Reduced apoptosis of monocytes and macrophages is associated with their persistence in wounds of diabetic mice

Abstract

Monocytes and macrophages (Mo/MΦ) rapidly accumulate in skin wounds after injury, then disappear as healing progresses. However, the mechanisms underlying their ultimate fate in wounds remain to be elucidated. Here, we show that apoptosis of Mo/MΦ parallels their reduction as wound healing progresses in non-diabetic mice. scRNAseq analysis confirmed enriched apoptosis GO pathways on day 6 post-injury in wound Mo/MΦ from non-diabetic mice. In contrast, there was significantly less Mo/MΦ apoptosis in wounds from diabetic mice, particularly in the pro-inflammatory Ly6C+ population, which may contribute to persistent Mo/MΦ accumulation and chronic inflammation. scRNAseq analysis implicated TNF, MAPK, Jak-STAT, and FoxO signaling pathways in promoting wound Mo/MΦ apoptosis in non-diabetic mice while cell proliferation related pathways appeared to be activated in diabetic mice. These novel findings indicate that reduced apoptosis is a contributor to persistent Mo/MΦ accumulation in diabetic wounds. These findings also highlight pathways that may regulate Mo/MΦ apoptosis during wound healing, which could be targeted to help resolve inflammation and improve healing.

Introduction

Monocytes and macrophages (Mo/MΦ) are essential contributors to all stages of wound healing, through enhancing the initial inflammatory response, preventing infection, removing dead cells, and promoting wound angiogenesis, granulation and closure [1]. Blood Ly6C+ Mo infiltrate into the wound in response to signals from damaged tissue, and subsequently proliferate, forming the primary pool of wound Mo/MΦ [1, 2]. During the resolution stage of healing, Mo/MΦ decrease in numbers, perhaps due to Mo/MΦ death, as indicated by appearance of macrophage arginase I in wound extracellular fluid [1, 3]. Another possible fate of wound Mo/MΦ is suggested by studies of peritoneal and lung MΦ, which appear to emigrate to the draining lymph nodes after resolution of inflammation [3]. However, much remains to be learned about the fate of Mo/MΦ during the resolution of inflammation during skin wound healing.

Apoptosis is critical for removal of different cell types from skin wounds without compromising the healing process [4]. Many factors and pathways, including TNF, Fas, PI3K-Akt, ER stress and hypoxia, influence Mo/MΦ apoptosis in pathophysiological conditions such as atherosclerosis and cancer [5]. However, the role of Mo/MΦ apoptosis and associated regulatory pathways have not been elucidated during skin wound healing. In addition, diabetic wounds are associated with persistent accumulation of inflammatory Mo/MΦ, which contributes to chronic inflammation and impaired healing [2, 6–8]. Whether apoptosis pathways are altered in diabetic wound Mo/MΦ also remains to be determined. Therefore, the purpose of this brief report was to ascertain whether apoptosis influences Mo/MΦ accumulation in wounds of non-diabetic and diabetic mice.

Materials and Methods

Animals and wound model

Male C57Bl/6 (non-diabetic) and BKS.Cg-Dock7^m +/+ Lepr^db/J (db/db, diabetic) mice (age 10–12 weeks, N = 4 – 6 per group/time point, The Jackson Laboratory, Bar Harbor, ME, USA) were housed in the animal facility of the University of Illinois at Chicago with free access to water and chow diet. Two full-thickness 8 mm diameter excisional wounds were created on the dorsal skin of mice as described [2]. Tegaderm was used to cover wounds until sacrifice.

Wounds were collected on day 3 and 6 post-injury using a 10 mm biopsy punch. Wounds were immediately dissociated into single cell suspensions and prepared for flow cytometry analysis as described [2]. All animal studies were approved by the Animal care and Use Committee of the University of Illinois at Chicago.

Flow cytometry

Cells isolated from skin wounds on day 3 and 6 post-injury were pre-incubated with anti-CD16/32 blocking antibody (Biolegend, clone S17011E). Surface antigens were labeled with anti-Ly6G-BV605 (clone 1A8), CD11b-APC/Fire750 (clone CBRM1/5), F4/80-PE (clone BM8), and Ly6C-Percp/cy5.5 (clone AL-21) antibodies (Biolegend, San Diego, CA, USA or BD Biosciences, San Jose, CA, USA). Next, cells were stained with FITC Annexin V Apoptosis Detection Kit with 7-AAD following manufacturer’s instructions (Biolegend). All samples were analyzed on either LSR Fortessa with HTS (BD Biosciences). Data were analyzed using FlowJo (FlowJo LLC).

For single cell RNA sequencing, single cells were pooled from skin wounds of two non-diabetic or two diabetic mice on day 6 post-injury as described above. Cells were first stained with Zombie Violet (Biolegend) to assess cell viability followed incubation with anti-CD16/32 blocking antibody. Next, cells were labeled with anti-CD45-FITC (Biolegend, clone 30-F11), Ly6G-PE/Dazzle^™ 594 (clone 1A8), and CD11b-APC/Fire750 (clone CBRM1/5). Target cells were sorted out as Zombie-CD45+CD11b+Ly6G- cells on BD FACSAria^™ III sorter (BD Biosciences).

Single cell RNA sequencing and analysis

Fresh sorted cells (viability > 85%) were captured and processed using the Chromium system (10X Genomics, San Francisco, CA, USA). Samples were sequenced on HiSeq^™ Sequencing Systems (Illumina, San Diego, CA, USA) with paired-end reads. Fastq files were generated and demultiplexed into single cells using Cell Ranger software (10X Genomics).

Data were filtered and pre-analyzed as described before [2]. Cells were clustered using k-means clustering and presented with 4 clusters by t-SNE plot. Differentially expressed genes per cluster were detected using the area under the receiver operating characteristic curve (AUROC), treating each gene as a classifier for each cluster. The Gene Ontology (GO) enrichment analysis of the top 200 expressed genes was conducted using The Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.8 (FDR < 5%).

Statistics

Data are expressed as mean ± SEM. Two-way ANOVA was used to test effects of time point and strain on Mo/MΦ numbers and percentages in different stages of apoptosis. Sidak’s multiple comparisons post-hoc test was conducted to detect specific pairwise differences between time points within strain or between strains within a given time point. A value of P<0.05 was considered statistically significant.

Data Availability

Single-cell RNA sequencing data in this study are available in Gene Expression Omnibus with the accession code GSE154400. All other relevant data are available from the corresponding authors.

Results and Discussion

Wound Mo/MΦ apoptosis is reduced in diabetic versus non-diabetic mice

Following tissue damage, wound Mo/MΦ accumulate primarily via infiltration of blood Ly6C+Mo and their subsequent proliferation [1–3, 9]. Numbers of these cells then decline during the resolution phase of normal healing, whereas they are maintained at high numbers in diabetic wounds [1, 2, 6–9]. Cells undergoing apoptosis can be detected by the binding of Annexin V to the cell surface; Annexin V binds phosphatidylserine, which is normally present only on the inner membrane [4, 5]. Therefore, we evaluated Mo/MΦ apoptosis/cell death during wound healing in non-diabetic and diabetic mice via co-staining of Annexin V and DNA dye 7-AAD by flow cytometry.

On day 3 post-injury, wounds of non-diabetic and diabetic mice contained comparable numbers of Ly6G-CD11b+ cells per mg of tissue; these cells were taken to represent the total pool of wound Mo/MΦ (Fig 1B). The number of Ly6G-CD11b+ Mo/MΦ tended to decrease in non-diabetic mice through day 10 post-injury but these cells were maintained at significantly higher numbers in diabetic compared to non-diabetic wounds. Next, we evaluated cells in the following categories: non-apoptotic (Annexin V-7-AAD−), early apoptotic (Annexin V+7-AAD−), and dead/late apoptotic (Annexin V+7-AAD+) (Fig 1A). In non-diabetic wounds, the percentage of non-apoptotic Ly6G-CD11b+ cells tended to decrease from day 3 through day 10 post-injury (Fig 1C), whereas the percentage of cells in early apoptosis tended to increase from day 3 to day 6 and then decrease on day 10 post-injury (Fig 1D). Finally, the percentage of dead/late apoptotic cells increased from day 3 to day 10 post-injury, with a significantly higher percentage on day 10 than on day 3 (Figure 1E). These data indicated that Ly6G-CD11b Mo/MΦ move through stages of apoptosis as healing progresses in non-diabetic wounds. In contrast, apoptosis did not show any significant changes in diabetic wounds over these time points (Fig 1C–1E) suggesting that apoptosis may be impaired in these wounds.

Figure 1. Wound Mo/MΦ undergo apoptosis during resolution phase of normal skin wound healing but apoptosis is reduced in diabetic wounds.

(A) Gating strategy for identifying apoptotic Mo/MΦ in skin wounds. (B) Numbers of Ly6G-CD11b+ Mo/MΦ in wounds on days 3 (n=6 per strain), 6 (n=6 per strain), and 10 (n=4 per strain), post-injury in non-diabetic (black bar) and diabetic (open bar) mice. (C-E) Percentages of cells in non-apoptotic (Annexin V-7-AAD−, C), early apoptotic (Annexin V+7-AAD−, D), and dead/late apoptotic stages (Annexin V+7-AAD+, E) for Ly6G-CD11b+ Mo/MΦ. (F) Numbers of pro-inflammatory Ly6G-CD11b+Ly6C+ Mo/MΦ in wounds on days 3 (n=6), 6 (n=6), and 10 (n=4). (G-I) Percentages of cells in non-apoptotic (Annexin V-7-AAD−, G), early apoptotic (Annexin V+7-AAD−, H), and dead/late apoptotic stages (Annexin V+7-AAD+, I) for Ly6G-CD11b+Ly6C+ Mo/MΦ. Data are mean ± SEM; Two-way ANOVA performed to evaluate effects of time point and non-diabetic/diabetic strains: ^a significant main effect of time point; ^b significant main effect of strain; and ^c significant interaction effect. Post-hoc testing indicated: *significant difference between time points within given strain. ^# significant difference between strains for given time point. P < 0.05 taken to indicate statistical significance.

Ly6C+ Mo/MΦ are typically considered pro-inflammatory and contribute to persistent inflammation in diabetic wounds [1, 2, 6, 8, 9]. Similar to Ly6G-CD11b+ cells, numbers of Ly6C+ Mo/MΦ per mg of tissue were comparable between non-diabetic and diabetic wounds on day 3 post-injury (Fig 1F). In non-diabetic wounds, the number of Ly6C+ cells decreased significantly from day 3 to day 10 post-injury, whereas Ly6C+ cells were maintained at significantly higher numbers in diabetic versus non-diabetic wounds through day 10 post-injury (Fig 1F). The percentages of non-apoptotic Ly6C+ Mo/MΦ were significantly different over time, tending to decrease particularly in non-diabetic wounds (Fig 1G). The percentages of cells in the early apoptotic stage were significantly different between strains, with diabetic mice tending to have higher levels (Fig 1H). Similar to what we observed in the total Mo/MΦ population, Ly6C+ Mo/MΦ from non-diabetic wounds exhibited significantly higher rates of apoptosis/cell death over time with the most striking differences on day 10 post-injury (Fig 1I).

Taking these data together, both the total Mo/MΦ population and the Ly6C+ subpopulation undergo increased apoptosis as healing progresses in non-diabetic wounds while apoptosis is maintained at lower levels in diabetic wounds, likely contributing to sustained pro-inflammatory Mo/MΦ accumulation in the latter wounds.

Single cell RNA sequencing analysis reveals enhanced expression of apoptosis-associated genes in Mo/MΦ from non-diabetic vs diabetic wounds

To begin to elucidate pathways that regulate Mo/MΦ apoptosis during wound healing, we performed single cell RNA sequencing (scRNAseq). For this analysis, we used live Ly6G-CD45+CD11b+ wound Mo/MΦ collected on day 6 post-injury. As shown in Fig 2A, four clusters were identified by k means clustering analysis in pooled Mo/MΦ from both strains. Cluster 1 was composed primarily of cells from non-diabetic mice (log2 enrichment score: non-diabetic: 0.31 vs diabetic: −0.24) while cluster 4 was overrepresented by cells from diabetic mice (log2 enrichment score: non-diabetic: −1.76 vs diabetic: 0.54). Next, we performed GO enrichment analysis of the top 200 expressed genes by AUC value. Interestingly, apoptosis/cell death related GO terms were highly enriched in cluster 1, which was predominantly composed of Mo/MΦ from non-diabetic mice (Fig 2B). This analysis is consistent with the flow cytometry results, that there was greater apoptosis/cell death in wound Mo/MΦ from non-diabetic mice compared to those from diabetic mice.

Figure 2. Single cell RNA sequencing analysis indicates enriched apoptosis pathways in a cluster of Mo/MΦ populated primarily by cells from non-diabetic wounds.

(A) tSNE plot of single cell RNA sequencing data using Mo/MΦ isolated from day 6 skin wounds from non-diabetic and diabetic mice reveals four distinct clusters with cluster 1 enriched with cells from non-diabetic wounds and cluster 4 enriched with cells from diabetic wounds. (B) GO terms related to apoptosis showed the highest enrichment in cluster 1. (C) Expression of genes associated with positive regulation of apoptotic process (GO:0043065) for each cluster was determined by AUC using ROC curve. Representative genes enriched in cluster 1 (top) and genes enriched in both cluster 1 and 4 (bottom) are highlighted. (D) Expression of genes associated with negative regulation of apoptotic process (GO:0043066) for each cluster was determined by AUC using ROC curve. Representative genes enriched in cluster 1 but downregulated in cluster 4 (top) and genes enriched in cluster 4 but downregulated in cluster 1 (bottom) are highlighted.

Next, we evaluated expression of genes associated with positive (GO:0043065) and negative (GO:0043066) regulation of apoptosis (Fig 2C and 2D). Among those genes associated with positive regulation of apoptosis, GO enrichment analysis indicated a group of genes involved in TNF, MAPK, Jak-STAT, and FoxO signaling pathways was highly enriched in cluster 1 (non-diabetic Mo/MΦ enriched), suggesting the activation of well-defined apoptosis-promoting signaling pathways in Mo/MΦ isolated from normally healing wounds in non-diabetic mice [5, 10–12]. Genes moderately enriched in both cluster 1 and 4 were mainly associated with Notch and HIF-1 signaling pathways (Fig 2C) [13, 14]. As for negative regulators of apoptosis, a group of genes linked to PI3K-Akt, NF-κB, and p53 signaling pathways were highly upregulated in cluster 1, indicating the activation of a complementary feedback loop in response to enhanced apoptosis of Mo/MΦ in wounds during the resolution phase of healing [15–17]. Additionally, among genes linked to negative regulation of apoptosis, a group of genes was downregulated in cluster 1 but upregulated in cluster 4 (Fig 2D). Interestingly, those genes were also related to cell proliferation by GO enrichment analysis [18]. This finding is in agreement with our previous report of advanced proliferation of wound Mo/MΦ in diabetic mice [9].

Thus, our scRNAseq data demonstrate that genes associated with apoptosis/cell death are preferentially expressed in a cluster of cells populated predominantly by non-diabetic wound Mo/MΦ and that TNF, MAPK, Jak-STAT, and FoxO signaling pathways may be involved in triggering the death of these cells.

Taken together, our data demonstrated persistent accumulation of Mo/MΦ, particularly Ly6C+ Mo/MΦ, in diabetic wounds on day 6 and/or 10 post-injury, which is in agreement with previous findings from our laboratory and others [2, 6–8]. In contrast, other studies have reported deficiency of CD68+ or F4/80+ MΦ accumulation early after wounding in diabetic mice [19, 20]. This disparity could be due to differences in markers used to identify macrophages, smaller wounds used in the previous studies, as well as our focus on later healing stages, during which persistent accumulation of MΦ accumulation has been reported in both our and the other studies [19, 20]. In addition, our flow cytometry and scRNAseq data indicated that reduced Mo/MΦ apoptosis in diabetic wounds is associated with their persistent accumulation. In contrast, another study reported that bee venom improved diabetic wound healing by protecting functional MΦ from apoptosis [21]. The difference in results could be due to the differences in macrophage populations assessed (isolated from tissue in our study, isolated from implanted polyvinyl alcohol sponges in previous study), methods of assessing apoptosis (Annexin-V and 7-AAD labeling in our study, JC-1 dye in previous study) and/or diabetes models used; db/db mice (model of type 2 diabetes) in the present study and streptozotocin injection (model of type 1 diabetes) in the previous study.

In conclusion, we report novel findings that apoptosis/cell death likely contributes to the loss of Mo/MΦ, including the pro-inflammatory Ly6C+ subset, as wound healing progresses in non-diabetic mice. In contrast, our data indicated reduced Mo/MΦ apoptosis/cell death in wounds of diabetic mice, which together with increased Mo/MΦ proliferation [2], may contribute to the persistent accumulation of pro-inflammatory Mo/MΦ and chronic inflammation in diabetic wounds. Moreover, scRNAseq analysis indicated that wound Mo/MΦ apoptosis was regulated by genes involved in the TNF, MAPK, Jak-STAT, and FoxO signaling pathways particularly in non-diabetic mice while cell proliferation-related pathways were activated in diabetic mice. Further studies should focus on mechanistic understanding of TNF, MAPK, Jak-STAT, and FoxO signaling pathways in promoting apoptosis/cell death during normal wound healing and the mechanisms underlying reduced apoptosis in diabetic wounds.

• Mo/MΦ apoptosis parallels their reduction in wounds of non-diabetic mice

• Reduced Mo/MΦ apoptosis linked to their persistence in wounds of diabetic mice

• TNF, MAPK, Jak-STAT, and FoxO pathways associated with wound Mo/MΦ apoptosis