Reactive Oxygen Species–Producing Myeloid Cells Act as a Bone Marrow Niche for Sterile Inflammation–Induced Reactive Granulopoiesis Free

Neutrophils are key players in innate immunity and host defense. During infection and inflammation, a large numbers of neutrophils are mobilized from the bone marrow (BM) to the circulation and then are recruited to affected tissues where they protect the host by recognizing, phagocytosing, and clearing invading pathogens. To compensate for their circulatory loss, BM granulopoiesis is enhanced during infection and inflammation.

Blood cells arise from self-renewing hematopoietic stem cells (HSCs) in the BM. Long-term HSCs first differentiate into short-term HSCs. These short-term HSCs then give rise to more differentiated nonrenewing multipotent progenitors, common myeloid progenitors (CMPs), and common lymphoid progenitors. CMPs gradually differentiate into megakaryocyte/erythroid progenitors (MEPs) and granulocyte-monocyte progenitors (GMPs). Although this classical hematopoietic hierarchy has long served as the conceptual framework for hematopoiesis research, recent studies using single-cell analyses indicate that progenitor populations, including multipotent progenitors, CMPs, and MEPs, are in fact heterogeneous and absent of mixed-lineage progenitors (1, 2). It also was reported that HSCs directly produce some self-renewing lineage-restricted progenitor cells (3).

Neutrophils are produced from GMPs through a series of developmental stages, including myeloblasts, promyelocytes, myelocytes, metamyelocytes, band neutrophils, and finally, mature segmented neutrophils (4). The process that maintains physiologic numbers of circulating neutrophils is known as steady-state granulopoiesis. The accelerated granulopoiesis that occurs during infection and inflammation is known as “emergency” granulopoiesis (5, 6). The two processes are regulated by distinct cellular mechanisms. For instance, steady-state granulopoiesis is regulated by the C/EBP-α, but not the /EBP-β, transcription factor (7, 8). In contrast, inflammation-induced accelerated granulopoiesis is largely controlled by C/EBP-β, but not C/EBP-α (8, 9).

Accelerated granulopoiesis is associated with microbial infection–elicited emergency granulopoiesis and sterile inflammation–initiated reactive granulopoiesis (10). The two processes are triggered by different stimuli. Emergency granulopoiesis is dependent on the presence of a disseminated microbial pathogen. The pathogen-induced upregulation of myeloid-differentiation pathways involves activation of TLR signaling in the progenitors (11–13), although a recent report suggests that TLR-independent pathways can also mediate hematopoietic stem and progenitor cell expansion (14). In contrast, sterile inflammation–associated reactive granulopoiesis is initiated by noninfectious stimuli, such as chemical agents (e.g., acid, TG, or alum), physical insults (e.g., trauma, surgery, burns, or radiation), or autoimmune disorders (e.g., lupus or rheumatoid arthritis). Due to the different upstream stimuli, there are also fundamental molecular differences between these two processes. For instance, the vaccine adjuvant alum induces reactive granulopoiesis in an IL-1R1–dependent manner (9). By activating IL-1R1–mediated signaling, alum elicits a transient increase in G-CSF production that mobilizes neutrophils from the BM. However, alum-induced accelerated granulopoiesis appears to be mediated by a density-dependent feedback that can sustain G-CSF levels (15). Nevertheless, LPS-induced emergency granulopoiesis, which mimics microbial infection, is totally independent of IL-1R1 signaling (13).

Microbial infection and sterile inflammation can accelerate granulopoiesis, suggesting that some molecular pathways might be shared between microbial infection–induced emergency and sterile inflammation–elicited reactive granulopoiesis. Extracellular granulopoietic factors, such as IL-6, IL-3, G-CSF, and GM-CSF, are implicated in reactive and emergency granulopoiesis (8, 10, 16–21). However, infection-induced emergency granulopoiesis and sterile inflammation–elicited reactive granulopoiesis are triggered by different sets of stimuli and are mediated by distinct upstream signals. Whether the induced accelerated granulopoiesis is mediated by a common intracellular pathway and the nature of such a pathway are poorly defined. We recently demonstrated that myeloid cell–derived reactive oxygen species (ROS) externally regulate the proliferation of myeloid progenitors in bacteria-elicited emergency granulopoiesis (22). In this study, we revealed that ROS also play a critical role in reactive granulopoiesis triggered by sterile inflammation. Although microbial infection–induced emergency granulopoiesis and sterile inflammation–elicited reactive granulopoiesis are triggered by different stimuli and mediated by distinct upstream signals, these pathways converge to a common point: NADPH oxidase–dependent ROS production by BM myeloid cells that accelerates granulopoiesis via upregulating PtdIns(3,4,5)P3 signaling. Collectively, we reveal a mechanism by which neutrophil homeostasis is rebalanced in response to acute sterile inflammation and provide insights into a novel hematopoietic niche that might serve as a useful target in several pathologies, such as chronic granulomatous disease (GGD), or immune reconstitution after BM transplantation.

X-linked CGD mice (on C57BL/6 background) (23) that contain disrupted alleles of the gene encoding gp91phox were purchased from the Jackson Laboratory. In this study, 8–12-wk-old male mice were used. In all of the experiments performed with CGD mice, we used C57BL/6 mice of the same age as wild-type (WT) controls. Every donor and recipient mouse used in the transplantation experiments was carefully genotyped for its CD45.1 and/or CD45.2 expression before the actual experiment. The numbers of mice analyzed per group for these transplant experiments are indicated in the figures; the data presented are a combination of three sets of separate experiments. Mice conditionally expressing EGFP (eGFP loxP/loxP) and myeloid-specific Cre mice were purchased from the Jackson Laboratory. Myeloid-specific EGFP gene expression was achieved by breeding EGFP mice with myeloid-specific Cre mice (LyzM-EGFP mice). All animal manipulations were conducted in accordance with the Animal Welfare Guidelines of the Children’s Hospital Boston. All procedures were approved and monitored by the Children’s Hospital Animal Care and Use Committee.

CGD or WT mice were left uninjected or were injected i.p. with 1 ml of 3% thioglycollate (TG; Fluka) in distilled water. The differential peritoneal cell count was determined by microscopic analysis of Wright-Giemsa–stained cytospins (24). At various times after injection, the mice were sacrificed, and inflammation-induced granulopoiesis was analyzed using the BM cells.

After anesthesia with ketamine hydrochloride (100 mg/kg i.p.) and xylazine (10 mg/kg i.p.), mouse trachea was surgically exposed, and a total volume of 50 μl of saline or hydrochloric acid (0.1 N HCl [pH 1]; Sigma-Aldrich) was instilled intratracheally to the left bronchus. Colloidal carbon (1%) was included in the instillate to indicate deposition (25). At the end of the experiments, mice were euthanized by CO₂.

Adult WT and CGD mice were anesthetized and immediately bled retro-orbitally into an EDTA-coated tube. Complete blood counts were performed using an automated hematology analyzer (Hemavet 850; Drew Scientific, Oxford, CT). For BM cells, the total cell counts were determined using a hemocytometer, and the differential cell counts were conducted by microscopic analysis of Wright-Giemsa–stained cytospin or FACS analysis using a FACSCanto II flow cytometer and FACSDiva software (BD Biosciences). The absolute number of neutrophils was determined based on the cytospin or FACS analysis.

Mice used for analysis were 7–9-wk-old males. Single-cell suspensions of BM were obtained by crushing both tibias and femurs using a mortar and pestle and filtering through 40-μm cell strainers. Erythrocytes in the sample were lysed with an ACK lysis buffer (Gibco BRL). The cells were washed with a buffer containing 2% FCS in PBS. The Abs used for flow cytometry included the following: allophycocyanin-conjugated lineage markers specific for CD3e (145-2C110), CD4 (RM4-5), CD8a (53-6.7), CD11b (M1/70), B220 (RA3-6B2), GR-1 (RB6-8C5), and Ter119 (TER119) (eBioscience, BioLegend, or BD Pharmingen). Other Abs included PC-Cy7– or FITC-conjugated Sca-1 (D7), allophycocyanin-conjugated c-kit (2B8), allophycocyanin-conjugated CD45.2 (104), PE-Cy5–conjugated CD3e (145-2C11), PE-conjugated CD45.1 (A20), PE-conjugated CD36/32 (93), and FITC-conjugated CD34 (RAM34). Unstained cells were used as negative control to establish the flow cytometer voltage setting, and single-color positive controls were used for adjustment of the compensation. The flow cytometric data were acquired using a FACSCalibur, and raw data were analyzed with FlowJo software (TreeStar). The HSC population was defined as Lin⁻IL-7Rα⁻cKit⁺Sca1⁺ cells; the hematopoietic progenitor cell population was defined as Lin⁻IL-7Rα⁻cKit⁺Sca1⁻ cells; the common lymphoid progenitor population was defined as Lin⁻IL-7Rα⁺cKit^intSca1^int cells; the CMP population was defined as Lin⁻IL-7Rα⁻c-Kit⁺Sca-1⁻FcγII/IIIR^intCD34⁺ cells; the GMP population was defined as Lin⁻IL-7Rα⁻c-Kit⁺Sca-1⁻FcγII/IIIR⁺CD34⁺ cells; and the MEP population was defined as FcγII/IIIR^lowCD34^lowc-Kit⁺Sca-1⁻IL-7Rα⁻ cells.

BM cells were resuspended in 3 ml of IMDM buffer and loaded on the top of Histopaque 1083 (Sigma-Aldrich) to prepare low-density BM cells. Briefly, cells were centrifuged for 25 min at 1700 rpm with the brake off. The intermediate cell layer was removed and transferred to a 50-ml tube. The cell suspension was centrifuged for 5 min at 1500 rpm, and the cell pellet was resuspended in 1 ml of PBS with 2% PBS. For sorting, cells were stained with the allophycocyanin-conjugated lineage-specific markers, PE-conjugated c-kit (2B8) and FITC-conjugated Sca-1 (D7). Lin^–c-kit⁺Sca-1⁺ (LSK) and Lin^–c-kit⁺Sca-1⁻ (LK) cells were sorted using a FACSAria II equipped with FACSDiva software (BD Biosciences).

BM cells (2 × 10⁴) from WT or CGD mice were seeded in semisolid MethoCult GF M3534 medium containing recombinant mouse SCF, recombinant mouse IL-3, and recombinant human IL-6 for detection of granulocyte/monocyte CFU (CFU-GM) (STEMCELL Technologies). The number of colonies that contained >50 cells was counted on day 7. Colony numbers from 20,000 BM-derived mononuclear cells (BMMCs) are shown. The size of the colonies was also measured on day 7. Colony morphology was scored based on STEMCELL Technologies criteria. l-Buthionine–sulfoximine (BSO; Sigma-Aldrich) was added to methylcellulose media at the indicated concentrations at the time of plating.

ROS accumulation in the BM during acute inflammation was measured in freshly isolated BM using an Amplex Red hydrogen peroxide assay kit. Amplex Red (Invitrogen), a H₂O₂-sensitive fluorescent probe, was prepared according to the manufacturer’s instructions. WT and CGD mice were injected i.p. with 1 ml of 3% TG in PBS. At the indicated times, mice were euthanized, and BM was prepared by spinning femurs and tibias with 100 μl of Krebs-Ringer phosphate buffer containing 5.5 mM glucose (pH 7.35). After further centrifugation (180 × g for 5 min) of the collected BM samples, the BM supernatant (extracellular ROS) was harvested, and 50 μl was assayed (in duplicated) in 96-well fluorescent assay plates (Thermo Fisher Scientific) containing 50 μl per well of Amplex Red solution with 0.2 U HRP. Fluorescence was recorded using a fluorometer (excitation, 540 nm; emission, 590 nm). The concentration of H₂O₂ was determined using a standard curve.

Cell proliferation was determined using a BrdU labeling kit (BD Biosciences) or a Click-iT Plus EdU Flow Cytometry Assay Kit (Invitrogen). Twenty-four hours before sacrifice, BrdU was administered by i.p. injection (2 mg per mouse in 200 μl of PBS) as a single dose. At the indicated time points, LSK, GMP, CMP, and MEP cells were sorted from BMMCs. Sorted cells were fixed in 70% ethanol overnight at −20°C, denatured in 2 N HCl/0.5% Triton X-100 for 20 min at room temperature, neutralized with 0.1 M sodium borate for 5 min, and stained with anti-BrdU Ab (BD Biosciences) for 30 min at room temperature in PBS/0.5% BSA/0.5% Tween 20. Cells were resuspended in 500 μl of PBS containing 10 μg of RNase A and 5 μg of propidium iodide, incubated for 30 min, and immediately analyzed by flow cytometry. The mean frequencies of BrdU⁺ cells in the HSC and each progenitor population were measured. For the EdU-incorporation assay, EdU was administered by i.p. injection (0.5 mg per mouse in 200 μl of PBS) as a single dose 24 h before sacrificing the mice. The isolated cells were prepared and stained following a protocol provided by the manufacturer (Invitrogen).

Neutrophil depletion was achieved by i.p. injection of anti-Gr1 mAb RB6-8C5 (200 μg/kg). The Ab was administered i.p. to obtain a sustained depletion over the first 48 h of the experiment. Differential WBC count using Wright-Giemsa staining was performed to confirm that the neutrophil depletion was successful.

Recombinant G-CSF (Amgen) was diluted in sterile PBS and was administered by s.c. injection (250 μg/kg body weight). BM cells and peripheral blood were collected 24 h following the G-CSF administration. H₂O₂ production in the BM was measured as described above. To neutralize G-CSF in vivo, mice were injected s.c. with 100 μg of anti-mouse G-CSF Ab (clone 67604; R&D Systems). Hematopoietic cell lineage analysis and H₂O₂ measurement were conducted 24 h following the Ab administration.

Age-matched C57BL/6 and CD45.1 mice were purchased from the Jackson Laboratory. Donor (CD45.1 mice) whole BM cells were prepared by spinning femurs and tibias under sterile conditions, and RBCs were lysed using ACK lysing buffer. LK progenitor cells were sorted using a FACSAria II equipped with FACSDiva software (BD Biosciences), as previously described (26). The transplantation was conducted using nonirradiated WT (CD45.2) and CGD (CD45.2) recipient mice. The donor LK cells (CD45.1, 2 × 10⁵) were transplanted into each nonirradiated WT (CD45.2) and CGD (CD45.2) recipient mouse via tail vein injection. To increase the efficiency of engraftment, LK cells were transplanted into recipient mice every 2 d for 1 wk. Hematopoietic chimerism was analyzed by FACS. In WT and CGD recipient mice, donor CD45.1 LK cells successfully engrafted with stable chimerism ∼ 0.4% over 6 wk. Acute peritonitis was induced using TG (3%) 6 wk after the first BM transplantation. The emergency granulopoiesis elicited by inflammation was assessed 36 h after the TG injection.

WT and CGD mice were treated with 3% TG (PBS as control) or 1 × 10⁶ Escherichia coli (PBS as control) i.p. for 24 h. LK cells were sorted using a FACSAria II cell sorter. To obtain enough materials for Western blotting, three mice were used for one data point in a single experiment. The sorted LK cells from each of the three mice were combined, and 500,000 LK cells were used for the assay. The cell pellets were lysed with 1× lysis buffer (30% 4× Invitrogen Nu-Page LDS buffer, 6% 2-ME, and 8% protease inhibitor mixture in PBS, 95°C). Cell extracts were resolved on Nu-Page 4–12% Bis-Tris gels, transferred onto polyvinylidene difluoride membranes, and immunoblotted against S473P-Akt Ab (1:1000) or Akt Ab (1:5000). Densitometry of the blots was performed using the ImageJ software Gel Analyzer plug-in. Phospho-Akt levels were then normalized based on total Akt levels (27). For phosphatase and tensin homolog (PTEN) oxidation analysis, the cell pellets were lysed with nonreducing LDS loading buffer. Cell extracts were then resolved on nonreducing SDS-PAGE. Reduced and oxidized PTEN could be detected using a specific PTEN Ab (28).

Mice were anesthetized with ketamine hydrochloride (100 mg/kg i.p.) and xylazine (10 mg/kg i.p.), and frontoparietal skull bone was exposed and prepared following previously established protocols (29). Two-photon microscopy on the calvarium BM was performed using an Olympus BX50WI fluorescence microscope equipped with a 20×, 0.95 NA objective (Olympus) and a Bio-Rad Radiance 2000MP Multiphoton system, controlled by LaserSharp software (Bio-Rad). For two-photon excitation and second harmonic generation, a Mai Tai Ti:Sapphire laser was tuned to a range of wavelengths (800–875 nm). The blood vesicles were labeled with tetramethylrhodamine-dextran (Invitrogen; 2,000,000 m.w.). Myeloid cells in the mice expressing EGFP (LyzM-EGFP) were detected via eGFP fluorescence, and bone was visualized by its second harmonic generation signal. To detect H₂O₂ in the BM, Peroxy Orange 1 (PO1; 0.1 mM in 100 μl of PBS, a generous gift from C. Chang, University of California, Berkeley) was injected i.v. Images were recorded every 30–40 s for 10 min. The generated sequences of image stacks were transformed into volume-rendered four-dimensional movies using Volocity (Perkin Elmer/Improvision), which was also used for tracking of cell motility in three dimensions.

Mice were perfused postmortem with 10 ml of paraformaldehyde-lysine-periodate fixative through the vena cava to achieve rapid in situ fixation and optimal preservation of the BM tissue. Femoral bones were isolated, fixed in paraformaldehyde-lysine-periodate for 4–8 h, rehydrated in 30% sucrose/PBS for 48 h, and snap-frozen in OCT (Tissue-Tek). Cryosections of nondecalcified whole longitudinal femoral bones were obtained using a Leica Cryostat and the CryoJane Tape-Transfer System (Leica Microsystems). BM sections were stained with the indicated Abs. DAPI (Invitrogen) staining was used for nuclear detection, and sections were mounted with VECTASHIELD mounting medium (Vector Labs) for immunofluorescence. High-resolution images of whole longitudinal immunostained femoral sections were obtained with a iCys Research Imaging Cytometer (Compucyte) equipped with four laser lines (405, 488, 561, and 633 nm) and four PMT detectors with bandpass emission filters at 450/40, 521/15, 575/50, and 650 LP.

WT and CGD mice were anesthetized and instilled with HCl, as described above. After 24 h, mice were euthanized by CO₂. The chest cavity was opened, and a catheter was tied to the trachea. Bronchoalveolar lavage was performed (1 ml of PBS/15 mM EDTA × 10) in each group. Bronchoalveolar lavage fluid (BALF) was centrifuged at 450 × g for 10 min, and the total and differential cell counts were determined from the pelleted cell fraction (25). The number of neutrophils in BALF was quantified by morphometric analyses of Wright-Giemsa–stained BALF cells. For morphometric examinations, investigators were blinded to the identities or the treatment of the mice.

Results are presented as means, with error bars indicating the SD. Differences between groups were assessed with the Student t test, unless noted otherwise. The p values < 0.05 were considered statistically significant. The normality of the data was confirmed by the Shapiro–Wilk normality test. All statistical tests were performed and graphics were created using GraphPad Prism (GraphPad, San Diego, CA) or SPSS Statistics (IBM, Armonk, NY) software.

To study reactive granulopoiesis during sterile inflammation, we used a mouse TG-elicited peritonitis model (27, 30). TG is a polysaccharide mixture that is commonly used to induce acute mild peritonitis. As a result of TG-induced acute peritoneal inflammation, neutrophils were mobilized from the BM, resulting in an elevated peripheral blood neutrophil count and a decreased BM neutrophil count (Fig. 1A, Supplemental Fig. 1A). After this initial immediate reduction in BM neutrophils, the BM neutrophil count gradually increased, indicating inflammation-induced granulopoiesis (Fig. 1A). To ensure that the TG-elicited effects were not due to endotoxin contamination, TG endotoxin levels were measured and were found to be much lower than those that would be expected to elicit a detectable response in mice (Supplemental Fig. 1B).

We next used FACS analysis to examine whether TG-induced sterile inflammation alters the number of hematopoietic progenitor cells. Similar to observations in bacteria-elicited emergency granulopoiesis (22), the percentage of GMPs (Lin⁻Sca-1^lowc-Kit⁺CD34⁺FcγR^high), but not CMPs (Lin⁻Sca-1^lowc-Kit⁺CD34⁺FcγR^low) or MEPs (Lin⁻Sca-1^lowc-Kit⁺CD34⁻FcγR⁻), increased significantly after TG challenge (Fig. 1B). Thus, TG-induced sterile inflammation specifically modulates myelopoiesis, and the increase in GMPs suggests that differentiation and proliferation of myeloid progenitor cells are specifically enhanced in TG-treated mice (Fig. 1B). Interestingly, and similar to previous reports (31, 32), TG-elicited inflammation also led to a significant expansion of LSK cells (Supplemental Fig. 1C). Consistent with specific GMP amplification, TG specifically augmented BrdU incorporation in GMPs, indicating an elevated proliferation rate in this population (Fig. 1C). Finally, BM from TG-treated mice contained more granulocytic and/or monocytic CFU (CFU-GM, CFU-M, and CFU-G) (>60/20,000 BM cells) than did untreated controls (∼40/20,000 BM cells). Colonies originating from TG-treated animals (>250 μm in diameter) were also significantly larger than control colonies (∼150 μm in diameter) (Fig. 1D). Collectively, these results indicate that TG-induced sterile inflammation can increase granulopoiesis.

ROS are key mediators of bacteria-elicited emergency granulopoiesis (22). To explore whether ROS also play a role in TG-induced reactive granulopoiesis, we first measured ROS levels in the BM using Amplex Red. During TG-induced acute sterile inflammation, H₂O₂ levels in BM extracellular space increased gradually, peaking by 48 h (Fig. 2A, Supplemental Fig. 1D). The results were similar when H₂O₂ was measured using another H₂O₂-specific probe, PO1, by multiphoton intravital microscopy of the BM microenvironment in live animals (Fig. 2B). The BM cavity in the frontoparietal skull was imaged and analyzed using multiphoton intravital microscopy 5 min after i.v. PO1 injection; mice challenged with TG for 24 h had significantly greater PO1 staining. Notably, the neutrophil count was reduced in mice treated with 5FU; however, the BM ROS level was not increased during the neutrophil-recovery phase (Supplemental Fig. 1E). Similarly, depletion of neutrophils with anti-Gr1 Ab failed to augment the BM ROS level (Supplemental Fig. 1F). These results suggest that the induction of BM ROS during inflammation is likely mediated by inflammatory factors, but is not simply due to rapid granulopoiesis caused by the stress of neutropenia.

Myeloid cells produce large amounts of ROS during acute infection and inflammation. These ROS are primarily generated by phagocytic NADPH oxidase (NOX2), and their classical function is to facilitate phagocytic killing of pathogens during infection (33, 34). To explore whether NOX2 is critical for TG-elicited ROS production, ROS levels were measured in TG-challenged CGD mice, in which the NADPH oxidase holoenzyme gp91 subunit is disrupted (23, 35). NOX2 deficiency significantly reduced BM ROS levels in TG-challenged mice (Fig. 2C, 2D). TG-induced ROS production was likely to be mediated by Gr1⁺ myeloid cells, because depletion of these cells with Gr1 Abs abolished TG-elicited ROS production (Supplemental Fig. 1D). Indeed, more than half of the cells in the BM were myeloid, with acute inflammation–elicited H₂O₂ production adjacent to these myeloid cells (Fig. 2E). We previously showed that every BM c-Kit⁺ progenitor cell is adjacent to at least one Gr1⁺ myeloid cell (22); thus, these progenitors are all surrounded by ROS in the BM.

NADPH oxidase–mediated ROS appeared to be essential for TG-induced reactive granulopoiesis. TG-elicited GMP (Fig. 3A, 3B) and LSK cell (Supplemental Fig. 2A) expansion was completely abolished in CGD mice. In addition, disruption of NADPH oxidase inhibited the TG-induced increase in the cell-proliferation rate in the GMP population (Fig. 3C). Consistently, CFU-GM assays showed that inhibition of NOX2-dependent ROS production significantly decreased the number of BM myeloid progenitor cells (Fig. 3D). Homeostatic signals were proposed to promote progenitor proliferation in response to the loss of BM neutrophils (15). We examined neutrophil counts in the peripheral blood (Supplemental Fig. 2B) and neutrophil recruitment to the inflamed peritoneal cavity (Supplemental Fig. 2C) shortly after the TG injection and did not detect any difference between WT and CGD mice. This suggests that NOX2-mediated ROS production was not required for neutrophil mobilization from the BM. Thus, the impaired TG-induced reactive granulopoiesis observed in CGD mice was not due to alteration of neutrophil mobilization and the subsequent homeostatic signals. A previous study showed that reducing ROS levels with ROS scavenger N-acetyl cysteine (NAC) in myeloid progenitor cells suppresses the colony-forming capability of the progenitors in vitro (22). We next used NAC to examine the involvement of ROS in reactive granulopoiesis. NAC treatment effectively reduced ROS levels in vivo in TG-challenged mice (Supplemental Fig. 2D) and suppressed TG-elicited GMP (Fig. 3E, 3F) and LSK (Supplemental Fig. 2E) cell expansion, confirming that ROS are required for TG-induced reactive granulopoiesis.

BSO, an inhibitor of GSH biosynthesis, was used to augment intracellular ROS levels in cultured myeloid progenitors (22). Myeloid progenitor cell proliferation is clearly enhanced by high ROS concentrations. BSO also elevated ROS levels in vivo. BSO treatment increased BM ROS levels >3-fold in WT and CGD mice (Fig. 4A). Consequently, GMP (Fig. 4B, 4C) and LSK (Fig. 4D) populations in BM were expanded significantly in WT and CGD mice, suggesting that the effect of BSO is independent of NADPH oxidase. In addition, NAC treatment counteracted the effect of BSO, confirming that the of action of BSO was indeed mediated by ROS induction. Importantly, BSO-induced granulopoiesis was largely comparable to TG-elicited reactive granulopoiesis, suggesting that TG-induced proliferation of myeloid progenitor cells may be solely mediated by ROS. Notably, BSO treatment did not induce neutrophil mobilization. The peripheral blood neutrophil count remained unaltered in the first 4 h after BSO treatment. BSO treatment increased the neutrophil count at a later time point (24 h) as a result of the proliferative response of progenitors (Fig. 4E). Finally, BSO alone did not induce G-CSF production. Thus, the BSO-induced increase in ROS levels in vivo appeared to be sufficient to elicit reactive granulopoiesis and neutrophilia independent of G-CSF (Fig. 4F).

We next explored the mechanism by which ROS mediate reactive granulopoiesis. One target of ROS-elicited protein modification is PTEN, a lipid phosphatase that negatively regulates PtdIns(3,4,5)P3 signaling. PTEN’s role in hematopoiesis and bacteria-induced emergency granulopoiesis is well documented (22, 36, 37). ROS can oxidize and inhibit PTEN activity and, consequently, upregulate the PtdIns(3,4,5)P3 signaling pathway in various cell types. Consistently, PTEN oxidation in LSK cells was detected on a nonreducing SDS-PAGE gel based on oxidation-induced mobility shift (28). TG treatment significantly increased the level of PTEN oxidation in LK cells (Fig. 5A). Disruption of NADPH oxidase prevented PTEN oxidation in vivo, confirming that this PTEN posttranslational modification event in LK cells was primarily mediated by NADPH oxidase in myeloid cells (Fig. 5B). It was reported that PTEN oxidation inhibits PTEN’s lipid phosphatase activity and, in doing so, upregulates the PtdIns(3,4,5)P3 signaling pathway. To establish whether this occurs in LK progenitor cells, the PtdIns(3,4,5)P3 signal was measured using Akt phosphorylation as a reporter (Fig. 5C). PtdIns(3,4,5)P3-dependent Akt phosphorylation in sorted WT LK progenitor cells was elevated >4-fold (30 h) in TG-induced sterile inflammation. This elevation was abolished in progenitor cells isolated from CGD mice (Fig. 5C). NAC inhibited TG-elicited upregulation of Akt phosphorylation, whereas BSO augmented Akt phosphorylation, even in the absence of TG challenge, indicating that ROS production was required and sufficient for upregulation of the PtdIns(3,4,5)P3 signaling pathway (Fig. 5D). Finally, treatment with a pan-PI3K inhibitor, LY294002, or a specific PI3Kδ inhibitor, IC87114, reduced TG-induced GMP expansion (Fig. 5E) and LSK cells (Supplemental Fig. 2F). This suggests that upregulation of PtdIns(3,4,5)P3 signaling was essential for reactive granulopoiesis. It is noteworthy that the PI3Kγ inhibitor AS605240 did not affect reactive granulopoiesis (Fig. 5E), confirming that PtdIns(3,4,5)P3 signaling in the progenitors was primarily maintained by PI3Kδ, which is downstream of receptor tyrosine kinases (38). Taken together, our results demonstrate that PTEN and PtdIns(3,4,5)P3/Akt signaling pathways are a major players in TG-induced and ROS-mediated reactive granulopoiesis.

ROS elevation during TG-induced sterile inflammation is predominantly mediated by NOX2, which is primarily expressed in BM myeloid cells. NADPH oxidase is gradually expressed during myelopoiesis. We showed previously that Gr1⁺ cells are uniformly localized in the BM, and Kit⁺ stem and progenitor cells are adjacent to Gr1⁺ cells (<50 μm) (22). Thus, the sterile inflammation–induced expansion and differentiation of myeloid progenitors can be regulated by ROS generated in the progenitor cells in an autonomous manner or by ROS generated by the surrounding Gr1⁺ relatively mature myeloid cells via a paracrine mechanism (Fig. 6A). To distinguish between these two possibilities, we conducted a BM transplantation experiment in which TG-elicited expansion and differentiation of transplanted WT or CGD progenitors were investigated in CGD or WT recipient mice (Fig. 6B). To examine the effect of BM myeloid cells on the proliferation of transplanted progenitor cells, nonirradiated mice were used as recipients. Disruption of NADPH did not affect the efficiency of donor cell engraftment. In WT and CGD recipient mice, CD45.1 LK cells successfully engrafted with stable chimerism ∼ 0.4% over 6 wk (Fig. 6B). The TG-induced augmented proliferation of transplanted WT progenitors was abolished in CGD recipient mice. In contrast, accelerated proliferation of transplanted CGD progenitors was detected in TG-challenged WT recipient mice (Fig. 6C, 6D). These results suggest that the TG-elicited reactive granulopoiesis was primarily dependent on ROS production in the recipient mice, demonstrating that BM myeloid cell–derived ROS externally regulate reactive granulopoiesis via a paracrine mechanism.

NOX2 may also be expressed in BM nonhematopoietic niche cells, although the level is low. Thus, some of the effects observed in CGD mice may be due to NADPH oxidase disruption in niche cells. To rule out this possibility, we examined the sterile inflammation–induced expansion of transplanted WT myeloid progenitors in lethally irradiated CGD recipients in which all of the hematopoietic cells were derived from WT donor hematopoietic stem/progenitor cells (Fig. 7A). In this set-up, TG-elicited reactive granulopoiesis was still observed in CGD recipient mice, indicating that NADPH oxidase in BM nonhematopoietic cells was not involved in ROS-mediated reactive granulopoiesis (Fig. 7B, 7C).

BM myeloid cell–derived ROS play an important role in eliciting reactive granulopoiesis in a TG-induced sterile peritonitis model. Although this system is a widely used and accepted acute sterile inflammation model, it has only limited clinical relevance. Thus, for clinical correlation and to test the generalizability of our findings, we next explored the role of ROS in regulating reactive granulopoiesis in the acid-elicited acute lung injury (ALI) model, another sterile inflammation model of greater clinical value.

Acid aspiration causes ALI. Acid aspiration in mice, which mimics acid-elicited human ALI, was induced experimentally by intratracheal instillation of HCl. Consistent with previous reports (39, 40), acid instillation resulted in the recruitment of significant numbers of neutrophils to the damaged lungs and caused acute sterile inflammation (Fig. 8A). Similar to TG-elicited inflammation, acid-induced inflammation was associated with augmented ROS production in the BM extracellular space. Acid-induced BM ROS production was significantly suppressed in CGD mice, suggesting that these ROS were generated by NOX2 (Fig. 8B). The severity of acid-induced lung inflammation measured by neutrophil recruitment was comparable between WT and CGD mice. Acid-induced ALI specifically expanded GMPs, an effect that was completely abolished in CGD mice and confirming that reactive granulopoiesis induced by acid-elicited ALI was a NOX2-dependent process (Fig. 8C). In addition, a lower percentage of Edu⁺ proliferating cells was present in the GMP population of acid-challenged CGD mice compared with WT mice (Fig. 8D). Finally, the CFU-GM assay showed that the number of myeloid progenitors in BM did not increase in CGD mice, again suggesting that inhibition of NOX2-dependent ROS suppresses reactive granulopoiesis in acid-elicited ALI (Fig. 8E). Taken together, our results demonstrate that ROS is the common pathway on which several clinically relevant noninfectious and infectious stimuli of emergency/reactive granulopoiesis converge, despite their distinct primary stimuli and recognition pathways.

At the early stage of inflammation, neutrophil mobilization from the BM leads to an immediate peripheral blood neutrophilia, which is followed by augmented BM granulopoiesis to compensate for their peripheral loss. In this article, we show that, although noninfectious stimuli-elicited reactive granulopoiesis and microbial infection-driven emergency granulopoiesis are triggered by different stimuli and there might also be fundamental molecular differences, NADPH oxidase–dependent ROS production is critical for both processes (Supplemental Fig. 3A). This also provides a novel therapeutic strategy for rebalancing neutrophil homeostasis in physiologic and pathologic conditions, such as immune recovery after chemotherapy or after BM transplantation.

The role of ROS in hematopoiesis is well documented (41–55). Most of these studies focused on ROS within stem or progenitor cells. In contrast, we show that microenvironmental, rather than endogenous, ROS contribute to stem and progenitor cell maintenance and proliferation during pathogen-induced emergency granulopoiesis and sterile inflammation–induced reactive granulopoiesis (Supplemental Fig. 3A). The BM is the major myeloid reservoir. A large amount of ROS is generated by Gr1⁺ myeloid cells in BM during infection or inflammation. The factors that induce NOX2 activation and ROS generation during infection and inflammation are still ill-defined. Proinflammatory factors, such as KC, MIP-2, IL-1, G-CSF, and TNF-α, can activate NADPH oxidase and elevate ROS production. ROS are essential intracellular signaling molecules. They often regulate the structure and function of target proteins via posttranslational thiol modifications, including glutathionylation, nitrosylation, and sulfenic acid and disulfide bond formation (56–58). There are numerous ROS-target proteins, such as protein kinases and phosphatases, actin, Ras GTPases, caspases, and transcription factors. In this article, we provide evidence that PTEN is a direct ROS target that plays a critical role in TG-induced reactive granulopoiesis. We demonstrate that oxidation-induced PTEN deactivation is essential, as well as sufficient, to generate ROS-elicited phenotypes. However, we cannot rule out the possibility that other ROS-mediated pathways may be involved in inflammation-induced emergency granulopoiesis.

G-CSF is a granulopoietic factor that is used clinically to produce granulocytes. G-CSF levels increase in infective and inflammatory conditions, and it is implicated in steady-state (59–61), pathogen-induced emergency granulopoiesis–triggered (11, 12), and sterile inflammation–elicited reactive granulopoiesis (9, 15). Intriguingly, TG-induced G-CSF expression (Supplemental Fig. 3B) and serum G-CSF levels (Supplemental Fig. 3C) were not suppressed in CGD mice, suggesting that G-CSF was not the major downstream ROS effector. Moreover, G-CSF signaling blockade with an anti–G-CSF Ab caused only a small reduction in BM H₂O₂ in TG-challenged mice. TG treatment still induced significant ROS production, even when G-CSF signaling was inhibited (Supplemental Fig. 3D). In addition, anti–G-CSF Ab treatment did not completely inhibit TG-elicited reactive granulopoiesis (Supplemental Fig. 3E). Finally, G-CSF–induced increases in ROS levels in BM were much smaller than those induced by TG (Supplemental Fig. 4A), but they still augmented granulopoiesis in CGD mice (22). Collectively, these results suggest that ROS and G-CSF regulate the expansion and differentiation of myeloid progenitors in parallel, further highlighting the significance of a G-CSF–independent ROS signal in granulopoiesis.

HSCs are well-known immediate targets of inflammatory signals (9). Infection and inflammation can modulate HSCs directly or indirectly (31). Consistent with this, we observed expansion of the LSK cell population in our TG-induced peritonitis model. However, unlike chronic inflammation–induced alterations in hematopoiesis, which are reported to be regulated by IFNs (31, 62–64), acute inflammation–induced LSK cell expansion is likely to be independent of IFN signaling because acute inflammation did not alter serum IFN (α, β, or γ) levels (Supplemental Fig. 4B). It appears that acute inflammation–induced LSK cell expansion is also controlled by BM ROS, but its physiologic significance is uncertain. It would be intriguing to see whether NADPH oxidase–dependent ROS production by BM myeloid cells plays a role in regulating HSC self-renewal and pluripotency. In addition, acute inflammation–elicited augmentation of cell proliferation is cell-type specific; the proliferation rate of CMPs and MEPs remained unaltered. The source of this specificity requires further clarification.

We thank John Lucky and John Manis for helpful discussions and Dr. Christopher J. Chang for providing PF6-AM dye.

The authors have no financial conflicts of interest.