Employing the Role of SLC40A1-Mediated Microglia Inflammatory Response in Diabetic Retinopathy Based on Single-Cell Transcriptomics Technology

Abstract

Diabetic retinopathy (DR) is one of the most severe complications of diabetes. Existing treatments are effective but still have many limitations. Thus, it is particularly crucial to uncover novel therapeutic targets and develop more efficacious treatment strategies. In this study, single-cell datasets of retinal and normal retinal tissues of DR patients were obtained from the Gene Expression Omnibus (GEO) database. Using cell annotation, Gene Ontology (GO) pathway enrichment analysis, and cell trajectory analysis, key cell subgroups and their enriched pathways affecting DR progression were selected. DR model mice was constructed. Co-localization between SLC40A1 and microglial cell marker IBA1 was assessed through immunofluorescence detection. qPCR, Western Blot, and ELISA were used to evaluate the expression levels of inflammatory factors. Bioinformatics analysis revealed that in retinal tissues of DR patients, SLC40A1+ microglia exhibited abnormal accumulation and were tightly correlated with inflammatory pathways. Animal experiments showed that the co-localization level of SLC40A1 with IBA1 increased in DR model mice. Further research revealed that knocking out the SLC40A1 gene in microglia can reduce the expression levels of pro-inflammatory factors, including IL-6, TNF-α, and IL-1β. It effectively alleviated the pathologic characteristics of DR. SLC40A1+ microglia modulated inflammatory responses to control DR progression, offering promising avenues for new therapeutic targets in DR treatment.

Introduction

Diabetic Retinopathy (DR), as one of the most common microvascular complications of diabetes, affects approximately 34.6% of individuals living with diabetes, with 30% progressing to varying degrees of visual impairment (Wang et al., 2024). The pathological characteristics primarily include capillary degeneration, loss of pericytes, and vessel leakage (Liu et al., 2023). At present, the therapeutic strategies for DR mainly focus on systemic treatment, vitrectomy, retinal photocoagulation, and intravitreal injection of anti-VEGF inhibitors (Yamato et al., 2024). While these treatments primarily address advanced DR, their efficacy is limited for early-stage disease. They can also result in many adverse reactions, including possible permanent loss of peripheral vision and worsening macular edema (Hu et al., 2024). Thus, exploring new therapeutic targets and developing more efficacious treatment strategies have become particularly urgent.

As intrinsic resident immune cells of the retina, microglia maintain homeostatic distribution under physiological conditions and participate in the dynamic equilibrium of the retinal microenvironment by constantly surveilling the local microenvironment, clearing metabolic waste, and stabilizing synaptic structures (Guo et al., 2022). Exposure to high glucose and oxidative stress triggers reactive activation of retinal microglia, accompanied by enhanced secretion of pro-inflammatory cytokines including IL-1β, IL-6, and TNF-α. Such activated microglia further exacerbate pathological progression of DR by disrupting the integrity of the blood-retinal barrier and inducing functional impairment of the neurovascular unit (Lei et al., 2025; Rudraraju et al., 2020). Nevertheless, single-cell transcriptomic datasets have uncovered that retinal microglia consist of functionally distinct subpopulations with divergent transcriptional signatures under pathological conditions, whose activation landscapes are far more intricate than those documented in preliminary investigations. Single-cell RNA sequencing profiling of retinal tissues harvested from DR patients enables stratification of microglia into discrete functional subsets with sharply differentiated transcriptional profiles. A fraction of these subsets engages in intercellular crosstalk with endothelial cells via the CSF1/CSF1R signaling cascade, which subsequently fuels the establishment of a pro-inflammatory microenvironment and pathological angiogenesis (Ben et al., 2024). In preclinical models of retinitis pigmentosa, a microglial subset marked by robust CCL7 expression initiates a self-amplifying inflammatory cascade through STAT1 pathway activation, which selectively mediates degenerative apoptosis of photoreceptors (Li et al., 2026). Collectively, identifying microglial subpopulations that undergo aberrant expansion and exert central pro-inflammatory effects during DR pathogenesis, alongside elucidating the molecular regulatory networks governing their activation, is critical for deciphering the pathogenic mechanisms underlying DR and designing targeted therapeutic interventions.

SLC40A1, also known as FPN1, MTP1 or IREG1, is a pivotal iron transporter that plays an essential role in maintaining iron homeostasis (Fang et al., 2024). Transcriptional regulation is influenced by multiple factors, including hypoxia, iron levels, and inflammatory signals (Feng et al., 2024). SLC40A1 exerts significant roles in the progression and development of various diseases through regulating ferroptosis pathways, such as diabetes progression, intervertebral disc degeneration, and polarized myocardial injury (Chen et al., 2024; Hao et al., 2021; Wang et al., 2023). Nevertheless, the biological function and regulatory mechanism of SLC40A1 in the DR remain unexplored.

In this study, we analyzed single-cell sequencing data from normal retina as well as those from DR patients. Results showed that SLC40A1+ microglia were notably upregulated in the retinal tissue of DR patients. Through animal model experiments, we revealed the biological function of this specific cell subgroup in promoting the progression of DR through its impact on inflammation. This discovery not only deepened our understanding of the pathogenesis of DR but also provided new potential targets for developing targeted therapeutic strategies.

Materials and Methods

Bioinformatics Analysis

Data Acquisition

Single-cell transcriptomic datasets associated with DR were retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/). Two datasets, GSE165784 and GSE245561, contained retinal single-cell profiles from DR patients, including 5 and 4 DR retinal specimens respectively; single-cell transcriptomic data of healthy human retinas were obtained from GSE202735. All retrieved datasets were utilized to construct single-cell expression atlases of normal and DR retinal tissues, followed by in-depth characterization of transcriptional signatures exhibited by microglial subpopulations under DR pathological conditions.

Single Cell Analysis

All single-cell transcriptomic raw matrices were processed via the R package Seurat (version 4.3.1). Independent quality control filtering was first implemented for each dataset to retain cells with feature counts ranging from 200 to 6000 and mitochondrial gene expression proportions below 25%. The NormalizeData function was applied to normalize raw expression matrices, while variable genes were identified through the variance-stabilizing transformation (vst) algorithm embedded in the FindVariableFeatures function. After dataset-wide scaling with ScaleData, principal component analysis (PCA) was performed using RunPCA, and optimal PCA dimensions were selected for subsequent dimensionality reduction and cell clustering based on principal component distribution patterns. RunUMAP was adopted to generate two-dimensional cellular visualizations, and unsupervised cell clustering was accomplished sequentially via FindNeighbors and FindClusters.

Cell type annotation was primarily accomplished by evaluating the characteristic marker gene expression patterns within distinct cell clusters, with auxiliary reference to well-documented retinal cell markers from peer-reviewed publications (Corano Scheri et al., 2023; Hu et al., 2022; Mullin et al., 2023) and the CellMarker 2.0 database. Following annotation of major retinal cell populations, microglial subsets were extracted separately for downstream subclustering analyses.

To comprehensively dissect transcriptional heterogeneity of microglia in DR lesions, microglial cells annotated from GSE165784, GSE245561 and GSE202735 were integrated into a unified analytical framework. The Harmony integration algorithm was employed to correct batch effects across distinct datasets within the low-dimensional embedding space, eliminating confounding inter-dataset variation that interferes with subpopulation discrimination. Re-clustering of integrated microglia was executed on batch-corrected embeddings using FindNeighbors and FindClusters, with UMAP projection adopted for intuitive visualization. Distinct microglial clusters were further subclassified into APOC1 + MGC, MKI67 + MGC, NAMPT+MGC, and SLC40A1 + MGC subsets according to the differential expression of signature genes unique to each cluster. Stacked grouped bar charts were generated to quantify and visualize the proportional composition of each microglial subpopulation in DR and normal samples.

Differential Expression and Functional Enrichment Analyses

The FindMarkers function within the Seurat toolkit was utilized to screen differentially expressed genes (DEGs) across discrete microglial subpopulations, with statistical significance assessed via the Wilcoxon rank-sum test; genes passing adjusted p-value thresholds were defined as statistically significant DEGs. Genes with prominent upregulation specifically enriched in SLC40A1 + MGC were extracted as core candidates for subsequent functional exploration. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were completed using the R package clusterProfiler. Enrichment outputs covered three hierarchical GO categories (biological process, molecular function, cellular component) alongside KEGG pathways, and top enriched functional terms were visualized in bubble plots.

Pseudotime Trajectory Analysis

Pseudotime trajectory reconstruction with Monocle2 was performed to delineate dynamic state transitions among distinct microglial subpopulations. A CellDataSet object was initially constructed based on integrated microglial expression matrices, and cellular ordering along developmental trajectories was defined using precomputed DEGs. Dimensionality reduction and trajectory layout construction further revealed the spatial distribution patterns of microglia across continuous pseudotime axes. Combined with the positional distribution of each subcluster along the trajectory and pseudotime progression gradients, dynamic expression trends of pivotal functional genes including SLC40A1, IFI27, and IGSF6 were tracked throughout the entire pseudotime continuum.

Model Construction for Diabetic Mice

Male wild-type C57BLKS/J mice (wild-type, 20–30 g, 8 weeks old) and C57BLKS/J db/db diabetic mice (40–60 g, 8 weeks old) were purchased from GemPharmatech LLC. (Jiangsu, China). Carrying a mutation in the leptin receptor (Lepr) gene, db/db mice spontaneously develop obesity and persistent hyperglycemia, making them a classic preclinical model for investigating DR. All animals were enrolled in experiments at 8 weeks of age, with fasting blood glucose and body weight monitored routinely. Successful establishment of diabetic phenotypes was confirmed when fasting blood glucose levels remained ≥13.9 mmol/L for two consecutive days (Zheng et al., 2024). To ensure full development of retinal pathological lesions induced by sustained hyperglycemia, all mice were longitudinally monitored from 8 weeks to 25 weeks of age over a total follow-up period of 17 weeks. Ocular globes and intact retinal tissues were harvested after euthanasia at the 25-week experimental endpoint. Therefore, 25 weeks of age was defined as the study’s terminal time point, and all retinal morphological alterations observed at this stage predominantly recapitulate DR-like injuries driven by long-standing diabetic conditions.

Microglia-specific SLC40A1 conditional knockout mice were generated via the tamoxifen-inducible Cx3cr1-CreERT2 Cre-loxP recombination system and provided by Shanghai Model Organisms Center, Inc. (China). Briefly, loxP sites were inserted flanking the functionally critical exons of the SLC40A1 locus to generate SLC40A1^flox/flox mice, which were subsequently crossed with Cx3cr1-CreERT2 mice. Under transcriptional control of the Cx3cr1 promoter, CreERT2 recombinase was selectively expressed within microglial and myeloid cell lineages. After induction, CreERT2 mediated deletion of the targeted SLC40A1 fragment flanked by loxP sites, yielding Cx3cr1-CreERT2; SLC40A1^flox/flox conditional knockout mice. To dissect the biological role of microglial SLC40A1 during DR progression, two experimental cohorts were established on the db/db diabetic background, with eight mice allocated to each group (n = 8). The db/db-NC group consisted of littermate control mice carrying the corresponding control genotype, while the db/db-SLC40A1^−/− group contained microglia-specific conditional SLC40A1 knockout mice generated via Cx3cr1-CreERT2-mediated recombination on a db/db background. To verify knockout efficiency, microglia were isolated from mouse retinal tissue at the experimental endpoint, and qRT-PCR was performed to measure SLC40A1 expression levels, confirming markedly reduced SLC40A1 expression in microglia of the db/db-SLC40A1^−/− group.

This study was approved by the Institutional Animal Care and Use Committee of Zhejiang Luoxi Medical Technology Co., Ltd., Hangzhou, China, Approval Number: LX4825010502. All mice were housed in specific pathogen-free barrier facilities maintained at a constant ambient temperature of 22 ± 2 °C under a standardized 12-hour light/dark cycle, with free access to food and water. Mice were randomly assigned to experimental groups. Each individual mouse served as one independent biological replicate, with eight biological replicates included per group. Three independent technical replicates were implemented for all qRT-PCR, Western Blot, and ELISA quantitative measurements. For histological and immunofluorescence (IF) assessments, representative retinal sections from each mouse were subjected to examination. Five non-overlapping fields of view were randomly selected on each section for quantitative analysis. All histological observations, image acquisition, and quantitative measurements were performed by investigators blinded to group assignments to minimize observer bias.

Hematoxylin and Eosin Staining (H&E)

Following fixation of mouse eyeballs in 4% paraformaldehyde solution for two days, they were successively subjected to dehydration treatments with a gradient concentration of ethanol before being transparentized with xylene and embedded in paraffin wax. Embedded ocular tissues were cut into 4-μm-thick sections, stained with H&E stain kit (C0105S, Beyotime, China). Briefly, after the sections underwent pre-treatment such as dewaxing and hydration, they were successively stained with hematoxylin for the cell nuclei and eosin for the cytoplasm. Finally, after being dehydrated, stained, and sealed, the changes in the tissue structure were observed under an optical microscope.

IF

Retinal sections were fixed in 4% paraformaldehyde for 10 min. After rinsing with PBS, sections were permeabilized with 0.1% Triton X-100 at room temperature for 10 min and blocked with 5% BSA for 1 h at room temperature. Double IF staining was performed on identical retinal slices to visualize the co-localization of SLC40A1 and the microglial marker IBA1. Sections were incubated simultaneously with primary antibodies against IBA1 (ab178846, Abcam, UK, diluted at 1:500) and SLC40A1 (26601-1-AP, Proteintech, China, diluted at 1:200) overnight at 4 °C. Following thorough PBS washing the next day, samples were incubated with Alexa Fluor 488-conjugated secondary antibody and Alexa Fluor 568-conjugated secondary antibody (both diluted at 1:500) for 1 h at room temperature in the dark. Alexa Fluor 488 was used to label IBA1-positive signals, whereas Alexa Fluor 568 visualized SLC40A1-positive signals. Nuclei were counterstained with DAPI, and images were captured under a fluorescence microscope. Five non-overlapping random fields were selected for quantitative analysis per section, and all images were acquired with identical exposure parameters to enable unbiased comparison of fluorescence intensity across experimental groups.

Quantitative Polymerase Chain Reaction (qPCR)

Total RNA was extracted from tissue samples using TRIzol reagent (R0016, Beyotime, China), and RNA concentration and purity were determined with a NanoDrop spectrophotometer. cDNA was reverse-transcribed from total RNA using the RevertAid First Strand cDNA Synthesis Kit (K1691, Thermo Fisher Scientific, USA). qPCR reactions were prepared with ChamQ Universal SYBR qPCR Master Mix (Q711-02, Vazyme, China) and run on an ABI 7500 real-time PCR system (Thermo Fisher Scientific, USA). The thermal cycling protocol was set as follows: an initial pre-denaturation step at 95 °C for 30 s, followed by 40 amplification cycles consisting of 95 °C denaturation for 10 s and 60 °C annealing/extension for 30 s. Melting curve analysis was performed upon reaction completion to verify the specificity of amplified products. All gene-specific primers were synthesized by a commercial vendor, and their specificity and amplification efficiency were validated via single peaks in melting curves, absent amplification signals in no-template controls, and standard curve profiling. The amplification efficiency of all primer pairs ranged from 90% to 110%. Relative gene expression levels were quantified using the 2^−ΔΔCt method and normalized to the internal reference gene β-actin. The primer sequences are listed in Table 1.

Table 1.

RT-qPCR primers.

Gene	Primer sequence (5′→3′)
SLC40A1	Forward primer: CTACTTGGGGAGATCGGATGT
	Reverse primer: CTGGGCCACTTTAAGTCTAGC
IL-1β	Forward primer: TGCCACCTTTTGACAGTGATG
	Reverse primer: TGATGTGCTGCTGCGAGATT
TNF-α	Forward primer: GATCGGTCCCCAAAGGGAT
	Reverse primer: CCACTTGGTGGTTTGTGAGTG
IL-6	Forward primer: GACAAAGCCAGAGTCCTTCAGA
	Reverse primer: TGTGACTCCAGCTTATCTCTTGG
GAPDH	Forward primer: ACCCTTAAGAGGGATGCTGC
	Reverse primer: CCCAATACGGCCAAATCCGT

Western Blot Analysis

Following protein sample separation via SDS-PAGE gel electrophoresis, the blotting procedure was initiated. Subsequently, a polyvinylidene fluoride membrane was placed in 5% defatted milk for a sealing period of 1.5 hours. Next, the membrane was incubated with various primary antibodies at 4 °C for 12 hours. The antibody included those targeting IL-6 (ab290735, Abcam, UK), TNF-α (ab183218, Abcam, UK), IL-1β (ab254360, Abcam, UK), SLC40A1 (26601-1-AP, Proteintech Group, Inc., China), and GAPDH (ab181602, Abcam, UK). Next day, the membrane was further incubated with the secondary antibody (ab6702, Abcam, UK). Following completion of incubation, samples were washed before undergoing a color reaction with ECL reagents. Visualization analysis was carried out using the ChemiScope 6000 Imaging System (China).

Enzyme-Linked Immunosorbent Assay (ELISA)

To evaluate the levels of inflammatory cytokines in the plasma of mice, ELISA kits (E-EL-M0044, E-EL-M3063, E-EL-M0037, Elabscience, China) were used for detection. Firstly, the blood samples were placed at room temperature, and then centrifuged at 1,000 g for 15 min to separate plasma from blood cells. Subsequently, plasma concentrations of IL-6, TNF-α, and IL-1β were assayed separately in strict accordance with the kit instructions.

Statistical Analysis

All animal and molecular experimental data were statistically processed using GraphPad Prism 8.0 software (GraphPad Software, USA). Measurement data were expressed as mean ± SD. Eight mice were assigned to each animal experimental group (n = 8), and each individual mouse represented one independent biological replicate. Three separate technical replicates were carried out for all qRT-PCR, Western Blot, and ELISA measurements.

A priori power analysis for sample size calculation was conducted using G*Power 3.1 software. All animal comparisons in this study involved two independent groups; thus two-tailed independent samples t-test was selected for power calculation with α = 0.05, statistical power = 0.80, and equal group allocation ratio (N2/N1 = 1). For comparisons between wild-type and db/db mice, the effect size Cohen’s d = 1.94 was estimated based on previously reported differences in inner nuclear layer thickness in db/db DR models, which indicated a minimum sample size of six animals per group. For comparisons between db/db-NC and db/db-SLC40A1^−/− groups, the effect size Cohen’s d = 4.53 was derived from published conditional knockout studies under diabetic backgrounds evaluating GFAP-positive cell counts, requiring at least three animals per group. Considering approximately 10% potential animal attrition, individual variability in diabetic pathological progression, and the need for multi-index detection, eight mice were ultimately allocated to each experimental group.

All animal experiments adopted two-group comparative designs, including comparisons between sham and db/db mice, as well as between db/db-NC and db/db-SLC40A1^−/− cohorts. Prior to statistical testing, the Shapiro–Wilk test was applied to assess data normality, and the Brown–Forsythe test was used to evaluate homogeneity of variances. Unpaired Student’s t-test was adopted for two-group comparisons if data followed normal distribution with equal variance; the Mann-Whitney U test was utilized for datasets violating normality or homoscedasticity assumptions. Effect sizes between two groups were reported as Cohen’s d alongside 95% confidence intervals where applicable. Exact p-values were presented in figures or figure legends, and p < 0.05 was defined as the threshold for statistically significant differences.

Bioinformatics analyses were implemented in R programming language with corresponding specialized packages. Differential expression analysis of single-cell transcriptomic data was performed using the FindAllMarkers and FindMarkers functions embedded in Seurat with the Wilcoxon rank-sum test, and the Bonferroni correction was applied to adjust p-values for multiple testing. GO enrichment analysis was completed via the clusterProfiler package, with p-values corrected by the Benjamini–Hochberg (BH) procedure. The screening thresholds were set as p valueCutoff = 0.05, p AdjustMethod = BH, and q valueCutoff = 0.5. Pseudotime trajectory analysis was conducted with Monocle 2. Adjusted p-values or q-values were used to judge statistical significance for bioinformatics outputs.

Results

Single-Cell Transcriptomic Analysis

Three single-cell RNA sequencing datasets associated with DR, namely GSE165784, GSE245561 and GSE202735, were retrieved from the GEO database. GSE165784 and GSE245561 contained DR retinal data, while GSE202735 corresponded to healthy retinal data. Quality control filtering, data normalization, dimensionality reduction, unsupervised clustering, and cell type annotation were executed separately for each dataset, followed by UMAP visualization to display the spatial distribution of distinct cell populations. UMAP plots revealed 16 unique cell clusters in GSE165784, 26 clusters in GSE245561, and 22 clusters in GSE202735, with discrete cell populations showing well-separated spatial profiles in low-dimensional embedding space (Figure 1(A,D,G)). Cell identity annotation was further accomplished according to the differential expression patterns of canonical marker genes within each cluster. The GSE165784 dataset primarily captured endothelial, fibroblast, microglial, monocyte, plasma cells, and T cells. Beyond these cell types, erythroid cells were additionally identified in GSE245561. The normal retinal dataset GSE202735 further contained rod cells, melanocyte, and neutrophil (Figure 1(B,E,H)). Bubble plots illustrated that CLDN5, VWF, and CDH5 were predominantly enriched in endothelial; COL1A1, COL1A2, COL3A1, and DCN were mainly expressed in fibroblast; C1QA, C1QC, APOE, and ITGAM were highly expressed in microglia; TRAT1, CD3D, CD3E, and IL7R were selectively enriched in T cells, which robustly validated the reliability of cell annotation (Figure 1(C,F,I)). Collectively, these findings delineated the single-cell landscape of major retinal cell subsets in both healthy and DR retinas, laying a solid foundation for subsequent investigation into microglial subpopulation shifts and their functional contributions to DR pathogenesis.

Figure 1.

Single-cell transcriptome data analysis. (A–C) UMAP plot of GSE165784 dataset, cell annotation UMAP plot, and bubble plot of marker genes used for cell annotation. (D–F) UMAP plot of GSE245561 dataset, cell annotation UMAP plot, and bubble plot of marker genes used for cell annotation. (G–I) UMAP plot of GSE202735 dataset, cell annotation UMAP plot, and bubble plot of marker genes used for cell annotation.

Modulation of Microglia in DR Progression

Drawing on the above single-cell annotation results and prior reports establishing a tight correlation between microglia and DR deterioration (Cowan et al., 2020), microglial cells were extracted from integrated datasets for in-depth subclustering analysis. Re-dimensionality reduction and clustering partitioned microglia into multiple discrete subpopulations with clearly demarcated distribution on UMAP projections (Figure 2(A)). Based on unique transcriptional signatures of each cluster, microglia were further stratified into four functional subsets: APOC1 + MGC, MKI67 + MGC, NAMPT+MGC, and SLC40A1 + MGC (Figure 2(B)). The APOC1 + MGC subset was characterized by elevated transcription of APOC1, HMOX1, MARCO and GAPDH. MKI67 + MGC displayed robust expression of MKI67 and TOP2A, indicative of potent proliferative capacity. NAMPT, ATP1B3, DOCK4, and ZNF331 served as signature genes for NAMPT+MGC, whereas SLC40A1 + MGC showed prominent upregulation of SLC40A1, IFI27, and IGSF6 (Figure 2(C)). Further comparison of the compositional proportions of distinct microglial subpopulations across groups revealed that the percentage of SLC40A1 + MGCs was higher in DR samples than in normal samples, whereas NAMPT+MGCs were more abundant in normal samples (Figure 2(D)). These data implied compositional remodeling of retinal microglia upon DR onset, with SLC40A1 + MGC representing a disease-specific activated microglial subset. Differential expression analysis further confirmed marked transcriptional upregulation of SLC40A1 in SLC40A1 + MGC, accompanied by elevated expression of immune activation markers IFI27 and IGSF6 (Figure 2(E)). Integrating subpopulation frequency shifts and signature gene profiles, our results identified a specialized microglial subset marked by high SLC40A1 expression in DR retinas, which exhibits augmented expression of immune-related mediators and may drive local inflammatory cascades in DR.

Figure 2.

Microglia regulation of DR progression. (A) UMAP plot of microglial subpopulations. (B) Cell annotation UMAP plot of microglia. (C) Marker genes used for microglia annotation. (D) Stacked bar chart showing the proportion of each microglial subtype in DR/HI. (E) Volcano plot of differentially expressed genes.

Analysis of Microglia Trajectories

To dissect the functional properties and progressive state transitions of SLC40A1 + MGC during DR pathogenesis, KEGG and GO enrichment analyses were performed on genes selectively upregulated in SLC40A1 + MGC. KEGG pathway enrichment revealed that differentially upregulated genes in this subset were significantly enriched in phagosome, antigen processing and presentation, oxidative phosphorylation, Th1 and Th2 cell differentiation, Th17 cell differentiation, and cell adhesion molecules (Figure 3(A)). Pathways governing antigen presentation, phagocytosis, and T helper cell polarization are functionally coupled to immune activation, suggesting that SLC40A1 + MGC orchestrate local retinal immune and inflammatory responses in DR. Meanwhile, enrichment of the oxidative phosphorylation pathway implied rewired energy metabolism within this activated microglial subset. Accumulating evidence has validated that ATP signaling modulates cellular proliferation, signal transduction, and inflammatory cascades (Madsen et al., 2023; Zhang et al., 2019). Alterations in oxidative phosphorylation-related pathways within SLC40A1 + MGCs may correlate with their inflammatory activation status.

Figure 3.

Analysis of microglia trajectories. (A) KEGG pathway enrichment results of genes specifically upregulated in SLC40A1 + MGC. (B) GO biological process enrichment results of genes specifically upregulated in SLC40A1 + MGC. (C) GO molecular function enrichment results of genes specifically upregulated in SLC40A1 + MGC. (D) GO cellular component enrichment results of genes specifically upregulated in SLC40A1 + MGC. (E) Distribution of microglial subpopulations on pseudotime trajectories. (F) Distribution of distinct microglial states on pseudotime trajectories. (G) Pseudotime distribution of microglia along pseudotime progression. (H) Expression trends of IFI27, IGSF6 and SLC40A1 with pseudotime changes.

GO biological process enrichment demonstrated that upregulated genes in SLC40A1 + MGC were predominantly implicated in antigen processing and presentation, antigen processing and presentation of peptide antigen via MHC class II, and antigen processing and presentation of exogenous peptide antigen (Figure 3(B)). GO Molecular Function enrichment highlighted prominent enrichment of terms including MHC class II protein complex binding, MHC protein complex binding, MHC class II receptor activity, immune receptor activity, and peptide antigen binding (Figure 3(C)). GO Cellular Component annotation indicated these genes primarily localize to MHC class II protein complex, MHC protein complex, vacuolar membrane, lysosomal membrane, and vesicle lumen (Figure 3(D)). Collectively, results across biological process, molecular function, and cellular component corroborate that SLC40A1 + MGC possess potent antigen-presenting and immune response capacities, functioning as a key pro-inflammatory microglial subpopulation in DR retinas.

Monocle2-based pseudotime trajectory reconstruction was next implemented to map dynamic state transitions of distinct microglial subsets throughout disease progression. Pseudotime ordering segregated microglia into three major trajectory branches with evident subset-specific spatial segregation (Figure 3(E,F)). NAMPT+MGC were predominantly distributed at the early pseudotime stages, whereas SLC40A1 + MGC accumulated at the late trajectory segments with partial overlap alongside APOC1 + MGC, indicating that SLC40A1 + MGC arises progressively as a disease-activated microglial phenotype during DR evolution. Continuous pseudotime gradient visualization further illustrated a directional shift of microglia from quiescent early states toward late-stage pathogenic branches, where SLC40A1 + MGC were predominantly localized (Figure 3(G)). Dynamic gene expression profiling along pseudotime revealed gradual upregulation of SLC40A1 and immune activation mediators IFI27 and IGSF6 as pseudotime advanced (Figure 3(H)). Combined with prior enrichment findings, these observations support a model wherein SLC40A1 + MGC facilitate DR progression via enhanced antigen presentation, immune receptor activity, and inflammation-associated signaling pathways.

High Expression SLC40A1 in Retinal Tissues of DR Model Mice

To characterize retinal pathological alterations and SLC40A1 transcriptional signatures under DR conditions, retinal phenotypes were compared between sham control and DR model mice, aiming to verify whether microglial SLC40A1 undergoes aberrant upregulation concurrent with exacerbated retinal inflammation in DR models. Results revealed marked microvascular anomalies and hemorrhages in the retina of DR model mice (Figure 4(A)). Histological examination by H&E revealed a significant increase in retinal thickness in DR model mice (Figure 4(B)). IF analysis indicated that the co-localization of SLC40A1 with the microglial marker IBA1 was increased in DR model mice. High-magnification images further showed that IBA1-positive signals were predominantly distributed in the cytoplasmic region surrounding DAPI-stained nuclei, whereas SLC40A1 exhibited a more peripheral, membrane-associated distribution within the same cells (Figure 4(C)). Lastly, qPCR and Western Blot analyses revealed that the expression levels of SLC40A1 and inflammatory factors IL-6, TNF-α, and IL-1β were appreciably upregulated in retinal tissue samples (Figure 4(D,E)). These findings further suggested that SLC40A1 + MG may play a pivotal role in the progression of DR by regulating inflammatory responses.

Figure 4.

High expression of SLC40A1 in retinal tissue of DR model mice. (A) Typical representative images of normal and DR eye fundus. (B) H&E stained typical images of retinal tissue from normal mice and DR model mice under 20× magnification (500 μm scale), 200× magnification (50 μm scale) and 400× magnification (20 μm scale). (C) Representative IF images showing SLC40A1, IBA1, and DAPI staining, including high-magnification views for visualization of their cellular and subcellular distribution (scale bars: 10 μm and 5 μm). (D) qPCR analysis of mRNA expression levels of SLC40A1 and inflammatory factors IL-6, TNF-α, and IL-1β. (E) Western Blot analysis of protein expression levels of SLC40A1 and inflammatory factors IL-6, TNF-α, and IL-1β. n = 8 mice per group, each mouse serves as an independent biological replicate. Three technical replicates were conducted for qRT-PCR and Western Blot detection. For H&E and IF staining, 3 representative retinal sections were selected from each mouse, and 5 non-overlapping random fields were analyzed on each section. Data are expressed as mean ± SD. Unpaired Student’s t-test was used for comparisons between two groups; the Mann–Whitney U test was adopted for data that did not conform to normal distribution or homogeneity of variance. *p < 0.05.

SLC40A1 + MG Mediating Inflammatory Responses to Promote DR Progression

Results in Figure 4 confirmed robust SLC40A1 upregulation and enhanced co-localization with the microglial marker IBA1 within retinas of DR mice. To establish a direct causal link between microglial SLC40A1 and DR pathological progression, microglia-specific SLC40A1 conditional knockout mice were generated on the db/db diabetic background, and efficient gene knockout was validated via qRT-PCR quantification (Supplementary Figure S1). Two experimental cohorts of DR model mice were established for comparative analysis: the db/db-NC group (NC) carrying the corresponding control genotype, and the db/db-SLC40A1⁻/⁻ group (SLC40A1⁻/⁻) with Cx3cr1-CreERT2-mediated microglia-specific conditional SLC40A1 knockout. Results revealed that in the SLC40A1^−/− group, the severity of retinal damage was alleviated compared to the control group (Figure 5(A)). Furthermore, compared to the control group, a reduction in thickness was exhibited in the retina of SLC40A1^−/− mice (Figure 5(B)). High-magnification images further confirmed the presence of IBA1 and SLC40A1 signals within the same DAPI-positive cells, while retaining their distinct cytoplasmic and peripheral distribution patterns (Figure 5(C)). These findings consistently indicated that reducing SLC40A1 expression in microglia had a positive impact on delaying the pathological progression of DR. Moreover, ELISA and Western Blot data revealed that the expression levels of inflammation factors IL-6, TNF-α, and IL-1β were significantly downregulated in the retina and plasma of SLC40A1^−/− mice (Figure 5(D,E)). Taken together, these findings suggest that SLC40A1 + MG may contribute to DR-associated retinal inflammation and pathological alterations through the mediation of inflammatory responses.

Figure 5.

Inflammation response promoted by SLC40A1 + MG in DR progression. (A) Typical representative images of eye fundus. (B) H&E staining of mice retinal tissue, 20× magnification (400 μm), 200× magnification (50 μm) and 400× magnification (20 μm). (C) Representative IF images showing SLC40A1, IBA1, and DAPI staining in retinal tissues from db/db-NC and db/db-SLC40A1^−/− mice, including high-magnification views for visualization of their cellular and subcellular distribution (scale bars: 10 μm and 5 μm). (D) ELISA detecting inflammatory factor concentrations. (E) Western Blot analysis revealing the protein expression level of SLC40A1 and inflammatory factors IL-6, TNF-α, and IL-1β. n = 8 mice per group, each mouse serves as an independent biological replicate. Three technical replicates were conducted for Western Blot and ELISA detection. For H&E and IF staining, 3 representative retinal sections were selected from each mouse, and 5 non-overlapping random fields were analyzed on each section. Data are expressed as mean ± SD. Unpaired Student’s t-test was used for comparisons between two groups; the Mann–Whitney U test was adopted for data that did not conform to normal distribution or homogeneity of variance. *p < 0.05.

Discussion

DR ranks among the most prevalent microvascular complications of diabetes mellitus and constitutes a leading cause of acquired vision loss among working-age populations worldwide (Li et al., 2024). As tissue-resident innate immune cells intrinsic to retinal tissue, microglia have long been recognized as central effector cells orchestrating inflammatory cascades in DR, yet their activation manifests as heterogeneous rather than uniform phenotypic shifts. Advances in single-cell transcriptomics have demonstrated that retinal microglia under pathological conditions comprise functionally distinct subpopulations with divergent transcriptional landscapes. Distinct microglial subsets engage signaling axes such as CSF1/CSF1R to fuel pro-inflammatory microenvironment formation and pathological angiogenesis (Ben et al., 2024; Liu et al., 2022). Against this backdrop, the present study performed high-resolution subclustering of retinal microglia isolated from DR lesions, identifying a disease-enriched microglial subset characterized by robust SLC40A1 expression, selective expansion in DR lesions, and prominent enrichment of immune activation signaling pathways. Subsequent validation via microglia-specific conditional knockout mouse models verified the pro-inflammatory function of this subset. Collectively, these findings indicate that pro-inflammatory microglial activation in DR is predominantly driven by this discrete subpopulation, with SLC40A1 serving as a reliable molecular marker to trace this pathogenic subset.

SLC40A1 encodes ferroportin, the primary iron efflux transporter, whose transcriptional abundance governs intracellular iron homeostasis, reactive oxygen species generation, and inflammatory signaling. A self-amplifying positive feedback loop exists between iron homeostasis and inflammatory cascades. Under pro-inflammatory stimuli, the nuclear factor-κB (NF-κB) signaling axis recruits histone deacetylases HDAC1 and HDAC3 to the SLC40A1 promoter, repressing its transcription and triggering intracellular iron retention within myeloid cells (Marques et al., 2025). Intracellular iron overload further stimulates reactive oxygen species production, which in turn hyperactivates NF-κB signaling and augments the secretion of pro-inflammatory mediators including IL-6 (Zhang et al., 2025). Consistent with this regulatory circuit, single-cell transcriptional profiling revealed selective expansion of SLC40A1 + microglia in DR retinal tissue, accompanied by robust enrichment of immune activation and inflammation-associated pathways. In db/db diabetic mice, co-localization between SLC40A1 and the microglial marker IBA1 was intensified. Conversely, microglia-restricted SLC40A1 knockout markedly reduced retinal expression of IL-6, TNF-α, and IL-1β, alleviating local retinal inflammatory infiltration and histological damage. Prior research has documented concurrent activation of iron metabolism cascades and NF-κB-mediated inflammatory signaling under the hyperglycemic and oxidative stress microenvironment of DR (Lu et al., 2024). This evidence implies that the iron homeostasis-inflammation positive feedback axis is persistently activated within SLC40A1 + microglia in diabetic retinas, sustaining their pro-inflammatory phenotype, though this mechanistic hypothesis requires further experimental validation.

Targeted knockout of SLC40A1 attenuated inflammatory cytokine release and mitigated retinal pathological lesions, confirming the involvement of this gene in maintaining the pro-inflammatory microenvironment of DR. Nevertheless, it remains unresolved whether SLC40A1 upregulation acts as a causal driver of inflammatory cascades or merely serves as a phenotypic biomarker of activated microglia. An alternative interpretation posits that elevated SLC40A1 partially reflects the activated state of microglia triggered by hyperglycemia, oxidative stress, or tissue injury, rather than functioning as an upstream initiator of inflammation. Retinal microglia can directly secrete mature IL-1β via danger-sensing receptors such as P2X7R (Campagno et al., 2025). The elevated levels of IL-1β, IL-6 and TNF-α observed in DR are regulated by multiple upstream signaling branches, among which SLC40A1 represents one iron metabolism-associated regulatory node. Subsequent tracking of alterations in iron homeostasis, oxidative stress, and inflammatory signaling in primary or sorted retinal microglia will help further clarify their functional roles in the inflammatory cascade of DR.

Although this study elucidates the pathogenic contribution of microglial SLC40A1 to DR progression, several inherent limitations should be acknowledged. First, all functional validation experiments were conducted exclusively in mice models without complementary verification using clinical human retinal specimens. While mice recapitulate partial retinal pathological hallmarks of diabetes, they fail to fully recapitulate the complex physiological and pathological milieu underlying human DR onset and advancement. Second, microglial identification in this study relied solely on IBA1 immunostaining. However, IBA1 is also expressed by infiltrating peripheral macrophages and other myeloid populations, precluding complete discrimination between resident retinal microglia and bone marrow-derived myeloid cells based on this single marker alone. Follow-up work incorporating markers (TMEM119, P2RY12, CD45, and CD11b), combined with flow cytometry or multiplex IF staining, will enable precise stratification of myeloid cell origins and comprehensive characterization of the composition of SLC40A1-positive myeloid subsets. Third, constrained by experimental logistics, retinal trypsin digestion assays were not performed in the current study, precluding direct quantitative evaluation of hallmark DR microvascular degenerative phenotypes including acellular capillary formation and pericyte loss. Further integrative investigations combining retinal trypsin digestion, quantitative vascular morphological analysis, and clinical sample validation are warranted to thoroughly delineate the regulatory impact of microglial SLC40A1 on microvascular impairment and inflammatory progression in DR.

Ethics Statement

This study was approved by the Institutional Animal Care and Use Committee of Zhejiang Luoxi Medical Technology Co., Ltd., Hangzhou, China, Approval Number: LX4825010502.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Data Availability Statement

Data will be made available on request from the corresponding author.