytokines in response to respectively allogeneic or autologous blasts (Figure 2C and D). Even if we cannot exclude the involvement of alloreactivity in the recognition of leukemic blasts by ILCs from healthy donors, our data suggest that ILCs from AML patients could be impaired in the production of IFN-γ or type 2 cytokines in comparison to ILCs from healthy donors. Of note, the capacity of ILCs from AML patients to produce TNF-α in response to autologous blasts suggest that ILCs might be able to control, through the TNF-α/TNF-R1 pathway, the leukemic cell growth/survival. 12

Finally, we compared the percentage of total circulating ILCs in AML patients (n= 9) responding to standard chemotherapy (2 cycles of daunorubicin/cytarabine) with that of patients at diagnosis and of healthy donors [F (2, 82)= 56.26; P< 0.0001]. The percentage of total ILCs is completely recovered in responding patients (Figure 3A). In addition, we asked if also ILC1 and ILC3 NCR+ cells, the subsets dysregulated at diagnosis, were recovered in responding patients. While ILC1 were still increased in comparison to healthy donors, ILC3 NCR+ cells were restored to normal levels [(F (2, 82 …