Chen et al. (2019) showed that 10 g/L (almost 50 mM) of Dgal-induced HUVECs can mimic accelerated senescence, and our previous study showed that 50 mM of Dgal-pretreated HUVECs can also mimic accelerated senescence (Hou et al., 2016). Therefore, we selected 50 mM of Dgal to establish a cell senescence model. The Cell Counting Kit-8 (CCK-8, E606335-0500, BBI Life Sciences) was used to detect cell proliferation; cells were cultured in a 96-well plate (Corning, United States) and exposed to Dgal (50 mM) for 46 h, after which the culture medium was removed. Then, 10 μL of the CCK-8 detection solution was added with 100 μL of endothelial cell medium (ECM). The background control was composed of 10 μL of CCK-8 and 100 μL of ECM without cells. After incubating the samples in a cell incubator at 37°C, the absorbance was detected at 450 nm using a microplate reader (Thermo Fisher Scientific, United States). The experiments were repeated at least thrice.

Plasma levels of H₂S (C57Bl/6 J, CSE^−/+, and CSE^−/− mice) and HUVECs were detected as described previously (Shen et al., 2011). All experiments were performed in accordance with the guidelines of the Ethics Committee of Experimental Research at Shanghai Children’s Hospital, School of Medicine, Shanghai Jiao Tong University. First, approximately 30 μL of blood and/or HUVEC lysates (0.5% ammonia was used to lyse the HUVECs) was incubated with 10 µL of ammonia (0.1%) and 80 μL of monobromobimane (MBB) for 40 min at room temperature. The reaction was terminated by adding 20% formic acid and analyzed using gas chromatography mass spectrometry (GC-MS). The experiments were performed in a double-blind manner for detecting the H₂S levels.

Total RNA of the HUVECs was extracted using RNAiso (TaKaRa, Beijing, China), and its integrity was evaluated using a Bio-analyzer 2100 system (Agilent Technology, CA, United States). A sequencing library was then constructed; the library was sequenced on an Illumina HiSeq 3500 platform to generate 150-bps-long paired-end reads. The reading counts for each sample were analyzed using HTSeq v6.0. The reads per kilo base million mapped reads (RPKM) were computed to estimate the gene expression levels. The gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment databases (http://www.genome.jp/kegg/) were used. Benjamini-corrected p < 0.05 was used as the cutoff value for significantly enriched biological processes, and the raw and clean data were obtained after filtering for quality control.

The HUVECs were fixed in 4% paraformaldehyde for 15 min. After blocking, the cells were stained with NRF2 (Proteintech, United States), CD31, γ-H2AX, proliferating cell nuclear antigen (PCNA) (Cell Signaling Technology, United States), or cystathionine γ-lyase (CTH) antibody (Abcam, United States). A secondary antibody (Invitrogen, United States) and DAPI were used for staining, and the cells were washed with a TBST buffer. Images were then obtained using a laser confocal microscope (Zeiss LSM710, Germany). Ten fields were randomly selected for each group, and all histological examinations were performed in a blinded manner.

The HUVECs were starved for 12 h with hydroxyurea (5 mM; Sigma-Aldrich, Germany) to inhibit cell proliferation before pretreatment with Dgal (25 mM) for 24 h. The monolayer was transfected with CSE green-fluorescent protein (GFP)/CSE-GFP-mut (CSE^D187A) or non-specific negative controls (GFP) and scraped with a 200-μL pipette tip to generate scratch wounds. The monolayer was then rinsed twice with ECM containing 1% fetal bovine serum. After a series of treatments including DL-propargylglycine (PPG, 2 mM), NaHS (50 μM), Nac (5 mM), vehicle (0.5% dimethyl sulfoxide; DMSO), Ly294002 (1 μM), and Ml385 (2 μM) for 24 h, the cells were photographed immediately and again after 24 h with a PE Operetta system (PerkinElmer, United States). All migration assays were performed in a blinded fashion and analyzed using ImageJ software.

The tube-like network formed on a Matrigel (BD Bioscience, United States) basement membrane matrix was added to a 24-well culture plate and incubated at 37°C until gelation (Wara et al., 2011). In brief, the HUVECs were transfected with CSE-GFP/CSE-GFP-mut or non-specific negative controls (GFP), cultured for 48 h, pretreated as described above for wound healing, and plated onto Matrigel at a rate of 30,000 cells/well. The branching points and tube lengths were quantitatively determined from five random microscopic fields.

Annexin V fluorescein isothiocyanate (FITC)/propidium iodide (PI) kits (Sigma-Aldrich, Germany) were used to detect apoptosis in the living cells (Hou et al., 2017). For the exogenous H₂S assay, the cells in the experimental group were pretreated with Dgal (25 mM) for 24 h and then treated with PPG, NaHS, or vehicle (0.5% DMSO) for another 24 h. The control group was treated with the vehicle (0.5% DMSO) for 48 h. For the endogenous H₂S assay, the cells were pretreated with Dgal (25 mM) and transfected with CSE-GFP or the non-specific negative control (GFP) for 24 h, before being treated with Ly294002, Ml385, or the vehicle (0.5% DMSO) for another 24 h. After treatment, the cells were collected, stained with Annexin V-FITC/PI, and analyzed by flow cytometry (FCM) using an Annexin V-FITC cell apoptosis detection kit according to manufacturer instructions.

Caspase-3, -8, and -9 protease assay kits (Thermo Fisher Scientific, United States) were used according to manufacturer instructions. The assay loading solutions for each substrate (caspase 3, 8, and 9) were prepared, mixed well, and incubated at room temperature for 30–60 min. The HUVECs were then seeded on a 96-well plate at a density of 2 × 10⁴ cells/90 µL. The cells were pretreated with Dgal (25 mM) and transfected with CSE-GFP or a non-specific negative control (GFP) for 24 h, followed by treatment with Ly294002, Ml385, or the vehicle (0.5% DMSO) for another 24 h. The fluorescence intensity was monitored using a fluorescence microplate reader at the following excitation/emission wavelengths: Caspase 3 = 535/620 nm (red); Caspase 8 = 490/525 nm (green); Caspase 9 = 370/450 nm (blue).

TEM (FEI Talos L 120C, Thermo Fisher Scientific, United States) was performed for morphological analysis at the Shanghai Institute of Precision Medicine, Ninth People’s Hospital, School of Medicine, Shanghai Jiao Tong University. For morphological TEM, acute isolated cardiomyocytes were fixed overnight with ice-cold 2.5% glutaraldehyde at 4°C. The ultrathin sections were then stained with uranyl acetate and lead citrate. Next, sections of 90–100 nm thickness were mounted on a 200-mesh copper grid and imprinted using an FEI Tecnai G2 Spirit transmission electron microscope (Thermo Fisher, MA, United States). Six to eight visual fields were randomly selected to analyze the number of mitochondria using ImageJ software. All the analyses were performed with observer blinding.

Mitochondrial fuel usage in the living cells was detected using the Seahorse XF cell Mito Stress Test Kit (Agilent Technologies, Santa Clara, CA, United States). The cells were transferred to a Seahorse XFp system and analyzed using Wave software (Agilent Technologies, Santa Clara, CA, United States). On the day prior to the experiments, the HUVECs (5000/well) were seeded in Seahorse XFp96 cell culture miniplates for 48 h, and a sensor cartridge was hydrated in the Seahorse XF calibrant overnight at 37°C in a non-CO₂ incubator. On the day of the experiments, the XF cell mito stress test medium (XF base medium, 1 mM pyruvate, 2 mM glutamine, and 10 mM glucose warmed to 37°C and adjusted to pH 7.4 with 0.1 M NaOH) was prepared. The cells were transferred and incubated with the appropriate assay medium for 1 h in a non-CO₂ incubator at 37°C. Pouches containing the compounds oligomycin, trifluoromethoxy carbonyl cyanide phenylhydrazone (FCCP), and rotenone/antimycin were incubated at room temperature for 15 min. Lastly, the compounds were resuspended with the prepared assay medium and diluted to obtain the following final concentrations: 10 μM oligomycin, 10 μM FCCP, and 5 μM rotenone/antimycin for the mito stress test.

Intracellular ROS levels were measured by dihydroethidium (DHE) staining (Sigma-Aldrich, Germany) and CellROX. For the DHE staining, the HUVECs were incubated with 1 μM DHE (dilution with phosphate-buffered saline (PBS)) for 30 min and observed under a laser confocal microscope (Zeiss LSM710, Germany) at excitation/emission wavelengths of 488/610 nm. For the CellROX assay, the HUVECs were incubated with 5 μM CellROX for 30 min and read on a FlexStation 3 multimode microplate reader (VWR Corporation, Radnor, United States). For the mitochondrial superoxide assay, the HUVECs were incubated with 5 μM MitoSox reagent (Invitrogen, United States) working solution for 30 min and observed under a laser confocal microscope at excitation/emission wavelengths of approximately 510/580 nm before being read on the FlexStation 3 microplate reader.

The HUVECs were seeded onto 3 mm sapphire discs; the sapphire disc was then placed with the cells facing up on a flat aluminum planchette, and the inner space of the other aluminum planchette (25 µm) was used as a cover. The space between the two aluminum planchettes was filled with 1-hexadecane. The samples were immediately frozen using an EM ICE high-pressure freezing machine (Leica, Germany) and rapidly stored in liquid nitrogen. The samples were first incubated in acetone containing 0.2% uranyl acetate (UA; −90°C, 48 h) and then at −50°C for 4 h before repeating the process. The samples were next incubated in acetone containing 0.2% UA (12 h) and then at −30°C for 4 h. Then, the samples were rinsed three times with pure acetone (15 min each) and gradually infiltrated in HM20 resin of grades 25%, 50%, 75%, and pure resin (1 h each) at −30°C. The samples were embedded in gelatin capsules after overnight infiltration in the pure resin and polymerized under UV light at −30°C (48 h) and 25°C (12 h). After polymerization, the samples were trimmed and ultrathin sectioned using a microtome (Leica UC7, Germany). Serial thin sections (100 nm thick) were collected on formvar-coated nickel grids and incubated in 0.01 M PBS (including of 1% bovine serum albumin, 0.05% Triton X-100, and 0.05% Tween 20) for 5 min. The sections were next incubated in the primary antibody CTH overnight at 4°C and then in the secondary antibody (goat anti-rabbit conjugated with 10 nm gold) for 2 h at room temperature. The sections were finally washed with PBS, dried at room temperature, and examined by TEM (FEI Talos L 120C, Thermo Fisher Scientific, United States).

The cytoplasmic and nuclear proteins were collected using a nuclear and cytoplasmic protein extraction kit (Beyotime Biotechnology, Nanjing, China) and quantified using a BCA kit (Shen Neng Bo Cai Corp., Shanghai, China). Whole proteins were lysed with a cell lysate buffer (containing radioimmunoprecipitation assay, protease inhibitors, and phosphatase inhibitors) and quantified. The proteins were resolved using sodium dodecyl-sulfate-polyacrylamide gel electrophoresis and transferred onto polyvinylidene fluoride membranes (Millipore, Bedford, MA, United States). The membranes were incubated with primary antibodies specific to NRF2, lamin B1, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH, Cell Signaling Technology, United States) at a dilution of 1:1,000 in a blocking buffer. The membranes were then incubated with secondary horseradish-peroxidase-conjugated antibodies (Cell Signaling Technology, United States) at a dilution of 1:2,000. After washing, the membranes were visualized using a chemiluminescent substrate (ECL) kit. The immunological band densities were analyzed via a scanning densitometer (GS-800, Bio-Rad Laboratories, Hercules, CA, United States) coupled with the Bio-Rad personal computer analysis software.

Tandem mass spectra were processed using PEAKS Studio version X (Bioinformatics Solutions Inc, Canada). The peptides were filtered with 1% FDR; the differently expressed proteins were filtered if they contained at least one unique peptide with significance over 13 (p < 0.05) with a fold change over 1.3. The results were expressed as mean ± standard error of the mean (SEM). Statistical analyses were performed using SPSS software, version 21.0 (SPSS, Inc., United States), and comparisons among the groups were performed by one-way ANOVA. The paired data were evaluated by two-tailed Student’s t-test. The data were considered to be statistically significant at p < 0.05.

Our previous study showed that H₂S levels decreased in the Dgal-induced accelerated mouse heart, liver, and kidney tissues (Wu et al., 2017) as well as decreased in the naturally aging mouse kidneys and plasma (Hou et al., 2016). In the present study, we found that the H₂S levels decreased in Dgal-induced accelerated senescence of endothelial cells in vitro (Figure 1A) and in TERT-knockout mouse (TERT-knockout model, telomerase-deficient mouse) plasma (Figure 1B). Therefore, the decrease in H₂S level may be a common phenomenon during aging. We then tested whether a Dgal-induced accelerated aging model could be constructed successfully. The HUVECs were positively stained with endothelium marker CD31 and DNA damage marker γ-H2AX after incubation with Dgal (25 mM) for 48 h (Figures 2A–C). The HUVECs also highly expressed aging markers, such as p16, p21, and p53, indicating that the accelerated aging cell model was built (Xu et al., 2018) (Figures 2D–F).

We profiled the Dgal-induced accelerated senescence of the HUVEC transcriptome using RNA-seq to assess the effects of CSE. The CSE mRNA levels increased significantly after CSE overexpression (Supplementary Figure S1A). We identified 162 differentially expressed genes, of which 113 (69.8%) were upregulated and 49 (30.2%) were downregulated after CSE overexpression (Supplementary Table S1 and Figure 3A). The H₂S levels increased significantly in the accelerated senescent HUVECs after CSE overexpression (Supplementary Figure S1B). KEGG enrichment analysis was performed to identify the potentially affected pathways; herein, we observed that the metabolic pathway, oxidative phosphorylation, pyruvate metabolism, and PI3K/AKT signaling pathway were enriched after infection of the HUVECs with the CSE adenovirus (Figure 3B) (p < 0.05). Similarly, GO enrichment analysis revealed that these differentially expressed genes were involved in proton-transporting ATPase activity, rotational mechanisms, ATPase activity, couples in transmembrane movement of ions, and ATPase-coupled ion transmembrane transporter activity (Figure 3C) (p < 0.05). Collectively, these data suggest that CSE plays a key role in endothelial metabolism.

We selected 10, 50, and 100 μM of NaHS as the exogenous H₂S donors and noted that 50 μM NaHS had better antiaging effects on Dgal-induced accelerated senescent HUVECs (Topf et al., 2018). Thus, we directly selected 50 μM NaHS. We then tested the effects of H₂S on the endothelial cells and whether they were regulated by the PI3K/AKT signaling pathway. PCNA staining was not significantly different between the control and Dgal groups (Figures 4A,B), while the H₂S-stimulated endothelial cell viability was dependent on PI3K/AKT and NRF2 signaling via the CCK-8 assay (Figure 4C; Supplementary Figure S1C). Moreover, CSE overexpression did not alter PCNA staining (Figures 4D,E) but rather promoted endothelial cell viability, in addition to not being dependent on PI3K/AKT and NRF2 signaling (Figure 4F). The exogenous H₂S also promoted angiogenesis in the accelerated-aging endothelial cells depending on PI3K/AKT and NRF2, which abated upon treatment with the CSE inhibitor, PPG, in the scratch wound healing assay (Figures 5A,B; Supplementary Figure S1D). CSE overexpression accelerated the aging endothelial cell migration, which was blocked by the PI3K/AKT and NRF2 inhibitors (Figures 5C,D). The angiogenesis promotion was also confirmed by the tube formation assay; H₂S promoted angiogenesis in the accelerated senescent endothelial cells, which abated upon treatment with the CSE inhibitor, PPG (Figures 6A,B). CSE overexpression promoted accelerated senescent endothelial cell tube formation and was blocked with an NRF2 inhibitor but not with the PI3K/AKT inhibitor (Figures 6C,D).

To validate the protective effects of the CSE-PI3K/AKT signaling axis, we measured the senescent endothelial cell apoptosis. We observed the antiapoptotic (including early, late, and total apoptosis) effects of H₂S in the HUVECs, which abated after PPG treatment (only affected late apoptosis but not early and total apoptosis), using FCM (Figure 7A). CSE overexpression decreased the apoptosis of accelerated senescent endothelial cells (including late and total apoptosis); late apoptosis was blocked by the PI3K/AKT and NRF2 inhibitors, whereas total apoptosis was blocked by the NRF2 inhibitor (Figure 7B). These effects were confirmed using the caspase family: the activities of caspase 3 and 9 decreased upon CSE overexpression; caspase 3 was blocked with the NRF2 inhibitor, caspase 9 was blocked with a PI3K/AKT inhibitor, and caspase 8 remained unaffected (Figures 7C–E).

We examined the mitochondrial ultrastructure and functions as mitochondrial dysfunction plays a crucial role in oxidative stress and senescence. TEM revealed that Dgal-treated HUVECs exhibited mitochondrial ultrastructural changes, including irregular arrangement and loss of cristae (Figure 8A), with a significantly increased proportion (approximately 40%) of mitochondrial ultrastructure disorders (Figure 8B). The seahorse assay showed that Dgal-treated HUVECs exhibited a significant decrease in mitochondrial respiration (ATP production, basal and maximal respiration). Respiration ability was further decreased in the PPG group, while the exogenous H₂S only alleviated maximal respiration (Figures 8C,D). Increases in ATP production and basal respiration were observed but were not statistically significant (Figures 8C,D). From the measurements of the endogenous H₂S and PI3K/AKT effects, CSE overexpression was observed to reverse Dgal-induced mitochondrial dysfunction in the HUVECs (ATP production, basal and maximal respiration), while these were blocked by the PI3K/AKT and NRF2 inhibitors (Figures 8E,F).

Mitochondrial dysfunction often leads to increased ROS (Newgard and Sharpless, 2013). DHE intensity was elevated in Dgal-induced accelerated senescence in the HUVECs relative to that in the control (Figures 9A,B, Supplementary Figure S2A). The DHE intensity of the PPG group was higher than that of the Dgal group but significantly decreased in the exogenous H₂S case (Figures 9A,B). Additionally, the DHE fluorescence intensity levels decreased upon CSE overexpression and were blocked by the PI3K/AKT and NRF2 inhibitors (Figure 9C). Similar protective effects were observed for both exogenous and endogenous H₂S in mitochondrial oxidative stress based on the MitoSox assay, while the MitoSox fluorescence intensity levels decreased upon CSE overexpression and were only blocked with the PI3K/AKT inhibitor but not the NRF2 inhibitor (Figures 9D–F, Supplementary Figure S2B). These data demonstrate that Dgal-induced oxidative burden is correlated with mitochondrial dysfunction.

The GO and KEGG analyses showed that the PI3K/AKT signaling pathway and ATPase activity-related signaling regulation were significantly affected by CSE overexpression in the accelerated-aging HUVECs (Figure 3). To determine the effects of CSE on the mitochondria in accelerated senescent HUVECs, we measured the expressions of the mitochondrial fusion regulators mitofusin2 (MFN2) and dynamin-1-like (DRP1), as well as the NRF2 and PI3K/AKT signaling pathways. The CSE protein expression decreased in the accelerated-aging HUVECs (Supplementary Figure S3A). MFN2 and NRF2 expressions decreased after incubation with Dgal for 48 h and were partially reversed upon treatment with exogenous H₂S (50 μM) (Figures 10A,B). For the endogenous H₂S, pretreatment with overexpressed CSE resulted in overactivation of PI3K/AKT (Figures 10C,D), which was in line with the RNA sequencing results (Figure 3). Additionally, we found that NRF2 expression increased after infection with CSE and decreased after treatment with the NRF2 and PI3K/AKT inhibitors (Figures 10E,F).

Our previous study showed that exogenous H₂S treatment induced NRF2 translocation from the cytoplasm to the nucleus (Hou et al., 2017). We also observed that the percentage of nuclear/whole cells increased after CSE overexpression and decreased upon treatment with the NRF2 and PI3K/AKT inhibitors (Figures 11A,B). Western blotting was performed to confirm the NRF2 translocation tendency (Figures 11C–E).

Next, we used immunofluorescence colocalization and immunoelectron microscopy to examine the translocation of CSE. As shown in Figure 12A, CSE colocalized with mitochondria (MitoTracker showed mitochondria), and this tendency was much greater in the Dgal + GFP-CSE group than in the control group. We observed more number of black/gold particles (CSE protein locations) in the mitochondria upon CSE overexpression in the accelerated senescent HUVECs (Figure 12B). Next, we overexpressed CSE during accelerated senescence of the HUVECs and performed mass spectrometry measurements; the corresponding results showed that TOMM70 increased significantly upon CSE overexpression, suggesting a close interaction between CSE and TOMM70 (Supplementary Table S2). Immunofluorescence co-immunoprecipitation (co-IP) confirmed that the expressions of TOMM20 and TOMM70 increased following CSE overexpression and that CSE was associated with both TOMM20 and TOMM70 (Figure 12C). To further evaluate the CSE protein function, we constructed a CSE-GFP-mut virus (CSE^D187A). The ability of CSE to catalyze H₂S production decreased after mutation of one of the important activity centers (CSE^D187A) (Supplementary Figure S1E). After pretreatment with CSE-GFP-mut, we found that migration promotion and tube formation had disappeared (Figures 13A–C) along with the antioxidative effect, as confirmed by the DHE and MitoSox assays (Figures 13D,E). The results of co-IP showed that TOMM20 expression increased, whereas TOMM70 expression remained the same following CSE-GFP-mut overexpression (Figure 13F).