Diabetes-Induced TREM2–Endothelial Cell Signaling Impairs Ischemic Vascular Repair

Abstract

Diabetes mellitus (DM) accelerates vascular diseases including peripheral arterial disease (PAD). Endothelial cells (ECs) and macrophages (MΦs) are important contributors to DM-associated vascular dysfunction, both individually and through reciprocal crosstalk. Although single-cell profiling has revealed the heterogeneity of ECs and MΦs, how this diversity translates into cell-cell interactions, and consequentially vascular function, remains unclear. We leveraged single cell RNA sequencing and spatial transcriptomics to profile human mesenteric arteries from non-diabetic donors and from donors with type 2 diabetes (T2D), generating a transcriptome and interactome atlas of diabetic vasculature. This analysis identified Triggering Receptor Expressed on Myeloid Cells 2 (TREM2) as one of the top T2D-induced genes in mononuclear phagocytes (MPs), with concomitant increases in TREM2 ligands in ECs. TREM2⁺ MPs exhibited foam cell–like features but acquired a pro-inflammatory gene profile in DM. Functionally, TREM2 inhibition in vitro attenuated pro-inflammatory responses in MPs and ECs, and enhanced EC migration. In streptozotocin- and high fat high sucrose diet-induced mouse models of diabetes with hindlimb ischemia (a model of PAD), TREM2 blockade using a neutralizing antibody improved perfusion recovery, whereas TREM2 activation with an agonist exacerbated ischemic injury. Analysis of clinical samples confirmed elevated EC-TREM2 signaling in human PAD, particularly in the setting of DM, highlighting its translational relevance. Collectively, our study presents an atlas of human diabetic vessels with single-cell and spatial resolution, identifying TREM2-EC interaction as a driver of diabetic vasculopathy and a potential therapeutic target in DM-associated PAD.

One-sentence summary:

Single-cell and spatial mapping reveals diabetes-driven TREM2-endothelial crosstalk that impairs vascular repair in peripheral arterial disease.

Editor’s summary

Diabetes can lead to peripheral artery disease (PAD), and how individual cell types across tissues contribute to disease progression has been difficult to understand. Malhi et al. focused on endothelial cells and macrophages from humans with and without type 2 diabetes to find that triggering receptor expressed on myeloid cells 2 (TREM2) was one of the top genes expressed in mononuclear phagocytes. This expression was associated with increased TREM2 ligands in endothelial cells. Mononuclear phagocytes with high TREM2 expression had proinflammatory gene expression profiles. In mouse models of diabetes with hindlimb ischemia, blocking TREM2 with an antibody improved perfusion recovery. In addition, clinical samples from patients with PAD had increased endothelial cell–TREM2 signaling. These data suggest that TREM2–endothelial cell interactions may drive diabetic vasculopathy. An accompanying Focus article by Chang et al. expands upon these findings and describes their relevance for translating the research advance. —Brandon Berry

Introduction

Diabetes mellitus (DM) is a global epidemic negatively affecting over half a billion adults and poses a major healthcare burden worldwide (1). DM predisposes and accelerates vasculopathies including coronary and peripheral arterial disease (PAD). The latter, which restricts perfusion to the limbs, is nearly four times more prevalent in individuals with DM. Yet effective therapeutic interventions remain suboptimal. Patients with diabetes-associated PAD (DM-PAD) experience particularly poor outcomes and are more likely to progress to chronic limb-threatening ischemia (CLTI) (2), reflecting the cumulative effects of inflammation and tissue injury (3). However, the mechanisms underlying DM-induced vascular dysfunction remain incompletely understood, which hampers the development of effective therapeutics for the management and treatment of diabetic vasculopathy.

Diabetic vasculopathy involves diverse vascular and immune cell types engaged in complex cellular communications and interactions. One of the earliest pathogenic events in DM is dysfunction of endothelial cells (ECs) lining the vascular lumen. Whereas quiescent ECs express low levels of adhesion molecules (e.g., vascular cell adhesion molecule 1 [VCAM1] and intercellular cell adhesion molecule 1 [ICAM1]), the inflammatory diabetic milieu cause ECs to become more adhesive, enabling the recruitment of monocytes, which adhere and transmigrate into the sub-endothelial layer to differentiate into macrophages (MФs) (4). These MФs, in turn, propagate the inflammatory response and vascular dysfunction (5).

Although EC-monocytes/MФs (also known as mononuclear phagocytes, MPs) are well recognized in diabetic vasculopathy, their communication are often viewed as unilateral. Emerging evidence reveals bi-directional communications between ECs and MPs: monocytes patrol the vascular endothelium and remove damaged cells (6, 7), and intimal resident MΦs maintain a non-thrombogenic endothelial state (8). Under hypoxia or ischemia, MΦs modulate angiogenesis and collateral arterial remodeling, possibly through vascular endothelial growth factors (VEGFs) and monocyte chemoattractant protein-1 (MCP-1) signaling (9, 10). In cancer, activated MΦs can secrete inflammatory mediators to disrupt EC tight junctions and increase EC permeability (11). Single-cell technologies have further uncovered remarkable heterogeneity among MΦs (12, 13), revealing subtypes featuring distinct gene signatures linked to tissue-damaging or -reparative functions (8, 14). Despite these insights, the specific ligands and receptors mediating EC-MP interactions and their diverse functional outcomes in diabetes remain poorly defined, representing a major gap in understanding diabetic vascular inflammation and dysfunction.

In this study, we sought to identify EC-MФ interactions that contribute to diabetic vasculopathies, with a focus on PAD. We hypothesized that diabetes reprograms EC–MΦ interactions within the arterial intima in a manner that promotes vascular inflammation and impairs reparative responses following ischemic injury. We used single cell RNA sequencing (scRNA-seq) and spatial transcriptomics of human mesenteric arteries from non-diabetic (ND) donors and donors with type 2 DM (T2D) to systematically map the EC-MP interactions within the arterial wall. Unbiased analyses identified Triggering Receptor Expressed on Myeloid Cells 2 (TREM2), encoding a phagocytic cell-surface receptor implicated in metabolic disease, atherosclerosis, cancer, and neurodegeneration, as one of the most upregulated genes in MPs in the diabetic intima, with concomitant increases in its putative endothelial ligands. We further hypothesized that TREM2 serves as a key mediator of dysfunctional EC-MP cross-talk in diabetic vasculature. We therefore examined whether diabetes induces TREM2-dependent transcriptional and functional changes in macrophages, whether TREM2 signaling influences endothelial inflammatory and reparative programs, and whether soluble TREM2 (sTREM2) contributes to EC dysfunction. Using complementary in vitro systems and diabetic mouse models of hindlimb ischemia, we investigated the functional consequences of modulating TREM2 signaling on vascular inflammation and ischemic recovery. Finally, to identify translational relevance, we validated whether TREM2-associated EC–MΦ signaling pathways identified in experimental models are similarly engaged in human PAD, particularly in the context of diabetes.

Results

Profiling the diabetic transcriptome in human mesenteric intima.

We began by profiling the DM-associated transcriptomic changes in vascular cells, through scRNA-seq of human mesenteric arteries from 10 deceased individuals with or without DM (5 non-diabetic and 5 T2D) from a tissue biobank previously described (15–20) (table S1, fig. S1). The reasons for using mesenteric arteries include tissue availability, a well-defined and large intima-media structure that facilitates EC profiling, and their established utility in studying DM-related vascular dysfunction, including alterations in gene expression, vascular resistance, and vasodilatory responses (21–23). In addition, mesenteric ischemic disease is frequently observed in patients with PAD (24), underscoring the clinical relevance of this vascular bed. The five non-diabetic individuals included one Hispanic and four Caucasian individuals, all men. The five individuals with T2D were four Hispanic individuals (one woman, three men) and one Caucasian man. Four individuals with T2D were obese (BMI above or equal to 30). Of note, the Hispanic individuals were untreated for their T2D, and the one Caucasian individual was noncompliant with treatment.

ScRNA-seq of whole blood vessels indicates that ECs represent 5–15% of the entire cell population (25–28). To enrich ECs and capture EC-interacting cells, especially MΦs, we isolated the inner layer of the arterial wall (including intima), using an intima scRNA-seq protocol that we described (16) (Fig. 1A). From the 10 arterial segments, ~14,000 cells were obtained, all of which passed quality control for scRNA-seq. Principal component analysis (PCA) of the batch-corrected intima scRNA-seq data revealed slight separation between non-diabetic and T2D samples, implying generally similar cell populations obtained from non-diabetic and T2D donors (Fig. 1B). Based on unsupervised clustering, these cells were separated into 21 clusters (fig. S2A). Monocytes (in potentially differentiating states) and MΦs were assigned collectively as mononuclear phagocytes (MPs). Annotation of all intima scRNA-seq cells using reported cell type marker genes (28) identified 6 major cell types: ECs (~34%, 2 clusters), MPs (~27%, 5 clusters), T/natural killer (NK) cells (~30%, 6 clusters), vascular smooth muscle cells (VSMC)/fibroblasts 2%, 2 clusters), and others (e.g., B cells) (Fig. 1, C and D). This cell composition is expected considering that the intima was dissociated and EC enriched.

Figure 1. Profiling of human mesenteric artery intima.

(A) Schematic of human mesenteric artery processing for scRNA-seq, created using BioRender.com. ND: Non-diabetes; T2D: Type 2 diabetes. (B) Principal component analysis (PCA) plot colored by diabetic state for the scRNA-seq analysis. (C) Uniform manifold approximation and projection (UMAP) showing major cell types and subclusters generated through Seurat. EC: endothelial cell; VSMC: vascular smooth muscle cell; MP: mononuclear phagocyte; T/NK: T/natural killer cell. (D) Dotplot of cell-marker genes (x-axis) across annotated cell population (y-axis). Circle color shows expression level; circle size shows % of cells expressing the gene. (E) Cell composition (%) of ND vs T2D donors (n=5, mean±SEM). (F) Representative CD31 (purple)/CD68 (yellow) co-staining. Red arrows indicate CD68-positive staining. Scale bar = 50 μm. (G) CD68 quantification in mesenteric artery intima (n=3 ND/T2D, mean±SEM). **p<0.01 vs ND (Student’s t-test).

Comparing the cell composition between non-diabetic and T2D, we observed no major difference in expression of cell marker genes in most clusters (fig. S2A). When we compared the distribution of the major cell types, although the percentages of ECs and T cells were not substantially different (39 to 32% for ECs, p=0.33 and 31 to 30% for T cells, p=0.44), that of MPs increased from 15% in non-diabetic samples to 31% in T2D samples (p=0.037) (Fig. 1E). We validated this at the protein level, observing more CD68 (an MP marker) positive staining in the T2D intima as compared to non-diabetic intima (Fig. 1F and G, fig. S2B). This is in line with the notion that DM induces inflammatory infiltration and MΦ accumulation in the sub-endothelial layer (4, 5).

Endothelial and MP diversity in the diabetic intima

We next investigated the DM-associated transcriptomic changes within ECs and MPs. We first focused on ECs. Two EC clusters were identified from the intima scRNA-seq, EC1 (79% of ECs) and EC2 (21% of ECs) (table S2). The proportion of EC clusters did not differ significantly between non-diabetic vs T2D samples. Whereas EC1 expresses higher levels of genes encoding extracellular matrix components (e.g., COL8A1, ELN), EC2 expresses higher levels of genes related to lipid metabolism (e.g., FABP4, FABP5) (fig. S3), similar to EC clusters previously described (25). We then analyzed differentially expressed genes (DEGs) by comparing T2D-EC vs ND-EC data using RePACT, a regression-based algorithm with improved sensitivity in identifying disease relevant genes even with a small number of samples (29). RePACT analysis revealed a difference in the transcriptome states between T2D-ECs vs ND-ECs (Fig. 2, A to C), with DEGs enriched for wound healing/angiogenesis, oxidative stress, and inflammatory responses (fig. S4, A and B). For example, genes promoting inflammatory response and endothelial-mesenchymal transition (EndoMT, e.g., CCN2, THBS1, FN1) were induced, whereas genes crucial for EC homeostasis and angiogenesis (e.g., NOS3, NRP1, PECAM1) were suppressed in T2D-ECs (Fig. 2C, fig. S4A, table S3). These changes are in line with the reported effect of DM on ECs (30), and validates the use of these human data to investigate DM-induced vascular changes.

Figure 2. DM-induced transcriptome shifts in EC and MP.

(A-C) ECs profiled by intima scRNA-seq. (A) PCA plot of non-diabetic (blue) or T2D (red) ECs plotted in the 3D PCA space (n=5). (B) Violin plots showing comparison of pseudo-disease T2D state of cells from each donor by RePACT analysis (n=5). Violins colored by diabetic status of the donor. Box plots within violins show median, upper and lower quartile of pseudo-disease state of cells. (C) Heatmap showing relative expression (Exp) of top 50 DEGs in T2D vs non-diabetic ECs, each row in the heatmaps represent the transcriptional changes from low-index pseudo-states to high-index pseudo-states. (D-G) MPs profiled by intima scRNA-seq (n=5). (D) Dotplot of Z-score-scaled expression of MP subtype markers in non-diabetic and T2D MPs, circle color indicating expression level and circle size denoting % of cells expressing the gene. (E) MP pseudotime trajectory (Monocle 3) superimposed on UMAP. (F) Percentage of non-diabetic vs T2D MPs in MP clusters (MP-C1 to C5). (G) Percentage of MP-C1 to -C5 in total MP population from non-diabetic and T2D samples.

We analyzed in parallel the MP transcriptome and identified 5 MP subclusters (MP-C1 to C5). MP-C5 was enriched for classical monocyte markers, including S100A8 and S100A9. MP-C3 displayed a mixed phenotypes encompassing features of foam cells, M1- and M2-like MΦs, resident MΦs, and a TREM2^Hi subset characterized by high expression of TREM2, SPP1 and CD9 (31). The remaining subclusters exhibited less distinct patterns (fig. S5A). We then evaluated diabetes-associated transcriptional changes across these MP populations. Globally, we identified 935 DM-induced DEGs, including 473 up- and 462 down-regulated (fig. S5, B and C). The top enriched pathways are involved in immune and inflammatory responses, as well as angiogenesis (fig. S5D). The common MP marker genes CD14 (for monocytes) and CD68 (for MΦs) showed a slight but insignificant increase in T2D (Fig. 2D). There was an apparent decrease in classical monocyte markers S100A8, and VCAN in T2D, suggesting more monocyte differentiation into MΦ. Whereas the M1-like MΦ markers IL1B and TNF, were increased, the M2-like markers ARG1 and RETN were decreased, indicative of a more pro-inflammatory state of MΦs under DM conditions. The marker genes of foam cells (e.g. FABP5 and APOE) and TREM2^Hi MΦs were also increased in T2D MPs (Fig. 2D).

Next, trajectory analysis using Monocle 3 was carried out to infer DM-associated cell state changes (32), which revealed a clear trajectory within the MP population (Fig. 2E). Specifically, MP-C3 showed the furthest progression through pseudotime, suggesting that this cluster represents an advanced T2D state. Although MP-C5 is more dominant in non-diabetic samples, MP-C1 to MP-C4 are more present in T2D (Fig. 2F). Among these, MP-C3 demonstrated the largest percentage increase in MPs in T2D vs non-diabetic (Fig. 2G), suggesting an important role of this cluster in DM-induced vascular inflammation and dysfunction. Pathway enrichment analysis of DM-associated changes in the MP clusters revealed that MP-C3 was enriched in genes involved in angiogenesis, hypoxia response, and nitric oxide (NO) biosynthetic process, all EC-relevant pathways, and pathways regulating inflammatory response and cytokine production (figs. S6 and 7).

We then determined the key DEGs within the MP-C3 contributing to the T2D state. To this end, we identified 65 genes (46 increased and 19 decreased) that were consistently differentially expressed between non-diabetic vs T2D samples in all the MP population as well as in MP-C3 (fig. S8A, table S4). Among these genes, the top ranked is ADAM28, encoding a metalloprotease with functions in cell adhesion, migration, proteolysis, and signaling (33). The second ranked gene is TREM2, encoding a promiscuous receptor primarily expressed on the surface of myeloid cells where it regulates immune responses (34). Other notable genes include Histamine Receptor H2 (HRH2) and plasminogen activator, urokinase (PLAU), to which antagonists have been developed for the treatment of DM (35) (fig. S8B).

Additionally, we identified extensive changes in the T, NK, and B cell transcriptome between non-diabetic and T2D samples (with 425, 325, 862 DEGs respectively) (fig. S9) and presented their pseudotime trajectories (fig. S10). Given the focus of the current study on ECs and MΦs, we did not perform further analyses on these cells.

Increased EC-MP interactions in T2D mesenteric artery

To identify EC-MP interactions with a spatial context, spatial transcriptome-seq using 10X Genomics Visium platform was performed. The platform captures RNA on a tissue slide (6.5 mm × 6.5 mm) consisting of ~5,000 spots, termed voxels (50 μm in diameter) with embedded spatial barcodes. We performed spatial transcriptome-seq on 4 mesenteric arteries (2 non-diabetic vs 2 T2D) (table S1). The vessel segments were from different individuals than those used for intima scRNA-seq. Whereas the non-diabetic arteries showed no obvious pathologies, the DM arteries showed intima-media thickening and increased MΦ infiltration (marked by CD68) (fig. S11). Through the Visium workflow, an average of 2,000 unique molecular identifier (UMI) counts from 60,000 reads per voxel and 17,000 genes per sample were captured (fig. S12).

We anchored the spatial transcriptome-seq data based on histology. As expected, VWF (an EC marker gene) was detected in the intima, where CD68 was also detected (Fig. 3A). These data indicate that the spatial transcriptome-seq data indeed retains the spatial features of the gene expression and captures ECs and MΦs in proximity. Given the 50 μm resolution of our spatial transcriptome-seq, which does not provide gene expression profiles at the single-cell level (fig. S13, A and B), we deconvoluted the spatial transcriptome-seq data using scRNA-seq data. For this purpose, we performed scRNA-seq of whole arteries, which have cellular composition matching that of spatial transcriptome-seq to enable effective data integration and spatial anchoring (fig. S13, C to F). Through the integrative analysis of ST- and scRNA-seq, each voxel was assigned probability scores for one of the five major cell types in the arterial wall namely ECs, VSMCs, MΦs, fibroblasts, and T cells. As an example, voxels with probability scores greater than 20% for ECs were defined as “EC-dominant voxels”. We then took these voxels and queried information on other present cells in EC proximity, defined as within 50 μm distance, to encompass potential intercellular interactions. In the non-diabetic arteries, the probability of MΦs in the ‘EC-dominant’ voxels was 16%, whereas in the T2D arteries the probability was 41% (Fig. 3B), suggesting an increase in EC-MΦ interactions in the T2D arteries.

Figure 3. Increased EC-MP and EC-TREM2 interactions in T2D arteries.

(A) SpatialFeaturePlots showing histology guided intima annotation (left). Middle and right spatial plots colored to show VWF and CD68 expression (Exp) in the intima via spatial transcriptome-seq. Scale bar = 500 μm. (B) Pie chart of probability of cell-type presence within ‘EC-dominant’ voxels in non-diabetic and T2D arteries, each segment represents the probability of a different cell type, computed through integrating spatial transcriptome-seq (n=2) and scRNA-seq (n=5). (C) Intercellular interactions comparison between non-diabetic vs T2D using CellChat (n=5), arrow thickness is proportional to the change in communication strength between conditions (red arrows, increased; blue arrows, decreased in T2D). (D) TREM2 expression along MP pseudotime (log scale). Each point represents a single cell, black line indicates the smoothed trend. (E) Expression of TREM2 ligands in non-diabetic vs T2D ECs, Circle color shows expression level; circle size shows % of cells expressing the gene. (n=5). (F) CellphoneDB analysis of bidirectional EC–MP ligand–TREM2 interactions, dot color indicates average expression level of interacting pair, and dot size denotes interaction significance (n=5). (G) TREM2 expression levels in spatial transcriptome-seq voxels stratified by dominant cell type determined through integrating spatial transcriptome- and scRNA-seq. (H) Immunohistochemistry of CD31 (purple), CD68 (yellow), and TREM2 (turquoise), red arrows indicate CD68-TREM2 co-positive spots. Scale bar = 50 μm. (I) Quantification of CD68-TREM2 co-staining in the intima (n=3). *p<0.05, ***p<0.001 vs non-diabetic (Student’s t-test).

To validate that the increased EC-MP interactions in T2D arteries occur in the intima and to identify ligands and receptors mediating these interactions, we performed CellChat and CellPhoneDB analyses with intima scRNA-seq data as shown in Figures 1 and 2. These bioinformatics tools allowed inference of cell-cell communication based on gene expression levels of ligand-receptor pairs (36, 37). The analysis revealed an overall increase in intercellular interactions in T2D as compared to ND, evidenced by both the number of interactions and the interaction strength (fig. S14A). EC interactions with MPs as well as with other immune cells (T/NK) were increased in T2D (Fig. 3C, fig. S14B). Focusing on the EC-MP interactions, we observed a marked increase in bidirectional communications in DM samples, characterized by enhanced ligand-receptor pairing particularly in EC to MP direction (fig. S15A) and enrichment of signaling pathways involving these interactions (fig. S15B). In the EC to MP direction, upregulated CCL14–CCR1, CXCL12–CXCR4, MIF–CD74+CXCR4, and THBS1-CD47 signaling would support monocyte recruitment, adhesion, and pro-inflammatory activation (38, 39). In the MP to EC direction, increased TNFSF12–TNFRSF12A and TGFB1–ACVR1B+TGFBR2 signaling suggested sustained EC activation and potential EndoMT (40), whereas IGF–IGF1R and VEGFB–VEGFR1 interactions can modulate EC metabolism and growth (41). These bidirectional EC-MP interactions outline an amplifying loop in which dysfunctional ECs recruit and activate MPs, which in turn reinforce EC chemokine and adhesion programs.

With our focus on EC-MP communication networks, we were particularly interested in TREM2, one of the most strongly induced MP genes in the diabetic mesenteric artery intima and encoding a surface receptor (Fig. 2D, fig. S8). Consistently, pseudotime analysis revealed a progressive increase in TREM2 expression in MPs along the DM trajectory (Fig. 3D). In parallel, several known TREM2 ligands, including clusterin (CLU), amyloid-b precursor protein (APP), apolipoprotein E (APOE) (34) were increased in T2D ECs (Fig. 3E, fig. S15C). Correspondingly, the inferred ligand-TREM2 receptor pairs CLU-TREM2 and APP-TREM2 were increased in T2D relative to non-diabetic samples (Fig. 3F). Deconvolved spatial transcriptome-seq data revealed that TREM2 expression was increased in T2D arteries, not only in ‘MP-dominant’ voxels as expected, but also in ‘EC-dominant’ voxels (Fig. 3G). This finding was corroborated by increased TREM2-CD68 co-staining within the intima layer of T2D arterial segments (Fig. 3, H and I). Together, these spatially resolved and single cell data indicate that TREM2-expressing MΦs engage in enhanced crosstalk with ECs under DM conditions.

Increased TREM2 expression by diabetic conditions in vitro and in vivo

TREM2 expression was found to be clearly increased in the MPs in the intima of human T2D mesenteric arteries and was highest in MP-C3 (Fig. 4, A and B). Extending this, we found that Trem2 was increased in the aortas of C57Bl/6 mice treated with streptozotocin (STZ) and db/db mice, models of T1D and T2D respectively (fig. S16). Moreover, THP-1-derived human MФs, human monocyte-derived MФs, and mouse RAW 264.7 MФs exposed to high glucose and tumor necrosis factor-α (TNF-α) (which mimic a DM condition in vitro) (15, 42), showed increased TREM2 expression as compared to the normal glucose supplemented with mannitol as osmolarity control (NM) (Fig. 4C to F, fig. S16C). Of note, the increased TREM2 was associated with increased phospho-Syk, indicative of more TREM2 activity (43) (Fig. 4G, fig. S16D).

Figure 4. TREM2 is increased in DM-MP and promotes inflammatory response.

(A) UMAP of TREM2 expression in non-diabetic and T2D donors in intima scRNA-seq (n=5), with cells colored by TREM2 expression level. (B) Dot plot of TREM2 expression in 5 MP clusters (C1–C5) of intima scRNA-seq, circle color shows expression level; circle size shows % of cells expressing gene (n=5). (C-E) TREM2 mRNA levels in THP-1-derived MΦs (C, n=6), human monocyte-derived MΦs (HMDMs) (D, n=3), and mouse RAW 264.7 MΦs (E, n=3) treated with 25 mM glucose, 5 ng/mL TNF-α (HT) or 25 mM mannitol (NM). (F, G) Immunofluorescence staining (F) of TREM2 (red) (with DAPI counterstain (blue)), scale bar = 100 μm, and immunoblotting (G) of TREM2 and p-Syk in THP1-derived MΦs as in (C). (H) Dot plots of foamy and inflammatory marker expression in TREM2⁺ MPs in non-diabetic and T2D intima, circle color shows expression level; circle size shows % of cells expressing the gene (n=5). (I-J) Microfluidic adhesion assay using CMFDA-labeled (purple) TREM2-high/low (TREM2^−Hi/Lo) myeloid cells from diabetic human spleens and HUVECs. Representative images with 100 μm scale bar (I) and quantification presented as a violin plot, midline denotes median and dotted line denotes upper and lower quartiles (J, n=3). (K-M) Adhesion assay with ECs cultured under HT and incubated with CMFDA-labeled (green) THP-1 monocytes with and without TREM2 knockdown (KD), representative images with 100 μm scale bar (in K) and quantification presented as a violin plot, midline denotes median and dotted line denotes upper and lower quartiles (in L, n=3) (M) Quantification of adhesion assay as in K but with THP-1 monocytes pre-treated with 200 ng/mL TREM2 neutralizing antibody or IgG isotype control (n=3). (N, O) Single-frequency label-free SRS imaging of lipid (CH₂, red) and protein (CH₃, blue) channels and corresponding lipid:protein ratio (rainbow) in THP-1-derived MΦs. Representative images (N) and quantification of lipid:protein ratio (O). Data represent mean±SEM from 3–4 independent experiments. *p<0.05, **p<0.01 (Student’s t-test in C-E, J, L, M; one-way ANOVA followed by Dunnett’s test in O).

Previous studies have revealed highly context-dependent roles of TREM2 and TREM2⁺ cells across different disease settings (34). To define their function in diabetic vasculature, we first characterized the expression profiles of TREM2⁺ MPs in the intima scRNA-seq data. In the non-diabetic intima, TREM2⁺ MPs closely resembled previously reported TREM2⁺ foam cell-like MФs (14, 31, 44), exhibiting higher expression of foam cell-associated genes (e.g., CTSB, CTSD, PLIN2, and LGALS3), lower expression of pro-inflammatory markers (e.g., IL1B, TNF, CCL3, and REL) and higher levels of anti-inflammatory markers (e.g., HMOX1, ANXA1, STAB1 and FN1) compared with the TREM2⁻ MPs (Fig. 4H, two lower rows). In contrast, within the diabetic intima, TREM2⁺ MPs retained the foam cell-like features but exhibited markedly increased expression of pro-inflammatory markers and reduced expression of anti-inflammatory genes (Fig. 4H, two upper rows), indicating a maladaptive inflammatory shift under DM conditions. In support of this idea, TREM2^−Hi myeloid cells isolated from the spleens of T2D donors (fig. S17) adhered markedly more to ECs in a microfluidic vessel-on-chip device (45), as compared to the TREM2^−Lo counterparts (Fig. 4I and J).

To directly test the role of TREM2 in MP function under DM, we inhibited TREM2 in THP-1 monocytes using TREM2 small interfering RNA (siRNA) or a neutralizing antibody (TREM2-Ab) (46, 47). The effect of TREM2 knockdown (KD) was confirmed by the decreased levels of TREM2 mRNA and protein (fig. S18, A to C) and that of TREM2-Ab was verified by the reduction of phospho-Syk (fig. S18, D and E). Both TREM2-KD and TREM2-Ab substantially decreased adhesion of monocytes to HT-exposed ECs (Fig. 4, K to M). Conversely, treating monocytes with a TREM2 agonist (48) significantly (p=0.0004) increased their adhesion to ECs in NM conditions (fig. S19).

Given the reported role of TREM2 in lipid metabolism and uptake in myeloid cells (49), particularly in foamy MΦs in atherosclerosis (44), we examined the effect of TREM2-KD in lipid uptake in the context of DM. Using stimulated Raman scattering (SRS) microscopy, which visualizes metabolites through label-free imaging (50), we observed increased lipid-to-protein ratio in THP-1-derived MΦs by HT as compared to NM, which was attenuated by TREM2-KD (Fig. 4N and O). Additionally, we performed bulk RNA-seq to identify the effect of TREM2-KD on the MΦ gene expression profile under HT. TREM2-KD resulted in 2752 DEGs (including 1146 up- and 1606 down-regulated), which were enriched for pathways involved in signal transduction, inflammation, angiogenesis and lipid metabolism (fig. S20, A and B). Several marker genes of foamy MΦs were downregulated (fig. S20C), and a list of chemokines enriched in DM-TREM2⁺ MPs, e.g., CCL2, CCL4, and CXCL2 (Fig. 4H) were also downregulated by TREM2-KD (fig. S20D). Apoptosis and proliferation pathways were also enriched in the DEGs (fig. S20B), but at the cellular function level, we did not observe significant difference in cell viability due to TREM2-KD (fig. S21A). These results suggest that TREM2 plays a crucial role in the pro-inflammatory response and lipid uptake of MPs under diabetic conditions.

TREM2 modulates EC function in diabetic conditions

Among all MP clusters identified from the intima scRNA-seq data from human mesenteric arteries, TREM2 is most abundant in MP-C3 (Fig. 4B), the cluster enriched for EC-relevant pathways, including angiogenesis, hypoxia response, and nitric oxide biosynthetic process (fig. S7). Indeed, we also found differential regulation of genes involved in angiogenesis upon TREM2-KD in bulk RNA-seq data (fig. S20B and E). Thus, we queried whether TREM2 in MΦs can reciprocally regulate EC function under DM conditions. Monolayers of human aortic ECs (HAECs) were treated with conditioned medium from THP-1-derived MΦs with or without TREM2-KD. The HAEC layer was scratched, and the rate of ‘wound’ closure was determined (Fig. 5A). Compared to conditioned medium from NM-treated MФs, conditioned medium from HT-treated MФs inhibited the wound closure of ECs, which was partially reversed by TREM2-KD (Fig. 5, B and C). However, the effect of TREM2-KD was not significant under NM. Similar data were obtained from human microvascular ECs (HMVECs) and human umbilical vein ECs (HUVECs) (fig. S22, A and B). Furthermore, conditioned medium from MФ under HT caused ECs to express higher levels of pro-inflammatory markers VCAM1, ICAM1, and MCP-1, which was abrogated with TREM2-KD (Fig. 5D, fig. S22C). Of note, addition of HT alone to the media did not affect EC wound closure (fig. S22, D and E), suggesting that the secreted factors from HT-treated MФs, rather than HT per se, impaired EC migration. Compared to the scrambled control, TREM2-KD decreased several pro-inflammatory cytokines, including interleukin-6 (IL-6), IL-7 and IL-8 (Fig. 5, E to G). Additionally, basic fibroblast growth factor (bFGF), a pro-angiogenic factor was slightly increased and an anti-angiogenic factor, IL-1RA was decreased by TREM2-KD (fig. S22, F and G).

Figure 5. TREM2 modulates EC function in diabetic conditions.

(A-G) Schematic of human EC scratch wound assay using conditioned medium from THP-1-derived MΦs with and without TREM2 knockdown under NM/HT, created using BioRender.com. (B, C) Representative images with 200 μm scale bar (B) and quantification (C) of scratch wound assays in human aortic ECs incubated with MΦ conditioned medium over 24-hour (h) time course (n=3). (D) qPCR of VCAM1, ICAM1 and MCP-1 in ECs treated with MΦ conditioned medium for 24 h (n=3). (E-G) IL-6, 7, and 8 levels in MΦ conditioned medium measured by Luminex (n=3). (H) ELISA quantification of soluble TREM2 (sTREM2) in conditioned medium from macrophages treated as in (A) for 24 or 72 h (n=3). (I) Scratch assays of HAECs treated with sTREM2 at 0, 100, 200 or 300 ng/mL. Representative images at 0 and 24 h with 200 μm scale bar. (J) Quantification of scratch assay in I (n=3). (K-M) qPCR analysis of HAECs treated with sTREM2 at 0, 100, 200 or 300 ng/mL showing expression of inflammatory/adhesion markers VCAM1 (in K), ICAM1 (in L) and MCP-1 (in M) (n=3). (N) sTREM2 concentrations in HT-MΦ-conditioned medium following singular or dual siRNA-mediated knockdown of ADAM10 and ADAM17 normalized to HT-scrambled (Scr) control (n=4). Data represents mean±SEM from 3–5 independent experiments. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 (ANOVA followed by Dunnett’s post-hoc test [C, D, H, J-N]; Student’s t-test [E-G]).

To further delineate the communication between TREM2 and EC, we investigated the role of soluble TREM2 (sTREM2), the proteolytic cleavage product of membrane TREM2 shown to stimulate inflammatory activation in microglia and monocytes/MФ (51). We found that HT progressively increased sTREM2 production by THP-1-derived MФ as compared to NM, which was diminished by TREM-KD (Fig. 5H). Incubation of HAECs with recombinant sTREM2 caused a dose-dependent reduction in EC migration and an increase in expression levels of inflammatory/adhesion molecules (VCAM-1, ICAM-1, MCP-1) even at baseline (Fig. 5, I to M), suggesting pro-inflammatory and anti-EC repair effects of sTREM2. Upstream of sTREM2, metalloproteases ADAM10 and ADAM17 were shown to mediate the cleavage of full length TREM2 and the shedding of sTREM2 (52). Indeed, ADAM10 and ADAM17 were upregulated under diabetic conditions in the intima of mesenteric arteries (fig. S23A). When either ADAM17 alone or both ADAM10 and ADAM17 were knocked down in HT-treated MФ, sTREM2 levels were reduced in the conditioned medium, suggesting that these proteases mediate the sTREM2 production in DM (Fig. 5N). Collectively, these data suggest that TREM2-mediated signaling from MPs promotes inflammatory responses and inhibits wound healing capacities of ECs.

TREM2 neutralization improves ischemic perfusion recovery in DM

Human mesenteric arteries have an established utility in studying DM-related vascular abnormalities (21). To validate the relevance of mesenteric arteries to limb muscle vasculature, we compared the EC transcriptome from mesenteric artery with our published scRNA-seq dataset of ischemic and non-ischemic muscle from three patients with CLTI (two non-diabetic and one T2D) (GSE227077) (53). We found 49 of the top 100 DM-associated DEGs in ECs of mesenteric intima showed concordant directional changes in PAD muscle, including canonical activation/inflammation markers (CCL2, SERPINE1, VCAM1, CCN2, THBS1, and FN1; fig. S24). In terms of TREM2 in MP, its DM-associated increase was also observed in ischemic muscles (GSE227077). Of note, the increased TREM2 in MP was also found in carotid plaque (GSE224273, GSE235437) (13), retina (GSE245561, GSE130636) (54, 55), and kidney (GSE131882), all related to diabetic vasculopathy (56) (S25, A to D). Moreover, expression profiles of the DM-TREM2⁺ MP in these tissues positively correlated with those in the mesenteric intima, with CLTI muscle showing the strongest correlation (R = 0.83; fig. S25, E to G). Among the correlated genes, TNF-α, a canonical inflammation driver, was consistently increased in MPs (and specifically TREM2⁺ MP) in mesenteric artery intima and CLTI muscles from T2D vs non-diabetic individuals (fig. S26A and B; Fig. 4H). ADAM10 and ADAM17, the proteases that cleave full length TREM2 to sTREM2, were upregulated under diabetic conditions in the intima of both mesenteric arteries and ischemic limbs (fig. S23). Furthermore, the enrichment of foam-cell markers and heightened inflammatory markers were also shared between DM-TREM2⁺ MP of mesenteric intima and their counterparts in ischemic muscles from both human and mouse (fig. S26, C and D). Together, these cross-dataset analyses show concordant diabetes-associated EC dysfunction and TREM2⁺ MP in mesenteric arteries and ischemic limb tissues, supporting mesenteric arteries as a relevant model for diabetic PAD.

Next, we investigated the role of TREM2 in vivo using two mouse models with DM and ischemic injury. In the first model, we induced hyperglycemia using STZ (which increases Trem2 in the aorta [fig. S16A]), followed by femoral artery ligation to induce hindlimb ischemia (HLI) (Fig. 6A, fig. S27A). scRNA-seq of the limb muscles revealed that Trem2 expression was increased in MФ in DM as compared to non-diabetic mice, and further increased in the ligated ischemic limbs as compared to the non-ischemic controls (Fig. 6A). In the diabetic mice, there was an increase in TREM2 protein and its co-localization with isolectin B4 ([IB4], EC marker) in the ischemic vs non-ischemic limb muscles (Fig. 6B, fig. S27B). We then treated these mice with a local intramuscular injection of TREM2-Ab (equivalent to in vitro) (Fig. 6C-1). Compared to the immunoglobulin G (IgG) control, diabetic mice receiving TREM2-Ab showed reduced TREM2 protein levels in the ischemic muscles (fig. S28, A and B) and enhanced recovery of perfusion in both sexes (Fig. 6, D and E, fig. S28C), coupled with an increase in vascular density in the ischemic muscle (Fig. 6, F and G). No significant sign of apoptosis or MФ proliferation due to the TREM2-Ab injection was observed (fig. S21, B and C). Of note, in non-diabetic mice, TREM2-Ab did not significantly affect the recovery rate (fig. S28E).

Figure 6. Induction of TREM2 in DM-PAD and its effect on ischemic recovery.

(A) Top: Laser speckle imaging in C57Bl/6 mice treated with or without streptozotocin (STZ) (DM vs ND), 5 days post hindlimb ischemia (HLI). White arrow indicates reduced flow in the ligated limb of DM mice. Bottom: MΦ Trem2 expression in non-ligated (NI) or ligated (I) limb muscles from non-diabetic and DM mice, profiled by scRNA-seq (n=2/group). Circle color shows expression level; circle size shows % of cells expressing the gene. (B) IB4 (green)-TREM2 (red) co-IF in NI vs I limb muscles from STZ-treated mice. Scale bar = 50 μm. (C) Schematic of DM models (STZ or high fat high sucrose [HF/HS]), HLI, and intramuscular injection of TREM2-antibody (-Ab), -agonist or respective controls (IgG or DMSO) post-op Day (d) 1, 4, & 8, created using BioRender.com. (D Representative perfusion images from mice that underwent the scheme in C-1. (E) Quantification (%) of flow recovery in D expressed as ligated/non-ligated limb ratios (n=8/group [7 male, 9 female]). (F) Representative images of IB4 staining (green) of ischemic muscles on post-op Day 6. Scale bar = 100 μm. (G) Quantification of IB4 staining in F (n=6). (H, I) Mice (3-month) fed HF/HS diet for 6-months, underwent bilateral HLI and received IgG (right limb) and TREM2-Ab (left limb) (as in C-2). Representative perfusion images (H) and quantification of ischemic/non-ischemic perfusion ratios (n=13/group [8 male, 5 female]) (I). (J, K) Mice underwent C-3 scheme. Representative perfusion images (J) and quantification (K) of ischemic/non-ischemic perfusion ratios (n=6 males/group) *p<0.05, **p<0.01, ***p<0.001 (Student’s t-test).

In the second model, we employed a high-fat/high-sucrose (HF/HS) diet-induced insulin resistance model that better represents T2D, which is more prevalent in humans (Fig. 6C-2). Consistent with the STZ model, TREM2-Ab also improved perfusion recovery, detected from Day 6 post-ligation (Fig. 6, H and I, fig. S28, F and G), with comparable effects in male and female mice (fig. S28D). Conversely, administration of a TREM2 agonist (Fig. 6C-3) reduced perfusion recovery in DM mice (Fig. 6, J and K), but had an insignificant effect in non-diabetic mice (fig. S28, H and I). These data indicate that Trem2 is increased by ischemic injury under DM condition and TREM2 inhibition promotes angiogenesis and ischemic recovery.

TREM2-EC interactions in human DM-PAD

To explore the disease relevance of TREM2 and its endothelial interactions in human PAD, we first consulted a published bulk RNA-seq data (GSE120642) of gastrocnemius muscle from 32 non-PAD control (C), 27 patients with intermittent claudication (IC) and 19 with CLTI (52). TREM2 was markedly increased in the skeletal muscle from the individuals with CLTI vs IC and C (Fig. 7A). We next measured plasma sTREM2 levels in an independent patient cohort including 21 individuals without PAD (C), 37 with IC, and 61 with CLTI. Consistent with the transcriptomic findings, sTREM2 concentrations were highest in patients with CLTI, exceeding those in both C and IC groups (Fig. 7B and table S4).

Figure 7. TREM2 and sTREM2 are increased in human ischemic and diabetic PAD.

(A) TREM2 mRNA levels from bulk RNA-seq of human muscle biopsies from non-PAD control (C, n=32) intermittent claudication (IC, n=27) or chronic limb-threatening ischemia (CLTI, n=19) donors. Quantification presented as a violin plot, midline denotes median and dotted line denotes upper and lower quartiles. (B) sTREM2 levels in 21 C, 37 IC and 61 patients with CLTI. (C) scRNA-seq analysis of TREM2 expression in MΦ from patients with CLTI. Circle color indicates expression level; circle size denotes % of cells expressing the gene. (D) TREM2 ligands in ECs in (non-)ischemic (NI/I) muscles from C. (E, F) CD31 (green)-TREM2 (red) co-IF in NI/I muscles from patients with CLTI (n=7). Representative images with 20 μm scale bar (E) and quantification by Pearson’s R values (F). (G) Plasma sTREM2 levels in patients with PAD, grouped by non-diabetic (n=56) and T2D (n=42). *p<0.05, ***p<0.001 (ANOVA with Dunnett’s post hoc [A, B]; paired [F] or unpaired [G] Student’s t test).

To further explore cell-level changes, we analyzed our CLTI scRNA-seq dataset (52). TREM2 expression was elevated in MФ from DM vs non-diabetic muscle (Fig. 7C). TREM2 levels followed the hierarchy: DM-ischemic > DM-non-ischemic > non-diabetic-ischemic > non-diabetic-non-ischemic (Fig. 7C), mirroring the pattern observed in the mouse DM-HLI model (Fig. 6A). In line with the increased levels of TREM2 in ischemic vs non-ischemic muscles, those of TREM2 ligands in ECs were also higher in the ischemic muscle (Fig. 7D). Furthermore, co-staining of TREM2 and CD31 in muscle samples from 8 patients with CLTI (table S5) showed greater spatial proximity in the ischemic sites, as compared to the non-ischemic sites, evident by the higher Pearson’s R values (Fig. 7, E and F), suggesting enhanced TREM2-EC interactions under ischemia. Finally, comparison of plasma sTREM2 levels in patients with PAD with and without DM revealed significantly (p=0.04) higher concentrations in DM-PAD relative to ND-PAD (Fig. 7G and table S4). Together, these findings demonstrate that TREM2-EC signaling is elevated in PAD, particularly in the context of diabetes, linking TREM2 expression and sTREM2 to disease severity and ischemic burden.

Discussion

Our findings align with the context-dependent roles of TREM2 and TREM2⁺/TREM2^−Hi cells across tissues (34). In the central nervous system, TREM2 can support debris clearance and neuroprotection in some contexts (57–59) but have also been linked to microglial senescence and Alzheimer’s disease (60). In metabolism, both overexpression and KO of TREM2 have been associated with obesity and insulin resistance (61–63). In cancers, TREM2⁺ tumor-associated MΦs promote tumor growth and immune evasion (64). In atherosclerosis, myeloid-specific TREM2 deletion attenuates plaque progression (44), although bone-marrow chimera studies showed that loss of TREM2 can increase necrotic core size without reducing overall plaque burden (65). In contrast, TREM2 agonism can expand plaque MΦs (66) and attenuate necrotic core formation (65, 66). Recent studies in diabetic nephropathy proposed a protective role of TREM2 and TREM2^−Hi MΦs (67, 68). The divergent effects of TREM2 in different organs and tissues may be due to differences in its interacting partners. Indeed, the communication between TREM2⁺ MФ with adipocytes (61, 62) and tubular epithelial cells (68) are likely different from that with ECs. Here, we find that in the context of DM-PAD, TREM2 signaling was tuned toward endothelial activation and inflammation. Our KD and blocking experiments indicate that TREM2 activation in diabetic MPs leads to increased monocyte-EC adhesion, elevated pro-inflammatory cytokines (e.g., IL-6, IL-7 and IL-8) production, and exacerbated EC inflammation. TREM2 signaling also modulates EC migratory and repair function in part through sTREM2 and other angiogenic factors, such as bFGF and IL-1RA. These effects were not observed under nondiabetic conditions, likely reflecting the low expression of TREM2 and its ligands. The exact endothelial ligands and downstream signaling of sTREM2 in ECs warrant future investigation. Together with the increased EC-MΦ bi-directional communication centered on chemokines, our findings support a model in which dysfunctional ECs recruit and activate MPs, which in turn release cytokines and sTREM2 to reinforce EC adhesion/inflammatory programs, a feedforward loop amplified by diabetes.

TREM2 has emerged as a therapeutic target, with both TREM2 agonists and antagonists being developed for use in treating Alzheimer’s disease and cancer (64, 69). In the context of PAD, the role of TREM2 had not been previously elucidated. We demonstrate the following: 1) antibody blockade of TREM2 improved ischemic recovery in DM mice but not non-diabetic mice; 2) TREM2 expression in muscle and sTREM2 in the circulation tracked with PAD severity (CLTI>IC>non-PAD); and 3) MΦ TREM2 and sTREM2 expression was higher in DM-PAD than nondiabetic PAD. These findings suggest that EC-TREM2 interactions contributes to DM-PAD. Future studies will elucidate whether this mechanism also plays a role in patients without diabetes. Our findings have clinical implications. Patients with DM-PAD experience poorer outcomes and microvascular dysfunction independently associates with disease progression and limb ischemic severity (2). Thus, targeting TREM2-mediated EC–MΦ crosstalk may help improve clinical endpoints relevant to limb salvage and recovery. Moreover, plasma sTREM2 may be useful as a circulating marker of endothelial/vascular dysfunction in PAD for risk stratification and outcome prediction. Additionally, our findings suggest caution when considering TREM2-enhancing therapeutics, particularly in individuals with existing DM and ischemic disease.

The present study has limitations. First, the profiling of human mesenteric arteries may not capture all DM-associated changes in other vascular beds, e.g., ischemic limb microvasculature. Although we validated findings in HAECs and HMVECs and observed cross-bed concordance, broader studies are warranted. Second, given donor background complexities, some observed T2D-associated changes could be influenced by comorbidities such as hypertension and hyperlipidemia. Moreover, although donors varied by race and sex, the small sample size precluded analysis of demographic-specific differences. Third, although commonly used, the HLI model causes acute arterial occlusion, thus does not allow delineation of the role of TREM2 in chronic ischemia underlying PAD. Additionally, intramuscular TREM2-Ab could affect cell types beyond MΦs. Nonetheless, our study provides an atlas of diabetic human arteries with single-cell and spatial resolution, identifies an important role of TREM2-EC interactions in diabetic vasculopathy, and lends support for biomarker development and targeted intervention of EC-TREM2 signaling to treat DM-PAD.

In this study, we leveraged scRNA- and spatial transcriptome-seq to map DM-associated transcriptomic changes in human arteries at the single-cell level within their native spatial context and to delineate EC-MФ communication. A standout finding was TREM2, identified as the top DM-induced MФ receptor and an exemplary demonstration of how DM-mediated MФ-EC crosstalk contributes to diabetes-associated vascular dysfunction. Through integrative analyses combining cell cultures, mouse models of DM-PAD, and samples from human donors with DM and PAD, TREM2 was identified as a critical mediator of EC-MФ communication. The activation of EC-TREM2 signaling in DM promoted vascular inflammation and impaired ischemic recovery. Analysis of clinical cohorts corroborated higher TREM2 expression, EC–TREM2 proximity, and plasma sTREM2 in patients with PAD—particularly those with diabetes—supporting translational relevance. These results provide a high-resolution resource for dissecting DM-associated vascular pathology and highlight a previously unknown role of TREM2-mediated EC-MФ communication in diabetic vasculopathy.

Materials and Methods

Study Design

This study was designed to define the role of TREM2-mediated EC–MΦ interactions in diabetic vasculopathies using human, mouse, and in vitro approaches. Human relevance was assessed using mesenteric artery and muscle samples from non-diabetic and diabetic donors, as well as plasma samples from patients with IC or CLTI. Mouse studies employed STZ- and diet-induced diabetes models combined with hindlimb ischemia to examine the role of TREM2 in on ischemic response and perfusion recovery. Mechanistic studies were performed using primary human macrophages, THP-1–derived macrophages, and human endothelial cells under normoglycemic or diabetic conditions with genetic or pharmacologic manipulation of TREM2 signaling.

ScRNA-seq and spatial transcriptomic analyses were conducted to identify cell-type–specific changes in TREM2 expression and ligand–receptor interactions, followed by functional validation using immunostaining, flow-based adhesion assays, scratch-wound assays, RNA quantification and metabolic imaging. Sample sizes were determined based on prior studies and feasibility and are reported in figure legends. Animals housed within the same cage were randomly assigned to treatment groups to minimize potential cage effects. Perfusion and imaging analyses were performed in a blinded manner. Statistical analyses were performed using appropriate parametric tests as indicated, and all individual-level data for experiments with n<20 are provided in Data file S1. Animal experiments were approved by the Institutional Animal Care and Use Committees at City of Hope (#18069 & #17010) and the Institutional Biosafety Committee (#16023).

Human Tissues

Studies on mesenteric arteries were conducted on deidentified human specimens obtained from the Southern California Islet Cell Resource Center at City of Hope. The research consents for the use of postmortem human tissues were obtained from the donors’ next of kin and ethical approval for this study was granted by the Institutional Review Board of City of Hope (IRB #01046). The selection of donors was based upon the Integrated Islet Distribution Program criteria. T2D was identified based on diagnosis in the medical records of the donors and/or as a glycated hemoglobin A1c (HbA1c) level of 6.5% or higher.

Peripheral blood mononuclear cells (PBMCs) and CD14⁺ monocytes were isolated from the peripheral blood of donors as described (42) (approved under IRB #00123 from City of Hope).

Skeletal muscle was obtained from individuals with CLTI undergoing lower-limb amputation in accordance with a research protocol approved by the Duke University Institutional Review Board (IRB Pro00065709). For plasma samples, eligibility criteria included patients undergoing vascular surgery procedures such as amputation, open vascular or endovascular procedures. This allowed recruitment of both patients with and without PAD.

Statistical analysis

Statistical analyses for data other than high-throughput sequencing were performed using GraphPad Prism with input from a Ph.D. level biostatistician from the Biostatistics Core at City of Hope. All data are representative of at least three independent (biological) replicates unless specifically noted. The sample size was precalculated to satisfy power requirements (with >85% confidence) for animal experiments. To accomplish randomization for imaging analysis, data from animal experiments, immunofluorescence, immunohistochemistry, and perfusion images, were sorted by a secondary blinded investigator prior to analysis. For sequencing experiments, genes were considered differentially expressed if their adjusted P value (q value) was below 0.05 and above a log fold chang (FC) threshold of 0.25. Exclusion criteria were defined a priori and limited to technical issues (namely sequencing that did not meet pre-defined quality control metrics, inadequate signal quality, or surgical procedures not completed as intended). Samples meeting these criteria were excluded prior to analysis. No data points were excluded based on experimental outcomes. Two-group comparisons were performed using unpaired or paired parametric Student’s t-test based on the normality of the data (specified in figure legends) and multiple-group comparisons were performed using analysis of variance (ANOVA) followed by post hoc tests as appropriate (specified in figure legends). P values less than 0.05 were considered statistically significant.

Supplementary Material

Materials and Methods

Figures S1 to S28

Tables S1 to S7

References (70–72)

Data file S1–S2

MDAR Reproducibility checklist

Data, code, and materials availability

All data associated with this study are present in the paper or supplementary materials. All high-throughput sequencing data used in this study have been deposited on GEO (GSE317257). All materials used or generated in this study are commercially available or will be supplied upon reasonable request.