Healthy donors were enrolled at Bambino Gesù Children’s Hospital, Rome, Italy. G-CSF mobilized donors received subcutaneous administration of G-CSF for five days (until apheresis) at the dose of 10⁻ ₁₂ μg/kg/day. Peripheral blood mononuclear cells (PBMCs) were obtained from G-CSF mobilized healthy donors (n = 12) and non-mobilized healthy donors (HD; n = 9). PBMCs were obtained after density gradient centrifugation (Ficoll-Lympholyte, Cederlane) as described before. Both G-CSF mobilized donors and HD gave their informed consent to participate in this study, which was approved by the Bambino Gesù Children’s Hospital (Rome, Italy) ethics committees (Prot. n. 1724/2018) and was conducted in accordance with the tenets of the Declaration of Helsinki.

NK cells and PMN-MDSCs were isolated from PBMCs of mobilized and non-mobilized donors using NK isolation kit and CD66b⁺ microbeads (purity >98%, data not shown) following manufacturer’s instruction (Miltenyi Biotec, Bergisch Gladbach, Germany). Before starting any experiment, we determined the purity of isolated cells by flow cytometry using anti-CD3-APC, anti-CD19-ECD, and CD56-PC7 (Beckman Coulter, Brea, CA) for NK cells. PMN-MDSCs were labeled with anti-CD3-AF700, anti-CD19-AF700, anti-CD11b-FITC, anti-CD33-PC7, anti-HLA-DR-PE, anti-CD14-ECD, anti-CD45-KrOr, and anti-CD66b-APC desiccated in the Duraclone custom design platform (Beckman Coulter, Brea, CA) adding anti-CD56-BV650 (BioLegend, San Diego, CA) and following manufacturer’s instruction. After the staining procedures, cells were acquired at Cytoflex LX and analyzed with Cytexpert software (v2.4, Beckman Coulter, Brea, CA). Freshly isolated NK cells were immediately used for functional studies and gene expression evaluation.

Cell cytotoxicity assays were performed using as target NALM-18 cell line (childhood B-cell acute lymphoblastic leukemia) or K562 cell line (erythroleukemia) and as effector cells (mobilized or not) NK cells at different Effector/Target (E/T) cell ratios. Killed cells were evaluated after 4 hours. At the end of the co-culture, the assay was stopped by chilling cells on ice, and Propidium Iodide (PI) was added to each sample immediately before acquisition in order to identify the percentage of target cell lysis, as previously described (29). For each set of experiment, all the acquisitions (5,000 target cells/sample) were performed within 20 min. Statistical analysis was performed using GraphPad Prism.

PMN-MDSC apoptosis was examined using Annexin V-FITC apoptosis detection kit (BD biosciences, Franklin Lakes, NJ, USA). PMN-MDSCs derived from non- and post-mobilized donors were separated and cultured in 24-well plates at a density of 1 × 10⁶ cells/well. Analysis of cell viability and apoptotic status was evaluated at different time points (from 1 to 6 days). PMN-MDSCs were incubated with 5 μl of Annexin V-FITC and 2 μl of PI at room temperature for 15 min. The PMN-MDSC apoptotic rate was measured using a flow cytometer (Cytoflex LX, Beckman Coulter, Brea, CA, USA). Chemotaxis of NK cells was measured by migration through a polycarbonate filter with an 8.0 μm pore size in 24-well trans-well chambers (Sarstedt, Nümbrecht, Germany). The assay medium consisted of Roswell Park Memorial Institute (RPMI) supplemented with 10% fetal bovine serum (FBS) derived from purified PMN-MDSCs non-mobilized and post-mobilization after 24 h of culture. 500 μl of assay medium was added to the lower chamber in the presence or not of CCL4 blocking antibodies (1.2 μg/ml; R&D Systems, Minneapolis, MN, USA). 500 μl of assay medium was used as a control for spontaneous migration. Then, 1 × 10⁵ NK cells were seeded to the upper chamber. After 1 hour of incubation at 37°C, NK cells that migrated to the bottom chamber were harvested and counted by flow cytometry (absolute count).

Total RNA extraction from purified NK cells and PMN-MDSCs was performed with miRNeasy mini kit, combined with on-column DNase I treatment, following the manufacturer’s protocol (Qiagen GmbH, Hilden, Germany). RNA concentration and purity were evaluated by spectrophotometric analysis (Nanodrop 2000; Thermo Fisher Scientific, Wilmington, DE). For gene expression analysis in NK cells and PMN-MDSCs, total RNA was reverse transcribed with random primers using Super Script IV first-strand synthesis system following manufacturer’s instructions (Thermo Fisher Scientific, Wilmington, DE, U.S.A.). Real Time PCR on NK cells was performed by 384-wells TaqMan array microfluidic cards with a custom configuration for the detection of selected genes implicated in NK cell biology (Thermo Fisher Scientific, Wilmington, DE, U.S.A.). Briefly, 200 ng of cDNAs for each sample was mixed with an isovolume of TaqMan Advanced Master Mix 2× and loaded in 384-wells cards (100 ng/channel). Real time PCR on PMN-MDSCs was performed in 20 µl of total volume with TaqMan™ Fast Advanced Master Mix (Applied Biosystems, Foster City, CA, U.S.A). The following TaqMan™ Gene Expression assays were used: MS4A4A (Hs00254780_m1), CLEC7A (Hs01902549_s1), CD177 (Hs00360669_m1), CCL4 (Hs99999148_m1), 18S (Hs99999901_s1). Reactions were carried out on a QuantStudio 12 Flex instrument using thermal PCR cycling conditions suggested by the manufacturer (Applied Biosystems, Foster City, CA, U.S.A). Data analysis was performed on Thermo Fisher Cloud with Design and Analysis New qPCR application (Thermo Fisher Scientific, Wilmington, DE, U.S.A.). PCR array data were analyzed with relative threshold algorithm and normalized using the mean of ACTB and HPRT expression, used as reference genes. Real time PCR on PMN-MDSC samples was analyzed with baseline threshold algorithm and 2^−ΔCt method. Data were normalized with 18S housekeeping gene.

The samples of non-mobilized (n = 5) and mobilized (n = 8) PMN-MDSCs from which we obtained enough amount of high quality total RNA were used as start material for the microarray gene expression analysis. Biotin labeled RNA was generated using the GeneChip 3′ IVT Plus labeling kit (Thermo Fisher Scientific, Wilmington, DE) according to the manufacturer’s instructions. The Biotin labeled RNA was subsequently hybridized to a HT U133 plus 16-array plate (Thermo Fisher Scientific, Wilmington, DE). The hybridization, wash, staining, and scanning procedures were done in a GeneTitan™ Instrument according to the manufacturer’s protocol (cat#: 00-0373; Thermo Fisher Scientific, Wilmington, DE). For labeling, hybridization, washing, staining, and scanning procedures we took advantage of Eurofins Genomics microarray service (Eurofins Genomics; Ebersberg, Germany).

The cel file output was used as input in the Partek Genomics Suite Software (Partek, St. Louis, MO) for generation of Robust Multi-Array Average (RMA) normalized data.

Data in Figures 1 , 2A were expressed as mean ± standard deviation (SD) or ± standard error of the mean (SEM) as indicated. For RNA yield evaluation, statistical significance was calculated using paired Student’s t-test. A p value ≤0.05 was considered statistically significant. DEGs between non-mobilized and mobilized PMN-MDSC groups were calculated by ANOVA one-way test. Lists of DEGs were filtered applying fold change (FC) >2 or <-2 (FC>10 or <-10 in Figure 3A ) and a p-value ≤0.05 with Benjamini–Hochberg (1995) False Discovery Rate (FDR) correction criteria. PCA, ANOVA test, unsupervised hierarchical clustering in Figure 3A , and box plots in Supplementary Figures 2A, B , were generated by Partek Genomics Suite (Partek, St. Louis, MO). Enrichment of GO BP terms was performed by GOrilla tool (30) applying a p-value threshold of 10-3, calculated according to the mHG or HG model. Graphical summarization of enriched GO BPs in a semantic treemap was performed by REVIGO allowing medium similarity and applying the simRel score (31). Two-way Anova was used for Figure 4A , Mann–Whitney test was used for Figures 3B , 4B, C .

As previously shown, the HSC mobilization regimen induces accumulation of PMN-MDSCs in the PB of G-CSF-mobilized donors. Moreover, these cells sharply inhibited the effector function of autologous and allogeneic NK cells in vitro (28). Based on these results, we verified whether the cytolytic activity of NK cells, isolated from the PB of mobilized donors, was affected by the concomitant presence of high numbers of PMN-MDSCs. Fresh NK cells were isolated from apheresis (post-mobilization) or from PB of unrelated HD, and their ability to kill NALM-18 tumor target cells was assessed. As shown in Figure 1A , NK cells isolated from mobilized donors showed a lower cytolytic activity as compared to HD NK cells. As previously demonstrated, PMN-MDSCs are capable, after 48 h of co-culture, to strongly impair the cytolytic activity of NK cells. In order to clearly demonstrate the impact of PMN-MDSCs on NK cell function, we further removed PMN-MDSCs after 48 h from NK/MDSC co-cultures and tested the cytolytic activity of these NK cells after additional 24 h of culture (referred as NK/MDSC 24 h alone; Supplementary Figure 1 ). The inhibitory effect of PMN-MDSCs was reversible, since the cytolytic activity of NK cells could be restored by removing PMN-MDSCs from co-cultures ( Supplementary Figure 1 ).

To support the functional data, we further investigated whether NK cells of mobilized donors were modified in their gene expression profiles. Thus, we analyzed by PCR array the expression of a wide panel of genes regulating cell function in NK cells purified from mobilized and non-mobilized donors. In particular, we evaluated genes considered important in the control of immune effector function of NK cells. As shown in Figure 1B , in comparison to NK cells from HD, “mobilized” NK cells revealed a decrease in the expression of some genes encoding for surface receptors involved in anti-tumor activity including NKp44 and CD94, or signal transduction molecules CD3ζ, ZAP70, DAP10, DAP12, or STAT family members, or effector molecules such as IFN-γ, TNF-α and GZMB. Interestingly, in NK cells from mobilized donors, we also found a higher expression of the CD9-encoding gene ( Figure 1B ). CD9 up-regulation was reported in NK cells converted into a pro-angiogenic, non-cytotoxic state upon exposure to TGFβ that, in turn, can be released by PMN-MDSCs as well (32, 33). Taken together, these results suggest that the effector function and the activation status of NK cells in the graft may be influenced by the presence of PMN-MDSCs.

Next, we verified whether the mobilization regimen induced changes in the molecular properties of PMN-MDSCs present in mobilized donors. First, we compared the frequency of PMN-MDSCs in G-CSF mobilized and non-mobilized donors. In G-CSF mobilized donors PMN-MDSCs were ≈20% while they were <5% non-mobilized donors (28). Their percentages increased to ≈40% in the collection bag (due to their enrichment related to the B- and αβ T-cell depletion; data not shown). Notably, RNA yield from freshly isolated PMN-MDSC in the donor graft was significantly higher than that of the same cells isolated from the same pre-mobilized donors (∼three-fold) ( Figure 2A ). The increased RNA content suggested an activated metabolic state in PMN-MDSCs of mobilized grafts. Thus, we investigated whether these cells displayed transcriptional features different from non-mobilized PB PMN-MDSCs. To this purpose, we performed a global gene expression profiling by microarray to compare eight samples of PMN-MDSCs isolated from PB of mobilized donors with a control group represented by five PMN-MDSC samples derived from non-mobilized HD. In the microarray, three out five of the control samples were derived, before mobilization, from the same donors analyzed in the mobilized group (samples c#1, c#,2 and c#3 versus m#1, m#2, and m#3 respectively; Figure 2B ). We used a three-dimensional Principal Component Analysis (PCA) plot to summarize the microarray data. This exploratory analysis showed that all the mobilized PMN-MDSCs samples finely clustered together and were clearly distinguished from non-mobilized PMN-MDSCs samples, indicating that, indeed, the two groups are characterized by distinct transcriptional profiles ( Figure 2B ).

We further investigated the differentially expressed genes (DEGs) in mobilized or non-mobilized PMN-MDSCs. To this end, we performed an Analysis of Variance (ANOVA) test, imposing a p-value ≤0.05 and a fold change (FC) ≥|2|. We found 4,625 differentially expressed probe sets displaying robust and statistically significant variations between mobilized and non-mobilized PMN-MDSCs. Among them, 2,589 probe sets were up-regulated and 2,036 down-regulated in mobilized versus non-mobilized PMN-MDSCs ( Figure 2C and Supplementary Table 1 ). Hierarchical clustering based on these significantly deregulated genes also confirmed a clear discrimination between mobilized and non-mobilized PMN-MDSC samples (data not shown). It should be noted that some important genes characterizing the PMN-MDSCs subset such as LOX1, S100A9, GAS6, TGFβ1 and IL-10 were expressed in both groups with no significantly different expression level, suggesting that mobilized PMN-MDSCs retain some features of typical PMN-MDSCs ( Supplementary Figure 2A ). Remarkably, in mobilized PMN-MDSCs we detected a differential expression of some relevant genes associated with cell proliferation and immunosuppressive function. In particular, MS4A4A (↑), MKI67 (↑), ARG1 (↑), CEACAM1 (↑), FKB5 (↑) genes were up-regulated whereas CCL4 (↓), IL7R (↓) were down-regulated in mobilized PMN-MDSCs ( Supplementary Table 1 and Figure 3A ).

A heat-map showing the most modulated genes in mobilized PMN-MDSCs is represented in Figure 3A . In this analysis, we filtered the list to DEGs with FC ≥ |10| to narrow down to the most deregulated genes. One of the most up-regulated genes in mobilized PMN-MDSCs was CD177 (FC = +27.4) ( Supplementary Figure 2B ), a surface membrane glycoprotein expressed in neutrophils and neutrophil precursors (34). Consistent with the wide up-regulation of other genes involved in MDSC proliferation and activation, this gene is involved in neutrophil proliferation and was found to be up-regulated in human MDSCs upon inflammation (35, 36). Furthermore, the tetraspan molecule MS4A4A was one of the most up-regulated genes in mobilized MDSCs (FC = +24). It has recently been reported that MS4A4A expression is restricted to human mononuclear phagocytes, where MS4A4A co-localizes with the β-glucan receptor dectin-1 (CLEC7A), mediates NK cell activation and exert a control of metastases in murine models (37). Our data indicate that Dectin-1/CLEC7A is down-regulated (FC= –11) in mobilized PMN-MDSCs. Since Dectin-1 signaling is crucial for mediating cell-to-cell contact and subsequent NK cell activation, the up-regulation of MS4A4A might have a different role besides inducing NK cell activation via MS4A4A-CLEC7A. A sharp increase of MS4A4A expression, usually associated with cells of the monocytic lineage, suggests that PMN-MDSCs may acquire unique molecular characteristics upon G-CSF induced mobilization. Conversely, one of the most down-regulated genes was CCL4, encoding for an inflammatory chemokine functioning as chemoattractant for different leukocytes, including NK cells ( Supplementary Figure 2B ). The above mentioned genes were selected for PCR validation on the same samples used in the microarray. Real time PCR analysis confirmed a striking differential expression of all the genes tested, confirming the reliability of the microarray approach ( Figure 3B ).

Transcriptome data indicated that mobilized PMN-MDSCs are characterized by an activated phenotype. Thus, we asked whether their distinct gene expression profile was associated with an altered cell survival. To this purpose, mobilized and non-mobilized PMN-MDSCs were cultured in vitro, and cell viability was evaluated for 6 days. Whereas non-mobilized PMN-MDSCs underwent rapid cell death upon in vitro culturing, mobilized PMN-MDSCs displayed a significantly prolonged survival ( Figure 4A ). Of note, after 24 h of in vitro culture, the large majority of mobilized PMN-MDSCs were represented by live cells, whereas their non-mobilized counterpart already showed an extensive fraction of apoptotic cells ( Figure 4B ).

Transcriptional profiles also suggested that mobilized PMN-MDSCs could have different immune properties as compared to the non-mobilized ones. Based on down-regulation of several chemokines including CCL4 (see Figure 3A and Supplementary Table 1 ), we tested whether mobilized PMN-MDSCs could exhibit different chemo-attractive properties on NK cells. Supernatants of non-mobilized and mobilized PMN-MDSC cultures displayed significant differences in the capability to attract NK cells in trans-well chambers system. In particular, supernatants of mobilized PMN-MDSCs exhibited a reduced ability to attract NK cells as compared to non-mobilized ones ( Figure 4C ). The use of a blocking antibody against CCL4 chemokine caused a slight decrease of NK cell migration in non-mobilized supernatants, indicating that CCL4 chemokine is involved, at least in part, in this process ( Figure 4C ).

In order to deepen the biological significance of DEGs identified in PMN-MDSCs of mobilized donors, we performed an enrichment analysis of biological processes (BPs) from Gene Ontology (GO) database. We separately analyzed the list of up-regulated (FC ≥2; p value ≤0.05) and down-regulated (FC ≤-2; p value ≤0.05) genes in mobilized MDSCs to find emerging BPs in these groups. A list of BPs significantly over-represented is shown in Figure 5 , whereas the complete list of enriched BPs in up-regulated and down-regulated genes is reported in Supplementary Tables 2 and 3 . This analysis clearly highlighted a striking predominance of BP associated with cell proliferation in the up-regulated genes in G-CSF mobilized PMN-MDSCs, including cell cycle process (GO:0022402), chromosome organization (GO:0051276) and segregation (GO:0007059). Another important group of BPs regards the immune defense compartment, in particular, the myeloid cell activation (GO:0002274) and other processes specifically associated with granulocytes, such as neutrophil degranulation (GO:0043312), activation (GO:0042119) and extravasation (GO:0072672) ( Figure 5A ). Conversely, down-regulated genes in mobilized PMN-MDSCs were enriched in some BPs associated with macromolecule metabolism, e.g. transcription DNA-templated (GO:0006351) and nucleic acid metabolic process (GO:0090304). Other significantly enriched BPs were associated to covalent chromatin modification (GO:0016569) and histone modification (GO:0016570) ( Figure 5B ). A summarization of the GO BPs significantly enriched in up-regulated and down-regulated gene lists was performed by REVIGO (31) to graphically represent in a semantic tree-map all the BPs identified ( Figure 5 ). Collectively, these data suggest that the mobilization regimen with G-CSF induced a transcriptional program promoting cell proliferation and activation of myeloid cells.