A late-lineage murine neutrophil precursor population exhibits dynamic changes during demand-adapted granulopoiesis

Homeostasis of neutrophils—the blood cells that respond first to infection and tissue injury—is critical for the regulation of immune responses and regulated through granulopoiesis, a multi-stage process by which neutrophils differentiate from hematopoietic stem cells. Granulopoiesis is a highly dynamic process and altered in certain clinical conditions, such as pathologic and iatrogenic neutropenia, described as demand-adapted granulopoiesis. The regulation of granulopoiesis under stress is not completely understood because studies of granulopoiesis dynamics have been hampered by technical limitations in defining neutrophil precursors. Here, we define a population of neutrophil precursor cells in the bone marrow with unprecedented purity, characterized by the lineage−CD11b+Ly6GloLy6BintCD115−, which we call NeuPs (Neutrophil Precursors). We demonstrated that NeuPs differentiate into mature and functional neutrophils both in vitro and in vivo. By analyzing the gene expression profiles of NeuPs, we also identified NeuP stage-specific genes and characterized patterns of gene regulation throughout granulopoiesis. Importantly, we found that NeuPs have the potential to proliferate, but the proliferation decreased in multiple different hematopoietic stress settings, indicating that proliferating NeuPs are poised at a critical step to regulate granulopoiesis. Our findings will facilitate understanding how the hematopoietic system maintains homeostasis and copes with the demands of granulopoiesis.


Results
Identification of CD11b + Ly6G lo Ly6B int CD115 − neutrophil precursor cells in the BM. To define murine neutrophil precursors in the BM, we analyzed lineage-negative (CD3ε − CD19 − NK1.1 − B220 − ) BM cells for expression of CD11b, CD115, Ly6B, and Ly6G by flow cytometry (Fig. 1a). Neutrophils expressing Ly6G at high levels in the lineage-negative populations were excluded and the remaining CD11b + cells were further analyzed based on the expression level of CD115 and Ly6B. We observed that Ly6B + cells could be divided into two distinct populations based on their CD115 expression level. The Ly6B hi CD115 + population were inflammatory monocytes, as previously reported 61 . The unknown Ly6B int CD115 − population expressed Ly6B at a level comparable to that of neutrophils, but expressed Ly6G at lower levels (Fig. 1b).
We compared results obtained using our method with previous analyses of immature and mature neutrophils based on Gr-1 and CD11b expression 6,54,58,59 (Fig. 1c). The Gr-1 hi population consists of mature neutrophils while Gr-1 int CD11b + cells are a population of immature neutrophil precursors. We observed that the Ly6B int CD115 − population was found within Gr-1 int CD11b + immature neutrophils. Therefore, we suspected that the Ly6B int CD115 − cells are neutrophil precursor cells and termed them NeuPs (Neutrophil Precursors). Remarkably, we noted that Gr-1 int CD11b + cells were not pure immature neutrophil precursors, but included inflammatory monocytes (Fig. 1c). Similarly, the gating scheme shown in a recent study attempting to dissect neutrophil developmental stages based on the differential expression of c-kit and Ly6G 52 was also unable to prevent CD115 + monocyte contamination in each precursor population ( Supplementary Fig. S1).
We sorted NeuPs from BM using multicolor flow cytometry ( Supplementary Fig. S2). Cytological analysis with May-Grünwald-Giemsa staining of neutrophils, monocytes, and NeuPs isolated from BM revealed that NeuPs are morphologically distinct from neutrophils and monocytes (Fig. 1d). NeuPs had either ring-shaped or peanut-shaped nuclei, whereas neutrophils had segmented lobular nuclei. Murine neutrophil or macrophage precursor populations in the BM have been reported to have ring-shaped nuclei 62 . The size and granularity of NeuPs were similar to those of neutrophils, but distinct from those of eosinophils ( Supplementary Fig. S3a). Little or no expression of the basophil markers, DX5 and Mar-1, or the eosinophil marker, Siglec-F, was detected on NeuPs ( Supplementary Fig. S3b). NeuPs express CD24, CD44, and CD172 at high level similar to BM neutrophils and monocytes ( Supplementary Fig. S3c).
Taken together, these findings demonstrate the successful development of a novel flow cytometry scheme for isolating a neutrophil precursor population with the lineage CD11b + Ly6G lo Ly6B int CD115 − , with minimal contamination by other types of myeloid cells.

NeuPs effectively differentiate into a neutrophil population in vitro and in vivo.
To firmly establish that NeuPs have the potential to differentiate into neutrophils, we sorted NeuPs or mature neutrophils from BM and cultured them in the presence of G-CSF. After 24 hours of culture, NeuPs expressed the mature neutrophil marker, Ly6G (Fig. 2a). Furthermore, the ring-shaped nuclei of NeuPs adopted the very dynamic, lobular form of nuclei of mature neutrophils (Fig. 2b). These data indicate that NeuPs did in fact differentiate into neutrophils. We further validated the neutrophil-differentiation potential of NeuPs in vivo. Since we could not detect transferred NeuPs after intravenous injection (data not shown), we injected NeuPs directly into the BM, which is expected to provide a niche for neutrophil differentiation (Fig. 2c). NeuPs sorted from CD45.1 mice were transferred to CD45.2 congenic mice, and NeuP-derived cells were detected and analyzed by staining for congenic markers. Most transferred NeuPs displayed high levels of Ly6G expression and no other lineage markers, such as the monocyte marker CD115 (Fig. 2d), indicating that NeuPs give rise only to neutrophils in vivo. Thus, these data further confirm that NeuPs are bona fide neutrophil precursors.
Functional analysis of NeuPs and NeuP-derived neutrophils. Next, we evaluated function of NeuPs and determined whether NeuPs differentiate into functional neutrophils by examining reactive oxygen species (ROS) production and phagocytic activity. Neutrophils can be induced to produce ROS by stimulating with complement and bacterial stimulation 2,63,64 . We evaluated ex vivo ROS production ability of NeuPs and neutrophils that were freshly isolated from BM under the treatment of complement and Escherichia coli. About 80% of mature neutrophils actively produced ROS under the stimulation, whereas around 15% of NeuPs produced ROS at lower levels ( Fig. 3a). We also treated G-CSF-cultured neutrophils and NeuP-derived neutrophils with or without E. coli. NeuP-derived neutrophils become as potent ROS-producing cells as neutrophils under bacterial stimulation (Fig. 3b). In contrast to ROS production, NeuPs displayed higher capacity to phagocytose fluorescein-conjugated beads than neutrophils ex vivo (Fig. 3c). After in vitro differentiation, about 35% of NeuP-derived neutrophils engulfed E. coli, a percentage slightly greater than that for mature neutrophils (Fig. 3d). These results confirm that NeuPs are capable of differentiating into fully functional neutrophils.
Previous reports have shown that GM-CSF can induce neutrophils to adopt dendritic cell (DC)-like characteristics 65,66 . To determine whether NeuPs could be similarly induced to display DC-like properties, we treated them with GM-CSF and examined expression of CD11c, a marker of DCs. After 4 days in culture with GM-CSF, about 80% of neutrophils and NeuPs were CD11c + (Fig. 3e). Some even expressed MHC class II molecules, indicating that GM-CSF can drive DC-like differentiation of NeuPs in a manner similar to its effects on neutrophils. However, G-CSF had no such effect on neutrophils or NeuPs (Fig. 3e). Interestingly, overall survival was better with GM-CSF than with G-CSF: a number of cells still survived after 6 days of GM-CSF-supplemented culture (data not shown).

Transcriptome analysis of NeuPs.
To further substantiate the identity of NeuPs and determine the NeuP-stage specific genes, we performed genome-wide expression profiling through RNA-sequencing of sorted NeuPs, monocytes and neutrophils, and evaluated the relationships among the three populations. A principal component analysis (PCA) of all genes with two major principal components showed that NeuPs were separated from neutrophils and monocytes by a substantial distance (Fig. 4a). Pearson correlation coefficients between transcription profiles of each sample were calculated and displayed as a heat map. Correlation coefficients between NeuPs and neutrophils ranged from 0.81 to 0.85, a range higher relative to that between NeuPs and monocytes (0.76-0.82), indicating that NeuPs are more closely related to neutrophils than to monocytes (Fig. 4b). A scatter plot analysis revealed that, among the ~11,000 mRNA transcripts analyzed, 2,482 and 2,840 genes were expressed higher in NeuPs by more than a 2-fold difference in expression compared to neutrophils and monocytes, respectively (Fig. 4c). A heat map analysis of differentially expressed genes (DEG) with a false discovery rate (FDR) of less than 0.05 revealed that NeuPs clustered more closely with neutrophils than monocytes (Fig. 4d). We also determined potentially NeuP stage-specific transcription factors by analyzing the expression pattern of transcription factors among the three populations ( Fig. 4e; and Table 1). These include transcription factors previously known to be critical for neutrophil development, such as C/EBPε 23,50 , GFI1 46,49 , HLX 67 , and MYB 55 . Some identified genes, such as Hmgb3, Ssrp1 and Foxm1, have not been previously associated with neutrophil development and may be novel NeuP-specific transcription factors. Foxm1 was known to regulate expression of Prom1 gene (also known as CD133) 68 . Prom1 (CD133) expression was exclusively high on NeuPs in our transcriptome analysis, so we examined whether CD133 can be used as a novel surface marker for NeuPs. CD133 was distinctly detected on NeuPs compared to neutrophils and monocytes in the flow cytometric analysis (Fig. 4f). Overall, NeuPs showed a closer relationship to neutrophils than monocytes, but exhibited unique gene expression profiles,  indicating the dynamic nature of the neutrophil development process. To verify RNA-seq results and further specify the stage of NeuPs, we confirmed the expression levels of three major granule genes-Mpo (myeloperoxidase, primary granule marker, enriched in promyelocytes), Ltf (lactotransferrin, secondary granule marker) and Mmp9 (matrix metallopeptidase 9, tertiary granule marker, enriched in neutrophils) 15 -by RT-PCR (Fig. 4g). NeuPs expressed Mpo and Ltf at high levels and showed lower expression of Mmp9 compared to neutrophils. Consistent with their morphology (Fig. 1d), transcriptome profiling also positions NeuPs between promyelocytes and neutrophils.
To explore the dynamics of gene expression and categorize genes according to the expression patterns during granulopoiesis, we combined our RNA-seq data with GMP and neutrophil gene expression data obtained from the ImmGen database ( Supplementary Fig. S4). A total of 9,158 genes were found in common between the two datasets and were used for further analysis. We calculated relative gene expressions normalized to that of neutrophils and categorized the genes into nine gene expression profile groups based on expression patterns along the transition from GMPs to NeuPs and neutrophils (Supplementary Table S1).
A gene ontology (GO) analysis of genes in groups 1-8 showing altered gene expression was performed using DAVID (Supplementary Fig. S5; and Supplementary Table S2). Expression of genes in groups 1-3 gradually decreased, whereas that of genes in groups 5-7 increased, along the course of neutrophil maturation. Genes in groups 4 and 8 were up-regulated and down-regulated, respectively, only in NeuPs. Overall, genes in groups 1-4 are mainly involved in cell cycle control and metabolic process related to the synthesis of genetic materials or cellular components. Groups 1-3 genes included those encoding transcription and translation machinery components, such as transcription factors, translation initiators, and aminoacyl tRNA synthetases. Although catalogued as metabolic genes, each group contained distinct gene sets. Group    synthesis and protein modification, whereas group 2 included genes for carbohydrate, lipid and protein anabolic and catabolic processes, and group 3 genes are involved in DNA replication and repair. Notably, group 4 included genes involved in stimulating cell division and coordinating cell cycle phases, such as checkpoint and chromatin segregation, implying that control of cell division is critical task at the NeuP stage. In contrast, genes in groups 5-8 are related to development of the immune system and immune responses. Notably, groups 5 and 6 included genes that function in innate immune signaling, such as Tlr (-2, -4, -6 and -8), Myd88, Ddx58, Dhx58, Nlrp3, and Oas family genes. Expression of cytokine genes, such as Il1b and Il15, was also upregulated as neutrophils mature. Collectively, the results of our GO term analysis reveal that, during maturation, neutrophil precursors come to express genes for immune-effector functions, concomitantly losing expression of genes for metabolism and proliferation, a transition that is highlighted at the NeuP stage.
Because increasing evidence suggests that epigenetic regulation plays a critical role in controlling hematopoiesis 69,70 , we selected and analyzed DEGs that serve as epigenetic regulators. Epigenetic regulators were detected in five groups (groups 1, 2, 4, 6, and 7). Each group included distinct types of chromatin remodelers and epigenetic (e) Heat map of genes encoding transcription factors whose expression showed more than a 3-fold change between cell types and less than a 2-fold difference between replicates. Pearson distance is used as a distance function. Genes were clustered into six groups and are listed in Table 1. (f) Expression level of CD133, novel candidate for NeuP specific surface marker, on NeuPs, neutrophils and monocytes were detected with FACS. (g) RT-PCR of genes for granule markers in NeuPs and neutrophils. This is a processed and cropped version. The original full-length gel image is presented in the Supplementary Figure S9. (h) Expression of neutrophil lineage genes (Elane, S100a8, Lyz2, and Ltf) among GMPs, NeuPs, and neutrophils. Relative expression compared to neutrophils is shown. (i-j) EYFP labeling of GMPs, NeuPs, neutrophils, and monocytes based on Lyz2 expression by using Lyz2-cre; ROSA-eYFP mice (i) or on S100a8 expression by using S100A8-cre; ROSA-eYFP mice (j). Percentage of eYFP-expressing cells, determined in comparison to WT mice, is shown. Data are representative of two independent experiments. regulators involved in histone modifications ( Supplementary Fig. S6). This analysis suggests the possibility that stages of granulopoiesis are determined by distinct types of epigenetic regulators; further investigation will be required to corroborate this notion.
Previous data showed that mice deficient of JDP2, a transcription regulator of the AP-1 family, exhibited accumulation of immature neutrophils and primary granule genes. JDP2 suppresses expression of ATF3 which negatively regulates neutrophil differentiation via inhibiting histone acetylation at the promoter of ATF3 71 . It was shown that JDP2 recruits HDAC families (HDAC1, 2-6, and 10) to the ATF3 promoter 72 . In our analysis, Jdp2 was detected as a neutrophil lineage specific transcription factor compared to monocytes (Table 1, cluster 1) and within group 7 (Supplementary Table S1) in which gene expression is highly upregulated in NeuPs and neutrophils compared to GMPs. In contrast, HDAC2 and 6, found within group 2, were highly expressed at the GMP stage and then significantly decreased at the NeuP stage. This inverse correlation in expression pattern of JDP2 and HDAC family proteins may explain the potential role of JDP2 in regulating neutrophil development at early stages of granulopoiesis.
Many genetically modified strains of mice that harbor genes encoding fluorescence reporter proteins have been used to track down specific lineages. Similarly, mice expressing cre recombinase together with a diphtheria toxin or its receptor have been developed in order to trace or conditionally delete cells of a target lineage 73 . Gene loci encoding S100A8 74,75 , Lyz2 76,77 , Ltf 78 , Elane 79 , and Ly6g 80 have been used for targeting neutrophils, although some other types of myeloid cells are labeled to a certain degree. For example, monocytes and macrophages are also labeled in Lyz2(LysM)-cre mice. Previously, it was shown that the labeling efficiencies vary among the neutrophil specific cre strains: S100A8(MRP8)-cre mice showed over 80% recombination efficiency in neutrophils whereas Lyz2-cre and GE (Elane)-cre mice showed 50-70% recombination rates 73 . We examined expression of the neutrophil specific genes in GMPs, NeuPs, and neutrophils from our analysis (Fig. 4h). Elane, found in group 1, is expressed 200-fold and 20-fold higher in GMPs and NeuPs than neutrophils, respectively. Expression of S100A8, Lyz2, and Ltf, categorized as group 7 genes, is much higher in NeuPs and neutrophils than in GMPs. NeuPs showed around 100-, 8-, 300-fold increase in expression of S100A8, Lyz2, and Ltf than GMPs, respectively. However, the three genes are expressed similarly between NeuPs and neutrophils. Thus, we suspected that NeuPs could be marked by a cre recombinase that is expressed under the control of S100A8, Lyz2 or Ltf. To test the possibility, we crossed Lyz2-cre and S100A8-cre mice with ROSA-eYFP mice in which a loxP-flanked transcription STOP sequence is followed by an eYFP-encoding cassette at the ROSA locus 81 and traced the history of Lyz2 and S100A8 expression by eYFP expression. In Lyz2-cre +/− ; ROSA-eYFP +/− mice, about 25% of NeuPs expressed eYFP (Fig. 4i). In accordance with Lyz2 expression that is higher in NeuPs than in GMPs (Fig. 4h), eYFP-expressing cells were around 10% more abundant in NeuPs compared to GMPs (Fig. 4i). Importantly, higher percentage of neutrophils in BM and blood were labeled than NeuPs. Furthermore, 40% of NeuPs and about 70% of neutrophils were eYFP-positive in S100A8-cre; ROSA-eYFP +/− mice, yet little GMPs were labeled by eYFP (Fig. 4j). These results corroborate that NeuPs precede neutrophils in development.
NeuPs have the proliferative potential, which is modulated during demand-adapted granulopoiesis. Our gene expression profiling of NeuPs revealed that genes regulating cell cycle are enriched in NeuPs ( Supplementary Fig. S5) and that NeuPs are developmentally located between proliferating promyelocytes and non-proliferating neutrophils (Fig. 4g). Therefore, we examined whether NeuPs have the proliferative potential by incorporation of EdU (5-ethynyl-2′ -deoxyuridine) in NeuPs, an analog of thymidine, as a marker of DNA synthesis. EdU was injected intravenously into mice 2 hours before harvesting BM cells. Compared to neutrophil and inflammatory monocyte populations, which had few EdU + cells, approximately 40% of NeuPs were EdU + , indicating that NeuPs are actively proliferating cells (Fig. 5a). Since the EdU + cells may have acquired EdU before NeuP stage, we first sorted NeuPs and then immediately labelled in vitro the proliferating NeuPs by incubating with EdU (Fig. 5b). We observed EdU-stained NeuPs in this setting, confirming NeuP's intrinsic proliferative potential. Similar results were obtained by staining for Ki-67, another marker of proliferating cells (Fig. 5c). As observed in proliferating early hematogenic precursors and neutrophil precursor cells 52,56 , NeuPs also expressed c-kit (Fig. 5d). Stress conditions that result in a severe reduction in neutrophils can trigger demand-adapted granulopoiesis 16,17 . Our isolation of highly pure neutrophil precursors enabled us to track dynamic alterations in neutrophil precursors in terms of number and proliferation potential during emergency granulopoiesis. Injection of G-CSF or LPS, mimicking emergency conditions, acutely enhances granulopoiesis 38 . Accordingly, we administered G-CSF to mice and examined NeuPs in bone marrow 24 hours later. G-CSF treatment doubled the number of bone marrow-resident NeuPs (Fig. 6a). No significant changes in the number of whole bone marrow cells or neutrophils were detected within this period (Supplementary Fig. S7a). Notably, the number of EdU + proliferating cells significantly decreased after G-CSF treatment (Fig. 6b,c). Next, we analyzed in vitro proliferation of NeuPs sorted from the G-CSF-or PBS-treated mice and observed significantly less EdU + NeuPs from G-CSF-stimulated mice, compared to PBS-treated control mice (Fig. 6d,e). This indicates that G-CSF-stimulation decreased the proliferative potential of the cells at the NeuP stage. Similarly, systemic LPS stimulation (intravenous injection) reduced the percentage of proliferating NeuPs 1 day after LPS stimulation (Fig. 6g,h), but we found an increase, but not significant, of bone marrow NeuPs at this time point (Fig. 6f). Of note, the number of whole bone marrow cells and neutrophils significantly decreased in LPS-treated mice compared with control mice (Supplementary Fig. S7b). 82,83 . NeuPs significantly increased in the bone marrow 48 hours after local infection, and the proportion of proliferating cells substantially decreased (Fig. 6i-k). However, NeuP cell numbers and the fraction of proliferating NeuPs did not change with systemic infection (Fig. 6i,j), but interestingly, the degree of EdU staining was decreased (Fig. 6k). The number of total bone marrow cells and neutrophils decreased in both cases, but more severely in systemic infection (Supplementary Fig. S7c). These results indicate that the behavior of neutrophil precursors varies dramatically depending on the type of infection.

Next, we examined whether local and systemic bacterial infection resulted in different NeuP dynamics. Footpad injection of Listeria induces local infection, whereas intravenous injection leads to systemic infection
We observed an increase in blood neutrophil numbers after ablation of monocytes/macrophages by treatment with clodronate-liposomes (data not shown). Thus, we wondered whether a non-infectious condition that might perturb the local bone marrow niche would affect granulopoiesis. To test the possibility, we treated mice with clodronate-or PBS (control)-liposomes and analyzed NeuPs. Clodronate-liposomes depleted circulating monocytes without significantly decreasing the number of total bone marrow cells or neutrophils ( Supplementary Fig. S7d). Bone marrow-resident NeuP numbers trended slightly higher, although this difference was not significant (Fig. 6l). The EdU + fraction of NeuPs, however, significantly decreased (Fig. 6m,n). Collectively, these results confirm that various conditions can trigger granulopoiesis and the changes in NeuPs are not uniform, suggesting a dynamic response.
The increased number of NeuPs under demand-adapted granulopoiesis may be due to the increase of influx from the upstream precursors. However, GMP did not show any difference both in cell numbers and EdU incorporation under G-CSF stimulation, compared to PBS-treated mice (Fig. 7a,b). Next, we examined whether the differentiation kinetics has changed. We performed a pulse-chase experiment with EdU by injecting EdU along with G-CSF or PBS treatment and let EdU-labelled precursors differentiate into neutrophils for 24 hours, rather than 2 hours, and compared the percentage of EdU-labelled neutrophils. We could see increase of EdU-incorporated neutrophil populations with G-CSF-, but not with PBS-treatment (Fig. 7c). This result indicates that neutrophil differentiation has been accelerated by G-CSF stimulation, which may contribute to the accumulation of NeuPs in G-CSF-stimulated mice.

Discussion
Here, we demonstrated that Ly6B, in combination with CD115, can be used to isolate and analyze neutrophil precursors, NeuPs, from bone marrow. Our results from flow cytometry, morphological analysis, gene expression profiling, in vitro culture, and in vivo transfer experiments provide convincing evidence that NeuPs are a population with the potential to proliferate in response to various stimuli and differentiate into functional neutrophils.
A transcriptome analysis of NeuPs revealed that this population is more closely related to neutrophils than monocytes. However, the gene expression profile of NeuPs is quite distinct from both, indicating that neutrophils undergo highly dynamic changes in gene expression during the maturation process. Combining microarray data for GMPs and neutrophils in the ImmGen database with our RNA-seq data, we determined stage-specific genes and functionally categorized gene expression patterns throughout granulopoiesis. This analysis identified many novel candidate genes, including transcription factors, that were previously uncharacterized in the context of neutrophil development and we are actively investigating their roles.
The importance of epigenetic regulation is an emergent theme in neutrophil development [69][70][71]84 . However, epigenetic factors that regulate granulopoiesis have not been determined. Our study identified potential NeuP-specific epigenetic regulators ( Supplementary Fig. S6). One such regulator, EZH2 is a core component of PRC2, which acts as a H3K27 methyl-transferase. Proper function of the PRC2 complex is known to play an essential role from the early stage of development by controlling plasticity and differentiation of stem cells 85 . EZH2 has been suggested to be a positive regulator of hematopoiesis 86 . Expression of the gene encoding SUZ12, another component of the PRC2 complex, was also found to be NeuP-specific. Whether these are truly stage-specific epigenetic regulators that are crucial for the transition to neutrophils is a topic of great interest and a focus of our current research efforts. We also found that different members within an epigenetic family, such as HDAC, CHD, and KDM families, are expressed in a stage-specific manner, providing a model in which each member within a family plays a distinct role at a different stage of differentiation, despite the overall structural and functional similarities of family members.
Our purification of NeuP made it possible to more precisely track the behavior of neutrophil precursors in multiple types of emergency circumstances, including G-CSF and LPS stimulation, Listeria infection, and myeloid ablation by clodronate-liposomes. Our data revealed a common behavior of neutrophil precursors. NeuP cell numbers increased in most cases, but the proportion of proliferating cells significantly decreased, except under systemic infection condition. We suspect that the NeuP population is a heterogeneous mixture of myelocytes and metamyelocytes, given that micrographs showed two morphologically distinct cell types (Fig. 1d). Myelocytes are more proliferative than metamyelocytes 52 , and if the differentiation balance between the two populations is shifted toward metamyelocytes during demand-associated granulopoiesis, the proportion of proliferating NeuPs would decrease, which is in line with our observations. We examined whether morphological shift (ring-vs peanut-shaped nuclei) occurred in the NeuP populations under G-CSF stimulation, but did not observe any significant alteration in the ratio of the two types of NeuPs. (Supplementary Fig. S8). A recent study demonstrated that ROS production by bone marrow myeloid cells is critical for controlling emergency granulopoiesis through induction of the expansion of early progenitors such as GMPs 87 . Notably, we observed a trend toward an increase in NeuP cell numbers under most test conditions, the exception being systemic bacterial infection. Hence, it is conceivable that granulopoiesis triggered by infection or cytokines leads to an increase in NeuP cell Scientific RepoRts | 7:39804 | DOI: 10.1038/srep39804 numbers, probably due to increased input from precursor cells prior to the NeuP stage and acceleration of the differentiation process, leaving fewer non-proliferating cells in the NeuP stage. However, we did not observe any difference in cell numbers and EdU incorporation of GMP (Fig. 7a,b) in our settings. Rather, we observed overall accelerated differentiation of neutrophils during the demand-adaptive granulopoiesis (Fig. 7c), supporting the possibility that altered EdU incorporation may be caused by differential modulation of residence time at a certain stage(s) of neutrophil differentiation. When the acceleration occurs and which stage of neutrophil differentiation is a critical rate-determining step is not clear. This will be answered when a methodology to definitely sort out neutrophil precursors upstream of NeuPs is devised. Although speculative, it is possible that blood cell precursors must "choose" between differentiation and proliferation along the developmental path in response to various environmental cues, and this leads to a tipping of the balance toward differentiation, a dynamic response that we consider constitutes the hematopoietic or granulopoietic triage process.
In summary, we have devised a new method for isolating neutrophil precursors using the surface marker Ly6B. To our knowledge, this is the first report of a neutrophil precursor population that has been isolated at high purity by flow cytometry. This CD11b + Ly6G lo Ly6B int CD115 − NeuP population is proliferative and has the potential to differentiate into neutrophils in vitro and in vivo. In demand-adapted granulopoiesis, NeuPs increased in numbers and were skewed toward differentiation rather than proliferation. Gene expression profiling analyses suggested numerous factors that might be critical for neutrophil differentiation at the NeuP stage. Collectively, our findings provide insight that could lead to a deeper understanding of the control of granulopoiesis and may also help improve strategies for treating diseases caused by uncontrolled granulopoiesis or myeloablation. Notably, a recent report has shown that immature Ly6G int cells with ring-shaped nuclei circulate in tumor-bearing mice and have tumor-promoting roles 88 . We suspect that these Ly6G int cells are related to NeuPs, a possibility that we are currently exploring.
Lyz2-cre (Lyz2 tm1 ( cre ) Ifo ) and ROSA-eYFP (Gt(ROSA)26Sor tm1 ( EYFP ) Cos ) were generous gifts from Drs. R.M. Locksley and C.A. Lowell (UCSF). S100A8-cre (Tg(S100A8-cre,-EGFP)1Ilw/J) 75   Intra-BM transfer. Intra-BM transfer of sorted precursors was performed as described previously 89 . Briefly, after a small incision was made around the knee to visualize the kneecap, cells were transferred into the bone cavity with a syringe. Sorting. Erythrocyte-depleted single-cell suspensions of BM cells, prepared as described above (flow cytometry), were blocked with anti-CD16/32. Whole BM suspensions were enriched for CD11b + cells by staining with a biotin-conjugated anti-CD11b antibody followed by incubating with streptavidin-microbeads (Miltenyi) and magnetic separation. Enriched cells were stained for surface molecules, and dead cells were excluded by staining with DAPI. Sorting was performed on an Aria II or Aria III system (BD Biosciences) using an 85-μ m nozzle.

Antibodies.
May-Grünwald-Giemsa staining. Cells were attached to glass slides using the Cytospin system (Thermo Scientific). Samples were air dried and then sequentially stained with May-Grünwald solution (Sigma-Aldrich) and Giemsa solution (Sigma-Aldrich). Images were captured with a Nikon Eclipse 80i equipped with a CCD camera DS-Ri1 (Nikon), and analyzed using Photoshop CS6 (Adobe).
ROS production assay. NeuPs or neutrophils were either sorted from the BM or cultured in vitro in the presence of G-CSF for 24 hours at 37 °C. Cells were stained with 5 μ M dihydrorhodamine 123 (Invitrogen). N-acetylcysteine (NAC) was treated while staining with DHR123 to show the negative staining. ROS production was stimulated by treating cells for 30 minutes with E. coli (DH5α , multiplicity of infection (MOI) = 10) alone for the sorted cells or with Low-Tox-M Rabbit Complement (Cedarlane) and E. coli (DH5α , MOI = 10) for the cultured cells. Dead cells were excluded with DAPI and stained cells were analyzed on an LSRFortessa (BD Biosciences); data were analyzed using FlowJo software (Tree Star).

Phagocytosis analysis.
NeuPs or neutrophils were either sorted from the BM or cultured in vitro in the presence of G-CSF for 24 hours at 37 °C. Sorted cells were incubated with FITC-conjugated microbeads (Polysciences, Fluoresbrite TM Plain YG 0.5 micron microspheres, 0.46 μ m) for 30 min at 37 °C. Cultured cells were stimulated with tdTomato-expressing E. coli (tdTomato-pcDNA3.1(+ )-myc-His C, DH5α ) for 30 minutes at a MOI of 10. Dead cells were excluded with DAPI and stained cells were on an LSRFortessa (BD Biosciences); data were analyzed using FlowJo software (Tree Star).

RNA-seq.
Total RNA was isolated using the RNeasy Mini kit and on-column DNase digestion (Qiagen).
The sequencing library was prepared using the TruSeq RNA Sample Preparation kit v2 (Illumina). The Illumina multiplexing adapters were ligated to the cDNA, and the fragments were amplified by polymerase chain reaction (PCR). Sequencing was performed in paired-end 100-base-pair reads using a HiSeq 2500 system (Illumina).
RNA was extracted with chloroform, and precipitated with isopropanol and ethanol. DNA was removed by treatment with RQ1 DNase (Promega). cDNA was synthesized with Superscript III Reverse Transcriptase (Invitrogen). Gene expression was confirmed by PCR using the following primers which produce distinct amplicons on genomic and cDNA template. EdU incorporation assay. For in vivo EdU incorporation assays, mice were intravenously injected with 0.5 mg EdU per mouse 2 hours before sacrifice. For in vitro EdU incorporation, sorted cells were cultured for 2 hours in the medium supplemented with 10 μ M EdU. For dead cell exclusion, a Live/Dead Fixable Violet Dead-Cell Stain kit (Invitrogen) was used. Cells were fixed with 3.7% formaldehyde and quenched with 10 mM glycine/phosphate-buffered saline (PBS). Fixed cells were permeabilized with 0.5% saponin/PBS. EdU was stained with 10 μ M azide-A488, 1 mM CuSO 4 , and 100 mM ascorbic acid in 0.5% saponin/PBS. Stained cells were analyzed on an LSRFortessa flow cytometer (BD Biosciences), and data were analyzed using FlowJo software (Tree Star).