Macrophage ferroptosis potentiates GCN2 deficiency induced pulmonary venous arterialization

Abstract

Pulmonary veno-occlusive disease (PVOD) is a fatal disease characterized by the remodelling of pulmonary veins and haemosiderin accumulation in macrophages. Although (General Control Nonderepressible 2) GCN2 deficiency has been reported in PVOD patients, the underlying mechanism by which GCN2 deficiency affects the pulmonary venous cells and the surrounding cells, remains unclear. Here, we perform immunohistochemistry and scRNA-sequencing analyses to show that macrophages are the major population affected by GCN2 deficiency and ferroptosis pathway-related genes are upregulated in lung macrophages of PVOD patients. Treatment with the specific ferroptosis inhibitor ferrostatin-1 (Fer-1) reverses the changes in haemodynamic indices observed in Eif2ak4^{K1488X/K1488X} hypoxia mice and PVOD model rats. Furthermore, GCN2 deficiency increases HMOX1 and iron levels to facilitate ferroptosis in macrophages, and enhances arterial marker expression in venous endothelial cells (VECs). Specifically, spatial transcriptome analysis shows increased expression of NRP1, KDR and EFNB2 through ETS1 in VECs from PVOD patients. Our findings suggest the potential of targeting macrophage ferroptosis as a therapeutic strategy for treating related vascular diseases, and of using NRP1/KDR/EFNB2 expression as a specific marker set for venous arterialization.

Fig. 1. Ferroptosis-related gene expression in lung macrophages is increased in PVOD patients with EIF2AK4^mut.

a DNA sequencing analysis of tissues from pulmonary veno-occlusive disease (PVOD) patients in this study confirms EIF2AK4 mutations. M: Male, F: Female. b Western blot showing the expression levels of GCN2 in Control or PVOD patient lung tissues; n = 5 individuals. c Representative images of MARCO (red) and GCN2 (green) immunofluorescence and DAPI (blue) staining of lung tissues from Control and PVOD patients. The data are representative images. White arrow: macrophages. Scale bar = 20 µm. n = 6 microscope fields from six individuals with similar results. d Representative images of lung tissues stained with haematoxylin and eosin (H&E). Red arrow: macrophages; Black arrow: pulmonary vein. Scale bar = 50 µm. Control, n = 6 individuals; PVOD n = 9 individuals with similar results. e UMAP projection showing the main cell type identified by integrated clustering analysis of scRNA-seq datasets from three Controls and three EIF2AK4^mut PVOD patients. f Bubble plot showing the incoming and outgoing interaction strengths for each subpopulation of immune cells in the Control and PVOD groups. The dot size represents the number of interactions. g GSEA of differentially expressed genes in alveolar macrophages, interstitial macrophages, and monocytes between the Control and PVOD groups. The ferroptosis-associated gene set was obtained from FerrDb. NES, normalized enrichment score. h Control and PVOD patient lung tissues were stained with Perls’ DAB to label ferric iron deposits. Scale bar = 50 µm. Control, n = 7 individuals; PVOD n = 9 individuals with similar results. i Heme iron levels in the lungs of Control and PVOD patients were measured. n = 5 individuals. j Non-heme iron levels in the lung were measured in Controls and PVOD patients. n = 5 individuals. k The MDA content in the lungs of Control and PVOD patients was measured. n = 5 individuals. l Representative images of IHC analysis of 4-HNE content in Control and PVOD patient lung tissues. Scale bar = 50 µm. The data are presented as the means ± s.e.m.; unpaired two-sided t test. Each dot represents an individual biological replicate, at least three independent experiments. P values are indicated in the figures. Source data are provided as a Source data file.

Fig. 2. Eif2ak4^{K1488X/K1488X} mice as a model for inducing pulmonary venous remodeling in PVOD.

a DNA sequencing analysis of Eif2ak4^{K1488X/K1488X} mice confirming the EIF2AK4 mutation. Western blot revealing decreased expression of GCN2 in Eif2ak4^{K1488X/K1488X} mouse lung tissues (n = 3 mice/group). b–d Wild-type (WT) and Eif2ak4^{K1488X/K1488X} (K1488X) mice were subjected to normoxia or hypoxia for 6 weeks the total pulmonary vascular resistance index (TPVRI) value, right ventricular systolic pressure (RVSP) and right ventricular hypertrophy (RVH) in the model mice (b male:n = 4 mice/group, female: n = 4 mice/group; c, d male:n = 5 mice/group; female: n = 5 mice/group). TPVRI: Normoxia K1488X vs Hypoxia K1488X, p = 3.5 × 10⁻⁹; Hypoxia WT vs K1488X, p = 7 × 10⁻⁷. RVSP: Normoxia K1488X vs Hypoxia K1488X, p = 2 × 10⁻⁸; Normoxia WT vs Hypoxia WT, p = 4.6 × 10⁻⁷. RVH: Normoxia K1488X vs Hypoxia K1488X, p = 7 × 10⁻⁸; Normoxia WT vs Hypoxia WT, p = 5 × 10⁻⁸. e Representative images of H&E-stained samples from the model mice. Scale bar = 100 µm. f Representative images of α-SMA (cyan), CD31(red), NR2F2 (yellow) and DAPI (blue) immunofluorescence staining of veins (with NR2F2 and DAPI co-staining), arteries and microwessels from WT and Eif2ak4^{K1488X/K1488X} mice under nomaxia and hypoxia. White scale bar = 50 µm. Yellow scale bar = 20 µm. White arrow: vessel. g Percent wall thickness of pulmonary arteries (n = 15 vessels of 7 mice/group), medial thickness of pulmonary veins (n = 15 vessels of 7 mice/group), and muscularization (%) of microvessels (n = 5 mice/group) in the model mice. Arteries: Normoxia K1488X vs Hypoxia K1488X, p = 6 × 10⁻⁸; Normoxia WT vs Hypoxia WT, p = 2 × 10⁻⁹. Veins: Normoxia K1488X vs Hypoxia K1488X, p = 6 × 10⁻⁹; Normoxia WT vs K1488X, p = 1 × 10⁻⁶; Hypoxia WT vs K1488X, p = 1 × 10⁻¹². Microvessels: Non WT vs Hypoxia WT, p = 9 × 10⁻¹¹; Full WT vs Hypoxia WT, p = 1 × 10⁻⁶. h Non-heme iron levels in the lungs of the mice were measured. n = 7 mice/group. i Heme iron levels in the lungs of the mice were measured. n = 7 mice/group. Normoxia K1488X vs Hypoxia K1488X, p = 2.9 × 10⁻⁵; Normoxia WT vs Hypoxia WT, p = 1.1 × 10⁻⁵. j Ferric iron deposits in the model mice were stained with Perls’ DAB. Scale bar = 50 µm. k MDA content in the lungs of the mice was measured. Normoxia WT vs K1488X, p = 5 × 10⁻⁵. l Representative images of 4-HNE staining in mice. Scale bar = 50 µm. The data are presented as the means ± s.e.m.; two-way ANOVA with Tukey’s multiple comparison test. Each dot represents an individual biological replicate, at least three independent experiments. P values are indicated in the figures. Source data are provided as a Source data file.

Fig. 3. Fer-1 reverses PVOD in Eif2ak4^{K1488X/K1488X} hypoxia-induced mice and MMC rats.

a Schematic diagram of the experimental design. Eif2ak4^{K1488X/K1488X mice} (K1488X) were subjected to hypoxic or normoxic conditions for 42 days and treated with vehicle or Ferrostatin-1 (Fer-1) from the start of the hypoxia treatment (prevention protocol) or beginning at 21 days (reversal protocol). b Assessment of TPVRI, RVSP and RVH in the mice (TPVRI male:n = 4,4,4,3 mice/group, female: n = 3,3,3,4 mice/group; RVSP male:n = 4,3,4,3 mice/group, female: n = 3,4,3,4 mice/group; RVH male:n = 4,4,4,3 mice/group, female: n = 3,3,3,4 mice/group). TPVRI: K1488X vs Hypoxia, p = 3.9 × 10⁻⁹; Hypoxia vs Prevention, p = 1.8 × 10⁻⁵; Hypoxia vs Reversal, p = 2 × 10⁻⁵. RVSP: K1488X vs Hypoxia, p = 6 × 10⁻⁸; Hypoxia vs Prevention, p = 9 × 10⁻⁶; Hypoxia vs Reversal, p = 2 × 10⁻⁵. RVH: K1488X vs Hypoxia, p = 9.7 × 10⁻⁷. c Representative images of H&E-stained lung sections from the mice described in a. Pulmonary veins (arrows), scale bar = 100 µm. The experiment was repeated independently five times with similar results. d Representative images of α-SMA staining in lung sections from the mice described in (a). Scale bar = 50 µm. e Percent wall thickness of pulmonary arteries (n = 15 vessels of 7 mice /group), medial thickness of pulmonary veins (n = 15 vessels of 7 mice/group), and muscularization (%) of microvessels (n = 6 mice/group) in the mice described in a. Arteries: K1488X vs Hypoxia, p = 5.8 × 10⁻⁷. Veins: K1488X vs Hypoxia, p = 7.5 × 10⁻¹²; Hypoxia vs Prevention, p = 1 × 10⁻¹⁰; Hypoxia vs Reversal, p = 1.7 × 10⁻⁹. Microvessels: Non K1488X vs Hypoxia, p = 2 × 10⁻¹¹; Hypoxia vs Prevention, p = 2 × 10⁻⁷; Hypoxia vs Reversal, p = 3.9 × 10⁻⁵. Full K1488X vs Hypoxia, p = 2 × 10⁻¹¹; Hypoxia vs Reversal, p = 9 × 10⁻¹⁰. f Schematic diagram of the experimental design. The rats were given 3 mg/kg mitomycin-C (MMC) (once per week for 2 weeks, i.p.) or vehicle for 35 days and Fer-1 or saline from the start of the MMC treatment (prevention protocol) or beginning at 21 days (reversal protocol). g Assessment of TPVRI, RVSP and RVH in the rats (n = 7 rats/group) described in f. TPVRI: Control vs MMC, p = 5.9 × 10⁻⁶; RVSP: Control vs MMC, p = 4 × 10⁻⁷; RVH: Control vs MMC, p = 5 × 10⁻⁵. h Representative images of H&E-stained lung sections from the rats described in (f). Pulmonary veins (arrows), scale bar = 100 µm. The experiment was repeated independently five times with similar results. i Representative images of α-SMA staining in the pulmonary arteries, veins and microvessels of the rats described in (f). Scale bar = 50 µm. j Surface open lumen/total area (%) of pulmonary arteries and veins (n = 15 vessels of 5 rats/group) and muscularization (%) of microvessels (n = 6 rats/group) in the rats described in (f). Arteries: Control vs MMC, p = 7 × 10⁻¹²; MMC vs Prevention, p = 4 × 10⁻⁹; MMC vs Reversal, p = 1 × 10⁻⁵. Veins: Control vs MMC, p = 8 × 10⁻¹²; MMC vs Prevention, p = 2 × 10⁻⁵; MMC vs Reversal, p = 1 × 10⁻⁶. Microvessels: Non Control vs MMC, p < 1 × 10⁻¹⁵; Mild MMC vs Reversal, p = 9.6 × 10⁻⁹; Moderate Control vs MMC, p < 1 × 10⁻¹⁵, MMC vs Reversal, p = 6 × 10⁻⁵; Severe Control vs MMC, p = 5.7 × 10⁻⁷. The data are presented as the means ± s.e.m.; one-way ANOVA with Tukey’s multiple comparison test. Each dot represents an individual biological replicate, at least three independent experiments. P values are indicated in the figures. Source data are provided as a Source data file.

Fig. 4. GCN2 deficiency enhances HMOX1 expression and promotes ferroptosis in macrophages.

a Volcano plot depicting the differentially expressed genes in alveolar macrophages from PVOD patients and Controls. The violin plot highlights HMOX1 expression, identified as significantly different. P values were determined via two-sided Wilcoxon rank-sum test with Bonferroni correction for multiple testing. p = 1.42e-686. b, c Western blot and quantification of HMOX1, FTH, FTL, and TFRC in Control and PVOD lung tissues. Data are presented as the means ± s.e.m. (n = 5 individuals); unpaired two-sided t test. d Representative images and quantification of MARCO(red) and HMOX1(green) immunofluorescence and DAPI (blue) staining of lung tissues from Control and PVOD patients. Scale bar = 20 µm. Data are presented as the means ± s.e.m. (n = 6 individuals); unpaired two-sided t test. HMOX1, p = 3.5 × 10⁻⁵. e, f Western blot of HMOX1, FTH, FTL, and TFRC in Control and MMC rat lungs. Data are presented as the means ± s.e.m. (n = 5 rats/group); unpaired two-sided t test. FTH, p = 8.39 × 10⁻⁵. g Representative images and quantification of MARCO(red) and HMOX1(green) immunofluorescence and DAPI (blue) staining of lung tissues from Control and MMC rats. Scale bar = 20 µm. Data are presented as the means ± s.e.m. (n = 5 rats/group); unpaired two-sided t test. h, i WT and Eif2ak4^{K1488X/K1488X} mice were subjected to hypoxia for 6 weeks. Western blot of HMOX1, FTH, FTL, and TFRC in mouse lungs. Data are presented as the means ± s.e.m. (n = 7 mice/group); two-way ANOVA with Tukey’s multiple comparison test. HMOX1, Hypoxia WT vs K1488X, p = 1.7 × 10⁻⁵; Normoxia K1488X vs Hypoxia K1488X, p = 2.3 × 10⁻⁶. j, k Western blot of HMOX1, FTH, FTL and TFRC in WT or Eif2ak4^{K1488X/K1488X} BMDMs treated with (FAC) or without (Control) 300 µM FAC for 48 h. Data are presented as the means ± s.e.m. (n = 7 mice/group); two-way ANOVA with Tukey’s multiple comparison test. HMOX1, K1488X Control (Con) vs FAC, p = 5 × 10⁻⁶; FTH, WT Con vs FAC, p = 2 × 10⁻⁹, K1488X Con vs FAC, p = 9.8 × 10⁻¹²; FTL, WT Con vs FAC, p = 9.7 × 10⁻¹¹, K1488X Con vs FAC, p = 5 × 10⁻¹¹; TFRC, WT Con vs FAC, p = 2 × 10⁻⁶, K1488X Con vs FAC, p = 2.7 × 10⁻⁶. l Cell viability was measured in WT and Eif2ak4^{K1488X/K1488X} BMDMs treated with the indicated concentrations of FAC for 48 h and in normal control (NC) and GCN2^−/− (gRNA) HT1080 cells treated with the indicated concentrations of FAC for 48 h. Data are presented as the means ± s.e.m. (n = 6 replicates); unpaired two-sided t tests. BMDM, FAC 300 µM, p = 1.3 × 10⁻⁶, FAC 500 µM, p = 8.3 × 10⁻⁵; HT1080, FAC 100 µM, p = 3 × 10⁻⁵, FAC 200 µM, p = 1.1 × 10⁻⁵. m Lipid peroxidation in BMDMs treated with FAC for 12 h, assayed using BODIPY 581/591 dye. Representative microscopy images are shown. Scale bar = 50 µm. The data are presented as the means ± s.e.m. (n = 15 microscope fields from 5 mice); two-way ANOVA with Tukey’s multiple comparison test. WT Con vs FAC, p = 9 × 10⁻⁸, K1488X Con vs FAC, p = 1 × 10⁻¹², FAC WT vs K1488X, p = 1 × 10⁻⁸. Each dot represents an individual biological replicate, at least three independent experiments. P values are indicated in the figures. Source data are provided as a Source data file.

Fig. 5. GCN2 deficiency enhances arterial markers expression in VECs and smooth muscle cell recruitment in the presence of iron.

a Violin plots highlighting the differentially expressed genes associated with arterial and venous identity in venous endothelial cells from human scRNA-seq data. P values were determined via two-sided Wilcoxon rank-sum test and marked on each panel. b Western blot of NRP1, EFNB2, p-ERK in NC or GCN2^−/− (gRNA) hUVECs treated with (FAC) or without (Control) 660 µM FAC for 72 h. Data are presented as the means ± s.e.m. (n = 4 replicates); two-way ANOVA with Tukey’s multiple comparison test. NRP1, NC Control (Con) vs gRNA FAC, p = 2 × 10⁻⁶. c Schematic of the Transwell coculture model. NC or GCN2^−/− hUVECs were seeded in the bottom compartment and treated with/without 660 µM FAC for 72 h, with SMCs seeded on top for 24 h. An 8-µm pore membrane facilitates migration; a 0.4-µm membrane tests proliferation. d Images and quantification of DAPI-stained SMCs cocultured with NC or GCN2^−/− (gRNA) hUVECs in an 8-μm pore size chamber in FAC-treated medium, indicating the migration ability of SMCs. Scale bar = 100 µm. Data are presented as the means ± s.e.m. (n = 15 microscope fields from three independent experiments); two-way ANOVA with Tukey’s multiple comparison test. Con NC vs Con gRNA, p = 3 × 10⁻¹⁰, gRNA Con vs FAC, p = 1 × 10⁻⁵. e Images and quantification of DAPI-stained SMCs cocultured with NC or GCN2^−/− hUVECs in a 0.4-μm pore size chamber in FAC-treated medium, revealing the proliferative ability of the SMCs. Scale bar = 100 µm. The data are presented as the means;± s.e.m. (n = 15 microscope fields from three independent experiments); two-way ANOVA with Tukey’s multiple comparison test. NC Con vs FAC, p = 5 × 10⁻⁵, gRNA Con vs FAC, p = 6 × 10⁻⁸. f Ratio of EdU-positive SMCs after treatment ± 660 µM FAC for 24 h. Data are presented as the means ± s.e.m. (n = 6 replicates); unpaired two-sided t test. g Representative images of hUVECs (green) cocultured with SMCs (PKH26, red) for 4 days. Scale bar = 100 µm. Quantification of total network area, total cord length, and branch points. Data are presented as the means ± s.e.m. (n = 8 microscope fields from four independent experiments); two-way ANOVA with Tukey’s multiple comparison test. Number of nodes, NC FAC vs gRNA FAC, p = 2 × 10⁻⁵, gRNA Con vs FAC, p = 2 × 10⁻⁶; Number of junctions, gRNA Con vs FAC, p = 4.5 × 10⁻⁵; Total length, gRNA Con vs FAC, p = 8 × 10⁻⁸; Total branching length, gRNA Con vs FAC, p = 3 × 10⁻⁷. Each dot represents an individual biological replicate, at least three independent experiments. P values are indicated in the figures. Source data are provided as a Source data file.

Fig. 6. Spatial transcriptomics reveals enhanced venous arterialization and ETS1-mediated gene regulation in PVOD lung vessels.

a H&E-stained images of 10X Visium spatial transcriptomics sections from Control (n = 2 individuals) and PVOD (n = 1 individual) lung tissues. The two control samples represent the upper and lower halves of the same slide (stitched together). Scale bar = 2 mm. b Spatial mapping of tissue region clusters (Alveoli, Bronchi, Vessel, Unspecified) on spatial transcriptomics spots from Control (left) and PVOD (right) lung samples. c Violin plot showing HMOX1 expression levels across tissue regions in Control and PVOD lung samples. P values were determined via two-sided Wilcoxon rank-sum test. d Violin plots depicting expression of arterial endothelial markers (KDR, CXCL12) and venous related marker (ACKR1) in vessel regions comparing Control and PVOD groups. P values were determined via two-sided Wilcoxon rank-sum test. e Violin plots showing arterial and venous endothelial gene set scores in vessel regions of Control versus PVOD samples. P values were determined via two-sided Wilcoxon rank-sum test. f Volcano plot of differentially expressed genes in vessel regions between PVOD and Control groups. P values were determined via two-sided Wilcoxon rank-sum test with Benjamini–Hochberg correction for multiple testing. Significance thresholds were set at |log2 fold change| > 0.5 and adjusted p-value < 0.05. The top 5 upregulated and top 5 downregulated genes are annotated in the plot. GO biological processes (g) and KEGG pathways (h) significantly enriched (FDR < 0.05) from upregulated genes in PVOD vessel regions. P values were calculated using the hypergeometric test with Benjamini–Hochberg correction for multiple testing. Ten relevant terms associated with pulmonary vascular disease are shown, ranked by combined score. Dot size represents the percentage of genes in the gene set, and dot color indicates –log10(FDR). i Volcano plot of transcription factor activity differences (z-score normalized AUC scores) between Control and PVOD vessel regions analyzed by the limma method. j Violin plot showing ETS1 AUC scores in Control and PVOD vessel regions. k Violin plot of ETS1 expression in venous endothelial cells from scRNA-seq data comparing Control and PVOD groups. P values were determined via two-sided Wilcoxon rank-sum test. l ETS1 transcription factor binding motif (metacluster_183.1) obtained from the cisTarget motif collection (v10nr_clust). m Spatial distribution of cell type proportions (EC_arterial, EC_venous, Macrophages, Muscular cells, Fibroblasts) inferred by RCTD deconvolution. Color intensity corresponds to the relative abundance of each cell type, with darker colors indicating higher proportions. n Heatmaps showing Pearson correlation between RCTD cell type scores and cell death pathway gene set scores in Alveoli (top) and Vessel (bottom) region of the PVOD lung sample (*P < 0.05, **P < 0.01, ***P < 0.001). P values are indicated in the figures. Source data are provided as a Source data file.

Fig. 7. Enhanced NRP1/KDR/EFNB2 expression as a marker set of venous arterialization of the VECs in PVOD patient lung.

a Representative images of NRP1/KDR/EFNB2 (red), NR2F2 (yellow), CD31 (green), and α-SMA (cyan) immunofluorescence and DAPI (blue) staining of lung tissues from Control and PVOD patients, with the boxed region magnified. Scale bar = 50 µm. Enlarged view scale bar = 10 µm. The data are representative images. Control, n = 7 individuals; PVOD n = 9 individuals with similar results. b qRT-PCR analyzed the expression levels of representative markers at various stages of VECs differentiation process. Data are presented as the means ± s.e.m. (n = 4 replicates); two-way ANOVA with Tukey’s multiple comparison test. CD31 VEC, p = 6 × 10⁻¹²; NR2F2 VEC, p = 4 × 10⁻¹⁰. c qRT-PCR analyzed the differential expression levels of candidate genes in Control-iPSC VEC (Con) and PVOD-iPSC VEC (PVOD). Data are the mean ± s.e.m. (n = 6), unpaired two-sided t-test. EFNB2 Con vs PVOD, p = 4.7 × 10⁻⁴; KDR Con vs PVOD, p = 8 × 10⁻⁶. d Immunofluorescence analysis of a venous marker (NR2F2) with DAPI counterstaining. Scale bar = 100 µm. The experiment was repeated independently four times with similar results. e Western blot showing the expression levels of NRP1, EFNB2, p-ERK, Total-ERK in Control-iPSC VEC (Con) or PVOD iPSC-VEC (PVOD). The experiment was repeated independently three times with similar results. f Immunofluorescence analysis of p-ERK with DAPI counterstaining in Control-iPSC VEC (Con) or PVOD iPSC-VEC (PVOD). Scale bar = 50 µm. Data are the mean ± s.e.m. (n = 6 replicates), unpaired two-sided t-test. Each dot represents an individual biological replicate, at least three independent experiments. P values are indicated in the figures. Source data are provided as a Source data file.

Fig. 8. Schematic summary.

Ferroptosis in lung macrophages with GCN2 deficiency upregulates HMOX-1, which releases iron to induce NRP1/KDR/EFNB2 expression through ETS1 transcription activation in VECs. Activated ERK drives venous arterialization and contributes to related vascular diseases. The image was drawn by Figdraw, with an authorization ID of SWURR94551.