Introduction

Myelodysplastic syndromes (MDS) are a group of clonal premalignant hematological diseases characterized by ineffective hematopoiesis, progressive peripheral blood cytopenias, increased risk of developing acute myeloid leukemia (AML), and poor overall survival1. Classification and risk stratification systems are constantly under revision for a better estimation of prognosis in MDS patients by combining clinical, cytogenetic, and molecular biology features2. Thus, the discovery of new biomarkers, like molecular data from targeted sequencing, has allowed classification of MDS patients based on genetic signatures, such as SF3B1 mutations in MDS with ring sideroblasts, and, for these reasons, these markers have been incorporated into prognostic scoring systems3. However, genetic alterations frequently found in MDS can be present also in other hematological disorders and in healthy individuals as clonal hematopoiesis is commonly seen with ageing4. Therefore, additional pathogenetic mechanisms are required for a dysplastic hemopoiesis, and immune dysregulation can initiate or support dyspoiesis, also in other non-malignant conditions known as idiopathic cytopenia of unknown significance (ICUS), including Chronic Idiopathic Neutropenia (CIN)5,6. These patients have persistent unexplained mild cytopenia(s) lasting for at least 4 months, without morphologic or cytogenetic evidence of MDS (or other underlying disorders), mild (< 10%) marrow dysplasia, and marrow blasts < 5%7.

Neutrophils, the most abundant circulating leukocytes, are mainly involved in pathogen clearance through three mechanisms: phagocytosis, degranulation, and release of neutrophil extracellular traps (NETs). Neutrophils are “pathogen killers” and can promote or suppress tumor growth and metastatic processes. Indeed, neutrophils can directly kill neoplastic cells and present tumor antigens to cytotoxic T lymphocytes (CTLs), or, conversely, they can impair CD8+ T and Natural Killer (NK) cells activity promoting tumor immune escape8. These different immunological functions mirror a great heterogenicity of neutrophil maturation and activation status9. Two circulating populations can be differentiated and separated by gradient centrifugation: low-density granulocytes (LDGs) and normal-density neutrophils (NDNs), based on their different stratification ability after Ficoll-Paque gradient separation10. Within each granulocyte population, there are several subsets with different biological activities, such as inflammatory and suppressive LDGs, and immature and proinflammatory NDNs. Moreover, inflammatory LDGs comprise of immature cells that are produced under emergency conditions, and pro-inflammatory LDGs, that are mature neutrophils with pro-inflammatory activities; conversely, suppressive LDGs include immunosuppressive granulocytes that reduce T cell proliferation and interferon(IFN)-γ production, and exhausted neutrophils11. Circulating LDGs can spontaneously release NETs12 and produce large amount of tumor necrosis factor (TNF)α, IFN-γ, and type I IFN cytokines, which are a hallmark of autoimmune disorders13. Several observations show the involvement of LDGs in immune dysregulation during autoimmune disorders with an activated phenotype13; while NDNs might exert immunosuppressive activities in an arginase-dependent manner8. LDGs might also contribute to endothelial cell dysfunction and vascular damage in autoimmune vasculitis, and can be hypo-responsive to anti-myeloperoxidase (MPO) antibodies despite their high surface expression14. MPO cytoplasmatic expression is associated with degranulation in neutrophils, and impairment in this function is a hallmark of MDS15. A recent study has described increased expression of neutrophil MPO in MDS which might be a highly specific neutrophil marker for differential diagnosis of hematologic malignancies15. Moreover, MPO is one of the major components of NETs together with neutrophil elastase, calprotectin, defensins, and DNA12. Immature NDNs are composed by immature cells produced in emergency conditions, and resting neutrophils, that are fully mature cells, while pro-inflammatory NDNs comprise cells that are primed for ROS, granules, and NETs after stimulation, and activated neutrophils, that are fully activated cells with release of high amount of ROS, granules, and NETs16. Previous studies have identified LDGs as a source of autoantigens in autoimmune diseases, such as systemic lupus or rheumatoid arthritis, and this population could represent up to 2/3 of circulating total granulocytes with heterogeneous subset distribution in each disorder9,14,17,18. For example, mature LDGs are the majority of circulating low-density neutrophils in ANCA vasculitis, while in systemic lupus pro-inflammatory LDGs secreting type I INFs are augmented14.

Based on recent findings in autoimmune disorders showing the involvement of pro-inflammatory LDGs in activation of type 1 helper responses and based on the predominant role of CTLs and type I immune responses in hematological malignancies19, we hypothesize that LDGs and NDNs might play a key role in MDS and AML pathophysiology. To date, there are no studies investigating frequency, composition, and functions of NDNs and LDGs in AML, MDS, and ICUS. We supposed that BM releases both mature high-density neutrophils and LDGs in MDS, especially in higher-risk diseases, in response to dyspoiesis or augmented destruction. However, under these stressful conditions, LDGs might be more prone to spontaneously produce type I IFN cytokines, thus sustaining pro-inflammatory environment and favoring clonal evolution.

Patients & methods

Patients

Whole peripheral blood (PB) specimens were collected in ethylenediaminetetraacetic acid (EDTA) tubes for immunophenotyping. Informed consent was obtained from all subjects and/or their legal guardian(s) in accordance with the Declaration of Helsinki and protocols approved by our local Ethics Committee ‘‘Campania Sud’’ (Brusciano, Naples, Italy; prot./SCCE n. 24988). A total of 102 consecutive subjects were enrolled from January 2020 to December 2024 who arrived at our observation for peripheral blood cytopenia(s) and were screened for hematological malignancies according to current guidelines at the Hematology and Transplant Center, University Hospital “San Giovanni di Dio e Ruggi d’Aragona” of Salerno, Italy. After work-up, patients received a diagnosis of AML (N = 12), MDS (N = 39), ICUS (N = 10), or other hematological conditions (others; idiopathic myelofibrosis, N = 5; immune thrombocytopenia, N = 5; chronic leukemias, N = 2; acquired aplastic anemia, N = 1; autoimmune hemolytic anemia, N = 1; blastic plasmacytoid dendritic cell leukemia, N = 1; monoclonal gammopathy of undetermined significance, N = 1; acute lymphoblastic leukemia, N = 4; leukocytosis of unknown origin, N = 3; and VEXAS syndrome, N = 1), and chemotherapy as per international guidelines. For AML patients, risk stratification was calculated according to 2017 and 2022 ELN system, while for MDS subjects to the Revised International Prognostic Scoring System (IPSS-R)20,21,22. International Working Group (IWG) consensus criteria were used to determine patients’ treatment response23. A group of healthy subjects was used as a control, after ruled out the presence of cytopenia(s) or other inflammatory/pathological conditions as per medical screening. Clinical characteristics at enrollment are summarized in Supplementary Table 1. Among MDS patients (N = 39; mean age, 69.2 years old; range, 27–89; M/F, 20/19), 24 of them (61.5%) received a diagnosis of MDS-LB, and 13 of MDS with increased blasts (MDS-IB1, N = 6; and MDS-IB2, N = 7). Four subjects were treated with hypomethylating agents, one with chemotherapy, and two with luspatercept or eltrombopag, while the remaining patients received supportive therapies, such as erythropoietin or red blood cell transfusion. All methods were carried out in accordance with relevant guidelines and regulations, and in accordance with protocols approved by local ethic committee (Ethics Committee “Campania Sud”, Brusciano, Naples, Italy; prot./SCCE n. 24988).

Flow cytometry

For neutrophil phenotype characterization, 50 µL of fresh EDTA whole PB was stained with the following antibodies according to the manufacturers’ instructions: CD56-ECD, CD45-Krome Orange, CD34-APC AlexaFluor700, CD19-PC5.5, CD11b-PC7, CD3-ECD, CD33-APC, HLA-DR-FITC, CD15-PE, CD14-APC AlexaFluor750, and CD16-Pacific Blue (all from Beckman Coulter, Milan, Italy). Blast phenotype was further studied for CD19-PC5.5, CD20-Pacific Blue, CD34-APC AlexaFluor700, CD56-ECD, CD5-PC7, CD117-PE, CD33-APC, CD16-Pacific Blue, CD11b-PC7, CD36-FITC, CD13-PC5.5, HLA-DR-FITC, CD64-ECD, CD4-PE, CD5-PC7, CD7-PC7, CD14-APC AlexaFluor750, CD10-PC7, CD15-PE, CD11a-FITC, CD11c-PE, CD45RA-FITC, CD45RO-ECD, CD61-FITC, CD42b-PE, TdT-FITC, and MPO-PE expression (all antibodies were from Beckman Coulter), as previously described2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24. Samples were incubated for 20 min at room temperature, red cells lysed using IO Test Lysing Solution (Beckman Coulter), cells were washed twice with phosphate-buffered saline (PBS) (Beckman Coulter), and then resuspended in 500 µL PBS for acquisition. Samples were acquired using a Navios/EX or a DxFlex cytometer (Beckman Coulter). Instrument daily quality control was performed using Flow-Check Pro Fluorospheres (Beckman Coulter). The Laboratory agrees to the UK NEQAS for Leucocyte Immunophenotyping international program for external quality controls. Compensation was monthly checked by a Beckman Coulter’s Specialist using Flow-Set and ClearLLab Compensation Beads in conjunction with the ClearLLab Compensation Kit (Beckman Coulter). Samples were acquired using the same PMT voltages, and at least 200,000 events were recorded. Post-acquisition analysis was carried out using Navios EX Software v2.0, or Kaluza Analysis Flow Cytometry Software v2.1.1 (Beckman Coulter).

Gating strategy

Gating strategy and representative dot plots of neutrophil subsets in each disease entity are represented in Supplementary Fig. 1 and Fig. 1, respectively. Cell populations were first identified based on linear parameter (forward scatter area, FSC-A) and CD45 expression, cells were gated, and CD19 and CD56/CD3 expression was investigated. On CD19-CD56-CD3- cells, DR-CD34- population was further identified and studied for CD33 and CD15 expression. On CD33+ CD15- neutrophils (normal density granulocytes, NDGs), maturation curve was investigated by CD16 and CD11b expression, and immature CD16-CD11b-, intermediate CD16-/+CD11b+, and mature CD16+CD11b+ neutrophils were identified (Fig. 1). On CD33+ cells, CD14 and CD15 expression was further studied, and CD15+CD14- LDGs were gated. Maturation was investigated using CD16 vs. CD11b expression, and CD15 vs. CD16. Immature CD16-CD11b-, intermediate CD16-/+CD11b+, and mature CD16+CD11b+ LDGs were identified, as well as immature CD15+CD16-, intermediate CD15+CD16dim, and mature CD15+CD16+ LDGs were gated. Lymphocytes were further studied for T (CD3 or CD5 or CD7, CD4, and CD8), B (CD19), and NK cell (CD56 and CD16) markers. CD33, CD14, CD11b, CD16, and CD56 expression was investigated on monocytes (Supplementary Fig. 2).

Fig. 1
figure 1

Representative neutrophil subset distribution in health and diseases. Representative plots for neutrophil subsets are reported for (A) healthy subjects, (B) idiopathic cytopenia of unknown significance (ICUS), (C) acute myeloid leukemia (AML), (D) low-risk myelodysplastic syndrome (MDS), and (E) high-risk MDS. Low-density granulocytes (LDGs in purple) and normal-density neutrophils (NDNs in dark blue) are gated based on CD33 and CD15 expression, and can be also identified based on linear parameter (side scatter area, SSC-A) and CD45 expression, as LDGs have a lower SSC-A and increased CD45 expression. On NDNs, maturation curve is assessed using CD11b and CD16, and immature CD11b-CD16-, intermediate CD11b+/-CD16-, and mature CD11b+CD16+ cells are identified. On LDGs, maturation curve is investigated based on CD16 expression on CD15+ granulocytes, and immature CD16-CD15+, intermediate CD16-/+CD15+, and mature CD16+CD15+ cells are gated.

Full size image

Plasma elastase measurement

A total of 49 specimens were employed for plasma elastase levels assessment by ELISA assay using the Human Neutrophil Elastase ELISA Kit for quantitative measurement of Human Neutrophil Elastase in EDTA plasma samples (ab270204; Abcam, Cambridge, UK), following manufacturers’ instructions. In details, plasma was obtained from 14 AML patients (M/F, 9/5; mean age, 67 years old; range, 44–91 years), 15 MDS (M/F, 8/7; mean age, 75 years old; range, 59–89 years), and 18 healthy subjects (M/F, 10/8; mean age, 60 years old; range, 30–82 years), and specimens were diluted 1:50 according to manufacturer’s instructions for elastase determination. A microplate reader (Infinite F200 PRO, Tecan Group Ltd., SW) was used for OD reading at 450 nm. All samples and calibrators were run in duplicate.

Statistical analysis

Data were analyzed using Prism (v.10.2.0; GraphPad software, La Jolla, CA, USA). For flow cytometry data, populations were reported as percent of positive cells. Parametric or non-parametric data distribution was first tested using the D’Agostino & Pearson, Anderson-Darling, Shapiro-Wilk, and Kolmogorov-Smirnov tests. Based on normality analysis results, Pearson or Spearman analysis was employed for studying correlations, or unpaired two-tailed t- or non-parametric Mann Whitney tests for two group comparison and Kruskal-Wallis test for three-group comparison were performed. Multiple linear regression was carried out for multivariate analysis. ROC and area under the curve (AUC) analysis was also performed to determine sensitivity and specificity of each biomarker. A post-hoc analysis was performed to calculate the power of our investigation, considering the number of patients with AML, MDS, ICUS, and healthy subjects (N = 78), and an alpha error of 0.05, and resulting in a power of 96%. RStudio (v.2022.07.1 + 554; RStudio software, Boston, MA, US) was employed for data visualization with correlograms using corrplot and PerformanceAnalytics packages. A P < 0.05 was considered statistically significant.

Results

Granulocyte terminal differentiation is impaired in AML and MDS

Immunophenotyping of granulocyte subsets was performed at diagnosis in 10 ICUS, five AML, and 28 MDS patients. Clinical characteristics at enrollment are reported in Supplementary Tables 1 and circulating granulocyte subset frequencies are summarized in Table 1. A group of 17 healthy subjects was studied to define normal ranges of granulocyte subpopulations. Overall, MDS and AML patients did not show clinically significant variations in frequencies of NDNs and their subsets compared to healthy individuals (all P > 0.05, except for total NDNs in MDS, median with IQR, 0 (3.2-0) vs. 0 (0–0.0%); P = 0.0433) (Fig. 2). Conversely, ICUS subjects had an increased proportion of immature (median with IQR, 25.3%, 0-50.3% vs. 0%, ICUS vs. healthy subjects; P = 0.0189), intermediate (median with IQR, 6%, 0-41.3% vs. 0%, ICUS vs. healthy subjects; P = 0.0119), and mature NDNs (median with IQR, 16.2%, 0-36.9% vs. 0%, ICUS vs. healthy subjects; P = 0.0015) compared to healthy subjects. For total LDGs, AML patients had significantly lower circulating frequency of this subset compared to healthy individuals (median with IQR, 42.8% [86.5–25.2%] vs. 90.1% [92.4–89.3%], AML vs. healthy subjects; P = 0.0254). In particular, AML subjects had significantly lower circulating mature LDGs (median with IQR, 74.3%, 67.3–89.2%) compared to healthy (median with IQR, 98.9%, 99.5–94.2%; P = 0.0064) and ICUS subjects (median with IQR, 99.1%, 99.5–96%; P = 0.0042). Conversely, AML patients at diagnosis displayed significantly higher frequency of circulating immature LDGs (median with IQR, 12.1%, 18.4–1.1%) compared to healthy and ICUS subjects (median with IQR, 0.1% [1.6 − 0.04%] and 0.2% [0.9 − 0.02%], respectively; P = 0.0152 and P = 0.0067, respectively), as well as increased intermediate LDGs (median with IQR, 13.2%, 17.2–6.3%) compared to ICUS (median with IQR, 0.4%, 2.9 − 0.2%; P = 0.0125) and healthy subjects (median with IQR, 0.7%, 1.8 − 0.2%; P = 0.0049). MDS patients at diagnosis had increased immature (median with IQR, 1.9%, 8.4 − 0.1%) compared to ICUS (P = 0.0484) and higher intermediate LDGs (median with IQR, 2.5%, 13.6–1.1%) compared to ICUS and healthy subjects (P = 0.017 and P = 0.0023, respectively), while mature LDGs were decreased (median with IQR, 94.9%, 98.3–78.4%) compared to ICUS (P = 0.0085) and healthy subjects (P = 0.0105). No differences were observed in circulating monocyte frequency and their subset distribution between groups, except for non-classical monocytes, which were slightly decreased in AML patients compared to healthy subjects (median with IQR, 0% [0.2-0%) vs. 0.3% [1.2 − 0.03%]; P = 0.0457).

Table 1 Granulocyte subset distribution at diagnosis.
Full size table
Fig. 2
figure 2

Granulocyte and monocyte subset frequency distribution. Percentages of total normal density neutrophils (NDNs), low-density granulocytes (LDGs), monocytes, and their subsets were calculated for healthy controls (HC), acute myeloid leukemia (AML), myelodysplastic syndrome (MDS), and idiopathic cytopenia of unknown significance (ICUS) patients. Data are shown as median with IQR. Kruskal-Wallis test was performed. *P < 0.05; **P < 0.01.

Full size image

When MDS patients were divided in low- and high-risk disease, no differences were observed in NDN and monocyte subsets compared to healthy subjects (all P > 0.05); conversely, LDGs were decreased in low-risk MDS (median with IQR, 81.7%, 85.9–55.9%) compared to healthy controls (median with IQR, 90.1%, 92.4–79.3%; P = 0.0355), as well as mature LDGs (median with IQR, 94.4% [98.2–87%) vs. 98.9% [99.5–94.2%], low-risk MDS vs. healthy subjects; P = 0.0149) (Supplementary Fig. 3). Moreover, intermediate LDGs were higher in both low- (median with IQR, 2.6%, 11.1–1.2%) and high-risk MDS (median with IQR, 2.1%, 8.6 − 0.1%) compared to healthy controls (P = 0.0025 and P = 0.0186, respectively).

Granulocyte terminal differentiation remains impaired after therapy in AML and MDS

To investigate prognostic relevance of NDNs and LDGs in hematological malignancies, circulating granulocyte subset frequencies were compared between subjects at diagnosis, in complete (CR) or partial remission (PR), or in progressive/stable disease (PD/SD) (Fig. 3). In spite of responsiveness to therapies, patients in CR/PR still displayed increased frequencies of total NDNs compared to healthy subjects (median with IQR, 8.2%, 17.4-0%, P = 0.0095), as well as immature NDNs (median with IQR, 12.4%, 69.2-0%; P = 0.0133), intermediate (median with IQR, 4.4%, 70.8-0; P = 0.006), and mature NDNs (median with IQR, 0.5%, 14.5-0%; P = 0.0259) (Fig. 3). Immature NDNs (median with IQR, 5.9%, 87.3-0%) and mature NDNs (median with IQR, 4.3%, 40.5-0%) were also augmented in patients with PD/SD compared to healthy subjects (P = 0.0397 and P = 0.0186, respectively). Similarly, intermediate LDGs were significantly increased in CR/PR patients compared to healthy controls (median with IQR, 3.8% [16 − 1.4%] vs. 0.7% [1.8 − 0.2%]; P = 0.0114).

Fig. 3
figure 3

Granulocyte subset frequency distribution at diagnosis and in treated patients. Percentages of total normal density neutrophils (NDNs), low-density granulocytes (LDGs), and their subsets were calculated for acute myeloid leukemia and myelodysplastic syndrome patients at diagnosis or after therapies in responders (complete remission [CR] + partial remission [PR]) or non-responders (progressive disease [PD] + stable disease [SD]). Data are shown as median with IQR. Kruskal-Wallis test was performed. *P < 0.05; **P < 0.01.

Full size image

Correlations with clinical features

To confirm the potential diagnostic and prognostic role of granulocyte subset immunophenotyping in AML and MDS, a multivariate analysis was first performed by investigating the impact of clinical and flow cytometric parameters on the percentage of leukemic cells or circulating CD34+ hematopoietic stem cells in AML and MDS cohort (Supplementary Table 2). Age at diagnosis was associated with higher blast count (estimate, 0.7294; 95% confidential interval [CI], 0.1017 to 1.357; P = 0.0321), as well as a diagnosis of high-risk MDS (estimate, -47.37; 95%CI, -80.40 to -14.33; P = 0.0164). Moreover, immature NDNs (estimate, -0.4862; 95%CI, -0.9185 to -0.0539; P = 0.0354), mature NDNs (estimate, -0.2765; 95%CI, -0.5379 to -0.01520; P = 0.0425), and non-classical monocytes (estimate, 4.345; 95%CI, 0.0094 to 8.680; P = 0.0497) were significantly associated with circulating blasts. A ROC curve analysis was also performed, showing no significant association of NDN and monocyte subsets in discriminating AML and MDS (all P > 0.05). Conversely, immature (AUC, 0.8471; P = 0.0208), intermediate (AUC, 0.9059; P = 0.0069), and mature LDGs (AUC, 0.9176; P = 0.0054) could be specific immunophenotypic markers in AML (Fig. 4), and immature LDGs also for differential diagnosis between AML and MDS, as no significant results were found in MDS (AUC, 0.6403; P = 0.1235). Conversely, intermediate (AUC, 0.793; P = 0.0013) and mature LDGs (AUC, 0.7319; P = 0.0109) were also associated with MDS. When divided based on clinical risk, only mature LDGs could specifically identify low-risk MDS (AUC, 0.7647; P = 0.0108), while not high-risk disease (AUC, 0.6872; P = 0.0997). Immature (AUC, 0.7708; P = 0.0272), intermediate (AUC, 0.7745; P = 0.0236), and mature NDNs (AUC, 0.7614; P = 0.0311) were associated with responsiveness to therapies, as well as intermediate LDGs (AUC, 0.8072; P = 0.0113). Conversely, only total LDGs could identify patients in PD/SD (AUC, 0.902; P = 0.0041).

Fig. 4
figure 4

Correlations with clinical parameters. (A) Percent of circulating blasts or (B) hemoglobin levels (Hb) were correlated with frequencies of low-density granulocytes (LDGs) and their subsets, and/or monocytes. Spearman correlation test was carried out and confidential interval is also shown. A ROC curve analysis for investigating the diagnostic power of intermediate and mature LDGs in (C) myelodysplastic syndromes (MDS) and (D) acute myeloid leukemia (AML) was performed. AUC, area under the curve. A P < 0.05 was considered statistically significant.

Full size image

Next, granulocyte subset frequencies were correlated with clinical parameters, such as hemoglobin levels, absolute neutrophil count, platelet count, WT1 copy number, and percentage of circulating leukemic cells or hematopoietic stem cells. Total LDGs were negatively associated with circulating blasts (r = -0.3931; P = 0.0008), while intermediate LDGs positively (r = 0.2494; P = 0.0402) (Fig. 4). Conversely, intermediate LDGs were negatively related to hemoglobin levels (r = -0.3022; P = 0.0249). We confirmed that increased WT1 copy number were associated with higher blast counts (r = 0.7764; P = 0.0003), as previously reported24. Moreover, monocyte frequencies were negatively associated with circulating blasts (r = -0.2806 and P = 0.0345), as well as intermediate monocytes with WT1 copy number (r = -0.4695; P = 0.0238). Classical monocytes were inversely related to intermediate monocytes in healthy (r = -0.93; P < 0.01), while intermediate monocytes were positively correlated with non-classical monocyte frequencies in ICUS (r = 0.74) (P < 0.05) (Fig. 5 and Supplementary Fig. 4). Moreover, intermediate monocytes were negatively associated with immature LDGs in healthy controls (r = -0.60; P < 0.01), positively with immature NDNs in ICUS (r = 0.71; P < 0.05), and proportionally associated with total LDGs in AML (r = 0.55). Mature NDNs in AML (r = -0.63) were inversely related to circulating levels of hematopoietic precursors (P < 0.05). Non-classical monocytes were positively associated with intermediate LDGs in AML (r = 0.60; P < 0.05) and negatively in healthy subjects (r = -0.52; P < 0.05), and negatively with mature LDGs (r = -0.57) in AML. Classical monocytes were also related to total NDNs (r = 0.57) or their immature forms (r = 0.54) in AML. Total monocytes were negatively related to total LDGs in ICUS (r = -0.92; P < 0.01). Immature LDG forms negatively correlated with mature LDGs (AML, r = -0.74; MDS, r = -0.53; and healthy, r = -0.81; P < 0.01) and immature NDNs in MDS (r = 0.66; P < 0.01), as well as intermediate LDGs with immature NDNs (AML, r = 0.72; and MDS, r = 0.62; P < 0.05).

Fig. 5
figure 5

Correlations using circulating granulocyte subsets. Circulating normal density neutrophil (NDN) and low-density granulocyte (LDG) subsets were correlated, also with circulating blasts or CD34 + hematopoietic stem cells in acute myeloid leukemia (AML), myelodysplastic syndrome (MDS), idiopathic cytopenia of unknown significance (ICUS), and healthy subjects. Values range between − 1 (red) and + 1 (blue). Related P values are reported in Supplementar Fig. 4.

Full size image

Plasma elastase levels are decreased in MDS patients

Finally, to assess a functional correlation between circulating neutrophil subset distribution and NET-related circulating proteins, as a surrogate a neutrophil activity, plasma elastase levels were assessed in a cohort of AML and MDS patients, and results were compared to healthy subjects. Moreover, elastase levels were also correlated with circulating neutrophil and monocyte subsets identified by flow cytometry immunophenotyping (Fig. 6). Proteins frequently present in NETs are alarmins of S100A family including the heterodimer calprotectin, or proteinase 3 (PRTN3)25, and circulating S100A8 is increased in the plasma of MDS patients especially low-risk MDS26. By searching into reported aptamer-based proteomics analysis of plasma obtained from AA patients1, other NET-related proteins, such as PRTN3, lactoferrin (LTF), transketolase (TKT), myeloperoxidase (MPO, the main NET-related component), and elastase were significantly increased in the plasma of healthy individuals compared to acquired aplastic anemia, a benign hematological condition characterized by bone marrow hypocellularity and peripheral blood cytopenia(s)19. For this reason, we chose elastase as a functional neutrophil marker. Notably, circulating plasma levels were significantly decreased in MDS patients compared to healthy subjects (mean ± SD, 290.1 ± 141.1 ng/mL vs. 432.4 ± 115.8 ng/mL; P = 0.0056), and to AML patients (mean ± SD, 413 ± 104.2 ng/mL; P = 0.0280). Moreover, classical monocytes were negatively associated with plasma elastase levels (r = -0.6970; P = 0.0306), as well as total LDGs (r = -0.7939; P = 0.0088). Conversely, all NDN subsets were positively related to circulating plasma elastase levels (total NDNs, r = 0.8740 and P = 0.0016; immature NDNs, r = 0.8193 and P = 0.0063; intermediate NDNs, r = 0.8467 and P = 0.0036; and mature NDNs, r = 0.7920 and P = 0.0107). In addition, a ROC curve analysis was also carried out to investigate the diagnostic potential of circulating plasma elastase in MDS and AML (Fig. 5E-F), showing a significant association in MDS (AUC, 0.7680; P = 0.0068), while not in AML (AUC, 0.5516; P = 0.6214).

Fig. 6
figure 6

Plasma neutrophil elastase in acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS). (A) Neutrophil elastase levels were measured in AML and MDS patients and compared to a group of healthy controls (HC). ANOVA test was performed, and data are shown as mean ± SD. *P < 0.05; **P < 0.01. (B) Percent of circulating classical monocytes, low-density granulocyte (LDG) subsets, or normal density neutrophil (NDN) subpopulations were correlated with plasma elastase levels. Spearman correlation test was conducted, and confidential interval is also shown. A ROC curve analysis for investigating the diagnostic power of plasma neutrophil elastase in (E) myelodysplastic syndromes (MDS) and (F) acute myeloid leukemia (AML) was performed. AUC, area under the curve. A P < 0.05 was considered statistically significant.

Full size image

Discussion

AML and MDS are heterogeneous hematological conditions characterized by ineffective hematopoiesis, differentiation block, and peripheral cytopenia(s) caused by leukemic cell proliferation and/or impaired maturation due to epigenetic, genomic, and phenotypic modifications27. The separation line between these two clinical entities is blurred and often overlaps, as the cut-off of 20% of blasts to define an AML or an MDS is more a classification exercise rather than a precise border. In addition, molecular alterations are frequently shared between these two disorders, and MDS can ultimately evolves into an AML28. This clinical and molecular overlap makes differential diagnosis and risk stratification of challenging, especially between high-risk MDS and AML. Therefore, new diagnostic and prognostic biomarkers are essential to integrate current classification and risk stratification systems, to better define prognosis and to promptly identify patients with a high risk of leukemic progression. In this study, we focused on circulating granulocyte subsets, including NDNs and LDGs with different immunomodulatory activities, in MDS, AML, and in subjects with cytopenia of unknown significance.

Neutrophils can exert immunomodulatory and immune surveillance activities or can conversely reduce anti-tumor immune responses by favoring cancer cell survival8,10. Within each granulocyte population, there are several subsets with different biological activities, such as inflammatory and suppressive LDGs, and immature and proinflammatory NDNs. To date, there is no information on granulocyte subset distribution and perturbation in hematological malignancies, including AML and MDS, or in clinical entities of unknown significance, such as ICUS. Moreover, there is no international consensus on surface or functional markers for LDG/NDN subsets identification, especially for utilization in routinely clinical practice8,9,12,13,14,15. In this observational real-life prospective study, we first optimized a 10-color staining for circulating granulocyte and monocyte subset immunophenotyping, that could be also used in clinical practice for rapid identification of principal peripheral blood populations, including CD33++CD14+CD11b++ monocytes, Lin-CD33+ granulocytes, CD33+dimCD45++CD11b++CD16+dim eosinophils, CD33+CD45dimCD11b+CD16- basophils, circulating hematopoietic stem (CD34+) and progenitor (CD34-CD117+) cells, and eventually CD45dim leukemic cells that would be further characterized. On granulocytes, LDGs and NDNs were identified based on CD15 expression, and their cytoplasmic complexity and partial/complete maturation could be also evidenced using linear parameter (side scatter area) and CD45 pan-leukocyte marker expression. Using our staining panel, on each granulocyte population, maturation curves could be studied using the standard CD11b vs. CD16 gating strategy for identification of immature CD11b-CD16-, intermediate CD11b+CD16+/-, and mature CD11b+CD16+ NDNs, and a CD16 vs. CD15 gating strategy for visualization of immature CD15+CD16-, intermediate CD15+CD16dim, and mature CD15+CD16+ LDGs. Moreover, by first gating on linear parameters (FSC-A vs. SSC-A), monocytes can be further characterized using CD14 and CD16 markers. A limitation of this panel was that CD56 was included within Lineage antibody (CD3, CD19, and CD56), used for rapid lymphocyte exclusion, and its abnormal expression on monocytes could not be investigated. Therefore, further implementation of our 10-color staining panel with the addition of CD56 as a separate marker could allow a deeper immunophenotyping of circulating cells and identification of dysplastic features on CD33++CD14+CD11b++ monocytes.

Neutrophil plasticity and heterogeneity in functions is currently a well-established concept that has replaced the prior understanding of a terminally differentiated innate immune cells that can only kill pathogens9,29. LDGs and NDNs can exert both immunomodulatory and pro-inflammatory activities, and during acute diseases LDGs and transcriptionally active myelocytes and metamyelocytes can exit the bone marrow under growth factor emergency stimulation and actively release azurophilic granule contents involved in inflammatory responses30. This condition is particularly frequent in ANCA+ vasculitis and systemic lupus, where more immature LDGs with high expression of serine proteases are present14. As observed in autoimmune diseases, immature forms of LDGs, especially intermediate CD15+CD16dim granulocytes, were increased in AML and MDS with a concomitant reduction of their mature forms compared to healthy and ICUS subjects. These immature forms could maintain a pro-inflammatory environment in the bone marrow31,32. In our study, LDGs were positively associated with normal hemoglobin levels; conversely more immature LDG forms with lower levels, while negatively related to percent of circulating blasts, as well as circulating monocytes. Indeed, granulocyte subset distribution seemed to vary based on disease severity, as mature LDGs were highest in healthy and ICUS subjects, while gradually decreased from low- to high-risk MDS, and were the lowest in AML patients, while intermediate LDGs were highly represented in AML and MDS patients. These perturbations confirmed that there might be a gradual switch from normal to dysplastic hemopoiesis to leukemic evolution, and that residual ineffective hemopoiesis could sustain a pro-inflammatory and pro-leukemogenic environment by activating emergency myelopoiesis with the release of immature forms of LDGs. Therefore, emergency hemopoiesis could be a first mechanism of the bone marrow to maintain normal peripheral blood counts in a setting of dysplastic ineffective hemopoiesis; however, this salvage process could be deleterious on long-term with the arise of leukemic clones and shutdown of immune responses. These mechanisms of emergency hemopoiesis remained active even after chemotherapy and CR achievement, probably because of immune surveillance and/or persistent drug-induced damage on the hematopoietic stem cell compartment. Those “dysplastic features” observed by optical microscopy in the bone marrow of AML and MDS patients in CR or PR could also include variations in granulocyte subset distribution with the prevalence of intermediate forms of LDGs and NDNs with more immunomodulatory functions rather than pro-inflammatory activities10. In previous studies, monocyte subsets have been found impaired in low-risk MDS, with reduced classical monocytes and increased intermediate monocytes, also showing altered transcriptome profiling and a with a higher susceptibility to produce TNFα in response to lipopolysaccharide33. In our study, we observed reductions in non-classical monocytes in AML patients and intermediate monocytes tended to be lower in AML and high-risk MDS patients. These differences might be due to the high heterogeneity of these conditions, even in low-risk diseases, characterized by various molecular signatures.

We have previously documented that mature CD11b+CD16+ granulocytes are significantly reduced in the marrow of high-risk myelodysplasia compared to low-risk MDS and that their frequency is inversely correlated with bone marrow CD3+CD56+ T regulatory lymphocytes, while not in AML34. Conversely, bone marrow CD3+CD56+ lymphocytes in AML are negatively related with WT1 expression levels, while not in MDS, suggesting a role of this T cell subset in immunosurveillance on dysplastic hematopoietic stem cells, resulting in reduced production of mature elements34,35. From our current study, we evidenced a more complex scenario in which ineffective hemopoiesis could simultaneously play the role of passenger and driver in immunosurveillance on bone marrow clones. Immature granulocytes could be released in response to ineffective hemopoiesis and exert their pro-inflammatory functions by priming Th1-mediated immune responses in early-stage disease, that in turns augments CD8+ T cell-mediated cytotoxicity on dysplastic progenitors, thus worsening ineffective hemopoiesis and increasing emergency processes with the release of immature granulocytes in positive feedback36. As a proof-of-concept of this hypothesis, we investigated variations in circulating levels of neutrophil elastase, a serine protease expressed in neutrophils and stored in their azurophilic granules, that can promote inflammation, bacterial infection progression, hypersecretion of mucus, cytokine production, and immune response and inflammation regulation37,38. Moreover, neutrophil elastase can influence tumorigenesis, as it can inactivate tumor suppressors or induce tumor growth and metastasis by regulating angiogenesis, and higher circulating and tissue levels are related to shorter overall survival38. Here, we showed that plasma elastase levels were significantly decreased in MDS patients compared to AML and healthy subjects, and were inversely related to total LDGs and classical monocytes, while positively associated with all NDN subsets, suggesting that reduced granulocyte functions could be present in MDS subjects thus promoting infectious disease susceptibility and impaired immune response regulation.

Finally, ICUS is defined as mild cytopenia lasting for at least 4 months, with no or only mild (< 10%) marrow dysplasia, bone marrow blasts < 5%, without clonal cytogenetic or molecular markers, and after exclusion of other diseases39. To date, clinical significance of these cytopenias is still unknown and under investigation. Here, we showed that there might be granulocyte subset distribution alterations, especially in NDNs, that might anticipate MDS development or sustain, for unknown causes, emergency hemopoiesis.

Our study has several limitations: (i) the short follow-up period with few observed progressions and deaths that could not allow a better investigation of the prognostic role of these granulocyte subset distribution variations in MDS and AML, or as a predictor of MDS evolution in ICUS subjects; (ii) the small number of subjects for each clinical entity, that could increase the risk of type II errors (false negatives), especially when clinically heterogenous diseases are investigated, such as MDS and AML. Indeed, this wide biological variability is due to patient-specific genomic and phenotypic signatures augments type II errors if subcategory sample size is not adequate; therefore, we could not exclude additional statistically significant differences between investigated groups, especially in MDS and AML entities (e.g., TP53 mutated diseases vs. wild type, or favorable vs. adverse-risk AML). (iii) The lack of sequential immunophenotyping for the entire cohort for a comparison before and after therapies; (iv) additional functional analyses should be performed to correlate impaired neutrophil subset distribution with their proinflammatory and pro-tumorigenic functions; and (v) lack of a priori power analysis and of a minimum detectable effect calculation.

In conclusion, we described for the first time, even in a small cohort of hematological patients, distributions and perturbations of LDG and NDN subsets in AML, MDS, and ICUS, suggesting that emergency hemopoiesis could be a first mechanism to sustain peripheral blood counts; however, the release of immature granulocytes with pro-inflammatory activities could maintain a pro-leukemogenic environment and immune selection of neoplastic clones, thus favoring clonal evolution. Moreover, we showed clinical feasibility and utility of a 10-color staining for comprehensive immunophenotyping of principal peripheral blood populations, including monocytes, granulocytes, and hematopoietic stem and progenitor cells, also allowing rapid identification of CD45dim leukemic cells. In addition, we demonstrated that LDGs and NDNs can be immunophenotyped even without prior gradient separation using fresh whole peripheral blood, as they displayed different morphological and phenotypical characteristics allowing their discrimination by linear parameters and CD45, CD33, and CD15 expression. However, further validation on larger cohort is needed to confirm our results.