Primary human umbilical vein endothelial cells (HUVECs) and culture medium (supplemented with or without 10% fetal bovine serum) were purchased from Allcells (Shanghai, China). Cells were cultured in gelatin-coated dishes at 37°C in an incubator containing 21% O₂ for normoxia and 8 or 10% O₂ for hypoxia, as well as 5% CO₂. Cells were used for experiments within six passages.

Confluent HUVECs were starved overnight and were scratched using a pipette tip for the wound healing assay. Markings were drawn on the culture dishes as reference points so as to make sure that the same visual field was photographed at 0 and 24 h. Typically, 8–12 visual fields were chosen randomly in one culture dish. Cells were photographed using an EVOS fl Microscope (Advanced Microscopy Group, Mill Creek, Washington) after incubation in normoxic or hypoxic conditions. The outline of the wound area (the area with no cells) was then drawn using Image J software to get the exact pixel coverage of each wound area. Then, the pixel coverage value of each wound area was converted to a standard value, shown as one data point, by being divided by a fixed value so as to normalize the average value of the control group to be 100.

siRNAs specific for human CSE, CBS, or MPST are as follows: human CSE siRNA: sense 5′-CUAUGUAUUCUGCAACAAATT-3′, antisense 5′-UUUGUUGCAGAAUACAUAGAA-3′; human CBS siRNA: sense 5′-CUCACAUCCUAGACCAGUATT-3′, antisense 5′-UACUGGUCUAGGAUGUGAGAA-3′; human MPST siRNA: sense 5′-GGAAUUCCGUUGACUUGUUTT-3′, antisense 5′-AACAAGUCAACGGAAUUCCTG-3′.

siRNA specific for human CSE, CBS, MPST, and scrambled siRNA as a negative control were purchased from Thermo Scientific-Ambion. HUVECs were transfected using Lipofectamine RNAiMAX (Thermo Scientific-Invitrogen) and OPTI-MEM (Thermo Scientific-Gibco) according to the instructions provided by the manufacturer and then the cells were cultured for another 48 h for further analyses.

Confluent HUVECs were starved overnight and were cultured in normoxic or hypoxic conditions. Then, cells were lysed in a 1 × sodium dodecyl sulfate (SDS) sample buffer (62.5 mM Tris-HCl pH 6.8, 10% glycerol, 2% w/v SDS). Bicinchoninic acid (BCA) reagent was purchased from Shen Neng Bo Cai Corp. (Shanghai, China) and the BCA method was used for protein concentration determination. Protein samples (50 μg) were subjected to electrophoresis in a 10% SDS-polyacrylamide gel, and then transferred to a polyvinyl difluoride membrane (Millipore-Upstate Biotechnology, Lake Placid, New York, USA). After blocking in 5% non-fat milk in Tris Buffered Saline with Tween-20 (TBST, 0.05% Tween-20) for 2 h at room temperature (RT), the membrane was incubated with primary antibodies at 4°C overnight. Anti-CBS and anti-β-actin were purchased from Proteintech Group (Chicago, IL, USA). Anti-CSE and anti-MPST were purchased from Santa Cruz Biotechnology (CA, USA). After washing the membranes using TBST and then incubating them with horseradish peroxidase-conjugated secondary antibodies for 2 h at RT, SuperSignal West Pico Chemiluminescent Substrate (Thermo-Pierce Biotechnology, Rockford, USA) was used for band detection. Band signals were then quantified using Smart viewer software.

Total RNA was extracted from cells using Trizol reagent. RNA was reverse-transcribed using a cDNA synthesis kit (Toyobo Life Science, Japan). Real-Time PCR was performed using a StepOnePlus Real-Time PCR Detection System (Applied Biosystems Inc., CA, USA). A total volume of 20 μL reaction mixture containing 1 μL cDNA, 10 μL SYBR Green PCR Master Mix (Toyobo, Japan) and 1.6 μL each primer (10 μM) was used for this assay. GAPDH was used as reference for normalization and the relative expression of mRNA was calculated according to the ΔΔCt method.

The primer pairs used for the Real-time PCR were as follows: for the human MPST gene, the forward primer was 5′-CGGAGTCTCCTCCCTTTGGT-3′, and the reverse primer was 5′-CCTCCCTAAGATGCAGCTCG-3′. For the GAPDH gene, the forward primer was 5′-TGCCCCCATGTTCGTCA-3′, and the reverse primer was 5′-CTTGGCCAGGGGTGCTAA-3′.

The human MPST mRNA 5′-UTR and 3′-UTR were amplified from HUVECs using the following primers: 3′-UTR, the forward primer was 5′-AGCTGGGCAGGACACAGG-3′, and the reverse primer was 5′-TTCCTCAAAAATAAAACAGAAAGGAGG-3′. 5′-UTR-exon1, the forward primer was 5′-GGAAGGCGCGGGCAGCAGCGGCTCCGAGTGGCCGCGGCGGTGGGCTGTGCC-3′, the reverse primer was 5′-ATTGGCGACCTGCAGCGGACCAAAGGGAGGAGACTCCGGCACAGCCCACCGCC-3′. 5′-UTR-exon2, the forward primer was 5′-CGCTGCAGGTCGCCAATATAAATGCTTGATGAG-3′, the reverse primer was 5′-CTGCGGGCCTCCACTTGCTGGGCAGG-3′. 5′-UTR-exon3, the forward primer was 5′-AAGTGGAGGCCCGCAGCCCGAGTGTCGCCGCC-3′, the reverse primer was 5′-GGCGGCGACACTCGGGCTGCGGGCCTCCACTT-3′. The 3′-UTR or 5′-UTR (exon1, exon2, and exon3 jointed together) was subcloned into the pGEM-T vector.

RNA transcripts were produced by T7 RNA polymerase (Roche, Basel, Switzerland) with Biotin RNA labeling Mix (Roche, Basel, Switzerland) in an 18 μl system. After incubation for 2 h at 37°C, 2 μL RNase-free DNase was added and incubated for 15 min at 37°C to remove template DNA. Then, 2 μL 0.2 M EDTA (pH = 8.0) was added to stop the reaction. Biotinylated RNA was added into the Structure Buffer (10 mM Tris pH = 7.0, 0.1 M KCl, 10 mM MgCl₂) to form the secondary structure.

The biotinylated RNA was bound to streptavidin-coated Dynabeads (Thermo Scientific-Invitrogen) according to the instructions provided by the manufacturer. Biotin-labeled antisense RNA was used as a control in all experiments. Magnetic beads were washed three times in 500 μL Wash Buffer I (10 mM Tris-Cl, pH 7.0, 1 mM EDTA, 2 mM NaCl) and resuspended in Wash Buffer I. Biotinylated RNA was incubated with magnetic beads at 4°C overnight. Biotinylated RNA-streptavidin magnetic beads were incubated with 1 mg cell extract with RNase inhibitor at RT for 1 h. The RNA/protein/bead complexes were washed three times using Wash Buffer II (10 mM Tris-Cl, pH 7.0, 1 mM EDTA, 100 mM KCl, 0.1% Triton X-100, 5% glycerol, 1 mM DTT). Then, proteins were eluted by boiling in 5 × SDS PAGE loading buffer and resolved using SDS-PAGE. Protein bands were silver-stained and cut for mass-spectrometric analysis (Lu Ming Corp. Guangzhou, China).

Protein bands from SDS-PAGE gels were excised. The gel fragments were cut into small pieces and decolored in 50 μL decoloring solution [15 mM K₃Fe(CN)₆ and 50 mM NaS₂O₃] at RT for 5 min. The gel pieces were dried in 50% acetonitrile (ACN) at RT for 30 min and then dried in 100% ACN at RT for 30 min. Protein was digested using 0.4 μg sequencing grade trypsin at 37°C for 16 h. The digested peptides were extracted from the gel pieces using Extraction Solution (water solution containing 5% trifluoroacetic acid and 67% ACN) at 37°C for 30 min and then applied to the on-line liquid chromatography (LC)-collision-induced dissociation (CID)-MS-MS analysis.

High-performance liquid chromatography was carried out to isolate trypsin-digested protein fragments on a 75 μm inner diameter 15 cm reverse-phase capillary column, 3.0 μm, 120 Å, C18 (ChromXP Eksigent) using a gradient from 5 to 35% buffer B (98% ACN, 0.1% formic acid) over a period of 90 min. Mass spectrometry was performed on a TripleTOF5600 system (AB SCIEX) using electrospray ionization. The optimum conditions: ion spray voltage, 2.5 kV; curtain gas, 30 psi; nebulizing gas (GS1), 5 psi; interface heater temperature, 150°C. For information dependent acquisition, survey scans were acquired in 250 ms. At most, 35 product ion scans were collected exceeding a threshold of 150 counts/s (2⁺~5⁺ charges). Total cycle time was set to 2.5 s. For CID, a rolling collision energy setting was applied to precursor ions. Proteins were identified using the MASCOT V2.3 search engine (Matrix Science Ltd., London, UK.). Search parameters were as follows: two missed cleavage sites; fixed modifications of Carbamidomethyl (C); partial modifications of Acetyl (Protein N-term), Deamidated (NQ), Dioxidation (W), Oxidation (M); precursor ion tolerance is ± 30 ppm, fragment ion tolerance is ± 0.15 Da.

All results are expressed as mean ± standard error of the mean (SEM). Statistical analyses were performed using GraphPad Prism statistical software 7.0 (Graphpad Software, La Jolla, California, USA). The Shapiro–Wilk normality test or the Kolmogorov–Smirnov normality test was used. Data were analyzed using unpaired t-test, Mann–Whitney test, one-way ANOVA followed by Tukey's multiple comparisons test or Kruskal–Wallis test followed by Dunn's multiple comparisons test wherever applicable. P < 0.05 was considered statistically significant.

After starving overnight, HUVECs were scraped using a pipette tip for the wound healing assay. Then, the cells were cultured in normoxic or hypoxic conditions with 8 or 10% oxygen levels. Results showed that hypoxia (10% oxygen level) significantly increased cell migration compared with cells cultured in normoxia (Figures 1A,B). However, more severe hypoxia (8% oxygen level) exerted no effect on cell migration (Figures 1C,D). Therefore, 10% oxygen level was chosen for the following experiments.

CSE protein levels were significantly decreased using RNA interference (Figures 2B,E). HUVECs transfected with negative control siRNA (NC) or CSE siRNA for 36 h were starved overnight, and then used in a wound healing assay. Results showed that CSE knockdown by RNAi (CSEi group) significantly increased the migration of endothelial cells compared with the NC group in both normoxia (NC normoxia vs. CSEi normoxia) and hypoxia (NC hypoxia vs. CSEi hypoxia; Figures 3A,B). These results suggest that the endogenous hydrogen sulfide produced by CSE might probably play an inhibitory role in the migration of vascular endothelial cells.

CBS protein levels were significantly decreased using RNA interference (Figures 2A,C). HUVECs transfected with negative control siRNA (NC) or CBS siRNA for 36 h were starved overnight, and then used in a wound healing assay. Results showed that CBS knockdown by RNAi (CBSi group) significantly decreased the migration of endothelial cells compared with the NC group in both normoxia (NC normoxia vs. CBSi normoxia) and hypoxia (NC hypoxia vs. CBSi hypoxia; Figures 3A,C). These results suggest that the endogenous hydrogen sulfide produced by CBS might probably play a stimulatory role in the migration of vascular endothelial cells.

MPST protein levels were significantly decreased using RNA interference (Figures 2B,F). HUVECs transfected with negative control siRNA (NC) or MPST siRNA for 36 h were starved overnight, and then used in a wound healing assay. Results showed that MPST knockdown by RNAi (MPSTi group) significantly decreased the migration of endothelial cells compared with the NC group in hypoxia (NC hypoxia vs. MPSTi hypoxia; Figures 3A,D). However, there was no difference between the NC group and the MPSTi group in normoxia (NC normoxia vs. MPSTi normoxia; Figures 3A,D). These results suggest that the endogenous hydrogen sulfide produced by MPST might probably turn into effect in promoting the migration of vascular endothelial cells once oxygen levels have decreased.

HUVECs in hypoxia migrated faster than those in normoxia after transfection with NC siRNA (NC normoxia vs. NC hypoxia; Figure 3). Knocking down MPST using siRNA could abrogate the hypoxia-induced increase in the migration of HUVECs (Figures 3A,D). However, after knockdown of CSE or CBS, the hypoxia-induced increase in endothelial cell migration persisted (Figures 3A–C). These results suggest that the endogenous hydrogen sulfide produced by MPST might probably mediate the hypoxia-induced increase in the migration of vascular endothelial cells, while hydrogen sulfide produced by CSE or CBS has no significant effect.

In order to exclude the impact of starvation time on the expression of CSE, CBS, and MPST, cell dishes were placed in the incubator for hypoxia treatment at different time points and harvested at the same time (8 or 24 h). Therefore, the cells were cultured in hypoxia for the times indicated in Figure 4. MPST protein levels were significantly increased after culturing in hypoxia for 4 or 16 h (Figure 4C). In addition, hypoxia could significantly decrease CSE protein levels at 8 h (Figure 4A), but had no effect on CBS protein levels (Figure 4B). These results suggest that the three endogenous hydrogen sulfide producing enzymes CSE, CBS, and MPST respond differently to hypoxia and the bi-phasic elevation of MPST implies a more complicated mechanism in the regulation of its protein level.

HUVECs starved overnight were treated using Trizol reagent for further real-time PCR analyses. Results showed that MPST mRNA levels were not increased at 4 h in hypoxia when MPST protein levels had already been elevated (Figure 5). These results suggest that the first-stage elevation of MPST protein levels was probably induced through a post-transcriptional mechanism. However, MPST mRNA levels were increased at 6 h in hypoxia (Figure 4), which suggests that the later-stage elevation of MPST protein levels might probably be induced at a transcriptional level.

To investigate the post-transcriptional regulation of MPST mRNA by hypoxia, the 5′UTR and 3′UTR of MPST mRNA with biotin tags were prepared. RNA pull-down experiments were performed by incubating biotin-tagged mRNA with lysates of HUVECs after culturing in normoxic or hypoxic conditions. The antisense sequence of the UTR region was used as a negative control. The mixture was then incubated with streptavidin-coated beads to obtain biotin-tagged RNA with binding proteins on it. SDS-PAGE with silver staining showed that many proteins bound with the 5′UTR, rather than the 3′UTR of MPST mRNA in normoxic conditions compared with their antisense RNA groups (Figure 6A). However, those binding proteins were significantly decreased after cells were cultured in hypoxia for 4 h (Figure 6B).

In order to identify the binding proteins of MPST mRNA which were significantly decreased at the protein level or dissociated from MPST mRNA, the protein bands indicated with arrows and numbers in Figure 6A were chosen for in-gel digestion using trypsin and then analyzed through LC-MS/MS. All proteins proposed by the mass spectrometry data and Mascot searching engine were ranked by score and are presented in the Supplementary Tables in the form of Excel files (candidates of bands no. 1 and 2 are shown in Supplementary Table 1; band no. 3, Supplementary Table 2; band no. 4, Supplementary Table 3; band no. 5, Supplementary Table 4; band no. 6, Supplementary Table 5; bands No. 7 and 8, Supplementary Table 6; band no. 9, Supplementary Table 7; band no. 10, Supplementary Table 8; bands no. 11, 12, and 13, Supplementary Table 9; band no. 14, Supplementary Table 10; band no. 15, Supplementary Table 11; bands No. 16, 17, and 18, Supplementary Table 12). Score and mass weight will be considered simultaneously in choosing the candidate RNA binding proteins of MPST mRNA, and whether hypoxia decreased the protein levels of candidates or inhibited the association of candidates with 5′UTR mRNA will be further investigated.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.