Enhanced Proliferation of Ly6C⁺ Monocytes/Macrophages Contributes to Chronic Inflammation in Skin Wounds of Diabetic Mice

Diabetes and related complications represent a growing public health crisis throughout the world (1). Among diabetes-related complications, nonhealing ulcers are a major concern, often leading to amputation and even death. The socioeconomic burden of chronic wounds is escalating, with annual Medicare costs as high as $13 billion (1, 2). Despite the heavy toll of chronic wounds, the causes of impaired healing in diabetes are not well understood (3).

Monocytes (Mo) and macrophages (MΦ) play critical roles in wound healing by regulating inflammation, removing debris and dead cells, and producing cytokines and growth factors that promote healing (3). However, a common characteristic of poorly healing diabetic wounds is chronic inflammation, associated with persistent accumulation of inflammatory Mo/MΦ (3). Our laboratory and others have shown that diabetes leads to dysregulation of myelopoiesis in bone marrow, which may contribute to persistent accumulation of inflammatory Mo/MΦ in skin wounds (4, 5). Moreover, our recent findings show that proliferation of skin wound Ly6C⁺F4/80^lo/− Mo/MΦ contributes to their accumulation during normal healing in nondiabetic mice (6). However, whether diabetes influences proliferation of inflammatory Mo/MΦ remains to be established.

Many factors, both cell intrinsic and environmental, have been reported to induce proliferation of Mo/MΦ, but these appear to be context dependent, varying between tissues and inflammatory conditions (6–9). In our recent study, the data indicated that the wound environment likely regulates Mo/MΦ proliferation (6). We screened a number of potential regulators that have been reported to influence Mo/MΦ proliferation, including IL-1β, IL-6, IL-4, and M-CSF (6–10). However, our data did not support a role for any of these candidates in regulating Mo/MΦ proliferation in skin wounds (6). Other cytokines including CCL2, CCL3, TNF-α, IL-10, and IL-33 have been reported to influence Mo/MΦ activity, although their role in regulating wound Mo/MΦ proliferation has not been explored (11–13). Thus, the factor(s) that regulate Mo/MΦ proliferation during wound healing remain to be elucidated.

The purposes of this study were to determine whether diabetes influences Mo/MΦ proliferation in skin wounds and to identify potential factor(s) that regulate Mo/MΦ proliferation in wounds. Indeed, we found that inflammatory Ly6C⁺F4/80^lo/− Mo/MΦ exhibit significantly higher rates of proliferation in wounds of diabetic compared with nondiabetic mice. Moreover, our data demonstrate a novel role for CCL2/CCR2 signaling in promoting proliferation of inflammatory Mo/MΦ in wounds.

Male C57BL/6 (nondiabetic controls, CD45.2), BKS.Cg-Dock7^m +/+ Lepr^db/J (db/db, diabetic), B6.SJL-Ptprc^a Pepc^b/BoyJ (CD45.1), and B6(C)-Ccr2tm1.1Cln/J (CCR2^−/−EGFP) mice (age 9–12 wk) were purchased from The Jackson Laboratory (Bar Harbor, ME) and housed in environmentally controlled conditions with a 12-h light/dark cycle. Water and food were available ad libitum. With the exception of the single cell RNA sequencing (scRNAseq), each experiment was performed at least twice with n = 3–6 mice per group/condition. To minimize bias, mice were randomly assigned to experimental groups, and the resulting samples were coded and analyzed in a blinded fashion. All animal studies were approved by the Animal Care and Use Committee of the University of Illinois at Chicago.

Age-matched nondiabetic and diabetic mice were subjected to excisional wounding (two wounds per mouse) of the dorsal skin with an 8-mm biopsy punch as previously described (6). Tegaderm (3M,1626W) was used to cover the wounds until tissue harvest.

Wound closure was evaluated by digital pictures using Fiji Image J. Skin wounds were collected using a 12-mm biopsy punch on day 3, 6, and 10 postinjury, followed by enzymatic digestion to obtain a single cell suspension for flow cytometry analysis or snap frozen in liquid nitrogen for protein measurements as previously described (6).

Skin wound, femoral bone marrow cells, and peripheral blood cells were used for flow cytometry analysis. For skin wound cells, Zombie Violet or Aqua (BioLegend, San Diego, CA) were used to assess cell viability. Cells from bone marrow and blood were not stained with Zombie dyes because of low percentage of dead cells (<3%). After Fc receptor blocking with anti-CD16/32 Ab (clone S17011E; BioLegend), skin cells were labeled with anti-Ly6G⁻Brilliant Violet (BV) 605 (clone 1A8; BioLegend), CD11b⁻ APC/Fire 750 (clone CBRM1/5; BioLegend), F4/80⁻PE (clone BM8; BioLegend), and Ly6C⁻Percp/cy5.5 (clone AL−21; BD Biosciences, San Jose, CA) Abs; cells from bone marrow and blood were stained with anti-Ly6G⁻BV605, CD11b⁻ APC/Fire 750, CD115⁻PE (clone AFS98), Ly6C⁻BV421 (clone AL−21; BD Biosciences), and/or cKit-Percp/cy5.5 (clone 2B8; BioLegend) Abs.

For proliferation/cell cycle analysis, all cells were fixed and permeabilized and then intracellularly labeled with anti-Ki67 Ab (ab15580; Abcam, Cambridge, MA) or its corresponding isotype control (ab171870; Abcam) followed by Alexa Fluor 488 anti-rabbit secondary Ab (ab150077; Abcam) incubation using BD Cytofix/Cytoperm Kit. Finally, cells were incubated with FxCycle Far Red (Thermo Fisher Scientific, Grand Island, NY) 30 min before acquisition following the manufacturer’s instructions.

For scRNAseq, single cells were collected and pooled from skin wounds of two nondiabetic or two diabetic mice on day 6 postinjury as described above. Target cell population was sorted out as Zombie⁻CD45⁺CD11b⁺Ly6G⁻ cells on BD FACSAria III sorter (BD Biosciences). Cells were first stained with Zombie Violet and then blocked Fc receptors by anti-CD16/32 Ab followed by incubation of anti–Ly6G-BV605, CD11b⁻ APC/Fire 750, and anti–CD45-FITC (clone 30-F11; BioLegend).

For adoptive transfer experiments, cells were collected from six long bones of each donor mouse (two femurs + two tibiae + two humeri). After Zombie Aqua staining and Fc receptor blocking, cells were labeled with anti–Ly6G-BV510 (clone 1A8; BioLegend), CD11b⁻ APC/Fire 750, CD115-PE, Ly6C-BV421, and cKit-Percp/cy5.5. Next, bone marrow immature Mo were sorted out as Zombie⁻Ly6G⁻CD11b⁺cKit⁻CD115⁺Ly6C^hi from CD45.1 mice or Zombie⁻Ly6G⁻CD11b⁺cKit⁻CD115⁺Ly6C^hi EGFP⁺ from CCR2 knockout EGFP mice on MoFlo Astrios EQ Sorter (Beckman Coulter, Brea, CA).

All samples were analyzed on either LSR Fortessa with high throughput sampler (BD Biosciences) or CytoFLEX S (Beckman Coulter) cytometers. Data were analyzed using FlowJo (FlowJo, Ashland, OR).

For single cell imaging, skin Ly6C⁺F4/80^lo/− Mo/MΦ and Ly6C⁻F4/80⁺ MΦ were prelabeled and FACS sorted. Then, 2000 cells from each population were collected on Amnis FlowSight (Luminex, Chicago, IL) with further confirmed surface labeling of either anti–Ly6C-FITC (clone AL−21; BD Biosciences) or anti–F4/80-PE (clone 1A8; BioLegend) together with DNA dye DRAQ5 (BioLegend). Cell images were captured using charge-coupled device camera with ×20 magnification on FlowSight and analyzed using IDEAS Software (Luminex). Five representative pictures of either Ly6C⁺F4/80^lo/− Mo/MΦ or Ly6C⁻F4/80⁺ MΦ were presented.

Cells from skin wounds of either nondiabetic or diabetic mice were collected and sorted as described above. Fresh cells with viability over 85% were resuspended in cold Dulbecco’s PBS (DPBS) with 10% FBS at a concentration of ∼200 cells/μl and then immediately processed using the Chromium platform (10x Genomics, San Francisco, CA) for high m.w. genomic DNA extraction, Gel Bead-In EMulsions generation, and barcoding. Library construction was prepared according to the manufacturer’s instructions and sequenced on HiSeq Sequencing Systems (Illumina, San Diego, CA) with paired-end reads. Fastq files were generated and demultiplexed into single cells using Cell Ranger software (10x Genomics).

Cell barcodes with <10% mitochondrial expression were removed from the dataset, and genes with fewer than 50 counts across all cells or expressed in fewer than 5% of the cells were removed. Gene expression was normalized to counts per million using trimmed mean of M values normalization with edgeR (14) and log scaled. Principal component analysis of cells was computed, and the top 10 principal components, which captured 47% of total variance, were used downstream for clustering analysis.

Cells were clustered using k-means clustering with 10 random initializations on a range of cluster numbers k (2–20). For each k, we evaluated the reproducibility of the repeated clustering runs by comparing the pairwise distance between clustering results, computed as the fraction of coclustered feature pairs between two clustering results relative to the number of coclustered feature pairs within each result individually. This difference was averaged across all result pairs for each value of k to evaluate cluster robustness. k = 4 was determined to be an optimal cluster count and used for downstream analysis.

Differentially expressed genes per cluster were detected using three statistical tests: Kruskal–Wallis, on log-scaled normalized gene expression levels; Fisher exact test, between the count of expressed versus unexpressed cells for each gene; and the area under the receiver operating characteristic curve, treating each gene as a classifier for each cluster.

Whole skin wounds from nondiabetic and diabetic mice were homogenized as previously described (6). Total protein levels were measured using Pierce Rapid Gold BCA Protein Assay Kit (Thermo Fisher Scientific). Protein levels of CCL2, CCL3, TNF-α, IL-4, IL-10, M-CSF, and IL-33 were determined using BioLegend LEGENDplex Kit. All samples were analyzed on CytoFLEX S (Beckman Coulter) cytometer.

On day 2 postinjury, recombinant mouse CCL2 protein (carrier-free; BioLegend) was administrated at a single dose of 60 ng per wound by intradermal injection at four equally spaced sites around the periphery of each wound in nondiabetic mice (15 ng/injection, n = 12 per group) (15). Control mice received an equal volume of DPBS injection (n = 12 per group). On day 3 postinjury, wounds treated with either CCL2 or DPBS were collected and processed as described above for flow cytometry analysis.

Bone marrow Ly6C^hi Mo from donor mice were collected as described above by FACS sorting. Approximately 1.5–2 million cells from each donor strain were resuspended in cold DPBS and injected intradermally into skin wounds of the recipient mice at four equally spaced sites by the edge on day 2 (cells combined from CCR2 knockout mice + CD45.1 mice to nondiabetic mice, n = 5 per group) or day 2 and day 5 (cells from CD45.1 mice to nondiabetic or diabetic mice, n = 3 per group). Approximately 18 h postadoptive transfer, skin wounds from recipient mice were collected on day 3 or day 6 postinjury and processed as described above for flow cytometry analysis to evaluate cell proliferation of the donor cells.

Data are expressed as mean ± SEM. Statistical significance of differences was evaluated by Mann–Whitney U test or ANOVA. A p value < 0.05 was considered statistically significant.

scRNAseq data in this study are available in Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) with the accession code GSE154400. All other relevant data are available from the corresponding authors.

Consistent with previous work by us and other groups, diabetic mice showed significantly delayed wound closure compared with nondiabetic controls (3, 6, 16). By day 10 postinjury, wounds in diabetic mice achieved an average of only ∼40% closure, whereas wounds in nondiabetic mice reached an average of ∼95% closure (Fig. 1A).

Next, we measured the accumulation and proliferation of two major Mo/MΦ populations in skin wounds by flow cytometry, namely live Ly6G⁻CD11b⁺Ly6C⁺F4/80^lo/− (inflammatory Mo/MΦ) and live Ly6G⁻CD11b⁺Ly6C⁻F4/80⁺ (mature MΦ) cells (Fig. 1B). Both cell populations accumulated in skin wounds by day 3 postinjury with significantly higher cell numbers in diabetic mice than in nondiabetic mice on day 6 and day 10 postinjury (Fig. 1C). Moreover, percentages of Ly6C⁺F4/80^lo/− Mo/MΦ were also significantly higher in diabetic wounds than in nondiabetic wounds on day 6 and day 10 postinjury, whereas percentages of Ly6C⁻F4/80⁺ MΦ were comparable between strains at all time points (Supplemental Fig. 1A).

Importantly, and consistent with our previous findings, there were significantly higher numbers and percentages of Ly6C⁺F4/80^lo/− Mo/MΦ in the S/G2/M phases of the cell cycle (Ki67⁺FxCycle⁺) compared with Ly6C⁻F4/80⁺ MΦ over the course of healing, in both diabetic and nondiabetic mice (Fig. 1D, Supplemental Fig. 1B). Although both diabetic and nondiabetic mice showed a peak of Ly6C⁺F4/80^lo/− Mo/MΦ in S/G2/M phases on day 6 postinjury, diabetic mice had a significantly higher peak of proliferating Ly6C⁺F4/80^lo/− Mo/MΦ, which lasted through day 10 postinjury (Fig. 1D, Supplemental Fig. 1B, 1C). In addition to extremely low numbers and percentages of Ly6C⁻F4/80⁺ MΦ in S/G2/M phases, there were no differences in proliferation of these latter cells at any time point between diabetic and nondiabetic mice (Fig. 1D, Supplemental Fig. 1B, 1D). Analysis of the forward and side scatter characteristics and morphology of these cells on day 3 and day 6 postinjury indicated that Ly6C⁺F4/80^lo/− Mo/MΦ were slightly smaller and clearly less granular than Ly6C⁻F4/80⁺ MΦ, indicating that the former exhibit a more monocytic morphology (Supplemental Fig. 1E, 1F). Note that the full cell cycle analysis of Ly6C⁺F4/80^lo/− MΦ/Mo and Ly6C⁻F4/80⁺ MΦ in both mouse strains can be found in the Supplemental Fig. 1C, 1D.

Taken together, these data indicate that wounds in diabetic mice contain higher levels of proliferating Ly6C⁺ Mo/MΦ compared with their nondiabetic counterparts.

To determine whether enhanced Mo/MΦ proliferation in skin wounds of diabetic mice is associated with altered expression of cell cycle genes in these cells, we performed scRNAseq analysis on Mo/MΦ isolated from wounds on day 6 postinjury, which we identified as the peak of proliferation within our time course.

Mo/MΦ from skin wounds of diabetic and nondiabetic mice were pre-enriched as live Ly6G⁻CD45⁺CD11b⁺ by FACS sorting before capturing. Data were prefiltered, which resulted in a total pool of 2904 cells (diabetic: 1767 versus nondiabetic: 1137). Using k-means clustering analysis on the combined pool of cells, we identified four clusters of Mo/MΦ. As shown in Fig. 2A–C, cluster 1 was enriched in cells from nondiabetic wounds with relatively high expression levels of several NLRP3 inflammasome-related genes, such as Cxcl2, Nlrp3, and Il1b (17, 18), along with other genes associated with inflammation such as Clec4d, Ets2, Vegfa, and Thbs1 (19–23). Cluster 2 contained similar levels of cells from both strains, but most genes were expressed only at low levels, suggesting an inactive cell population. In contrast, cluster 3, which was enriched in cells from diabetic mice, expressed genes associated with anti-inflammatory functions, including Coro1a, Cst3, Ly86, and Pld4 (24–28), as well as markers for other myeloid cells, such as Cd52 and H2-DMa (29–31). Importantly, cluster 4, which was populated primarily by cells from diabetic mice, contained cells that expressed high levels of several genes related to cell proliferation, apoptosis, oxidative stress, energy homeostasis, and acute inflammation, such as Ppia, mt-Nd1, Pfn1, Calm2, Pltp, and Prdx1 (32–39). Thus, cluster 4 may represent a subset of proinflammatory Mo/MΦ undergoing proliferation and/or apoptosis that warranted further investigation. Interestingly, differential gene expression analysis indicated that common M1 and M2 markers were not specifically expressed in any one cluster, and most clusters coexpressed M1 and M2 markers (Supplemental Fig. 3), indicating that these clusters do not fit nicely into the M1/M2 classification scheme.

Next, we compiled a list of genes associated with progression and regulation of the cell cycle from previous publications and analyzed them in our scRNAseq dataset (40). As shown in Fig. 2D, the majority of proliferation-related genes were highly expressed in cluster 4, which was significantly enriched in cells from diabetic mice (diabetic 88% versus nondiabetic 12%). There was significant upregulation of checkpoint genes of S phase (Mcm2, Mcm3, Mcm4, Mik67, Mre11a, Msh2, and Rad17), G2 phase or G2/M transition (Ppm1d), and M phase (Cdk2, Rad21, Ran, Shc1, Smc1a, Stag1, Stmn1, and Terf1). Moreover, expression of most cell cycle–promoting genes was also enhanced in cluster 4, including Skp2, Cdk4, E2f3, and Tfdp1. Interestingly, along with the upregulation of these positive regulator genes of the cell cycle, genes associated with cell cycle arrest/negative regulation were also upregulated, suggesting that feedback inhibition may be taking place simultaneously with the promotion of proliferation in this population.

Taken together, the scRNAseq data revealed four Mo/MΦ populations with different phenotypes on day 6 postwounding, of which cluster 4 was populated primarily by cells from diabetic mice and showed a high expression of genes related to cell cycle, consistent with enhanced proliferation of wound Mo/MΦ in diabetic mice.

Wound Mo/MΦ are derived primarily from infiltrated blood Mo, which, in turn, are generated in the bone marrow (3). Thus, we explored Mo proliferation in bone marrow and peripheral blood during wound healing. We first assessed two Mo populations in bone marrow: Ly6G⁻CD11b⁺cKit⁻CD115⁺Ly6C^hi and Ly6G⁻CD11b⁺cKit⁻CD115⁺Ly6C^lo cells (Fig. 3A). Consistent with previous reports (5, 6), Ly6C^hi cells were the dominant Mo population in bone marrow in both mouse strains (Fig. 3B). At steady state, diabetic and nondiabetic mice had comparable levels of Ly6C^hi Mo at baseline, whereas injury increased these cells in diabetic mice on day 3 postwounding, resulting in a significantly higher percentage compared with the nondiabetic controls. Meanwhile, levels of mature Ly6C^lo Mo remained stable and did not differ between mouse strains throughout the healing process (Fig. 3B).

Next, at baseline, both Ly6C^hi and Ly6C^lo Mo proliferate in bone marrow with larger percentages of Ly6C^hi cells in the S/G2/M phases (Ki67⁺FxCycle⁺) compared with Ly6C^lo cells. In response to skin injury, both diabetic and nondiabetic mice showed a trend toward lower percentages of Mo in S/G2/M phases in bone marrow on day 3 postwounding and then slowly recovered toward baseline levels on day 10. However, there were no significant differences between diabetic and nondiabetic mice in the percentage of Ly6C^hi Mo in S/G2/M phases at any time point. Diabetic mice tended to have lower percentages of Ly6C^lo Mo in S/G2/M phases compared with nondiabetic mice, and this difference reached statistical significance on day 6 postwounding (Fig. 3C, Supplemental Fig. 2A, 2B).

At baseline, percentages of peripheral blood Ly6C^hi and Ly6C^lo Mo did not differ between diabetic and nondiabetic mice. However, after injury, diabetic mice maintained relatively high percentages of Ly6C^hi Mo throughout the healing process, whereas nondiabetic mice showed a decrease especially on day 3 postinjury, which resulted in significantly lower percentages of Ly6C^hi Mo in nondiabetic compared with diabetic mice on day 3 and day 6 (Fig. 3E). Different from skin wounds and bone marrow, virtually no blood Mo were found in S/G2/M phases of the cell cycle in either diabetic or nondiabetic mice; these cells were found primarily in the G1 phase (Ki67⁺FxCycle⁻) (Fig. 3F, Supplemental Fig. 2C).

Thus, whereas bone marrow Mo actively proliferate before and after skin wounding, mobilization into peripheral blood results in transition to the G1 phase. Subsequent infiltration into the wound then triggers reentry into the S/G2/M phases. Combined, these data suggest that environmental factors drive the proliferation of Mo/MΦ in diabetic skin wounds.

To establish the importance of the wound environment in driving Mo/MΦ proliferation, we used a local Mo adoptive transfer approach that involved isolating Ly6C^hi Mo from the bone marrow of CD45.1 congenic mice and injecting into the wounds of diabetic and nondiabetic mice that express CD45.2 on day 2 or day 5 postinjury. The proliferative status of the transferred cells was then assessed by flow cytometry 18 h later on day 3 or day 6 postinjury.

Importantly, Ly6C^hi Mo transferred to day 6 diabetic wounds showed a significantly higher percentage of cells in the S/G2/M as compared with those transferred to nondiabetic wounds on day 6 postinjury while having comparable percentages on day 3 postinjury (Fig. 4B). Interestingly, some Ly6C^hi Mo differentiated to Ly6C⁻ Mo/MΦ in the wound environment (Supplemental Fig. 4A), and those Ly6C^hi Mo derived Ly6C⁻F4/80⁺ MΦ in skin wounds maintained some proliferative capability as percentages of these latter cells in S/G2/M phases were higher than what we previously found in skin wounds (Figs. 1D, 4B). In short, these experiments confirm that wound environmental factors likely contribute to the increased Mo/MΦ proliferation in diabetic mice.

We previously screened a number of cytokines that have been reported to play a role in the regulation of Mo/MΦ proliferation in other tissues and pathological conditions, such as IL-6, IL-4, M-CSF, and IL-1β (6). However, our data did not support a role for any of these cytokines in regulation of Mo/MΦ proliferation in skin wounds. In this study, we expanded our screen to other potential regulators of proliferation, including TNF-α, IL-10, CCL2, CCL3, and IL-33, which have been reported regulating proliferation or other functions of Mo/MΦ (6–9, 41).

Among the proteins screened, we found that levels of CCL2, which is known to recruit Mo/MΦ to damaged tissue, was significantly upregulated postwounding and higher in wound homogenates from diabetic mice on day 6 and day 10 compared with nondiabetic controls (Fig. 4C). In contrast, although skin injury altered levels of TNF-α, IL-10, CCL3, IL-33, M-CSF, and IL-4 in wounds, there were no significant differences between nondiabetic and diabetic mice (Supplemental Fig. 4B–G).

To explore whether CCL2 promotes Mo/MΦ proliferation during wound healing, we treated wounds in nondiabetic mice with a single dose of 60 ng CCL2 or equal volume of PBS on day 2 postinjury and assessed Mo/MΦ proliferation 24 h later. As expected, CCL2 treatment induced a 1.6-fold higher accumulation of Mo/MΦ (CD11b⁺Ly6G⁻) in the wounds compared with PBS treatment (data not shown). Importantly, in response to CCL2 stimulation, Ly6C⁺F4/80^lo/− Mo/MΦ showed a significant increase in the percentage of cells in S/G2/M phases compared with PBS treatment (Fig. 5B).

Next, because CCR2 is the major receptor for CCL2 signaling, we performed adoptive transfer experiments with Ly6C^hi Mo from CCR2^−/− EGFP mice, in which an EGFP sequence was inserted into the translation initiation site of the Ccr2 gene, thereby abolishing Ccr2 gene expression, to investigate the effect of loss of CCL2/CCR2 signaling on wound Mo/MΦ proliferation. Ly6C^hi Mo were collected from bone marrow from CCR2^−/− EGFP and CD45.1 (control CCR2^+/+) mice, and equal numbers from each donor strain were transferred to wounds of C57BL6 mice on day 2 postinjury. Proliferation of both CCR2^−/− and CCR2^+/+ Ly6C^hi Mo were then analyzed at 18 h posttransfer. As shown in Fig. 5D, live CCR2^−/− or CD45.1⁺ cells were both observed in the wounds of recipient mice, indicating successful tracking of these cells. At baseline prior to transfer, percentages of bone marrow Ly6C^hi Mo in S/G2/M phases were comparable between CCR2^−/− and CCR2^+/+ mice. However, after adoptive transfer to wounds, CCR2^+/+ cells showed a significant increase in the percentage in S/G2/M phases, whereas CCR2^−/− cells did not, indicating that proliferation was blocked by the loss of CCR2 signaling (Fig. 5D).

In summary, our data demonstrate that the wound environment contributes to increased proliferation of Ly6C⁺ Mo/MΦ postinjury and that CCL2/CCR2 signaling is a key factor in promoting Mo/MΦ proliferation during skin wound healing in mice.

We investigated whether factors in the diabetic wound environment enhance Mo/MΦ proliferation, thereby contributing to persistent accumulation of these cells and impaired healing. Our data indicate that diabetic mice indeed exhibit higher rates of proliferation in skin wound Ly6C⁺F4/80^lo/− Mo/MΦ compared with nondiabetic controls. In addition, unbiased analysis of scRNAseq data revealed the presence of a Mo/MΦ cluster that exhibited upregulation of genes associated with progression and regulation of cell cycle. Notably, this Mo/MΦ cluster is populated primarily by cells from diabetic mice. Moreover, consistent with our previous study (6), Ly6C^hi Mo proliferate in bone marrow of both diabetic and nondiabetic mice, whereas in peripheral blood, Ly6C^hi Mo are maintained in the G1 phase. Regarding the mechanism underlying increased proliferation in wounds of diabetic mice, diabetic mice showed higher CCL2 protein levels in skin wounds compared with nondiabetic controls, and local wound treatment in nondiabetic mice with CCL2 protein increased proliferation of Ly6C⁺F4/80^lo/− Mo/MΦ. Importantly, adoptive transfer experiments using Ly6C^hi Mo lacking CCR2 showed that CCL2/CCR2 signaling is required for proliferation of these cells in the wound environment. Together, these data demonstrate that the diabetic wound environment enhances Mo/MΦ proliferation, which likely contributes to persistent accumulation of these cells and impaired healing in diabetic mice. These data also demonstrate a critical role of the CCL2 signaling pathway in driving Mo/MΦ proliferation during skin wound healing, in addition to its known function as a chemokine.

Diabetes leads to impaired wound healing in both humans and in animal models (3, 42). Associated with impaired healing, we found persistently high levels of Ly6C⁺F4/80^lo/− Mo/MΦ in wounds of diabetic mice, which are thought to be proinflammatory cells (3, 42), whereas mature Ly6C⁻F4/80⁺ MΦ levels were comparable between diabetic versus nondiabetic mice. Several previous studies demonstrated that mature, resident MΦ can proliferate, contributing to their accumulation in the lung, aorta, peritoneal cavity, and adipose tissue (7–9, 43). In contrast, we previously reported that inflammatory Ly6C⁺F4/80^lo/− Mo/MΦ but not mature Ly6C⁻F4/80⁺ MΦ proliferate in skin wounds, and in the current study, we demonstrate that wounds in diabetic mice contained persistently higher levels of proliferating Ly6C⁺F4/80^lo/− Mo/MΦ compared with nondiabetic controls. We also provide evidence that skin Ly6C⁺F4/80^lo/− Mo/MΦ exhibit a more Mo-like morphology than Ly6C⁻F4/80⁺ MΦ cells. Dixit et al. (10) and Rolot et al. (44) have also reported that Ly6C⁺ Mo/MΦ proliferate in a urinary tract infection model and hepatic Schistosoma mansoni infection model in mice, respectively (6). Thus, the proliferation of different Mo/MΦ subsets may depend on tissue and/or pathological context.

We and others have reported that transition from a proinflammatory Mo/MΦ phenotype to a prohealing phenotype is critical for normal wound healing, whereas diabetes leads to dysregulation of this process that leads to persistent accumulation of proinflammatory Mo/MΦ and impaired healing (3, 45). scRNAseq allows for a more comprehensive and unbiased description of cell heterogeneity than methods used previously by us and others to characterize Mo/MΦ phenotypes. Therefore, we conducted scRNAseq analysis in live Ly6G⁻CD45⁺CD11b⁺ cells isolated from wounds of diabetic and nondiabetic mice at the time point of peak wound Mo/MΦ proliferation. Using k-means clustering, cells were separated into four clusters, but no cluster exhibited clear M1 or M2 phenotypes, indicating that this classification system is not sufficient to characterize skin wound Mo/MΦ. Nonetheless, cells in cluster 4 exhibited a high expression of genes related to checkpoints and regulators of cell cycle, consistent with an actively proliferating Mo/MΦ population. Notably, this cluster was characterized by a higher abundance of cells from diabetic wounds compared with nondiabetic controls. Together, data from flow cytometry and scRNAseq demonstrated that a subset of Mo/MΦ proliferate in skin wounds and that this subset in increased in diabetic mice.

Importantly, we identified a novel regulator of Ly6C⁺ Mo/MΦ proliferation in skin wounds. First, we confirmed our previous suggestion that the wound environment regulates Mo/MΦ proliferation (6) using adoptive transfer of Ly6C^hi Mo from nondiabetic mice; nondiabetic Mo proliferated at a higher rate when transferred into wounds of diabetic mice than when transferred into wounds of nondiabetic mice. As our previous data did not support a role for IL-1β, IL-6, IL-4, nor M-CSF in regulating proliferation of Mo/MΦ in skin wounds (6), we expanded our search to other cytokines/chemokines that play important roles in regulating Mo/MΦ during wound healing including TNF-α, IL-10, CCL2, CCL3, and IL-33.

Among the cytokines we screened, CCL2 is a well-characterized chemokine that attracts Mo/MΦ to inflammatory sites, including skin wounds (15, 46). In our study, CCL2 protein levels were elevated in skin wounds following injury, and diabetic mice had significantly higher CCL2 protein levels in wounds compared with nondiabetic controls. In addition, administration of a single dose of exogenous CCL2 directly to wounds enhanced proliferation of Ly6C⁺F4/80^lo/− Mo/MΦ compared with controls treated with PBS. Interestingly, CCL2 has previously been reported to promote proliferation of adipose tissue F4/80⁺ MΦ and several other types of cells in mice (9, 47–50). In addition to chemotaxis and proliferation of Mo/MΦ, CCL2 can affect the function of many other major cell types in wounds, including fibroblasts and endothelial cells (51–54). Therefore, to avoid off-target effects of CCL2/CCR2 signaling in our studies, we used adoptive transfer of Ly6C^hi Mo from CCR2^−/− EGFP mice to skin wounds to determine whether CCL2/CCR2 signaling promotes Mo/MΦ proliferation. Importantly, transferred Ly6C^hi Mo lacking CCR2 showed lower proliferation compared with CCR2 competent cells. Together, these data show that CCL2 is a key regulator of Mo/MΦ proliferation during skin wound healing and that persistently high levels of CCL2 and the resulting sustained proliferation of inflammatory Mo/MΦ contribute to chronic inflammation and impaired wound healing in diabetic mice.

Finally, evidence in the literature suggests that wound Mo/MΦ are derived primarily from Ly6C^hi Mo produced in the bone marrow, which are mobilized into the blood and then infiltrate into damaged tissue (42). Diabetes leads to the alteration of several critical steps in this process, including myeloid lineage progenitor bias in bone marrow, which, in turn, promotes excess accumulation of Mo/MΦ and prolonged inflammation in wounds (5, 42). In the current study, in agreement with previous studies, we detected significantly higher levels of Ly6C^hi Mo in bone marrow and the peripheral blood of diabetic mice following skin injury (5, 42). Moreover, similar to what we reported previously for nondiabetic mice, Ly6C^hi Mo proliferated in bone marrow but not in peripheral blood of diabetic mice (6). However, proliferation of bone marrow Ly6C^hi Mo did not differ between diabetic and nondiabetic mice. Combined with our previous studies, our data suggest that diabetes affects myelopoiesis but not Mo proliferation in bone marrow.

One limitation of this study is that we used a single time point for the scRNAseq analysis. Therefore, we are planning future studies to generate a time course of scRNAseq data for Mo/MΦ to produce a more comprehensive picture of Mo/MΦ characteristics over the course of skin wound healing. Once we have an unbiased atlas of Mo/MΦ phenotypes over the course of healing, we can start to better understand the regulation of cell cycle in these cells as well as other cellular functions. Another limitation is that other molecules in addition to CCL2 could promote Mo/MΦ proliferation in wounds. Although we did not detect significant differences of either IL-4 nor M-CSF levels in skin wound homogenates between nondiabetic and diabetic mice and these cytokines did not increase after injury, previous studies have implicated these cytokines in MΦ proliferation (7), and lack of differences in measured protein levels does not necessarily mean lack of a role in MΦ proliferation. Finally, our study did not distinguish between resident versus infiltrated Mo/MΦ in the skin wound, although current understanding states that Mo/MΦ accumulate in skin wounds mainly by the recruitment of inflammatory Ly6C⁺ Mo from the circulation (55).

In conclusion, we have demonstrated that Ly6C⁺ Mo/MΦ proliferate at a higher rate in wounds of diabetic compared with their nondiabetic counterparts. Moreover, we demonstrate that CCL2, known primarily for its chemoattractant function, promotes Ly6C⁺ MΦ/Mo proliferation, and sustained high levels of CCL2 in diabetic wounds likely drives proliferation of these inflammatory cells, contributing to chronic inflammation and impaired healing. Therefore, future studies should focus on the potential for inhibiting the enhanced Ly6C⁺ Mo/MΦ proliferation in diabetic wounds as a potential therapeutic approach to dampen chronic inflammation and improve healing of diabetic wounds.

We thank Dr. Luisa A. DiPietro and Dr. Giamila Fantuzzi for input on aspects of presentation of this manuscript.

The authors have no financial conflicts of interest.