Abstract

   Atherosclerosis (AS) is a major contributor to morbidity and mortality
   worldwide. However, the molecular mechanisms and mediator molecules
   involved remain largely unknown. Copper, which plays an essential role
   in cardiovascular disease, has been suggested as a potential risk
   factor. Copper homeostasis is closely related to the occurrence and
   development of AS. Recently, a new cell death pathway called
   cuproptosis has been discovered, which is driven by intracellular
   copper excess. However, no previous studies have reported a
   relationship between cuproptosis and AS. In this study, we integrated
   bulk and single-cell sequencing data to screen and identify key
   cuproptosis-related genes in AS. We used correlation analysis,
   enrichment analysis, random forest, and other bioinformatics methods to
   reveal their relationships. Our findings report, for the first time,
   the involvement of cuproptosis-related genes FDX1, SLC31A1, and GLS in
   atherogenesis. FDX1 and SLC31A1 were upregulated, while GLS was
   downregulated in atherosclerotic plaque. Receiver operating
   characteristic curves demonstrate their potential diagnostic value for
   AS. Additionally, we confirm that GLS is mainly expressed in vascular
   smooth muscle cells, and SLC31A1 is mainly localized in macrophages of
   atherosclerotic lesions in experiments. These findings shed light on
   the cuproptosis landscape and potential diagnostic biomarkers for AS,
   providing further evidence about the vital role of cuproptosis in
   atherosclerosis progression.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s43556-023-00131-5.

   Keywords: Atherosclerosis, Atherosclerotic cardiovascular disease,
   Cuproptosis, Bioinformatics

Introduction

   Over the past 20 years, cardiovascular disease (CVD) has gradually
   eroded and burdened healthcare systems, and is the leading cause of
   morbidity and mortality worldwide [[50]1]. AS is the primary cause and
   pathologic foundation of CVD [[51]2], and the pathophysiological
   mechanisms responsible for AS are multifactorial, including
   inflammation, oxidative stress, dysregulated lipid metabolism, and
   transition metal ion metabolism [[52]3–[53]6]. Despite these promising
   studies, however, CVD pathogenesis is still poorly understood. Research
   is needed to identify new pathways and reveal the underlying mechanism
   in AS that could be targeted therapeutically to reduce the global
   burden of AS.

   Copper is one of the essential trace elements involved in various
   important biological processes in the human body. In addition, studies
   show that copper ions are closely related to AS development. Elevated
   levels of copper ions in the serum and plaques of atherosclerotic
   patients have been demonstrated [[54]3]. The data from the Giovanni
   Tarantino group shows that copper bioavailability is also a central
   factor in early atherosclerosis, as evidenced by the altered serum
   copper bioavailability predicting early atherosclerosis as the main
   cardiovascular risk in obese patients with a low prevalence of
   comorbidities [[55]7]. Excess copper can be deposited in lysosomes,
   leading to cell death in human umbilical vein endothelial cells
   (HUVECs). Treatment with the copper ion chelator tetrathiomolybdate
   (TTM) alleviates this effect [[56]8, [57]9]. Wei et al. found that the
   use of copper chelating agents and copper ion carrier inhibitors can
   reduce the development of atherosclerotic lesions in mice [[58]10], and
   Völker et al. proposed that copper ions also can induce inflammatory
   reactions and promote AS-like neointima formation in rats [[59]11].
   Moreover, copper ions have also been shown to play important roles in
   the oxidative modification of low-density lipoprotein (LDL) and vessel
   wall homeostasis [[60]12, [61]13]. Therefore, an imbalance in copper
   homeostasis can lead to the progression of AS [[62]14, [63]15].

   Recently, a new cell death pathway called cuproptosis has been
   discovered [[64]16]. Cuproptosis relies on mitochondrial respiration
   and involves the direct binding of copper to the lipoacylated
   components of the tricarboxylic acid (TCA) cycle, resulting in toxic
   protein stress and ultimately inducing cell death [[65]16]. Elevated
   levels of copper ions and copper derivatives are essential for
   cuproptosis; however, a relationship between cuproptosis and AS has not
   been reported.

   In the present study, we downloaded atherosclerotic disease datasets
   from GEO, screened out the differential cuproptosis-related genes
   (DE-CRGs), and investigated the potential mechanisms by bioinformatics
   analyses. We provided evidence that suggests cuproptosis may be
   involved in AS, located the potential target genes in vascular smooth
   muscle cells and macrophages of atherosclerotic lesions, and evaluated
   their diagnostic value.

Results

Identification and correlation analysis of candidate DE-CRGs

   The overall design of this study is displayed in Fig. [66]1a. To
   investigate the differential expression of CRGs in atherosclerotic
   plaques, we curated a catalog of 16 genes (CDKN2A, FDX1, DLD, DLAT,
   LIAS, GLS, LIPT1, MTF1, PDHA1, PDHB, ATP7A, ATP7B, DBT, GCSH, DLST, and
   SLC31A1) that function closely with cuproptosis from the published
   literature [[67]16]. Differential expression analysis (AS vs. normal)
   identified 10,226 differentially expressed genes (DEGs) with thresholds
   of |log2FC|≥ 1 and adjusted p < 0.05 in [68]GSE97210. By combining the
   results of 16 CRGs and DEGs in the [69]GSE97210 datasets (normal vs. AS
   plaque, n = 3), we were able to obtain nine DE-CRGs (Fig. [70]2a).
   Compared with the normal control, five genes were downregulated, and
   four genes were upregulated in atherosclerotic plaques, as marked in
   the volcano plot (Fig. [71]2b). The expression of the nine DE-CRGs was
   visualized in a heat map (Fig. [72]2c). Pearson’s correlation
   coefficient was used for the analysis, which showed a significant
   correlation between these nine genes (Fig. [73]2d). To summarize, AS
   patients showed significant dysregulation of CRGs in arterial tissues
   when compared to normal individuals.

Fig. 1.

   [74]Fig. 1
   [75]Open in a new tab

   a The overall design of this study

Fig. 2.

   [76]Fig. 2
   [77]Open in a new tab

   Analysis of differentially expressed genes (DEGs) related to
   cuproptosis and functional enrichment analysis. a Venn diagram showing
   the cuproptosis-related genes in the [78]GSE97210 datasets. b The
   identified cuproptosis-related DGEs displayed in a volcano plot,
   upregulated (red) and downregulated (blue). c Heat map of the
   cuproptosis-related genes in the [79]GSE97210 datasets. d Correlation
   analysis of candidate cuproptosis-related genes in [80]GSE97210
   datasets

Functional enrichment analysis and PPI network construction with candidate
DE-CRGs

   Based on GO and KEGG pathway analyses, we explored the potential
   biological functions and pathways of the nine DE-CRGs. For GO analysis,
   molecular function (MF) results showed that dehydrogenase, transferase,
   and copper ion transmembrane transporter activities were significantly
   demonstrated in the DE-CRGs (Fig. [81]3a). Biological pathway (BP)
   analysis indicated enriched primary terms relevant to pyruvate-related
   enzyme synthesis processes, copper ion import, and protein lipidation
   (Fig. [82]3b). In the cellular component (CC) study, the screened
   DE-CRGs were significantly enriched in the mitochondrial matrix,
   pyruvate dehydrogenase complex, and late endosome (Fig. [83]3c).
   Moreover, investigation of the KEGG pathway analysis primarily
   suggested that these genes were involved in metabolic pathways, carbon
   metabolism, and the TCA cycle (Fig. [84]3d). Subsequently, to explore
   the interaction between the nine DE-CRGs, the genes were submitted to
   STING to search for protein–protein interactions and construct a PPI
   network. The results showed a total of nine nodes and twelve edges, and
   the proteins were highly connected in the PPI network (Fig. [85]3e).
   Taken together, these findings demonstrate that candidate DE-CRGs may
   play a significant role in the progression of AS.

Fig. 3.

   [86]Fig. 3
   [87]Open in a new tab

   Functional enrichment analysis and protein–protein interaction (PPI)
   network. a-c Gene Ontology (GO) enrichment analysis of DEGs of
   [88]GSE97210 datasets. a Biological process (BP); b Cellular component
   (CC); c Molecular function (MF). d Kyoto Encyclopedia of Genes and
   Genomes (KEGG) pathway enrichment analysis of DEGs of [89]GSE97210
   datasets. e Protein–protein-interaction (PPI) network of the
   cuproptosis-related genes in [90]GSE97210 datasets

Screening of Hub DE-CRGs

   Given the limitations of single data sources, we introduced another
   dataset [91]GSE43292 (normal vs. AS plaque, n = 32), as a validation
   set to compare the gene expression levels of the nine DE-CRGs in two
   datasets. We used the same expression, significance trend (P < 0.001),
   and Random Forest (RF) algorithm to identify the key genes. The results
   revealed that 4 cuproptosis-related genes, GLS, FDX1, SLC31A1, and
   GCSH, have the same expression and significance trend (P < 0.001) in
   plaque vs. normal control in the [92]GSE97210 and [93]GSE43292. Among
   them, FDX1 and SLC31A1 were upregulated, and GLS and GCSH were
   downregulated (Fig. [94]4a-b). The RF model was used to conduct deep
   learning on the sample data of [95]GSE97210 and [96]GSE43292.
   Considering the score distributions of RF and the gene functions, FDX1,
   SLC31A1, and GLS were identified as hub genes (Fig. [97]4c). To further
   validate the accuracy of our previous analysis, we selected three
   stages of atherosclerotic datasets: [98]GSE132571 (normal vs.
   atherosclerotic endothelial, n = 6), [99]GSE28829 (early vs. advanced
   plaques, n = 10), and [100]GSE120521 (stable vs. unstable plaques,
   n = 3). Comparing gene expression levels, GLS remained significantly
   downregulated across the three datasets, and FDX1 and SLC31A1 were
   consistently upregulated in advanced and ruptured atherosclerotic
   plaques (Fig. [101]4d-e). This suggested that GLS may influence all
   stages of AS, while a higher level of FDX1 and SLC31A1 expression could
   predict more severe kinds of atherosclerotic plaques. Together, FDX1,
   GLS, and SLC31A1 were identified as the hub DE-CRGs in cuproptosis and
   AS.

Fig. 4.

   [102]Fig. 4
   [103]Open in a new tab

   Screening of Hub differentially expressed cuproptosis-related genes
   (DE-CRGs). a-b Gene expression levels of nine DE-CRGs were compared in
   AS (Plaque) and control (Normal) groups with the Student’s t-test in
   [104]GSE97210 datasets (a) and [105]GSE43292 datasets (b). Significance
   markers: ****P < 0.0001, ***P < 0.001, **P < 0.01 and *P < 0.05. c
   Importance ranking of genes in a random forest (RF) model based on
   datasets [106]GSE43292. d Volcano plots displaying the expression of
   Hub genes in [107]GSE132651 datasets, [108]GSE28829 datasets, and
   [109]GSE120521 datasets. Upregulated (red) and downregulated (blue) or
   vice versa. e Gene expression level of the three Hub DE-CRGs was
   compared in normal coronary endothelial (control) and atherosclerotic
   coronary endothelial cells, early and advanced plaques, and stable and
   unstable plaques with Student’s t-test in [110]GSE132651 datasets,
   [111]GSE28829 datasets, and [112]GSE120521 datasets. Significance
   markers: ****P < 0.0001, ***P < 0.001, **P < 0.01 and *P < 0.05 vs.
   control group

The diagnostic value of FDX1, GLS, and SLC31A1 in AS

   ROC analysis and the area under the ROC curve (AUC) were performed in
   different datasets to evaluate the diagnostic value of three hub
   DE-CRGs in AS. ROC analysis demonstrated that FDX1, GLS, and SLC31A1
   accurately predicted AS in [113]GSE43292 (Fig. [114]5a). Furthermore,
   the overall inspection of the three ROC curves showed the same trend as
   the Student’s t-test. In [115]GSE132561, [116]GSE28829, and
   [117]GSE120521, the ROC curves of FDX1, GLS, and SLC31A1 also verified
   that both had a high value for classification of the atherosclerosis
   stage (Fig. [118]5b-d). The AUCs for FDX1, SLC31A1, and GLS with the
   progression of atherosclerotic plaques showed significant performance
   prediction for samples (Fig. [119]5a-d). Thus, those hub genes had a
   potential diagnostic value for AS.

Fig. 5.

   [120]Fig. 5
   [121]Open in a new tab

   The diagnostic value of 3 Hub DE-CRGs in AS. a-d The diagnostic power
   of FDX1, GLS, and SLC31AL in atherosclerosis by ROC curve and
   intergroup overall inspection of ROC analysis in datasets [122]GSE43292
   (a), [123]GSE132651 (b), [124]GSE28829 (c), and [125]GSE120521 (d)

Differential expression of FDX1, GLS, and SLC31A1 in human arterial tissues
between atherosclerotic patients and healthy controls

   We examined FDX1, GLS, and SLC31A1 expression by immunohistochemical
   (IHC) staining. A total of 16 coronary artery sample sections were
   selected in each group, including eight from healthy subjects and eight
   atherosclerotic patients. Compared with the control group, FDX1
   expression was increased (Fig. [126]6a, P < 0.01). The expression of
   GLS showed a significant decrease (Fig. [127]6b, P < 0.01). SLC31A1
   showed upregulation in atherosclerotic plaques (Fig. [128]6c,
   P < 0.01). A negative control of IHC was presented in Supplementary
   Fig. [129]1A-C. In conclusion, the results of IHC were consistent with
   the analysis of the datasets, demonstrating that FDX1, GLS, and SLC31A1
   were expressed significantly differently in AS plaques compared to
   normal tissues.

Fig. 6.

   [130]Fig. 6
   [131]Open in a new tab

   Immunohistochemistry for FDX1, GLS and SLC31A1. a-c Immunohistochemical
   staining of FDX1 (a), GLS (b), and SLC31A1 (c) in human normal coronary
   arterial tissues sections (Control) and AS tissues (Plaque), and mean
   density of FDX1 + (% area), GLS + (% area), SLC31A1 + (% area). Scale
   bar: 200 µm and 50 µm. ***P < 0.001, **P < 0.01 and *P < 0.05 vs.
   control group. n = 8

Verification of cellular localization of FDX1, GLS, and SLC31A1 in
atherosclerotic plaque

   Identification of cell clusters in AS samples and tracking the cellular
   localization of screened hub genes proceeded as follows. First, we
   utilized the Seurat pipeline for single-cell analysis of AS sample in
   dataset [132]GSE155512, which was performed on the carotid arteries of
   atherosclerotic patients. After data screening, Seurat object
   construction, quality control, dimensionality reduction, clustering,
   and cell annotation, the carotid intimal component of atherosclerotic
   patients were classified into seven large clusters of cells. We
   included vascular smooth muscle cells (VSMCs), mesenchymal stem
   cell-derived chondrocytes (MSCDCs), human umbilical vein endothelial
   cells (HUVECs), lymphatic endothelial cells, monocyte-derived
   macrophage cells, CD4 + T cells, and common myeloid progenitor (CMP) in
   the datasets (Fig. [133]7a). Next, we observed the clustering locations
   and average expression of the three hub genes on the T-SNE
   classification map and found that GLS was expressed in multiple cell
   clusters but concentrated in regions of VSMC distribution. SLC31A1 was
   more clearly expressed in the regions distributed by macrophages, while
   FDX1 was uniformly distributed in different atherosclerotic cell lines
   (Fig. [134]7b-e). Using double immunostaining in human plaques, we
   observed that the expression of smooth muscle α-actin (SM α-actin) in
   control and lesions was colocalized with GLS staining, and SLC31A1 also
   coincided with the fluorescence of macrophage localization marker CD68
   (Fig. [135]8a-b). This result is consistent with the above
   bioinformatics inference.

Fig. 7.

   [136]Fig. 7
   [137]Open in a new tab

   Cell localization and expression of FDX1, GLS, and SLC31A1 in AS
   single-cell data samples. a Classification of cell populations and
   location of specific genes in datasets [138]GSE155514. b-d Dot plots
   (in red) showing the expression of three hub genes on the t-SNE map:
   FDX1 (b), GLS (c), and SLC31A1 (d). e Dot plot showing the expression
   of three hub genes in each cluster. Deeper and larger dots indicate
   higher expression levels

Fig. 8.

   [139]Fig. 8
   [140]Open in a new tab

   Verification of cellular localization of GLS, and SLC31A1 in
   atherosclerotic plaque. a-b Immunofluorescence staining of GLS (a) and
   SLC31A1 (b) in sections of normal human coronary arterial tissues
   (Control) and AS tissues (Plaque). GLS and SLC31A1 are shown in red,
   while SM α-actin and CD68 are shown in green. The black box indicates
   the 40 × fluorescence shooting area. White arrows point to colocalized
   cells. Scale bars are 200 µm and 75 µm. IgG was used as a negative
   control. c-d Western blot analysis of GLS protein expression in human
   vascular smooth muscle cells (HVSMCs) and mice primary vascular smooth
   muscle cells (MVSMCs), and SLC31A1 protein expression in THP-1 and
   RAW264.7 cells. The data were reported as mean ± SD (n = 4).
   ***P < 0.001, **P < 0.01 and *P < 0.05 vs. control group. n = 4

   For further support, we measured GLS and SLC31A1 protein expression
   in vitro. In line with these findings, compared to the control group,
   the expression of GLS was significantly decreased in the ox-LDL-treated
   human VSMCs (HVSMC) and mouse primary VSMCs (MVSMCs) (Fig. [141]8c).
   SLC31A1 had higher expression in the ox-LDL-treated human and mouse
   macrophages (Fig. [142]8d). These results indicate that FDX1, SLC31A1,
   and GLS are critical factors in AS, and cuproptosis may contribute to
   the development of AS.

Discussion

   Cuproptosis is defined as a nonapoptotic and copper-dependent cell
   death pathway. Copper can bind to lipoylated components of the TCA
   cycle, which leads to proteotoxic stress and cell death [[143]16].
   Copper plays an essential role in cardiovascular disease and has been
   suggested as an influence risk factor [[144]17, [145]18]. Copper levels
   are known to be elevated in atherosclerotic lesions [[146]3]. Indeed,
   copper was associated with an increase in the ten-year coronary risk
   and an increased risk of mortality from cardiovascular causes
   [[147]17–[148]19]. Copper deficiency can induce cardiovascular
   diseases, while copper excess can mediate various types of cell death,
   such as autophagy, apoptosis, cuproptosis, pyroptosis, and cardiac
   hypertrophy and fibrosis. Clarifying the correlation between
   cuproptosis and AS may provide novel biomarkers and ideas for diagnosis
   and therapy.

   FDX1, an iron-sulfur protein, participates in a wider spectrum of
   metabolic processes by the TCA cycle, including iron-sulfur cluster
   biogenesis, bile acid synthesis, vitamin D synthesis, and
   steroidogenesis [[149]20, [150]21]. Recent studies point out that FDX1
   is the key regulator in cuproptosis, which can directly bind to copper
   and activate the proteotoxic stress [[151]17, [152]18,
   [153]20–[154]25]. Indeed, genome-wide association studies identified
   FDX1 as a susceptibility loci in the coronary artery disease [[155]26].
   Our results show it is similarly elevated in atherosclerotic lesions
   and has potential diagnostic value for AS. Single-cell sequencing
   suggested that FDX1 has a widespread distribution in different
   atherosclerotic cell populations. These results indicate that FDX1 may
   be one of the critical target genes that regulate cuproptosis in AS.

   One of the crucial mechanisms of cuproptosis is intracellular copper
   accumulation. The copper transporters SLC31A1 to input copper from
   circulation and maintain copper homeostasis of most cells [[156]27].
   Nadina Stadler et al. detected elevated levels of total copper in
   carotid lesions [[157]3]. Here, we detected upregulated SLC31A1 in
   human atherosclerotic plaques. These results may suggest that excess
   copper in the plaque was mediated by SLC31A1. Extensive
   neovascularization is a classical histological feature of an unstable
   plaque [[158]28, [159]29]. Das et al. demonstrated that SLC31A1 could
   sense reactive oxygen species (ROS) to induce angiogenesis [[160]30].
   In AS, macrophage accumulation is directly linked to the
   destabilization and rupture of plaque, causing acute atherothrombotic
   events [[161]31]. Our study shows that SLC31A1 was specifically
   upregulated in advanced plaque or unstable plaque, which are rich in
   macrophages. Furthermore, we also found ox-LDL-induced SLC31A1
   upregulation in macrophages. Taken together, it would be reasonable to
   believe that SLC31A1 could induce macrophage cuproptosis to
   promote plaque instability.

   GLS is the enzyme that catabolizes glutamine to glutamate and ammonia.
   GLS-driven catabolism of glutamine serves not only for energy
   production but also for the maintenance of the redox state. GLS breaks
   down glutamate production of glutamine, which in turn leads to
   increased synthesis of intracellular glutathione (GSH) to resist
   ROS-induced damage and protect cells from copper toxicity
   [[162]32–[163]35]. Indeed, the current study showed the downregulation
   of GLS transcripts in human plaque tissue compared with normal vessels,
   and double-fluorescence labeling confirmed that GLS was localized in
   VSMCs. These results suggest that excessive copper destroys the
   GLS-driven oxidation–reduction system of VSMCs, which aggravates the
   toxic damage of copper, thus promoting AS.

   In summary, we have first identified three CRGs, FDX1, SLC31A1, and
   GLS, in the development of AS. Our results also suggest FDX1, SLC31A1,
   and GLS as potential diagnostic biomarkers for AS, providing more
   insights into the vital role of cuproptosis in AS. However, the studies
   on cuproptosis and AS are still in their infancy. More extensive
   studies are still needed to determine the detailed molecular mechanism
   of cuproptosis, thus providing more evidence for cuproptosis in the
   prevention and treatment of AS.

Materials and methods

Data collection and acquisition of cuproptosis-related genes (CRGs)

   The bulk sequencing data and single-cell transcriptome data were
   collected from the GEO database with series numbers [164]GSE97210,
   [165]GSE43292, [166]GSE120521, GSE 132,651, [167]GSE28829, and
   [168]GSE155512. Patient information from the GEO datasets used is
   available on the GEO website ([169]https://www.ncbi.nlm.nih.gov/geo/).
   A total of 16 CRGs, retrieved from the published articles, were
   obtained for subsequent analyses [[170]16].

Identification of differentially expressed CRGs (DE-DRGs)

   The “limma” and “edgeR” packageswere used to calibrate the microarray
   data and identify the differentially expressed genes (DEGs) between the
   atherosclerotic plaques and normal arterial intimae in [171]GSE97210
   datasets [[172]36, [173]37], which characterized the mRNA profiles in
   advanced AS or normal intima tissues. The mRNAs that met the defined
   criteria were considered as DEGs (adj.pval < 0.05, |logFC|> 1). Then,
   the intersecting genes between DEGs were defined as the DE-CRGs. By
   using the OmicStudio tools at [174]https://www.omicstudio.cn/tool
   [[175]38], the DEGs were displayed in volcano plots, and the DE-CRGs
   were visualized in a heatmap. The number of CRGs was shown in a Venn
   diagram. Correlation analysis of the differential expression of CRGs in
   [176]GSE97210 datasets was based on the “corrplot” package [[177]39].

Functional annotation, pathway enrichment, and construction of
Protein–Protein Interaction (PPI) network

   Gene Ontology (GO) biological process and Kyoto Encyclopedia of Genes
   and Genomes (KEGG) annotation were performed by the DAVID website
   [[178]40], and the “ggplot2” package was used to render the results of
   the GO biological process and KEGG pathway in the bar plots. The STRING
   database was employed to analyze the interactions of the DE-CRGs and
   then used to construct and visualize the PPI network.

Screening cuproptosis-related candidate biomarkers

   The expression levels of DE-CRGs in the control group and AS group were
   analyzed by Student’s t-test in datasets [179]GSE97210 and
   [180]GSE43292, with a P value less than 0.001 considered to be
   significant. To evaluate the diagnostic value of DE-CRGs for AS, a
   Random Forest (RF) classification model was implemented with the
   package “randomForest”. Thereafter, three Hub genes with significant
   expression of the same trend in the two datasets and with high
   confidence scores in the RF model were selected for further analysis.
   Receiver operating characteristic (ROC) curves of three Hub genes were
   plotted, and areas under the ROC curve (AUC) were calculated using the
   “pROC” package [[181]41]. The three microarray datasets of
   atherosclerotic plaque ([182]GSE132651: normal vs. atherosclerotic
   endothelial; [183]GSE28829: early vs. advanced plaques; and
   [184]GSE120521: stable vs. unstable plaques) were then used to verify
   the expressions and diagnostic value of the Hub genes.

Cluster analysis for single-cell transcriptomic datasets

   The data matrix of sample [185]GSM4705589 from [186]GSE155512 datasets
   was transformed into a “Seurat object” using the package “Seurat”
   [[187]42]. The quality was controlled by sequencing gene count and
   expression and mitochondrial gene percentage. The "NormalizeData"
   function was applied to standardize the expression matrix, and the
   "RunPCA" function was used for PCA dimensionality reduction analysis.
   Inputting the first three principal components, we subsequently
   performed density clustering to identify the composition of vascular
   endothelial and atherosclerotic plaques and visualized it through
   t-distributed statistical neighbor embedding (t-SNE). Moreover, the
   package of “SingleR” was used for the cell-type annotation [[188]43].

Histological, immunohistochemical, and immunofluorescence analyses of
arterial tissues

   Autopsy specimens were collected from accidental death cases and from
   individuals with varying degrees of AS lesions in the aorta, and were
   obtained from the University of South China Forensic Identification
   Center. Additional details are provided in Supplementary Table [189]1.
   The research protocol was approved by the Medical Ethics Committee of
   the University of South China (Approval number: USC202209HT17). Tissue
   samples were embedded in paraffin blocks for histology studies. For
   immunofluorescence, after deparaffinization and dehydration, endogenous
   peroxidase was blocked with 3% hydrogen peroxide, and non-specific
   binding was blocked with 10% fetal bovine serum (FBS). Primary antibody
   was added for incubation overnight at 4℃. The sections were incubated
   with polymerized horseradish peroxidase (HRP)-anti-rabbit IgG and
   visualized using DAB. The integrated optical density (IOD) and area of
   five fields were randomly selected from each sample to calculate the
   mean density, which was used as the final value of each sample. For
   co-localization of immunofluorescence, after deparaffinization and
   dehydration, the sections were blocked with 10% FBS, followed by
   incubation with primary and secondary antibodies to target gene and
   marker proteins. The nuclei were counterstained with
   4',6-Diamidino-2-phenylindole (DAPI), followed by fluorescence
   microscopy. The same concentration of isotype control IgG was used as a
   negative control. Antibodies against FDX1(ZEN GIO, RRID: [190]R23394),
   GLS (HUABIO, ET1611-5), SLC31A1 (Bioss, bs-100773R), α-SMA
   (Proteintech, 67,735–1-gl), and CD68 (Abcam, ab201340) were used.

Cell culture conditions

   Human aortic SMCs (CRL-1999) were purchased from BeNa Culture
   Collection (BNCC, Beijing, China) and authenticated by the supplier.
   Human THP-1 monocytes were purchased from ATCC. RAW264.7 macrophages
   were obtained from the Chinese Academy of Sciences. Mouse primary
   vascular smooth muscle cells (VSMCs) were extracted from freshly
   dissected thoracic aortic segments of male four- to six-week-old C57/B6
   mice: First, samples of thoracic aortic segments isolated from mice
   were initially digested in enzyme mixtures (Worthington, [191]LS003570,
   [192]LS004174, LS00227). Then, the adventitia and intima were removed,
   leaving only the media portion of the vessel. Finally, the remaining
   vascular tissue was cut up and digested again with mixed enzymes to
   isolate primary mouse VSMCs. Both human and mouse VSMCs were cultured
   in DMEM/F-12 medium (Gibco, C11330500BT) containing 10% FBS (Gibco) and
   1% penicillin/streptomycin (Biosharp, Hefei, China). at 37℃ with 5%
   CO[2]. Cells were starved for 8 h and then treated with oxidized
   low-density lipoprotein (ox-LDL) (50 mg/ml) for 24 h.

Western blot

   Protein expression was detected by Western blotting. Proteins extracted
   from cells were processed with RIPA lysis buffer, and concentrations
   were measured using the BCA Protein Assay Kit (Thermo Scientific,
   23,227). After ultracentrifugation at 4℃, the supernatant was added to
   the loading buffer and loaded onto 8% SDS-PAGE gels for electrophoresis
   and transfer topolyvinylidene difluoride (PVDF) membranes. The
   membranes were blocked with 5% skim milk, incubated with primary
   antibody (overnight at 4℃), then washed three times in Tris Buffered
   Saline with Tween (TBST), and incubated with secondary antibody (2 h at
   room temperature). The bands were visualized by Enhanced
   Chemiluminescence (ECL) substrate. Antibodies against GLS (HUABIO,
   ET1611-5), SLC31A1 (HUABIO, HA720005), and GAPDH (Proteintech,
   10,494–1-AP) were used.

Statistical analysis

   All obtained images were submitted to Image-Pro Plus software (version
   6.0) and Image J software for quantification, and statistical analyses
   were performed using GraphPad Prism software (version 9.0). All
   bioinformatics analyses used R software (version 4.2.0). The
   Shapiro–Wilk normality test was used to analyze the normality of the
   data distribution. Descriptive data are expressed as mean ± SD. The
   differences between the two groups were compared with a Student's
   t-test, and a one-way analysis of variance (ANOVA) was performed to
   compare multiple groups. Pearson correlation analysis was used for
   normally distributed variables, and Spearman correlation analysis was
   used for non-normally distributed variables. Results were considered
   statistically significant when P values were less than 0.05.

Supplementary Information

   [193]43556_2023_131_MOESM1_ESM.pdf^ (5MB, pdf)

   Additional file 1: Fig. 1. Negative control of Immunohistochemistry.
   A-C Immunohistochemical staining and corresponding negative control for
   FDX1 (A), GLS (B), and SLC31A1 (C) in human coronary arterial tissue
   sections.
   [194]43556_2023_131_MOESM2_ESM.pdf^ (151.5KB, pdf)

   Additional file 2: Table 1. Patient details. Arterial plaques were
   graded according to the Oxford grading system. All human arterial
   samples were obtained with informed consent, and the procedures were
   performed in accordance with institutional guidelines.

Acknowledgements