Persistent Immune Stimulation Exacerbates Genetically Driven Myeloproliferative Disorders via Stromal Remodeling Free

Alterations in the bone marrow osteoblastic and vascular niches that nurture hematopoietic stem cells (HSC) are associated with the outgrowth of myeloid cells within a biased leukemic niche (1, 2). Such alterations encompass the adaptation of specific mesenchymal cell populations and the disruption of normal gradients in structural and secreted proteins, including matricellular proteins, cytokines, and chemokines (3). A key regulator of bone marrow stromal changes is the secreted protein acidic and rich in cysteine (SPARC; ref. 4), a matricellular protein in which up- and downregulation has been associated with hematologic neoplasms (2, 5).

Along with remodeling of the bone marrow mesenchymal architecture, myeloid clone expansion changes the cytokine profile of the hematopoietic milieu. Increased levels of proinflammatory cytokines including TNF, IL6, and IL8 have been reported in patients with myeloproliferative neoplasms, myelodysplastic syndromes (MDS), and myeloid leukemia with an unfavorable predicted disease course (6, 7). The induction of an inflammatory state in the myelopoietic parenchyma is mutually linked with the status of the stromal architecture. A defective ECM organization resulting from SPARC deficiency favors the over-activation of myeloid cells (8) and granulocyte immunogenic death via the formation of neutrophil extracellular traps (NET; ref. 5). NETotic death is characterized by the expression of IFN response-related transcriptional programs associated with the induction of inflammation and exacerbation of autoimmune disease, in which NETs contribute as sources of autoantigens and costimulatory signals (9–13). A link has been demonstrated between dysregulated immune stimulation and the occurrence of myeloid malignancies. Large-scale epidemiologic studies have shown that the history of chronic autoinflammatory or autoimmune diseases is correlated with increased incidence of MDS or acute myeloid leukemia (AML; ref. 14); however, the role of systemic immune activation in the pathobiology of myeloid malignancies remains unknown. The highest risk of development of MDS and AML has been reported in patients with autoimmune diseases in which NETs have a recognized pathogenic role, for example, systemic lupus erythematosus and antineutrophil cytoplasmic antibody (ANCA)-related vasculitis (9, 15).

In this study, we investigated whether a persistent immune stimulation can change the bone marrow stromal microenvironment and affect hematopoietic precursor proliferation. We identified a novel regulation of myeloid cell expansion involving the crosstalk between the immune cells and stromal components of the bone marrow microenvironment. Such crosstalk is proposed to be relevant in explaining the link between systemic immune activation and myeloid neoplasms.

hMPR8-NPMc⁺ transgenic mice (B6. Tg<hMRP8-NPMc>Ptprc<b>tg/J), hereafter referred as NPMc, were obtained by backcrossing (C57BL/6J × CBA) F1-hMRP8-NPMc mice to C57BL/6 (B6. Ptprc<b>). The bone marrow of Trif^–/– mice was kindly provided by Dr. Garlanda (Humanitas Clinical and Research Center, Milan, Italy). All experiments involving animals described in this study were approved by the Ministry of Health (authorization number 10/11). The protocol describing the immunization of WT and NPMc mice with NET-loaded DCs is described in the Supplementary Data.

Histologic, histochemical, and immunofluorescence analyses of murine and human bone marrow samples were performed as previously described (2, 5) and detailed in the Supplementary Data. For in situ histopathologic analyses of human bone marrow specimens, the bone marrow biopsies from 12 patients with autoimmunity-associated peripheral cytopenias were selected from the archives of the Department of Hematology, University of Foggia, Foggia, Italy, and the Department of Health Sciences, University of Palermo, Palermo, Italy. Patient characteristics are listed in Supplementary Table S1.

Briefly marrow and spleen cell suspensions were stained with a pool of antibodies for Lin positive markers, including CD3, CD11b, CD45R, Ly6G, CD4, CD8, and Ter-119, and stem cell and progenitor cell markers, including CD117 (c-Kit), CD34, and CD16/CD32. Progenitors were identified according to their expression of CD34 and CD16/CD32 within the gate of Lin^–CD117⁺ cells. Bone marrow, spleens, and PBL samples were also evaluated for granulocyte enrichment by staining cell suspensions with mAbs against CD11b, Ly6G, and CD45, followed by FACS analysis. All antibodies were purchased from eBioscience. Samples were acquired on an LSR II (BD Biosciences).

Inflammatory PMNs (IFNγ/TNF-treated), prone to dye of NETosis, were obtained from agar plugs subcutaneously injected into C57BL/6 mice, as previously described (9). In particular, PMN were seeded onto coated tissue culture dishes in IMDM 2% FCS, allowed to adhere for 30 minutes and added with myeloid dendritic cell (mDC; 1:1; PMN:DC) for 16 hours. During this period NETs are induced and transfer their component to mDCs (9). Dendritic cell (DC) were isolated from the coculture via positive selection, counted and injected intraperitoneally (2.5–5 × 10⁶ cells, once a week for a total of six injections) into naïve mice. Three months after the last immunization, mice underwent a boosting immunization. Autoimmunity was evaluated starting 1 month after the boost by measuring the titer of ANCA and dsDNA autoantibodies using specific ELISA kits (MPO-ANCA IgG, ELISA Kit from Cusabio and mouse anti-dsDNA antibody ELISA Kit from Alpha Diagnostic International).

For ex vivo isolated neutrophils, spontaneous NETosis was evaluated by seeding bone marrow-derived PMNs onto poly-d-lysine–coated glasses in 2% FCS-IMDM in presence or absence of PMA (16). To quantify NETosis and to distinguish NETosis from apoptosis, the cell-impermeable DNA dye Sytox green, was added to PMNs such to measure the size of the released nuclear contents from micrographs using the software-assisted technique, according to Metzler and colleagues (17). To assess NET formation in vivo in murine and human bone marrow biopsies, confocal microscopy analysis was performed by staining with mAbs to murine or human-MPO (from Millipore and Novocastra, respectively) and anti-Histone H3 (citrulline R2+R8+R17, from Abcam), followed by staining with Alexa Fluor 546-conjugated goat anti-rabbit or goat anti-mouse secondary antibodies (Invitrogen Molecular Probes). Nuclei were counterstained with Draq5 or Dapi. To evaluate nucleophosmin (NPM) relocalization onto the NET threads, PMA-induced NETs (obtained from splenic WT- and NPMc-PMN treated 16 hours with PMA) were sequentially stained with mAb to nucleophosmin (18) and MPO (Millipore) or PR3 (Millipore). The NET-DNA was counterstained with Draq5.

Lin^–c-Kit⁺ cells were FACS-sorted from the bone marrow of NPMc or control mice (BD, FACSAria), stained with carboxyfluorescein succinimidyl ester (CFSE) and added to PMNs undergoing NETosis [inflammatory PMNs from agar plugs (9) or PMA-treated PMNs]. The coculture was performed in StemSpan H3000 medium (Stem Cell Technologies), without adding any other supplement. After 72 hours of coculture, cells are harvested and the c-Kit⁺ cell fraction, which comprised c-Kit⁺Sca⁺GR-1^– early progenitors, c-Kit⁺Sca^–GR-1^– committed precursors (mostly GMP), and c-Kit⁺Sca^–GR-1⁺ myeloblasts, was analyzed for proliferation. Proliferation was read out by measuring dilution of the CFSE dye (19).

Gene expression profile (GEP) analysis was performed in a panel of primary hematopoietic malignancies, including AML (AN = 100), chronic myelogenous leukemia (CML, N = 50), MDS (N = 50), and in a panel of healthy bone marrow samples (N = 30), which were randomly selected among those included in a large multicentric GEP study available on GEO database (GSE13159; refs. 20, 21). We also analyzed 461 AML cases for which gene expression data and information on the NPM1 status (mutated vs. wild type) were available (GEO Access: GSE6891; refs. 22, 23) and 15 AML cases included in a phase II clinical trial testing the combination of lenalidomide and cytarabine (Len-Ara) and for which our group previously generated GEP (24). Each case was assigned to either group 1 or 2 based on the expression of the NET-related inflammatory signature. A detailed description of the analysis performed is provided in the Supplementary Data.

To investigate how systemic autoimmunity could affect bone marrow stroma remodeling, naïve C57BL/6 mice were repeatedly immunized by intraperitoneal injections of mDC loaded with NET components (Fig. 1A and Supplementary Data). As previously shown (9), immunized mice developed serum autoantibodies and ANCA-associated autoimmune vasculitis in the kidney and lung parenchyma (Fig. 1B; Supplementary Fig. S1A and S1B). Compared with naïve mice, the bone marrow of immunized mice showed a significant decrease in the levels of type I collagen and SPARC, as assessed by semiquantitative PCR (Fig. 1C), IHC (Fig. 1D), and in situ immunofluorescence (Supplementary Fig. S1C). The decrease in type I collagen and SPARC was particularly evident in CD29⁺ stromal cells lining the osteoblastic (OB) niche, which is enriched in c-Kit⁺ precursors (Fig. 1F). Moreover, stromal and endothelial bone marrow cells showed increased C1q expression, which was associated with C5a (Fig. 1G–I), suggesting activation of the complement cascade. Finally, a trend in the induction of P-selectin expression was also observed in the bone marrow of immunized mice (Supplementary Fig. S1D–S1F).

Foxp3⁺ Treg cells are typically either depleted or skewed toward inflammatory Th1 or Th17 cells under autoimmune conditions (25). In the bone marrow of naïve mice, nearly 30% of CD4⁺ T cells were Foxp3⁺ Treg cells, a frequency much higher than that of the other lymphoid organs, where Treg cells accounted for 5% to 15% of total CD4⁺ T cells (Supplementary Fig. S1G and S1H). In contrast, in the bone marrow of immunized mice the frequency of Treg cells was significantly reduced (Fig. 2A). Also reduced in this subset was the expression of OX40 (26), a marker of activation (Fig. 2A). On the contrary, the expression of CD25 and OX40 on Teff cells was increased (Fig. 2B) together with their production of IFNγ and TNF (Fig. 2C and D). In immunized mice, TNF and IFNγ were also induced in CD45⁺ cells other than T lymphocytes (Fig. 2E).

To test the existence of a link between the TNF- and IFNγ-rich inflammatory milieu and stromal ECM modifications, bone marrow-MSCs were isolated from the bone marrow of naïve mice and stimulated 24 hours with IFNγ, TNF, or both. Semiquantitative PCR analysis showed that these stimuli, mostly when in combination, significantly decrease the expression of col1a1 and Sparc in bone marrow-MSCs (Fig. 2F and G).

Collectively, these results show that systemic immune activation modifies the bone marrow stromal microenvironment and changes the immune cell composition from immune-regulatory to proinflammatory.

TNF, IFNγ (9), and complement factors (27) are major drivers of NET formation. In a context of defective ECM molecules (collagens and SPARC), their NET-inducing capacity increases because of the lack of inhibitory signals that are provided by collagens on activated PMN (5). Therefore, we investigated whether induction of systemic autoimmunity affected NET formation by bone marrow cells. Immunized mice showed increased NET formation, as assessed ex vivo by measuring the capacity of bone marrow–derived polymorphonuclear leukocytes (PMN) to undergo spontaneous NETosis (Fig. 2H and I) as well as in situ by immunoflourescence (Fig. 2J) and confocal microscopy (Fig. 2K).

We then investigated the capacity of NETs to affect the proliferation of bone marrow hematopoietic cells. Lin^–c-Kit⁺ (LK) hematopoietic cells were sorted from the bone marrow, stained with carboxyfluorescein succinimidyl ester (CFSE) and cocultured with NETs in conditions enabling HSC maintenance and differentiation toward myeloid lineages. As a source of NETs, we used PMNs isolated from in vivo inflammatory foci that undergo spontaneous NETosis dependent on TNF and IFNγ priming (9). In some experiments, PMA-treated NETotic PMNs were used as positive controls (28). Starting from 72 hours of coculture, NETotic, but not apoptotic, necrotic, or naïve PMNs, significantly enhanced the proliferation of c-Kit⁺ bone marrow cells (Supplementary Fig. S2A). An in depth analysis showed that NET specifically enhanced the proliferation of c-Kit⁺Sca⁺GR-1^– early progenitors (LSK), c-Kit⁺Sca^–GR-1^– (LK) committed precursors (mostly GMP), and c-Kit⁺Sca^–GR-1⁺ myeloblasts (Fig. 2L, representative gating strategy in Supplementary Fig. S2B). Moreover, NETs sustained the expression of Sca-1 in c-Kit⁺Sca⁺Gr-1^– early progenitor cells, which is crucial for HSC self-renewal and proliferation (Fig. 2M).

Investigating the mechanism involved in NET-induced proliferation of c-Kit⁺ cells, we first excluded the involvement of TLRs as sensor of the NET DNA-backbone (Supplementary Fig. S2C and S2D). Considering that the proliferation of c-Kit⁺ cells induced by NETs was diminished, but not abrogated, by the interposition of a transwell membrane (Supplementary Fig. S2E and S2F), we evaluated the role of soluble mediators. Stem cell factor (SCF) and IL6 are known to promote hematopoietic cell renewal and clonogenic capacity through NF-κB (29). MAbs blocking either the IL6 receptor (15A7) or c-Kit (ACK2) strongly inhibited NET-induced proliferation of c-Kit⁺ hematopoietic cells, an effect magnified by the simultaneous blockade of both receptors (Fig. 2N). Pharmacological inhibition of NF-κB had a comparable effect (Fig. 2O), being NF-κB the main transducer of IL6R on hematopoietic precursors (30). These data indicate that NET inflammatory pressure exerted on c-Kit⁺ hematopoietic cells through IL6 and SCF release confers a proliferative advantage to hematopoietic precursors.

We then investigated whether NET-induced stimulation of hematopoietic precursors could impact on the proliferation of myeloid cells under myeloproliferative conditions in vivo. NPMc transgenic mice carry the human mutated NPMc under a myeloid-specific promoter and are prone to develop an indolent myeloproliferation, in which mature myeloid cells expand without progressing towards myelodysplasia or leukemia (31). Induction of systemic autoimmunity in NPMc transgenic mice elicited modifications in the bone marrow microenvironment that mirrored those observed in immunized WT mice (Supplementary Figs. S3 and S4) and included in situ formation of NETs (Supplementary Fig. S5A).

Compared with naïve NPMc mice, a significantly higher percentage of immunized NPMc mice developed a myeloproliferative disorder at approximately 12 months of age (60% vs. 25%, P < 0.05). In these mice, bone marrow histopathology revealed an exacerbated myelodysplastic/myeloproliferative phenotype with a left-shift of myelopoiesis, clustering of atypical myeloid precursors, dysmegakaryopoiesis, and signs of intravascular hematopoiesis (Fig. 3A). Paralleling the exacerbated myeloproliferation the bone marrow of immunized NPMc mice showed a significant difference in the density of bone marrow erythroid colonies highlighted by TER-119 immunostaining (Fig. 3B).

Consistently, the frequency of cytoplasmic NPM-expressing elements and of proliferating Ki67-expressing myeloid elements in the bone marrow of immunized NPMc mice was significantly higher as compared with naïve NPMc controls (Fig. 3C–E). Moreover, bone marrow smears of immunized NPMc mice showed an increase in immature forms of myeloid cells with atypical morphology (Fig. 3F, red arrows), a concomitant decrease in segmented mature granulocytes (black arrows), and an increase in lymphoid elements (green arrows). Flow cytometry analysis of the bone marrow showed a significant expansion in the fraction of hematopoetic precursors (LK and LSK) and committed progenitors (GMP, MEP), megakaryocytes and megakaryocyte precursors in immunized NPMc mice. Conversely, the total number of B220⁺ B cells decreased (Fig. 3G; Supplementary Fig. S5B for representative gating strategy).

Dysplastic features were also found in megakaryocytic and granulocytic cells within extramedullary splenic hematopoietic foci, which were expanded in immunized NPMc mice (Supplementary Fig. S6A). Indeed, the frequency of cytoplasmic NPM-expressing elements and of proliferating Ki67-expressing myeloid elements was increased in the spleen of immunized NPMc mice (Supplementary Fig. S6B–S6E). Flow cytometry analysis of spleen myeloid populations showed a significant expansion in the amount of LK cells and of granulocyte–monocyte progenitor (GMP) in immunized NPMc mice (Supplementary Fig. S6F–S6I). Finally, c-Kit⁺ blasts had a higher frequency in the blood of immunized NPMc mice, a finding corroborated by the enrichment in morphologically immature circulating myeloid cells (Fig. 3H and I).

To provide an alternative approach testing the impact of persistent immune stimulation on the take of an established acute leukemia, we adopted the C1498 murine leukemia model (32). C57BL/6 mice were either immunized or not with DCs loaded with NET components and 1 month after the last immunization they were injected intra-bone with 2 × 10⁵ GFP-labeled C1498 cells. Twenty-eight days after cell injection, mice were sacrificed and evaluated for the presence of C1498 in the bone marrow by FACS. Immunization significantly promoted the leukemic cell take in the bone marrow with the increased fraction of C1498 cells inversely correlating with that of B220⁺ resident cells, which were outcompeted (Supplementary Fig. S7). These results indicate that systemic and persistent immune activation can exacerbate the phenotype of a genetically driven myeloproliferative spur.

Considering the data obtained in immunized NPMc mice, we hypothesized that NETs could confer a stronger proliferative advantage to hematopoietic precursors of NPMc rather than WT genotype. Unexpectedly, the presence of the NPMc mutation did not confer an intrinsic proliferative advantage to c-Kit⁺ cells in response to NET stimulation (Fig. 4A). In line, LSK cells from NPMc and WT mice cocultured with NETs in the absence of any growth factor (i.e., SCF or IL3) were equally able to survive in culture and to reconstitute lethally irradiated mice (Fig. 4B).

We then investigated whether bone marrow myeloid cells from NPMc and WT mice were differently prone to NET formation. To this end, bone marrow PMNs isolated from NPMc and WT mice were seeded onto poly-d-lysine–coated slides and treated with IFNγ plus C5a or PMA. In both conditions, PMNs from NPMc mice showed a significantly increased NET formation in response to inflammatory stimuli (Fig. 4D and E). Moreover, NETs from NPMc mice were more efficient in enhancing proliferation of c-Kit⁺ cells (Fig. 4F).

As NPM may act as an alarmin and contribute to systemic inflammation, for example, during sepsis (33), we hypothesized that the NPMc mutation, which causes cytoplasmic nucleophosmin localization (34), could be also responsible for its reallocation onto NETs, thereby influencing its availability as a NET-associated alarmin. IHC analyses performed on bone marrow sections showed that NPM, which is almost absent in mature bone marrow-PMNs of WT mice, is detectable in the cytoplasm of bone marrow PMNs from NPMc mice (Fig. 4G). In these mice, immunofluorescence on bone marrow sections using an anti-NPM Ab detecting both normal and mutated NPM, also highlighted NET threads (Fig. 4H). Moreover, at confocal microscopy, ex vivo NET from NPMc mice displayed NPM relocalized onto the NET DNA threads (Fig. 4I), where it associated with classical NET cytoplasmic proteins, such as MPO (Fig. 4J).

To test the ability of extracellular NPM to act as an alarmin for hematopoietic cells, LK cells were cultured in the presence of recombinant human and mouse NPM. As a result, proliferation of c-Kit⁺ cells exposed to extracellular NPM was significantly enhanced in a dose-dependent manner and this effect could be inhibited by NF-κB blockade (Fig. 4K and L). Recombinant NPM mimicked TNF and IFNγ inflammatory stimuli in downmodulating the expression of ECM molecules in bone marrow-MSCs (Fig. 4M), a condition that promotes NET formation, delineating a self-sustaining inflammatory loop in AML carrying NPM1 mutation.

To establish the translational significance of findings in experimental models, we performed studies in patients with systemic autoimmunity and in myeloid malignancies. Analysis of the bone marrow of patients with unexplained peripheral cytopenias and signs of systemic autoimmunity (AUC; Supplementary Table S1) showed mild alteration in granulocytic and megakaryocytic differentiation that did not configure impaired hematopoiesis or frank myelodysplasia when compared with normal controls (not shown). No significant differences were found in the overall amounts of bone marrow-infiltrating T cells; however, a trend towards the reduction of Foxp3⁺ Treg cells was observed in the bone marrow of patients with autoimmune disorders (Fig. 5A and B). Most importantly, in these same patients, NETs were identified in situ by staining for MPO and DAPI (Fig. 5C). In patients with circulating autoantibodies (Supplementary Table S1), NET extracellular MPO threads were associated with C1q deposition (Fig. 5D).

We then investigated the presence of NETs in the bone marrow of AML patients carrying mutations in the NPM1 gene. Immunofluorescence analysis for the chromatin marker Histone-H3 and either MPO or NPM was performed on 20 cases of AML with mutated (N = 10) or WT (N = 10) NPM1. We found that, in human AML cases, NETs were detected as Histone-H3⁺ threads associated with MPO in cases in which MPO expression was preserved (Fig. 5E). As shown in transgenic NPMc mice, in the bone marrow of patients with NPM1-mutated AML, NET figures could be identified as characterized by the peculiar NPM localization onto the DNA thread marked by Histone-H3 (Fig. 5E). Quantification of NET figures as highlighted by Histone H3 and NPM or MPO on 5 high-power magnification microscopic fields, revealed that a higher frequency of NET figures could be identified in NPM1 mutated cases (Fig. 5F; Supplementary Table S2). These data suggest that formation of NETs in the bone marrow observed in experimental systemic autoimmunity can be also observed in spontaneously occurring human autoimmune pathologies and occurs within the bone marrow environment of AML with mutated NPM1.

We then asked whether the inflammatory profile related with NETosis could be differently enriched at the transcriptional level in human myeloid malignancies.

To this end, we designed an ad hoc gene signature representative of the pathways activated in the bone marrow upon immunization and implicated in NET induction (Supplementary Table S3; ref. 5). The signature, thereafter referred to as NET-related inflammatory signature, was analyzed in a large panel of primary hematopoietic malignancies, including AML (N = 100), chronic myelogenous leukemia (CML, N = 50), MDS (N = 50), and in healthy bone marrow samples (N = 30), obtained from a large multicentric GEP study (see the Supplementary Information). Unsupervised hierarchical clustering based on the expression of the NET-related inflammatory signature separated AML samples from CML, MDS, and healthy bone marrow samples (Supplementary Fig. S8A). When a supervised approach was used to identify, within the NET-related inflammatory signature, the most suitable genes to differentiate the various diseases (ANOVA, P < 0.05, fold change > 2), AML was again separated from the other samples (Supplementary Fig. S8B).

Considering the variability in the NET-related inflammatory GEPs in AML, we investigated the AML cases in depth in order to understand the contribution of inflammation-related genes in this pathologic setting. We identified 20 genes (corresponding to 27 probe sets) belonging to the NET-related inflammatory signature that were differentially expressed in AML versus healthy bone marrow samples (t-test, P < 0.05, fold change >2; Supplementary Table S4). On the basis of the expression of these genes, AML cases could be divided into two clusters: cluster 1 and cluster 2 (Supplementary Fig. S8C). A supervised analysis (t test, P < 0.05; multiple testing correction: Bonferroni family wise error rate) was then performed to identify differentially expressed genes in cluster 1 and cluster 2. We identified 592 probe sets that enabled efficient discrimination between the two clusters (Supplementary Table S5; Fig. 6A). When studied by gene set enrichment analysis (GSEA) the corresponding 440 genes turned out to be significantly associated, as expected, with inflammation and immune response programs as well as with other relevant cellular processes and pathways, including some related with signal transduction, hematopoietic and leukemic stemness and, most notably, NPM1 signaling (Supplementary Table S6; Fig. 6B). This expression pattern was not found to be related to cytogenetic features in AML. Indeed, the two AML clusters comprised all different karyotypic groups (Supplementary Fig. S8D). Furthermore, the 17 genes (corresponding to 25 probe sets) belonging to the NET-related inflammatory signature, which were differentially modulated across the cytogenetic groups, as identified by supervised analysis (ANOVA, P < 0.05, fold change >2), were unable to clearly discriminate between AML cases with different karyotypic abnormalities (Supplementary Fig. S8E and S8F).

We extended the supervised analysis to an independent panel of 461 AML cases for which gene expression data, information on NPM1 mutational status, and clinical follow-up were available (GEO Access: GSE6891; ref. 23). Again, the identified NET-related inflammatory signature defined two AML clusters in which mutated and wild-type NPM1 cases were significantly segregated (χ², P < 0.001; Fig. 6C). Similar results were obtained by performing hierarchical clustering based on the expression of the 440 genes (Supplementary Table S5) discriminating the two AML clusters: NPM1-mutated and NPM1-wild type cases were significantly divided (χ², P < 0.001; Supplementary Fig. S8G). This pattern was independent from karyotype, in particular cases with normal karyotype were homogeneously distributed in the two clusters (χ², P = 0.19). Consistently, we found that differentially expressed genes in mutated versus wild-type NPM1 cases were significantly enriched in the inflammatory response, IFNγ response, IL6 signaling, and complement cascade programs at gene set enrichment analysis (Supplementary Table S7). Conversely, when GEPs of cases presenting with somatic mutations affecting leukemia-associated genes other than NPM1 (CEBPA, IDH1, IDH2, KRAS, NRAS, and FLT3-ITD) were investigated, IDH1-mutated cases were found to display significant enrichment in genes belonging to the NET-related signature, even if only very few cases (N = 3) among those of the analyzed panel had isolated IDH1 mutations.

To assess the potential clinical impact of NET-related inflammatory features, we applied our signature to a cohort of high-risk AML patients (N = 26; complete responders = 14; nonresponders = 12) treated with the combination of chemotherapy and the immunomodulatory drug lenalidomide (24). The molecular signature associated with complete response (24) significantly overlapped with the NET-related inflammatory signature (χ², P = 0.03), being also enriched in genes characteristic of NPM1 mutated AML. In the same cohort of patients, the NET-related inflammatory signature enabled subdivision of the AML cases into two groups with significantly different overall survival (median estimated survival 1 month vs. 20 months; 95% confidence interval, 0.140–6.88/1.86–45.11 months; P < 0.001; Fig. 7A). A minimal signature consisting of three genes (Fig. 7B) belonging to the NET-related inflammatory signature, namely IRF7, C5AR1, and NF-KBIA, were able to discriminate patients with a significantly better or worse outcome (P = 0.03; Fig. 7C). Taken together, these data showed that, although not associated with clinical outcome in a cohort of AML patients treated with conventional chemotherapy, the expression pattern of genes related with NETs was associated with the clinical response and survival of AML patients receiving the immunomodulatory agent lenalidomide.

We showed that structural changes in the bone marrow stroma are involved in shaping the myeloproliferative bone marrow responses to systemic persistent immune stimulation. Immunization-induced autoimmunity was found to elicit downregulation of SPARC and type I collagen in the bone marrow stroma as well as to promote local inflammatory conditions associated with T-cell activation and NETosis.

Within the bone marrow, NET sustained the proliferation of hematopoietic precursors via the release of the hematopoietic cytokines SCF and IL6 involving NF-κB.

The discovery of NET-induced expansion of hematopoietic precursors prompted us to investigate whether the transcriptional profiles of various human myeloid malignancies were differently enriched in a NET-related gene signature. We found that AML were divided into two main clusters according to the inflammatory imprint. The inflammatory cluster segregated according to NPM1 mutation-associated molecular programs. The NPM1 mutation is the most frequent genetic alteration identified in acute myeloid leukemia (AML) accounting for about 30% of all cases. This mutation defines a subgroup of AML with distinctive clinicopathologic features and it is also associated with a unique GEP (18). The role of NPM1 mutation in leukemogenesis in vivo has not been fully elucidated. The low genetic complexity of NPM1-mutated AML suggests that this mutation requires additional stimuli, which might also be provided by the inflammatory microenvironment, to promote malignant clone progression (18).

In this context, a large study involving 1,540 patients has shown that the impact of NPM1 mutation on transformation become manifested only in presence of initiating mutations in epigenetic pathways, such as in DNMT3, TET2, IDH1/2 genes (35).

This makes no surprise that in the different transgenic (31) or conditional mouse model (36, 37), lacking addition mutations, a frank leukemia was rarely observed, although an indolent myeloproliferation, perturbed megakaryopoiesis, or MPD have been reported. However, investigating whether an inflammatory bone marrow environment was able to aggravate the indolent myeloproliferative phenotype of NPMc transgenic mice we found that features of dysplastic changes and increased extramedullary hematopoiesis were induced in NPMc immunized mice. The NPMc mutation cooperated with NET-mediated stimulation of c-Kit⁺ precursors according to the propensity of NPMc PMN to NETosis and because of the alarmin function of the NET-complexed cytoplasmic mutated NPM. This may explain, at least in part, why the NET-related inflammatory signature was found to be enriched in AML with mutated NPM1. Accordingly, the mutated NPM, although aberrantly translocated to the cytoplasm, conserves its functional domains, including the histone-binding domains (38), which may explain its affinity for histone-rich NET DNA threads and its related pro-alarmin function. This suggests that NPMc-bearing clones may benefit from the presence of NET-triggering conditions, which may be irrelevant for clones with t(3,5) translocation (NPM1-MLF1) or loss of chromosome 5, which result in the loss of the DNA-binding domain or in the complete loss of NPM (39).

We discovered a novel role for NETs in stimulating hematopoietic precursor proliferation, also extending the influence of this form of myeloid cell death to hematopoiesis and related disorders. The identification of AML subgroups characterized by a NET-related inflammatory signature and bone marrow NET induction claims for the investigation of immunomodulatory agent-containing treatments to be adopted in combination with conventional therapy in selected patients.

No potential conflicts of interest were disclosed.

Conception and design: C. Tripodo, A. Liso, G. Visani, M.P. Colombo, S. Sangaletti

Development of methodology: C. Tripodo,

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Tripodo, C. Chiodoni, P. Portararo, B Cappetti, L. Botti, A. Gulino, A. Isidori, A. Liso, G. Visani, M.P. Martelli, B. Falini, P.P. Pandolfi, S. Sangaletti

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Tripodo, A. Burocchi, P.P. Piccaluga, C. Chiodoni, A. Gulino, G. Visani, M.P. Martelli, B. Falini, S. Sangaletti

Writing, review, and/or revision of the manuscript: C. Tripodo, A. Burocchi, P.P. Piccaluga, A. Isidori, A. Liso, M.P. Colombo, S. Sangaletti

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases):

Study supervision: M.P. Colombo, S. Sangaletti

The authors thank Prof. Federica Sallusto (Institute for Research in Biomedicine, Università della Svizzera Italiana) and Prof. Niccolò Bolli (Fondazione IRCCS Istituto Tumori Milano and University of Milan) for critical reviewing of the manuscript. The authors also thank Professor Maurilio Ponzoni (Università Vita- Salute San Raffaele) and Professor Umberto Gianelli (Fondazione IRCCS Ca' Granda - Ospedale Maggiore Policlinico) for providing human tissue samples. The authors thank the Conventional and Confocal Microscopy Facility for confocal microscopy images acquisition, the Cell Sorting Facility for cell sorting, and Mrs. Ester Grande for administrative assistance.

This work was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC grant number 12810 to S. Sangaletti; Program Innovative Tools for Cancer Risk Assessment and Diagnosis, 5 per mille number 12162 to C. Tripodo and M.P. Colombo; investigator grant number10137 to M.P. Colombo), and the Italian Ministry of Health (GR-2013-02355637; to S. Sangaletti).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.