Sodium bisulfite (NaHSO₃, 243973-100G, Sigma) and sodium sulfite (Na₂SO₃, S4672-250G, Sigma) were freshly mixed in a 1:3 ratio (pH 7.4), which was used as the SO₂ donor. Tamoxifen (579002, Sigma, United States) was used to induce AAT1 knockout in the animal models. The primary reagents for cell culture included fetal bovine serum (FBS) (10099-141, Gibco, United States), penicillin-streptomycin solution (PS, 100×, 15140-122, ThermoScientific, United States), glutathione (LG, 100×, 25030081, Thermo Scientific, United States), and trypsin-containing EDTA (0.25%, 25200-056, ThermoScientific, United States). The primary antibodies used in this study were AAT1 (SAB2500473, Sigma, United States), β-actin (TA-09, Zsbio, China), Tp53 (2524, CST, United States), and p21^Cip/Waf (ab109199, Abcam, United States). Lentivirus-delivered Rat AAT1 shRNA (VB151210-10015, Cyagen, China) or scrambled shRNA (VB151214-10025, Cyagen, China) was used for the experiments. DAz-2 (13382, Cayman, United States) and phosphine-biotin (13581, Cayman, United States) were used for the protein sulphenylation assays.

A7r5 (rat VSMCs) was purchased from the National Experimental Cell Resource Sharing Platform (1101RAT-PUMC000219, China) and cultured at 37°C in a 5% CO₂ incubator using Dulbecco’s modified eagle’s medium (DMEM) containing 10% FBS, 1% PS, and 1% LG.

At the appropriate cell density, A7r5 was infected with lentivirus containing AAT1 shRNA or scramble shRNA at a multiplicity of infection (MOI) of 10 in the presence of polybrene (1 μg/mL). After 48 h, the culture medium was replaced with a puromycin (1 μg/mL)-containing medium to establish a stably transfected cell pool. After 14 days of screening, stable cell lines were used for subsequent experiments.

Stably transfected cell lines were divided into three groups: Scramble, AAT1 shRNA, and AAT1 shRNA + SO₂ groups. Cells stably transfected with scramble shRNA were used as controls. For the AAT1 shRNA + SO₂ group, cells were treated with 100 μM SO₂ donor for 4 days. For the scrambled and AAT1 shRNA groups, the cells were treated with equal volumes of saline. The AAT1 shRNA sequences (5′-3′) used in this study are as follows: TCG​AAT​TGG​AGC​TGA​CTT​CAG​ATA​CT. The scramble sequences (5′-3′) used in this study are as follows: GCA​CTA​CCA​GAG​CTA​ACT​CAG​ATA​GTA​CT. All experiments were performed in duplicate and repeated independently three times (n = 3).

This study protocol received approval from the Animal Experimentation and Welfare Committee at Peking University First Hospital, with the approval number J2022051. All procedures involving animals adhered to the ethical guidelines established by the Peking University First Hospital for the management of animal research.

The 2- and 24-month-old C57BL/6J male mice used in this study were purchased from Sibeifu (Beijing, China), and the aortic tissues were collected to detect protein levels of Tp53, p21^Cip/Waf and AAT1. The 2-month-old mice were served as controls. VSMC-AAT1-KO mice were constructed using Cre/LoxP technology as described in previous study (Huang et al., 2021). Briefly, SM22α-Cre transgenic (SM22α-CreERT2⁺) mice and AAT1 flox/flox (AAT1 f/f) mice were purchased from Guangzhou Cyagen (Guangzhou, China). SM22α-CreERT2⁺ and AAT1 f/f mice were hybridized to generate SM22α-CreERT2⁺ AAT1 f/f mice. VSMC-AAT1-KO mice were obtained from 8-week-old male SM22α-CreERT2⁺ AAT1 f/f mice by intraperitoneal injection of tamoxifen (2.5 mg/day, dissolved in ethanol/peanut oil [volume ratio 1:19]) for 5 consecutive days at 3-day intervals. Littermate AAT1 f/f mice were served as control. At the end, the aortic tissues of 5-month-old mice (AAT1 f/f and VSMC-AAT1-KO) were collected post euthanasia by excessive anesthesia. n = 6 independent biological samples.

Eight-week-old C57BL/6J mice were selected for aortic ring culture experiments. The murine aorta was bluntly detached and placed in sterile phosphate-buffered saline (PBS), with the adventitia of the aorta and connective tissues stripped. The aorta was cut into 3-mm rings and transferred to complete DMEM containing 10% FBS, 1% PS, and 1% LG and cultured at 37°C in a 5% CO₂ incubator. The following day, the rings were divided into control, D-gal and D-gal + SO₂ groups. In the control group, the rings were cultured in complete DMEM. In the D-gal group, the rings were cultured in complete DMEM containing 7.5 g/L D-gal. In the D-gal + SO₂ group, the rings were cultured in DMEM containing 7.5 g/L D-gal, and 100 μM SO₂ donor given twice daily. The medium was changed every 2 days. Aortic rings were collected after 7 days of induction for subsequent experiments. n = 6 independent biological samples.

Cells were fixed at room temperature for 15 min. After PBS washing, the working solution was applied and incubated overnight at 37°C without CO₂. The solution was then removed, followed by washing with PBS for three times. The cell observations were performed using a light microscope. A blue color indicates a positive signal. All experiments were performed in duplicate and repeated independently three times (n = 3).

A7r5 cells were lysed to extract total RNA with TRIzol reagent. cDNA synthesis was performed using 2 μg of total RNA and M-MLV reverse transcriptase. RT-qPCR was conducted using 2× Go Taq qPCR Master Mix on a 7500 Fast Real-Time PCR System (Applied Biosystems), with β-actin serving as the control. The primers used in this study were as follows: Rat-β-actin-F 5′-ACC​CGC​GAG​TAC​AAC​CTT​CTT-3′ and Rat-β-actin-R 5′-TAT​CGT​CAT​CCA​TGG​CGA​ACT-3′; Rat-IL1β-F 5′-CAC​CTC​TCA​AGC​AGA​GCA​CAG-3′ and Rat-IL1β-R 5′- GGG​TTC​CAT​GGT​GAA​GTC​AAC-3′; Rat-IL6-F 5′-CCC​ACC​CTC​CAA​CAA​AGA​TT-3′ and Rat-IL6-R 5′- GCT​CCA​GAG​CAG​AAT​GAG​CTA -3′. All experiments were performed in duplicate and repeated independently three times (n = 3).

Arotic tissues were formalin-fixed, embedded in paraffin, 4 μm sectioned, and dehydrated with xylene and graded ethanol concentrations. Antigen repair was performed using Tris-EDTA (pH 9.0) for 15 min. After permeabilization, the sections were incubated overnight with Tp53 (dilution 1:200), p21^Cip/Waf (dilution 1:200), γ-H2AX (dilution 1:200) antibodies. After incubation with the secondary antibody, the slices were sealed with a DAPI-containing sealer. Red (Tp53 and p21^Cip/Waf) or Green (γ-H2AX) fluorescence observed using Leica TCS SP8 MP FLIM indicated the expression of target protein. n = 6 independent biological samples.

Cells were fixed with 4% formaldehyde at room temperature for 15 min, permeabilized with pre-cooled methanol, and blocked with 5% bovine serum albumin. Subsequently, cells were incubated with primary antibody against IRF1 and Flag overnight at 4°C. After washing, cells were incubated with the corresponding secondary antibody and mounted with a DAPI-containing dye. Target protein expression was visualized as red fluorescence using a Leica TCS SP8 MP FLIM microscope. All experiments were performed in duplicate and repeated independently three times (n = 3).

A7r5 cells were lysed in a non-denaturing lysis buffer (Applygen, Beijing, China) containing 5 mM DAz-2 for 20 min. The supernatant was collected after centrifugation at 16,000 g for 4 min at 4°C and incubated with gentle shaking at 37°C for 2 h. DAz-2-labeled lysates were reacted with 250 μM phosphine-biotin (13581, Cayman, United States) at 37°C for 2 h. Neutravidin ultralink resin (53150, Thermo Fisher Scientific, United States) was added at a volume ratio of 1:10 and incubated overnight with shaking at 4°C. The suphenylated proteins enriched with resin were mixed with a 2× non-denaturing sample buffer and boiled at 100°C for 10 min. The supernatant was collected by centrifugation at 5,000 g for 10 min and subjected to Western blot. All experiments were performed in duplicate and repeated independently three times (n = 3).

Human IRF1 wild-type (WT) and C83S mutant plasmids were constructed (Transgene, China). When the A7r5 cells reached approximately 50%–60% confluence, cells were transfected with plasmids via Lipofectamine™ 3000 (L3000075, Thermo Fisher, United States), followed by medium replacement 6–8 h post-transfection.

The protein expression of AAT1, GAPDH, Tp53, p21^Cip/Waf, and IRF1 was detected using Western blot. The mouse aortic tissue was homogenized in the lysis buffer. Cells were also lysed in lysis buffer for 30 min at 4°C. The total protein concentration was quantified using the BCA method. An equal amount of protein was taken for electrophoresis, and the membrane was subsequently blocked and incubated with primary antibody overnight at 4°C. Following incubation for 1 h with secondary antibodies at room temperature, the membrane was washed with PBST and processed on a FluorChem M MultiFluor System (Protein Simple, United States) with enhanced chemiluminescence solution (MA0186, Meilunbio, China). n = 6 independent biological samples in animal studies and n = 3 independent experiments performed in duplicate in cell studies.

The original single-cell transcriptome data of aged aortic tissue used in this study is publicly available that can be found in the GEO database (GSE164585) (Xie et al., 2022). Aortas of 4-, 26-, and 86-week-old C57/BL6J mice were analyzed using single-cell RNA sequencing. Dimensional reduction was conducted using t-distributed stochastic neighbor embedding (UMAP), executed through the “RunUMAP” functions. Cluster identification was performed by the “FindClusters” function, resulting in the delineation of 16 distinct clusters from the comprehensive cell population. The “FindAllMarkers” function, employing the Wilcoxon rank-sum test, was used to identify signature genes for each cluster, enabling the assignment of cellular identities, which included various cell types such as B cells, fibroblasts, monocytes, and others.

DEGs from scRNA-seq were identified via “FindAllMarkers” or “FindMarkers” using the “Wilcox” test, yielding “p_val_adj” with Bonferroni correction and “avg_logFC.” Biomarkers were selected from DEGs with p_val_adj ≤0.05 and avg_logFC ≥0.5. GO enrichment analysis was conducted through Metascape and GO software. REVIGO summarized redundant GO terms based on semantic similarity, and non-redundant terms were visualized in Cytoscape, with nodes representing GO terms and 3% pairwise similarities as edges.

Transcription factor prediction was performed using CheA3 online tools (https://maayanlab.cloud/chea3/).

Statistical analysis was performed with SPSS (version 17.0; IBM, United States) and GraphPad Prism 8.0 (GraphPad Software Inc.), and results were represented as mean ± standard deviation (SD). When the variance was uniform and satisfied the normal distribution, an independent sample t-test was used to compare the two groups. One-way ANOVA followed by Bonferroni post-doc analysis was used to compare the difference among multiple groups. Statistical significance was defined as p < 0.05.

We initially measured the expression of senescence markers and SO₂-generating enzymes in the aortic tissues of young (2-month-old) and aged (24-month-old) mice. The results showed that, compared to young (2-month-old) mice, senescence markers Tp53 and p21^Cip/Waf were increased in the aortic tissues of aged (24-month-old) mice. However, the protein expression of AAT1, a key enzyme in SO₂ generation, showed a significantly reduced expression in the aortic tissues of aged (24-month-old) mice (Figure 1A).

To identify the cell populations responsible for reduced AAT1 expression in aortic tissue during aging process, we performed dimensionality reduction and clustering analysis of single-cell transcriptomic data from 4-week-old, 26-week-old, and 86-week-old mice. As a result, we finally grouped cells into 16 distinct clusters, comprising dendritic cells (DC), fibroblasts, B cells, CD8⁺ T cells, CD4⁺ T cells, lymphatic endothelial cells, macroECs, mastocytes, mesothelial cells, microECs, monocytes, neuron cells, neutrophils, pericytes, plasmocytes, and VSMCs (Figure 1B). We next performed gene expression mapping, which revealed that AAT1, the SO₂-generating enzyme, was mainly reduced in VSMC from aged (86-week-old) mice (Figure 1C). Moreover, AAT1 expression was upregulated in mastocytes and pericytes from aged (86-week-old) mice (Figure 1C).

Therefore, we detected the expression of senescence markers Tp53, p21^Cip/Waf and γ-H2AX in the aortic tissues of 5-month-old AAT1 f/f and VSMC-AAT1-KO mice. As expected, the expression of Tp53, p21^Cip/Waf, and γ-H2AX was elevated in the aorta of 5-month-old VSMC-AAT1-KO mice compared to 5-month-old AAT1 f/f mice (Figures 1D–F), suggesting that the downregulation of endogenous SO₂/AAT1 pathway might be associated with vascular aging.

To confirm the importance of endogenous SO₂/AAT1 pathway in VSMC senescence, we generated VSMCs with stable knockdown of AAT1 by transfecting AAT1 shRNA, and cells stably transfected with Scramble shRNA were used as controls. AAT1-knockdowned VSMCs were then treated with or without SO₂ donor. As expected, senescence markers Tp53 and p21^Cip/Waf protein levels increased in AAT1-knockdowned VSMCs, which were rescued by SO₂ donor supplementation (Figures 2A, B). Moreover, SA-β-gal activity increased in AAT1-knockdowned VSMCs, which was rescued by SO₂ donor supplementation. These findings suggested that AAT1 knockdown led to VSMC senescence.

To determine the protective effects of endogenous SO₂ on vascular aging, aortic rings treated with DMEM medium containing 7.5 g/L D-gal were used as a vascular aging model and incubated with or without an SO₂ donor. Aortic rings treated with DMEM medium were used as controls. Compared to the control group, D-gal significantly upregulated Tp53 expression in aortic rings, and supplementation with the SO₂ donor reduced Tp53 protein expression (Figure 2C). These results suggested that sufficient SO₂ protected against vascular aging.

Next, we explored the potential mechanisms underlying the inhibitory effect of the endogenous SO₂/AAT1 pathway on VSMC senescence. First, we analyzed the DEGs in VSMC clusters of aortic tissue from 4- and 86-week-old mice using single-cell RNA sequencing (4-week-old mice as controls), which revealed 641 upregulated and 596 downregulated genes (Figure 3A). KEGG enrichment analyses of these DEGs showed that the upregulated pathways were mainly related to cell senescence (Figure 3B). GO enrichment analyses of these DEGs showed that the upregulated pathways were mainly related to MHC class I complex (Figure 3D), which is also a hallmark of aging (Suryadevara et al., 2024). Whereas the downregulated pathways in KEGG enrichment analyses were mainly related to VSMC contraction and extracellular matrix remodeling (Figure 3C). GO enrichment analyses of these DEGs showed that the downregulated pathways were mainly related to muscle contraction, which is also a marker of vascular aging (Figure 3E).

We then used the CheA3 online tool to predict the potential transcription factors that regulate genes involved in cell senescence and found that IRF1 might play an important role in VSMC senescence (Figure 3F). IRF1 target genes include IL-1β and IL-6, which are SASP factors. The mRNA levels of IL-1β and IL-6 increased in the AAT1-knockdowned VSMCs compared with the scramble group but reduced after SO₂ donor supplementation (Figure 3G). Furthermore, IRF1 nuclear translocation increased in AAT1-knockdowned VSMCs, which reduced after SO₂ donor supplementation (Figure 3H). These results suggested that the endogenous SO₂/AAT1 pathway inhibited IRF1 nuclear translocation and downregulated IRF1-targeted senescence-associated gene expression in VSMCs.

Various kinds of protein modification were involved in aging-related vascular diseases (Zu et al., 2010; Zou et al., 2018; Zhao et al., 2021; Zou et al., 2023). Previous studies have shown that SO₂ regulates protein function via sulphenylation (Huang et al., 2021). Moreover, cysteine 83 (C83) in IRF1 mediates its affinity with DNA (Antonczyk et al., 2019; Jefferies, 2019). Therefore, we hypothesized that SO₂ could inhibit IRF1 transcription activity and its downstream inflammation-mediated VSMC senescence by sulphenylating IRF1 at C83. First, we constructed Flag-IRF1-WT (Flag-tagged wild-type IRF1) and Flag-IRF1-C83S (mutation at cysteine 83 by serine) plasmids and detected the IRF1 suphenylation. Cells transfected with the corresponding plasmids treated without SO₂ as controls. The results showed that the SO₂ donor could sulphenylate IRF1 in VSMCs transfected with the Flag-IRF1-WT plasmid. However, in VSMCs transfected with the Flag-IRF1-C83S plasmid, supplementation with SO₂ failed to increase the sulphenylated IRF1 levels (Figure 4A). These results indicated that SO₂ sulphenylated IRF1 at C83 in VSMCs.

Meanwhile, in VSMCs transfected with Flag-IRF1-WT plasmid, AAT1 knockdown enhanced IRF1 nuclear translocation compared to the scramble group (control), which could be attenuated by the SO₂ donor supplementation (Figure 4B). However, in VSMCs transfected with Flag-IRF1-C83S plasmid, neither AAT1 knockdown nor SO₂ donor supplementation affected IRF1 nuclear translocation (Figure 4B). These results revealed that sulphenylation of IRF1 at C83 is essential for SO₂-inhibited IRF1 nuclear translocation (Figure 4B).

Furthermore, we examined whether C83 was a target of SO₂ for inhibiting the senescence phenotype of VSMCs. In VSMCs transfected with Flag-IRF1-WT plasmid, AAT1 knockdown increased IL-1β and IL-6 mRNA levels (Figure 4C), elevated Tp53 and p21^Cip/Waf protein levels (Figure 4D), and enhanced SA-β-gal activity (Figure 4E) compared to the scramble group, while SO₂ donor supplementation downregulated the levels of IRF1 target genes, SASP factors and SA-β-gal activity compared to AAT1-knockdowned group. However, in VSMCs transfected with Flag-IRF1-C83S plasmid, neither AAT1 knockdown nor SO₂ donor supplementation affected the VSMC senescence phenotype (Figures 4C–E). These results further confirmed that endogenous SO₂ could inhibit VSMC senescence phenotype through IRF1 sulphenylation at C83.