Diabetes Inhibits Gr-1⁺ Myeloid Cell Maturation via Cebpa Deregulation

Diabetes-induced chronic inflammation underlies multiple secondary pathologies including the foot ulcer (rev. in 1), which can be modeled using the Lepr^db/db mouse (2). In both humans and mice (3), diabetes impairs the orderly transition from an early, inflammatory phase of repair to a later, prohealing stage, preventing timely healing. Excessive numbers of myeloid cells are found in the wound (2,4), and they display a persistent proinflammatory, M1-like (classical) or M1/M2 (mixed) phenotype, instead of a clear transition to an M2-like (alternative) state (5,6).

Gr-1⁺CD11b⁺ cells are deregulated by diabetes and include monocytes, granulocytes, and precursors. They home to injury sites after mobilization from the bone marrow (BM), neutralize pathogens early in healing, and promote angiogenesis later in repair (7). Wound recruitment is controlled by the chemokine receptors CCR2 and CxCR2 (8,9), which are expressed by Ly6C⁺ monocytes (10) and Ly6G⁺ granulocytes (11), respectively. Both subsets are required for timely healing, and CCR2⁺ monocytes play a crucial role in wound angiogenesis (8,9). However, diabetes significantly expands the Gr-1⁺ population in the blood (7) and elevated numbers are recruited to and retained in the db wound (7). They show an altered monocyte/granulocyte differentiation capacity, possess aberrant chemotactic and adhesive properties, and fail to promote wound angiogenesis (7). The impact of diabetes on individual Gr-1⁺ subpopulations is unknown.

Chronic inflammation perturbs hematopoiesis, possibly via increased myeloid cell turnover at inflammatory sites and a correspondingly elevated demand (12). Stable intrinsic defects induced in myeloid cells by diabetes prior to wound recruitment promote chronic inflammation and impair healing (5). Persistent Ly6C/Gr-1 and diminished F4.80 expression in the db wound suggest that one facet of the aberrant myeloid cell phenotype is persistent immaturity (5,6). It is unknown whether this is prespecified, but peripheral blood myeloid cells from db patients have impaired effector functions, including bacterial killing and phagocytosis, that are characteristic of immaturity (13). Candidate mechanisms underlying intrinsic defects in differentiation include altered histone modifications at gene promoters of important myeloid differentiation genes, as found at inflammatory genes in vitro in response to high glucose (14). However, data from primary cells are limited and the promoters of genes involved in myeloid differentiation remain unexamined.

Here, we show that diabetes inhibits Gr-1⁺ cell maturation in the wounds of Lepr^db/db mice and that cellular changes underlying this defect occur before cells exit the BM. Diabetes induces altered expression of multiple chromatin-modifying enzymes, particularly histone deacetylases (HDACs), and myeloid differentiation genes, including the key myeloid transcription factor Cebpa. Diminished Cebpa promoter acetylation accompanies Cebpa repression and causes misregulation of target genes including Ccr2 and Cxcr2 and aberrant differentiation of BM-derived (BMD) Gr-1⁺ cells. Cebpa upregulation by granulocyte colony-stimulating factor (G-CSF) restores regulation of chemokine receptor expression and promotes maturation of db-derived Gr-1⁺ cells, thus rescuing the db phenotype. Administration of G-CSF to db wounds accelerates wound healing in vivo, suggesting that myeloid cell immaturity contributes to impaired wound healing associated with diabetes.

Mice were housed at the University of Manchester animal care facility. All procedures were approved by the local ethical review committee and the U.K. Home Office under the Animals (Scientific Procedures) Act, 1986. Lepr^db/db and Lepr^db/+ mice were purchased from Harlan (Oxfordshire, U.K.). Heterozygous animals were crossed to the C57BL/6-Tg(CAG-EGFP)1Osb/J line, which ubiquitously express eGFP (The Jackson Laboratory), and then backcrossed to produce Lepr^db/db animals carrying the eGFP transgene to obtain GFP-labeled myeloid cells. Animals were used between 8 and 16 weeks of age and were age and sex matched to controls. Full-thickness excisional dorsal wounds (8–10 mm) were made and animals killed at time of tissue collection. Treated wounds received 500 ng G-CSF or PBS on days 0, 2, 4, 6, and 8 after wounding. Wound closure was determined as percentage of original area. Full details of wounding procedure and analysis can be found in Supplementary Experimental Procedures.

Antibody clones used for myeloid cell isolation, chromatin immunoprecipitation (ChIP), and immunofluorescence are listed in the Supplementary Experimental Procedures.

Gr-1⁺ cells were isolated from a single-cell suspension of BM flushed from femurs and tibiae or from dissociated wound tissue prepared as previously described (5) either by flow cytometry or using the bead-based FlowComp Flexi system (Life Technologies). Full details of the wound tissue dissociation protocol are given in Supplementary Experimental Procedures.

Ethics approval for this study was granted by the North West–Greater Manchester West Research Ethics Committee (reference 12/NW/0775), and informed consent was given by all participants prior to blood sampling. A full list of inclusion and exclusion criteria can be found in Supplementary Experimental Procedures. Participants were male or female and aged between 18 and 65 years. Granulocytes were isolated from 15 mL peripheral blood by centrifugation with Histopaque 1119 (Sigma) and washed three times in PBS before RNA isolation.

Myeloid cells were cultured in DMEM supplemented with l-glutamine, 4,500 mg glucose/L (or 1,000 mg glucose/L for low-glucose experiments), 110 mg sodium pyruvate/L, 10% FBS, and 1% penicillin-streptomycin plus appropriate cytokines as stated below. Transient transfection of Gr-1⁺ cells was performed using the Nucleofector system (Amaxa) and is described in Supplementary Experimental Procedures.

Flow cytometric analysis of CCR2 and CxCR2 expression was performed on 5 × 10⁵ BMD Gr-1⁺ cells isolated from one or two mice per condition before and after culture with or without 100 ng/mL granulocyte macrophage colony-stimulating factor receptor (GM-CSF) or 50 ng/mL G-CSF (Peprotech). Experiments were performed three times independently. Gr-1⁺ cytospins were prepared with fresh or cultured cells for staining in fresh 5% Giemsa (Sigma) after methanol fixation. Full details of fixation, staining, and imaging are given in Supplementary Experimental Procedures.

The FAM poly caspases assay (Molecular Probes) was used to measure apoptosis by cells incubated for 1 h or 48 h and harvested simultaneously. Cell division was assayed by membrane labeling with eFluor 450 (eBioscience). Fold difference in FAM or eFluor 450 median fluorescence intensity (MFI) between 1-h and 48-h samples as analyzed by flow cytometry was used to quantify apoptosis or cell division, respectively. For cell survival, GFP⁺Gr-1⁺ cells were imaged every 24 h for 3 days and cells present in five fields of view per condition were scored in Adobe Photoshop. Phagocytosis assays were performed using db-derived Gr-1⁺ cells incubated overnight in the presence or absence of 50 ng/mL G-CSF and then allowed to phagocytose Alexa Fluor 488–labeled zymosan particles (Life Technologies) for 1 h. Full protocol details are given in Supplementary Experimental Procedures.

Total RNA was isolated using Trizol and treated with TURBO DNase I, and the concentration was determined by Nanodrop. cDNA synthesis was performed with Bioscript using 0.1–1 µg total RNA for analysis of gene expression with the Taqman system, using Hist2h2aa1 as a reference gene to calculate relative expression of genes assayed. Full protocol details and primer/probe sets are listed in Supplementary Experimental Procedures.

Next-generation sequencing (NGS) was performed on two pools of five nondiabetic (ndb) or six db mice using Gr-1⁺ cells isolated from BM or dissociated wound tissue using the FlowCompFlexi system. Total RNA was isolated as described above. Two libraries were generated for each of the four conditions (ndb BMD, db BMD, ndb wound derived, and db wound derived) and sequenced on an Illumina HiSEq. 2000 (GATC Biotech). Full details are provided in Supplementary Experimental Procedures.

Genes showing fold change >1.5 between ndb and db, with expression level >1 fragment per kilobase of transcript per million mapped reads (FPKM) and P < 0.05, were regarded as differentially expressed. Gene enrichment analyses were performed using DAVID, version 6.7. The gene list for the gene ontology (GO) term “myeloid cell differentiation” (GO:0030099) was obtained using AmiGO 2, version 2.2 (amigo.geneontology.org), compared with the list of genes differentially expressed found in both BM- and wound-derived Gr-1⁺CD11b⁺ cells using Python, version 3.3.3 (Python Software Foundation), and used to populate Table 1. Chromatin regulators were identified using the functional annotation tool of DAVID to interrogate the differentially expressed gene set (P < 0.05) between ndb and db BM-derived Gr-1⁺CD11b⁺ cells. Hits were then filtered for fold change (>1.6) and expression level (>2 FPKM) and listed in Table 2. Further details are given in Supplementary Experimental Procedures.

Five to six ×10⁶ Gr-1⁺ cells, obtained from a pool of two animals, were crosslinked for 5 min at room temperature and ChIP assays performed as previously described (15). Antibodies used were acetylated H3 Lys 27 (H3K27Ac) (ab4729; Abcam) and IgG. Primer sequences and full protocol details are given in Supplementary Experimental Procedures.

Macrophages were differentiated from total BM on coverslips in presence of macrophage colony-stimulating factor receptor for 7 days and fixed in 4% paraformaldehyde for staining with anti-HDAC4 and an appropriate Alexa Fluor 488–conjugated secondary antibody prior to mounting in Prolong with DAPI. Imaging and quantification details are given in Supplementary Experimental Procedures.

Paired or unpaired t tests with one or two tails as appropriate were performed using GraphPad Prism. Equal variance was not assumed. Hypergeometric tests were performed in Microsoft Excel.

We and others have shown that monocytes and macrophages in the db wound express persistently high levels of the inflammatory monocyte marker Ly6C and low levels of the macrophage marker F4.80 late in healing, consistent with an immature phenotype (5,6). CxCR2 expression remains uncharacterized. To explore this further, we examined CCR2 and CxCR2 expression in wound-derived Gr-1⁺ cells. NGS (data set available online [Supplementary Experimental Procedures]) found Gr-1⁺ cells isolated from day 3 (D3) db wounds express significantly fewer Ccr2 and Cxcr2 transcripts than Gr-1⁺ cells from the D3 ndb wound (Fig. 1A) (P < 0.001 and P < 0.05, respectively). This is despite the significantly elevated proportion of Gr-1⁺ cells in the db wound at both D3 (Supplementary Fig. 1A) (P < 0.05) and D7 (Supplementary Fig. 1B) (P < 0.01), as previously noted (7), and indicates that expression of chemokine receptors by Gr-1⁺ cells is impaired. Flow cytometric analysis confirmed that in the db wound, the number of Gr-1⁺CCR2⁺ cells is approximately threefold lower than in the ndb wound at both D3 (Fig. 1B) (P < 0.05) and D7 (Fig. 1C) (P = 0.06), providing further evidence that chemokine receptor expression by myeloid cells is deregulated in the db wound.

We previously showed that hyperpolarization of db wound-resident myeloid cells occurs before recruitment and not solely due to signals received in the wound (5). To determine whether misexpression of chemokine receptors is similarly prespecified, we analyzed expression of Gr-1, CCR2, and CxCR2 by BMD Gr-1⁺ cells by flow cytometry.

We divided Gr-1⁺ cells into four subsets based on CCR2 and CxCR2 expression, designated “P1–P4” (Fig. 2A), each with distinct morphologies (Supplementary Fig. 2A). P1, henceforth “Gr-1⁺ precursors,” is a large population of blasts, promyelocytes, and eosinophils and some monocytes expressing variable levels of Gr-1 and neither chemokine receptor. P2, henceforth “Gr-1⁺ monocytes,” contains Gr-1^intermediate monocytic cells, consistent with the reduced affinity of anti–Gr-1 for Ly6C (16). They are CCR2⁺ but CxCR2⁻. P3 comprises 1% of cells in unwounded mice and contains neutrophils and monocytes of intermediate maturity. P4, henceforth “Gr-1⁺ neutrophils,” is Gr-1^high, CxCR2⁺, and CCR2⁻. Morphologically neutrophilic, they represent the largest Gr-1⁺ subset. In unwounded mice (Fig. 2B), diabetes does not significantly alter P2 or P4 but expands P1 (P < 0.05). In wounded db animals (Supplementary Fig. 2B), an expansion in P1 (P < 0.01) is accompanied by a significant contraction of P4 (P < 0.01). No overall expansion in the Gr-1⁺ BM population is seen (7). As CxCR2 expression correlates positively with Gr-1 expression and neutrophil maturity (11), these data indicate that diabetes inhibits neutrophil maturation, promoting an expansion of the Gr-1⁺ precursor subset, and that wounding compounds this effect.

To further explore the misexpression of maturation and lineage-specific genes by db-derived BM Gr-1⁺ cells, we analyzed expression of the early hematopoietic markers CD34, Sca-1, and Kit and the myeloid marker CD11b. As anticipated, >90% of Gr-1⁺ cells coexpressed CD11b and no difference in expression was seen when ndb- and db-derived cells were compared (Supplementary Fig. 3A). A small CD34⁺ subset (Supplementary Fig. 3B) and a small Sca-1⁺Kit⁺ subset (Fig. 2C) were detected within the Gr-1⁺ population, but again, no differences between ndb- and db-derived cells were observed. However, a sizeable Gr-1⁺Kit⁺ Sca-1⁻ subset was detected (Fig. 2C) that was significantly expanded in the db-derived population (P < 0.05), while the Gr-1⁺Sca-1⁻Kit⁻ population contracted (P < 0.01). Staining Gr-1⁺Kit⁺Sca⁻ cells for CxCR2 and CCR2 revealed that ∼6% expressed CCR2, indicating that a small proportion of these cells are monocytes (Fig. 2D). As confirmed by morphological analysis of Giemsa-stained cells (Fig. 2C), however, the majority are CCR2⁻ granulocytes and their precursors, which express varying levels of CxCR2 (Fig. 2D). Compared with the ndb-derived population, significantly more db-derived Gr-1⁺Kit⁺Sca-1⁻ CCR2⁻ cells also expressed CxCR2, while markedly fewer cells belonged to the CxCR2⁻ precursor subset (Fig. 2D). These data indicate that Gr-1⁺CxCR2⁺ granulocytes isolated from the db BM express inappropriately high levels of the progenitor marker Kit, providing further evidence that they are immature.

We next examined Gr-1, CCR2, and CxCR2 MFI, which indicates cell-surface marker expression by individual cells. The db-derived Gr-1⁺CD11b⁺ population showed significantly reduced surface Gr-1 expression (Supplementary Fig. 3C) (P < 0.05). Likewise, db-derived Gr-1⁺ monocytes had significantly less surface CCR2 (Supplementary Fig. 3D) (P < 0.05) and db-derived Gr-1⁺ neutrophils had significantly less surface CxCR2 (Supplementary Fig. 3E) (P < 0.01). Reduced surface CCR2 and CxCR2 may explain the impaired chemotaxis of db-derived Gr-1⁺ cells (7). In combination with reduced Gr-1 MFI, lower CxCR2 MFI demonstrates that diabetes impairs neutrophil maturation.

To show that lower Gr-1 MFI reflects neutrophil immaturity and not an expanded Gr-1^intermediate monocyte population, we examined the morphology of Giemsa-stained cells. This confirmed that Gr-1⁺ cells are predominantly granulocytes (Supplementary Fig. 3F) and that the Gr-1⁺ monocyte population is not expanded in db BM (Supplementary Fig. 3G). As the expanded Gr-1⁺ precursor population includes eosinophils, we verified that diabetes does not skew the neutrophil/eosinophil balance (Supplementary Fig. 3H). Furthermore, we found that diabetes does not alter the proportion of immature blasts and promyelocytes relative to more mature myelocytes and metamyelocytes (Supplementary Fig. 3I). Although gross morphological analyses do not allow for detailed assessment of maturity, together these data support the conclusion that db-derived Gr-1⁺ neutrophils are immature.

CCR2 and CxCR2 expression by Gr-1⁺ cells can be used as a readout in a culture maturation assay. We predicted that the proportion of Gr-1⁺ cells within the Gr-1⁺ neutrophil subset would increase due to upregulation of CxCR2 as neutrophils mature. Conversely, we expected the Gr-1⁺ monocyte population to shrink because CCR2 is downregulated upon monocyte maturation (17). We therefore compared chemokine receptor expression in ndb- and db-derived BMD Gr-1⁺ cells after 48 h in culture (Fig. 3A). As anticipated, the Gr-1⁺ monocyte subset significantly decreased as a proportion of the total ndb-derived Gr-1⁺ population (Fig. 3B) (P < 0.001), while the Gr-1⁺ neutrophil subset expanded (P < 0.05). Strikingly, the db-derived Gr-1⁺ population showed no expansion of the Gr-1⁺ neutrophil subset and no contraction of the Gr-1⁺ monocyte subset, demonstrating that neutrophilic and monocytic db-derived Gr-1⁺ cells regulate their chemokine receptor expression inappropriately and mature abnormally.

We next assayed survival, division, apoptosis, and transdifferentiation of cultured Gr-1⁺ cells to exclude the possibility that a shift in population accounted for differences in CCR2 and CxCR2 expression (e.g., increased proliferation or apoptosis in CCR2⁺ or CxCR2⁺ cells). The morphology of Giemsa-stained Gr-1⁺ cells was examined and quantified after 48 h in culture. The monocyte/granulocyte balance was unaltered with culture and did not differ between db and ndb populations (Supplementary Fig. 4A), indicating that transdifferentiation did not cause the increase in CxCR2⁺ and decrease in CCR2⁺ cells after culture. Likewise, although some cultured Gr-1⁺ cells changed morphology and developed protrusions after 72 h, morphology generally did not differ in db-derived cells (Supplementary Fig. 4B) and at no point was cell number significantly altered from controls (Supplementary Fig. 4C). eFluor 450–labeled db- and ndb-derived cells underwent an ˜0.5-fold change in MFI during 48 h in culture, indicating that Gr-1⁺ cells from db and ndb mice divide at similar rates (Supplementary Fig. 4D). Analysis of apoptosis indicated no significant differences between the two populations (Supplementary Fig. 4E). This is further evidence in support of a preprogrammed failure of db-derived cell maturation.

We analyzed our NGS data set from BM- and wound-derived Gr-1⁺ cells isolated from ndb and db mice to identify intrinsic factors that may contribute to the impaired differentiation phenotype. Enrichment analyses of loci with significantly differential reads were performed using DAVID (18–20). Functional annotation analyses showed that the GO term “regulation of myeloid cell differentiation” (GO:0045637) was significantly enriched for genes differentially expressed between ndb- and db-derived Gr-1⁺ cells from both the BM and the wound (BM, P = 3.4 × 10⁻⁴, Benjamini correction for multiple comparisons P = 3.7 × 10⁻²; wound, P = 9.9 × 10⁻³, Benjamini correction for multiple comparisons P = 2.0 × 10⁻¹). This deregulation of gene expression may underpin aberrant myeloid cell differentiation in the db environment. We identified 48 genes associated with the GO term “myeloid cell differentiation” (GO:0030099) that were deregulated in both the db BM and wound (Table 1). Cebpa is a key master regulator of myeloid cell differentiation and granulocyte maturation (21); thus, it was selected for functional analysis. Cebpa mutations, promoter methylation, and histone deacetylation are associated with an immature myeloid phenotype, e.g., in acute myeloid leukemia (22–24). Notably, Cebpa deficiency causes myeloid cell immaturity (25).

Like many genes, acetylation of histone H3 at its promoter regulates Cebpa transcription (24). Diabetes-induced histone modifications at promoters of regulators of inflammatory gene expression contribute to the altered myeloid phenotype (26). However, the impact of diabetes on the promoters of genes driving myeloid differentiation was previously unknown. We therefore performed ChIP at the Cebpa proximal promoter in Gr-1⁺ cells (Fig. 4A) to examine the effect of diabetes on H3K27Ac, a mark of active chromatin that is extensively modulated during myeloid differentiation (27). We found significantly diminished H3K27Ac at the Cebpa promoter in db-derived Gr-1⁺ cells (Fig. 4B) (P < 0.05), consistent with the reduced mRNA level seen in our NGS data set and Cebpa deregulation.

CEBPA targets in human neutrophils were recently identified (28). We examined whether this gene set was significantly deregulated in db-derived Gr-1⁺ cells. We used the Mouse Genome Informatics database to identify the murine homologs of 30 of the 33 downstream target genes described in that study and found that 22 of 30 were deregulated in Gr-1⁺ cells (Supplementary Table 1). This is significantly higher than expected by chance as indicated by a hypergeometric test (P = 2.8 × 10⁻⁵).

Interestingly, Ccr2 and Cxcr2 are also known targets of CEBPA in myeloid cells (29,30). Both genes are significantly repressed in our db-derived Gr-1⁺ population (Supplementary Fig. 3). In addition, CEBPA activates CSf2rα, which encodes GM-CSF receptor α (GM-CSFRα) (31) and is downregulated by 0.74-fold (P = 0.08) in db-derived Gr-1⁺ cells in our NGS data set. For examination of the impact of CSf2rα transcript deficiency on Gr-1⁺ cells’ ability to respond to GM-CSF, the ligand for GM-CSFR that promotes monocyte maturation, we repeated our culture maturation assay with GM-CSF–cultured cells. Gr-1⁺ cells should downregulate CCR2, as they mature in response to GM-CSF. We found that the ndb-derived Gr-1⁺ monocyte population showed a significant contraction in size in response to GM-CSF (Fig. 4C) (P < 0.05), while db-derived Gr-1⁺ monocytes again failed to respond. These data suggest that aberrant expression of Cebpa induced by diabetes results in dysregulation of key maturation-promoting pathways, including GM-CSF signaling, and impaired myeloid differentiation and maturation.

To determine whether deregulation of Cebpa expression extended to db patients, we used quantitative RT-PCR (qRT-PCR) to assay Cebpa expression by peripheral blood granulocytes isolated from patients with T1D or T2D and healthy control subjects (Supplementary Fig. 5A). As in our mouse model of T2D, we found that granulocytes from patients with T2D expressed less Cebpa than granulocytes from healthy volunteers (P = 0.11). However, granulocytes from patients with T1D showed little difference in Cebpa expression. We confirmed this by analyzing NGS data from healthy individuals and patients with T1D that was available on the GEO database (Supplementary Fig. 5A). Again, granulocytes from patients with T1D did not differ in Cebpa expression levels compared with healthy control subjects. This suggests that repression of Cebpa as a mechanism for inhibition of myeloid maturation is restricted to T2D. We also measured the effect on murine macrophage Cebpa expression of differentiation from ndb- and db-derived BM precursors in the presence of “high” and “low” glucose concentrations in an in vitro hyperglycemia model (Supplementary Fig. 5B). We found that growth medium glucose concentration did not affect Cebpa expression in macrophages, indicating that factors other than hyperglycemia may underlie the Cebpa repression seen in the db mouse model of patients with T1D and T2D.

As we saw significantly decreased acetylation of the Cebpa promoter in db-derived myeloid cells compared with ndb-derived cells, we investigated whether the db environment deregulated HDACs. Culture in high-glucose medium to mimic the effects of hyperglycemia has shown that glucose induces changes in the expression of chromatin-modifying factors in primary monocytes or myeloid cell lines (14). However, investigations of primary myeloid cells from db patients or mouse models have been very limited in scope.

Using our NGS data set, we identified 17 chromatin regulators that were significantly deregulated in db-derived myeloid cells compared with their ndb-derived counterparts (Table 2), including Hdac4. Transcriptional repression mediated by HDAC4 promotes a myeloid leukemia phenotype showing impaired myeloid maturation (32). HDAC4 is associated with increased production of reactive oxygen species and monocyte adhesion to smooth muscle cells in response to tumor necrosis factor (TNF) (33). We therefore examined HDAC4 expression in myeloid cells.

Macrophages cultured from db-derived BM appear to express more Hdac4 transcript than macrophages from ndb mice (P = 0.089) (Fig. 5A). Importantly, immunofluorescence indicated that the HDAC4 nuclear-to-cytoplasmic ratio was significantly elevated in db-derived macrophages (P < 0.05) (Fig. 5B). As HDAC4 nuclear localization is required for function (34), this suggests higher HDAC4 activity in db-derived myeloid cells.

G-CSF induces Cebpa expression via SHP2 and promotes granulocyte differentiation (24). To establish whether G-CSF treatment could correct the db phenotype with respect to CCR2 and CxCR2 expression, we performed an additional maturation assay. Freshly isolated Gr-1⁺ cells were treated with G-CSF for 48 h, and CCR2 and CxCR2 expression was analyzed by flow cytometry (Fig. 6A). In both ndb- and db-derived cultures, the proportion of Gr-1⁺ monocytes significantly decreased (P < 0.05) and the proportion of Gr-1⁺ neutrophils significantly increased (P < 0.01). This indicated that diabetes-induced defects in maturation capacity could be rescued by G-CSF treatment. We also verified by qRT-PCR that G-CSF induced Cebpa in db-derived Gr-1⁺ cells to levels similar to those in untreated ndb-derived Gr-1⁺ cells (Fig. 6B). Finally, to ascertain that the G-CSF–mediated rescue of db myeloid maturation was specifically due to Cebpa upregulation, we transiently transfected db-derived Gr-1⁺ cells with a Cebpa expression vector. Again, CxCR2 expression was significantly increased in Cebpa-transfected cells compared with control transfected cells (Supplementary Fig. 6A). This confirms that inducing Cebpa expression in db-derived Gr-1⁺ cells promotes maturation as assessed by chemokine receptor expression.

Like myeloid cells from db patients (35), Gr-1⁺ cells from Lepr^db/db mice have an impaired capacity for phagocytosis (36). To examine whether Cebpa induction by G-CSF could rescue this defect, we compared phagocytosis of Alexa Fluor 488–labeled zymosan particles by db-derived Gr-1⁺ cells cultured for 24 h in the presence and absence of G-CSF. Compared with cells cultured in the absence of G-CSF, we found a small but significant increase in phagocytosis in G-CSF–cultured Gr-1⁺ cells (Supplementary Fig. 6B), indicating that G-CSF–mediated upregulation of Cebpa promotes functional maturation of db-derived myeloid cells.

Finally, to determine whether G-CSF improves diabetic wound healing in vivo, we applied G-CSF to dorsal excisional wounds made on diabetic mice. Wounds were treated every 48 h for the first 8 days after wounding, and wound area was compared with that of PBS-treated controls (Fig. 6C). Mean area was consistently lower for G-CSF–treated wounds compared with controls, indicating that G-CSF accelerates wound closure.

Chronic inflammation affects myeloid cell development from the initial production of multipotent progenitors in the BM through their differentiation and recruitment to sites of inflammation. Colitis models show disrupted hematopoiesis, including alterations in hematopoietic stem and progenitor cell (HSPC) proliferation and myeloid/lymphoid differentiation capacity, and shifts in the relative abundance of specific progenitor cells in the BM (12). Diabetes causes systemic chronic inflammation (37), and T2D was recently shown to induce widespread changes to the BM stem cell niche that diminish numbers of early hematopoietic progenitors and deregulate miRNAs induced during the macrophage inflammatory response (38,39). However, the full extent of the effect of diabetes on hematopoiesis is unknown.

We previously demonstrated that the Gr-1⁺CD11b⁺ myeloid progenitor population is deregulated by diabetes in the BM, blood, and wound. Db-derived BM Gr-1⁺ cells perform poorly in functional assays (7,36) and do not promote angiogenesis in vivo (7). Their altered differentiation capacity is shown both in their skewed colony formation (7) and in their hyperpolarization in response to stimuli (5). Wound resolution requires myeloid cells (40), but the excessive numbers of defective cells found in the db wound are believed to impair healing (41). In addition to persistent expression of M1 “proinflammatory” markers late in healing, gene expression analysis indicates that macrophages do not downregulate Gr-1/Ly6C or upregulate F4.80 in the db wound (5,6), consistent with persistent myeloid cell immaturity in the db wound.

We recently provided novel evidence that wound macrophage polarization is prespecified, occurring prior to recruitment through unknown mechanisms (5). Consequently, and because of increasing evidence that myeloid cells in the db wound are functionally immature, we used BMD Gr-1⁺ cells to examine how diabetes inhibits myeloid maturation and whether this was specific to the monocytic or granulocytic Gr-1⁺ subset.

We divided Gr-1⁺ cells into four subpopulations based on CCR2 and CxCR2 expression. Chemokine receptor expression patterns were indicative both of cell maturity and of granulocytic/monocytic morphology. Within the Gr-1⁺ population, CxCR2 expression is on a continuum, with the highest expression levels corresponding to the most morphologically developed neutrophil granulocytes (11); when ndb Gr-1⁺ cells are removed from the BM and allowed to mature, the proportion of cells expressing CxCR2 increases. CCR2⁺ cells form a discrete population of monocytes; during maturation, the proportion of the ndb Gr-1⁺ population that expresses CCR2 decreases, consistent with a recent description of CCR2 expression as a marker of immature monocytes and CCR2 downregulation as a requirement for maturation (17). Strikingly, we found that in Gr-1⁺ cells from db mice, neither CxCR2⁺ neutrophils nor CCR2⁺ monocytes regulated their chemokine receptor expression appropriately: there was no significant change in the proportion of the population expressing either marker, which indicates that diabetes inhibits maturation of both monocytes and neutrophils. Interestingly, we also saw that Kit expression was significantly elevated in db-derived Gr-1⁺ cells and more particularly in the CxCR2⁺ subset. This provides further evidence of granulocyte immaturity.

NGS data from ndb- and db-derived Gr-1⁺ cells showed that diabetes results in the deregulated expression of a battery of myeloid differentiation factors and chromatin-remodeling enzymes. This strongly suggests that diabetes causes widespread disruption to myeloid differentiation, with chromatin remodeling as a potential mechanism. Excitingly, we found that the key myeloid transcription factor CEBPA is repressed in myeloid cells in both the Lepr^db/db mouse model of T2D and peripheral blood neutrophils from patients with T2D. CEBPA is a crucial factor in myeloid fate choice and granulocytic differentiation that was recently shown to be significantly upregulated during monocyte-to-macrophage differentiation (21,27). Cebpa was therefore a promising candidate gene for investigation. We found significantly reduced transcript abundance and significantly diminished H3K27Ac at the Cebpa promoter, indicating that the db environment repressed Cebpa transcription and that this was mediated in part by aberrant chromatin remodeling. CEBPA target gene expression was also significantly disrupted. In particular, CEBPA activates both Ccr2 and Cxcr2 (29,30), and we found that individual db-derived Gr-1⁺ cells expressed significantly lower levels of both markers. A further CEBPA target, Csf2ra (42), encodes the α subunit of the receptor for GM-CSF, a cytokine critical for myeloid maturation. Its downregulation in diabetes was associated with failure of db-derived Gr-1⁺ cells to mature properly in response to GM-CSF. Crucially, upregulation of Cebpa in response to exogenous G-CSF treatment of db-derived Gr-1⁺ cells was sufficient to rescue maturation, as measured by CCR2/CxCR2 receptor regulation, thus conclusively demonstrating that Cebpa repression by diabetes underlies the immature phenotype of Gr-1⁺ cells. Treatment of db wounds with G-CSF enhanced wound closure, although effects were modest. It is possible that targeting myeloid cells prior to their recruitment to the wound may be more efficacious in accelerating healing.

CEBPA is known as a key regulator of myelopoiesis (21), but it was recently identified as a “pioneer factor” that specifies myeloid cell fate in early progenitors in combination with endogenous PU.1/Spi1 (43) and is sufficient to drive lymphocyte-to-macrophage transdifferentiation (44). CEBPA binds “closed” chromatin at sites that will act as enhancers, flagging regulatory DNA to chromatin-modifying enzymes that subsequently “open” chromatin by depositing histone marks, thus establishing lineage-specific transcriptional programs that promote differentiation (43). CEBPA functions thus in lipopolysaccharide (LPS)-induced macrophage activation, where it is recruited to new enhancer sites along with the nuclear factor-κB subunit p65 (45). Absence of CEBPA in a myeloid cell line wipes H3K9Ac and H3K4me3 from target gene promoters and represses gene expression (29), while Cebpa knockout induces aberrant expression of chromatin-remodeling enzymes and altered patterns of histone modification in HSPCs (46). CEBPA-driven transdifferentiation of lymphocytes to macrophages is also accompanied by downregulation of chromatin-modifying factors (44). Notably, in our data set, Cebpa repression in db-derived cells was accompanied by diminished H3K27Ac at the gene promoter and widespread changes to expression of chromatin-remodeling enzymes including the HDAC family. The potential impact of Cebpa deregulation by diabetes therefore extends across all stages of myeloid differentiation.

How diabetes induces Cebpa repression is unclear, but the Toll-like receptor (TLR)4 pathway is a promising avenue for further investigation. Myeloid cells signal through TLR4 in response to ligands including bacterial LPS, causing nuclear translocation of the nuclear factor-κB subunit p65 and the expression of proinflammatory cytokines including TNF (47). TLR4 activation and ligands are elevated in patients with diabetic (48,49). TLR4 activation in myeloid cells results in suppression of CEBPA and its target genes (50), while in HSPCs, low-dose LPS, absence of CEBPA, and streptozotocin-induced diabetes are associated with an expanded long-term hematopoietic stem cell population (51–53). Together, these data suggest that examining the connection between chronic TLR4 activation in diabetes and its consequences for Cebpa expression and myeloid maturation may be important. ChIP-seq of db-derived myeloid progenitors for CEBPA occupancy of promoters and enhancers would be particularly informative.

Here, we have shown that Cebpa is repressed by diabetes in a myeloid population spanning myeloblasts to mature monocytes and granulocytes, leading to target gene repression and impaired cellular maturation that likely synergize with other intrinsic defects in recruited myeloid cells, such as altered noncoding RNA expression, to contribute to chronic inflammation in the diabetic wound. In light of the indispensible role CEBPA plays in lineage specification and myeloid differentiation, this is an extremely significant finding that represents a large step forward in our understanding of how diabetes deregulates the myeloid phenotype. Elucidation of the underlying mechanisms by which this occurs is crucial both to further our understanding of chronic inflammation and for the appropriate targeting of therapies to treat diabetes and its complications.

Acknowledgments. The authors thank Mike Jackson for flow cytometry assistance, Roger Meadows for microscopy assistance, Marzieh Kamjoo for animal genotyping, Thomas Bleazard for help with Python, and Mat Hardman and Matt Ronshaugen for useful discussions (all from the University of Manchester).

Funding. The Bioimaging Facility microscopes used in this study were purchased with grants from the Biotechnology and Biological Sciences Research Council (BBSRC), the Wellcome Trust, and the University of Manchester Strategic Fund. Funding for this study was provided by the Wolfson Charitable Trust, The Healing Foundation, and the BBSRC.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.W. designed experiments, performed experiments, analyzed data, and wrote the manuscript. T.T. designed experiments, performed experiments, and analyzed data. T.U. performed experiments. S.A. designed experiments and analyzed data. N.B. designed experiments and analyzed data. K.A.M. designed experiments, performed experiments, analyzed data, and wrote the manuscript. K.A.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the European Molecular Biology Organization Molecular and Cellular Basis of Regeneration and Tissue Repair meeting, Sant Feliu de Guixols, Spain, 6–10 September 2014, and the Gordon Research Seminar and Conference on Tissue Repair and Regeneration, New London, NH, 6–7 June 2015.