Induced cell-autonomous neutropenia systemically perturbs hematopoiesis in Cebpa enhancer-null mice

Pioneering transplantation studies showed that long-term hematopoietic stem cells (LT-HSCs) possess a multilineage differentiation potential toward all hematopoietic lineages after myeloablative conditioning.^1,2  Bone marrow (BM) differentiation occurs along a continuum of cellular states, progressing in a trajectory from LT-HSC toward cell-lineage specific progenitors.^3-6  Hematopoietic stem and progenitor cells (HSPCs) forming these trajectories are interconnected by transcription factor networks that drive differentiation.⁷  Despite their role in differentiation, it remains unclear how transcription factors protect HSCs from exhaustion to preserve BM integrity.⁸ 

The myeloid lineage transcription factor C/EBPα, encoded by Cebpa, has been studied extensively to understand its role in myeloid differentiation. The expression of Cebpa in myeloid cells is specifically controlled by the +37 kb enhancer (+42 kb in humans).^9,10  Genetic knockout of Cebpa¹¹  or its +37kb enhancer^9,12  in vivo (both referred to as Cebpa null mice) causes neutropenia concomitant with reduced LT-HSC numbers. These studies describe C/EBPα as one of the major myeloid regulators that interconnect HSCs with myeloid progenitors. In addition, it has been suggested that C/EBPα has a dual role in LT-HSCs: maintaining LT-HSC quiescence by repressing the self-renewal¹³  and proliferative¹⁴  gene expression programs while simultaneously priming early myeloid genes.¹¹ 

One major technical limitation in these studies is the low resolving power of the technologies used to study rare cell types such as LT-HSCs. Bulk genome-wide transcriptomics measures gene expression signatures at the population level, thereby masking the presence of any rare and transient cell state of physiological importance in the BM. This limitation has been overcome by high-resolution single-cell technologies combined with lineage tracing or in vivo barcoding. Emerging findings from studies in native hematopoiesis^4,15,16  place LT-HSCs as a separate and an occasional contributing entity to myelopoiesis in contrast with a continuous HSC to myeloid state, therefore questioning the role of Cebpa as a myeloid priming factor in HSCs.

Here we investigated whether LT-HSC loss in Cebpa null mice is the cause or consequence of neutropenia. We hypothesize that either (1) LT-HSCs harboring an active Cebpa locus are lost upon enhancer deletion, leading to myeloid trajectory shutdown, and ultimately neutropenia or, (2) myeloid-biased progenitors expressing Cebpa are lost upon enhancer deletion, causing neutropenia, which systematically disturbs and depletes the LT-HSC pool. To address this question, we combined previously published single-cell datasets from wild-type (WT) BMs with bulk-cell transcriptomics from the +37kb enhancer-deleted mice. Furthermore, we transplanted Cebpa enhancer-deleted cells to study the possible systemic effects on hematopoiesis of the host. Using these approaches, we conclude that LT-HSCs do not express detectable levels of Cebpa in unperturbed hematopoiesis, and Cebpa null-induced neutropenia systemically disturbs LT-HSC quiescence, leading to HSC depletion, BM hypocellularity, and severe dysplasia.

Total sample RNA was extracted using Trizol with Genelute LPA (Sigma) as a carrier, and SMARTer Ultra Low RNA kit for Illumina Sequencing (Clontech) was used for cDNA synthesis according to the manufacturer’s protocol. cDNA was sheared with the Covaris device and processed according to the TruSeq RNA Sample Preparation v2 Guide (Illumina). Amplified sample libraries were subjected to paired-end sequencing (2 × 75 bp) and aligned against mm10 using TopHat.¹⁷  Gene expression levels were quantified by the fragments per kilobase of exon per million fragments mapped (FPKM) statistic as calculated by Cufflinks¹⁸  in the RefSeq Transcriptome database.¹⁹  Read counts were determined with HTSeq-count²⁰  and subsequently used for differential expression analysis in DESeq2,²¹  with default parameters in the R environment. Multiple testing correction was performed by the Benjamini-Hochberg procedure on the calculated P values to control the false discovery rate (FDR).

For gene set enrichment analysis (GSEA), a ranking metric was defined for each gene as the log10 of the adjusted P value calculated by DESeq2 with the sign of the log2 fold change. The ranked gene list was tested against a customized version of the C2 MSigDB collection, incorporating datasets on HSC quiescence from the literature^3,22-24  (supplemental Table 1).

A compendium of previously published mouse BM single-cell RNA-Seq datasets (Fluidigm C1 platform) was assembled to evaluate the expression of key progenitor genes of interest described^25-27  (available at: http://www.altanalyze.org/ICGS/Public/Mm-Grimes-Fluidigm-Panorama/User.php). Specifically, WT in vivo mouse SLAM, LSK, GMP, CMP, and lineage-negative Sca1+ CD117+ cells from BM were selected, with labels derived from the noted prior studies (GSE70245, GSE141472). Data were analyzed with RSEM to estimate TPM for all genes as previously described.²⁸  The gene expression data of selected genes were visualized in the python package plotly or GraphPad Prism, represented as log2 TPM values.

For the analysis in supplemental Figure S1, published single-cell RNA-seq data of HSPCs from the BM of 10 female 12-week-old C57BL/6 mice were retrieved.²⁹  Raw counts were downloaded from the GEO database (GSE81682), with labels derived from the broad gating strategy used in the original study (available here: http://blood.stemcells.cam.ac.uk/data/all_cell_types.txt). Data were imported and processed using the Seurat R package.³⁰  Cells with fewer than 200 000 total counts, fewer than 4000 detected genes, or more than 10% mitochondrial reads were excluded. Expression data were log-normalized, and the 5000 most variable features were selected for dimensionality reduction with UMAP. Cebpa expression was projected on the UMAP and compared with the expression of other selected genes (Mecom, Hlf) to investigate its association with different cell populations.

The strains of +37kb enhancer-1.2kb and +37kb enhancer-1.15kb-deleted mice were generated using zygotes derived from C57/BL6 mice by CRISPR/Cas9 editing and maintained as previously described.⁹  The CRISPR/Cas9 single guide RNAs were directed against the +37kb enhancer of Cebpa, using the following sequences:

• 5′: TGAAGCCTACACTACTTTGT and AGAGGTAGGAACTCCATTCC

• 3′: AGAGCCTCGCTCAAGCCCAT and TTGAGACATCTGGTAACCTT

Recipient mice were exposed to a 5.5 Gy of γ radiation. Given that the native +37kb^HOM mouse exhibits a 3- to 5-fold increase in the cKIT+ lineage-negative progenitor fraction of the BM, we transplanted 1 million WT CD34.2 total BM cells and 250 000 of HOM CD34.2 cells to compensate for the fold difference. Nontransplanted mice were sacrificed for FACS analysis between 4 and 8 weeks of age. For transplantation experiments, BM from 4 week old WT or +37kb enhancer-deleted CD45.2 mice were harvested in phosphate-buffered saline (PBS)/5% FCS and injected intravenously in tails of (CD45.1) female mice. All mice were sacrificed in a CO₂ chamber. Animal studies were approved by the Animal Welfare/Ethics Committee of the EDC in accordance with legislation in The Netherlands (approval No. EMC 2067, 2714, 2892, 3062).

Flow cytometry was carried out on the LSRII and the FACSCanto II (BD Biosciences). FACS AriaIII (BD Biosciences) was used for cell sorting, using the following fluorescent antibodies: markers for mature hematopoietic cells (CD11B APC/GR1-FITC/B220-PE/CD3 PB); pan-hematopoietic marker (CD45.1 PE; CD45.2 APC-CY7); LIN biotinylated cocktail (CD11B, GR-1, B220, and CD3) streptavidin-Pacific Orange; LSK (cKIT-APC/SCA1-PB); LT-HSCs (CD48-FITC/CD150-PE-CY7). Lineage-negative selection was carried out using a cocktail of antibodies targeting antigens expressed on mature hematopoietic cells, including CD11B, GR-1, B220, and CD3. All antibodies were purchased from BD Biosciences or Biolegend (supplemental Table 2). The LSK population was gated from live (DAPI^-), CD45+, lineage^-cKIT⁺ SCA-1⁺ BM cells, and the sorted LSK fraction was collected in 500 μl PBS with 5% FCS, spun down and resuspended in 800 μl of Trizol and used for RNA-seq.

To study the causal relationship of neutropenia and loss of LT-HSCs, we investigated the allelic dosage effect of Cebpa enhancer deletion on the numbers of neutrophils and LT-HSCs using Cebpa +37kb enhancer heterozygous (+37kb^HET)-deleted and +37kb homozygous (+37kb^HOM)-deleted mouse strains.⁹  The neutrophil (Mac1+Gr1+) frequency and absolute numbers in the peripheral blood (PB) and BM of the +37kb^HET and the +37kb^HOM mice correlated with their monoallelic and biallelic enhancer deletion, respectively (Figure 1A-D). Therefore, 50% enhancer activity reduced the neutrophil output by approximately half (median frequency: 44.6%) (Figure 1C), while a full enhancer deletion completely abrogated neutrophil production (median frequency 0%). The reduction of neutrophils occurring in the +37kb^HOM mice was confirmed by hematoxylin and eosin (H&E) staining of BM sections (Figure 1E) and by the decreased frequency of S100A8 cells (Figure 1F). To investigate whether enhancer deletion affects the LT-HSC pool, we studied the LT-HSC population using the SLAM CD150 and CD48 markers (LT-HSCs: LSK/CD150+CD48-) in the +37kb^HET and +37kb^HOM mice. The LT-HSC pool was not affected in the +37kb^HET mice despite the 50% enhancer dosage (Figure 1G,H). Although there was a near 2-fold reduction in LT-HSC numbers, this difference was not statistically significant. In contrast, a 10-fold reduction of LT-HSC numbers at 100% enhancer dosage reduction was observed in the +37kb^HOM mice (Figure 1G,H). The fact that monoallelic enhancer deletion does not affect LT-HSC numbers, but reduces the neutrophil compartment by half, suggests that Cebpa only becomes expressed in neutrophil-primed progenitors and is inactive in LT-HSCs. Thus, the quantitative changes in the LT-HSC pool of the +37kb^HOM mice could be explained by an indirect effect of complete ablation of neutrophils, whereas partial neutropenia in the +37kb^HET is insufficient to inflict these changes. Alternatively, myeloid progenitors may be more sensitive to reduced C/EBPa levels than LT-HSCs, which would only become depleted upon biallelic deletion of the +37 kb enhancer.

To study Cebpa expression along differentiation from LT-HSCs toward multipotent and myeloid progenitors, we used single-cell RNA sequencing datasets generated by a Fluidigm-based platform. In total, 1110 BM cells were analyzed (27 libraries, ∼50 cells per library, and >2 million reads/cell on average) and assigned to a specific BM cell type. The identified cellular states were classified into LT-HSC (SLAM) population, HSPC-1 and HSPC-2 (both LSK), and multilineage progenitors, which constitute a mixture of bipotent and unipotent restricted myeloid progenitors, as previously described.^25-27  To investigate at what differentiation stage Cebpa becomes expressed, we first used known gene markers associated with HSC quiescence, namely Hlf³  and Mecom,³¹  and early myeloid differentiation (Elane, Mpo) (Figure 2A). We found a positive correlation between Hlf and Mecom, while Cebpa negatively correlated with Hlf/Mecom (Figure 2B). Accordingly, Cebpa expression was almost negligible in LT-HSCs (detected in 0.8% of LT-HSCs), while it appeared in a subset of HSPCs expressing early myeloid lineage genes, indicating priming of the myeloid lineage at very early stages of differentiation (Figure 2C). The detection limit of this technique is 0.25 transcripts per million (TPM), which could possibly exclude cells with very low but present levels of Cebpa. However, these results were confirmed in another dataset²⁹  generated by a different single-cell sequencing strategy (supplemental Figure S1). Altogether, our single-cell analysis does not support a cell-autonomous role for Cebpa in LT-HSCs. Thus, the reduction in LT-HSC numbers observed in the Cebpa null mice seems to occur systemically as a consequence of neutropenia.

The results so far indicate that the enhancer deletion decreased the pool of LT-HSCs through a systemic effect in the presence of neutropenia. To study mechanisms linking the 2 observed events, we applied RNA sequencing on sorted HSPCs using LSK markers on +37kb^WT (n = 3) and +37kb^HOM (n = 3) BM cells. We confirmed that +37kb enhancer deletion in HSPCs reduces Cebpa expression relative to WT controls (Figure 3A). Differential expression analysis identified dysregulated genes related to neutrophil differentiation (S100a9, Camp) and HSC quiescence (Hlf) (Figure 3B; supplemental Table 3). Differentially expressed genes in +37kb^HOM LSKs were further investigated by GSEA (supplemental Tables 4 and 5). Genes associated with neutrophil ontogeny were downregulated in +37kb^HOM LSK cells (Figure 3C), indicating an early block in myeloid differentiation. Interestingly, we found an early myeloid gene set (Cpa3, Mpo, Cd48) (Figure 3D) to be upregulated in +37kb^HOM HSPCs, suggesting that an early myeloid-biased population upstream of Cebpa-HSPCs is primed for myelopoiesis. Based on these findings, we conclude that Cebpa-HSPCs are intermediate progenitors that link early myeloid-biased HSPCs to myeloid-committed progenitors, and they represent the cell of origin for the induced neutropenia in Cebpa null mice.

To investigate pathways related to LT-HSC depletion, we retrieved datasets from published hematopoietic studies and pooled them with the MSigDB datasets.^3,23,32-34  Of the most significantly (FDR <0.025) enriched pathways, 6 gene sets were enriched for loss of HSC quiescence and exhaustion (Figure 3E). The identified gene sets (Figure 3F; supplemental Figure S2A-F) included transcription factors related to HSC dormancy (Hlf, Mecom, Tcf15),^3,31,35  HSC retention factors (Ptpn21, Cxcr4),^36,37  cluster of differentiation (Cd) markers for HSC regeneration and engraftment (Cd81, Cd274),^38,39  and the polycomb chromobox proteins for HSC self-renewal regulation (Pbx6, Pbx7) (gene lists in supplemental Table 1). Deregulation of ribosomal genes (Figure 3G) in our dataset is in accordance with the loss of HSC quiescence. Thus, LT-HSC loss in the presence of neutropenia may result from quiescence exit and subsequent exhaustion. Altogether, the transcriptomic analysis of sorted LSK cells from +37kb^HOM mice suggests that myeloid priming in HSPCs occurs before Cebpa activation and that a neutrophil differentiation block in these progenitors potentially leads to HSC quiescence exit.

Next, we studied the systemic effects of the +37kb^HOM BM on normal hematopoiesis in recipient mice. A sublethal irradiation approach was used to overcome the low survival rate caused by the weak chimerism known to occur when transplanting Cebpa null BM cells into lethally irradiated recipients.¹¹  We used donor BM cells from 2 different +37kb^HOM mouse strains to ensure that the observed phenotype was not caused by an off-target effect. The strains differed in the genomic size of enhancer deletion generated by CRISPR, ie, +37kb^HOM deletion of 1.15kb (CD45.2 +37kb^HOM-1.15kb deletion) or +37kb^HOM deletion of 1.2kb (CD45.2 +37kb^HOM-1.2kb) (supplemental Figure S3A,B). Sublethally irradiated recipient (CD45.1) mice were transplanted with total BM of CD45.2 +37kb^WT (n = 8), CD45.2 +37kb^HOM-1.15kb (n = 7), or CD45.2 +37kb^HOM-1.2kb (n = 8) mice (Figure 4A). As expected, PB samples withdrawn 12 weeks after transplantation showed a weaker chimerism in mice transplanted with +37kb^HOM BM compared with those transplanted with WT BM (Figure 4B). For cellular reconstitution, we determined the frequency of myeloid cells (Mac1+Gr1+), B cells (B220+), and T cells (CD3+) by flow cytometry for mice transplanted with +37kb^WT or +37kb^HOM BM cells (Figure 4C-F; supplemental Figure S4A-C).

Populations from the T-cell, myeloid, and B-cell lineages were present in the recipients from the 3 cohorts, indicating that hematopoiesis is functional in the recipient (Figure 4D upper panel and 4F upper panel). However, +37kb^HOM donor cells (Figure 4E lower panel and Figure 4F lower panel) lacked myeloid reconstitution as compared with +37kb^WT (Figure 4C lower panel and Figure 4D lower panel). Therefore, transplanted +37kb^HOM BM cells in a WT niche did not recover neutrophil differentiation, confirming that neutropenia in +37kb^HOM enhancer-deleted mice is cell-autonomous. In the +37kb^WT donor cell compartment, there was a large contribution of B cells, as previously observed in other transplantation experiments.^40-42  Of note, the presence of B cells and T cells from the +37kb^HOM donors in both the native and the transplanted mice also argues against a strict need for Cebpa in LT-HSCs since they were able to survive and differentiate into the lymphoid lineage.

Despite the low chimerism, the CD45.2 +37kb^HOM BM cells persisted and survived for at least 10 months in the CD45.1 recipient but eventually declined to almost 0% (median = 0.89%) on the day the mice were sacrificed (supplemental Figure S5A). Therefore, cells from the blood and BM samples analyzed were mostly derived from the recipient mice. We grouped the recipient mice based on their phenotype severity, ie, +37kb^HOM-mild (n = 8) or +37kb^HOM-severe (n = 6, namely mice # 2,5,7,9,12, and 17). The +37kb^HOM-severe mice represented 40% of all the transplanted mice that showed physical weakness (squinting eyes, hunched posture, and social isolation), which had to be sacrificed (supplemental Figure S5B). One of the mice died before it could be analyzed. BM hypocellularity and pancytopenia were the primary hallmarks of the +37kb^HOM-severe group (Figure 5A; supplemental Figure S5C). Hypocellular BMs showed remodeling of blood vessels surrounded by a high degree of immature hematopoietic cells characterized by incomplete differentiation (Figure 5B). The +37kb^HOM-mild group showed lower white blood cell (supplemental Figure S6A) and platelet counts (supplemental Figure S6C) compared with WT controls, whereas hemoglobin levels were comparable to those of the +37kb^WT group (supplemental Figure S6B). BM cellularity in the +37kb^HOM-mild mice varied from normocellular to hypocellular. Both the +37kb^HOM-severe and the +37kb^HOM-mild cohorts showed abnormal megakaryocytes with dysplastic features (Figure 5B). These data are in line with the dysplastic megakaryocytes found in the native +37kb enhancer-deleted mice (Figure 1E; supplemental Figure S7), which might possibly explain the abnormal PB platelet counts in the recipient mice (supplemental Figure S6C). In conclusion, a significant number of mice transplanted with +37kb^HOM BM exhibit perturbed hematopoiesis featuring hypocellularity and peripheral cytopenia, incomplete differentiation, and severe dysplasia.

Next, we sought to study whether the differences between the +37kb^HOM-severe and the +37kb^HOM-mild phenotypes involve maturation defects of myeloid cells or LT-HSCs in the BM of recipient mice. The +37kb^HOM-severe-transplanted mice showed a marked reduction in the numbers of recipient (CD45.1) Mac1+Gr1+ cells (Figure 5C), which was confirmed by the decrease in S100A8 protein-expressing cells (Figure 5D). In contrast, the +37kb^HOM-mild mice showed normal neutrophil numbers compared with the WT controls (Figure 5C). The +37kb^HOM-severe mice also showed a reduction in the CD45.1 HSPC and LT-HSC numbers (Figure 5E,F), whereas the LT-HSC numbers in the +37kb^HOM-mild mice were comparable to those of the +37kb^WT group. Thus, the findings in the +37kb^HOM-severe support the notion that Cebpa null-induced neutropenia triggers LT-HSC loss in a cell nonautonomous manner, as observed in 37kb^HOM native mice.

C/EBPα is indispensable for myeloid lineage formation and differentiation.^13,43  Induced genetic defects in the Cebpa locus of mouse models^9,12,44-46  or CEBPA coding mutations detected in human acute myeloid leukemia (AML) specimens^47,48  are all associated with myeloid differentiation abnormalities. The expression of CEBPA is also frequently repressed in leukemia by transcriptional or posttranscriptional mechanisms.⁸  Notably, CEBPA is silenced by oncoproteins targeting the +42 kb enhancer (homologous to +37 kb in mouse), including RUNX1-RUNX1T1⁴⁹  and EVI1.⁵⁰  Earlier mouse studies revealed an indispensable role for Cebpa in differentiating common myeloid progenitors into granulocyte-monocyte progenitors. The profound neutropenia observed in our +37 kb enhancer-deleted model confirms the absolute requirement for Cebpa in myelopoiesis. We observed a 3-fold increase in Cebpb expression in cKit+ Sca1- Cd34+ myeloid progenitors, but this was insufficient to compensate for the loss of Cebpa. Although C/EBPbeta can substitute for C/EBPalpha during hematopoiesis when knocked into the Cebpa gene locus,⁵¹  it cannot fully replace C/EBPalpha when expressed from its native locus.

Several studies hinted toward an intrinsic role for Cebpa in myeloid lineage priming of LT-HSCs.^11,12  However, the bulk-sequencing methods used in these studies do not meet the resolution required to dissect the heterogeneity of the HSPC compartment. Using single-cell RNA sequencing, we showed that Cebpa is barely detectable in LT-HSCs during steady-state hematopoiesis. In contrast to previous studies, we found that Cebpa licenses myeloid-primed HSPCs downstream of LT-HSCs for neutrophil lineage differentiation. This suggests that other transcription factors account for the myeloid lineage bias in LT-HSCs⁵²  and activate the neutrophil lineage trajectory through binding the Cebpa +37kb enhancer in a subset of HSPCs.^9,53  Although Fluidigm C1 exhibits better sensitivity⁵⁴  and fewer dropout events^55,56  than droplet-based approaches, it remains possible that LT-HSCs with very low Cebpa levels (below the detection threshold of 0.25 TPM) exist. The physiological relevance of such low transcript levels, which could be attributable to pervasive transcription, is unknown. It is equally possible that the few Cebpa-expressing LT-HSCs identified are the result of technical noise or phenotypic misclassification.

In the enhancer-deleted models, LT-HSC loss was proportional to the degree of neutropenia, suggesting a causal relationship between them. Although an alternative explanation could be that LT-HSCs are less sensitive to Cebpa levels than myeloid progenitors, the normal production of lymphoid cells indicates that LT-HSCs remain viable in the absence of Cebpa. This is further supported by the almost complete absence of Cebpa-expressing LT-HSCs in healthy BM. Critically, LT-HSCs were also lost in mice transplanted with +37 kb^HOM BM: 40% of the transplanted mice showed this severe phenotype, while the other 60% showed signs of myeloid differentiation defects and dysplasia, suggesting they were in an early stage of the same pathological process. These effects can only be explained by systemic consequences triggered in the presence of neutropenia. Even in a hybrid model in which a few LT-HSCs express Cebpa, the evidence gathered in this study supports the hypothesis that the major LT-HSC depletion in Cebpa null mice is an indirect consequence of neutropenia.

To our knowledge, none of the previously reported neutropenia mouse models^26,57  demonstrated a depletion of LT-HSCs. A likely explanation is that the differentiation block in these other models (such as Sbds⁵⁷  and Gfi1⁵⁸  mutants) often occurs at a late stage of neutrophil differentiation. Niches supporting a late stage of neutrophil differentiation are located distant from LT-HSC niches. Given that the differentiation block in Cebpa null BM occurs in early progenitors located proximally to LT-HSC niches, the proposed systemic effect of neutropenia onto the LT-HSCs might be specific to this model. Fifteen percent of patients with congenital neutropenia develop LT-HSC clonal BM conditions such as myelodysplasia and AML,^59,60  suggesting LT-HSC impairment may also occur in human neutropenia. Thus other models of neutropenia are required in order to understand the mechanisms that lead to hematopoietic insufficiency in the presence of neutropenia, such as Cebpa-+37kb-Enh(f/f);Mx1-Cre mice.¹²  The advantage of those mice with an inducible +37kb enhancer-deletion system is that one could avoid enhancer deletion during embryogenesis.

Loss of HSC quiescence is one of the hallmarks identified at the transcriptional level in the +37kb^HOM enhancer-deleted mice. The activation of compensatory mechanisms forcing neutrophil differentiation in myeloid-biased progenitors is a potential underlying cause. For example, increased production of granulocyte colony-stimulating factor (G-CSF) in the absence of peripheral neutrophils is seen in patients with neutropenia.^61,62  G-CSF stimulates granulopoiesis through binding to G-CSF receptor-expressing progenitors⁶³  and activates HSCs through attenuation of the Cxcr4-Cxcl12 retention factors expressed on HSCs and BM stromal cells, respectively.^64,65  Therefore, the neutropenia-GCSF-HSC activation loop can eventually lead to exhaustion and consumption of the LT-HSC pool. Other potential causes that might lead to LT-HSC quiescence exit include metabolic stress caused by impaired differentiation,⁶⁶  emergency myelopoiesis due to infections acquired in the absence of neutrophils,⁶⁷  or the lack of mature neutrophils in the BM niche that support HSC quiescence.⁶⁸ 

Myeloid niches in the BM have been reported to be located spatially distant from niches occupied by HSCs⁶⁹  or lymphoid progenitor populations.⁷⁰  In addition, the dendritic, neutrophil, and monocyte lineages are also distantly located from each other and organized in different sinusoid niches. Therefore, it is likely that the neutrophil-primed progenitors derived from both donor and recipient share a common environment that is potentially disturbed upon transplantation of +37kb^HOM BM. The question that remains to be answered is how transplanted +37kb^HOM myeloid progenitors impair the differentiation process of the host. Transcriptome analysis shows that myeloid progenitors (cKit+ Sca1- Cd34+) derived from the +37kb enhancer-deleted model are metabolically reprogrammed and exhibit downregulation of the oxidative phosphorylation (OXPHOS) pathway (supplemental Figure S8A; supplemental Table 6), as compared with normal myeloid progenitors that are dependent on oxidative and mitochondrial metabolism.⁷¹  Shutting down OXPHOS eventually activates the glycolytic pathway (known as the Warburg effect), under the control of nuclear factors such as Mycn,^72,73  a critical metabolic regulator involved in cell competition. In line with this, we find Mycn expressed at high levels in the +37kb^HOM myeloid progenitors (supplemental Figure S8B). Such metabolic changes impact cell-to-cell communication processes required for normal differentiation programs,⁶⁶  which may partially explain the acquired neutropenia in the recipient when transplanted with +37kb^HOM BMs.

Our study suggests that prolonged neutropenia induced perturbations in localized myeloid niches and further caused systemic BM changes resulting in LT-HSC loss, BM hypocellularity, and severe dysplasia. Although the underlying mechanisms remain unclear, we hypothesize that Cebpa null progenitors acquire metabolic reprogramming that impairs differentiation and disturbs HSC quiescence. Functional studies are required to investigate the intracellular role of Cebpa in controlling the metabolic pathways related to neutrophil differentiation and further elucidate how Cebpa null progenitors are metabolically reprogrammed to inflict systemic changes in the BM. From a clinical perspective, this phenomenon may explain how metabolic stress on HSCs might give rise to BM clonal disorders in a subset of congenital neutropenia patients. Therefore, our study sets a paradigm about the underlying mechanisms involved in the progression of neutropenia to clonal BM disorders.

The authors are grateful to Michael Vermeulen for the flow cytometric analysis and to their colleagues from the Erasmus Animal Facility, who took care of the mice used in this study and conducted regular bleeding to obtain blood cells. They would also like to thank Stefan Erkeland, Zhen Ping, and Jacqueline Feyen for their valuable input. The work was funded by grants and fellowships from the Dutch Cancer Society (EMCR 2015-7935) (R.D., R.A., R.M.L., M.H., and L.S.) and the Lady Tata Foundation (R.A.).

Contribution: R.A., R.M.L., and R.D. designed the research and wrote the paper; R.A., M.H., L.S., N.S., and E.B. performed research; and R.A., R.M.L., R.H., N.S., R.K.S., and L.G. analyzed data.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Corresponding author: Ruud Delwel, Department of Hematology, Erasmus University Medical Center, Dr Molewaterplein 40, 3015 GD Rotterdam, The Netherlands; email, h.delwel@erasmusmc.nl.