2.1 High expression of RTN3 is associated with arterial hypertension

Our previous study reported that overexpression of RTN3 may partly lead to HTG and obesity, which are important factors in the development of secondary hypertension.^{12, 17} To further investigate the association between RTN3 and hypertension, we enrolled 18 healthy controls and 56 primary hypertension patients with normal triglyceride levels (<1.7 mmol/L) and body mass index (18.5–24 kg/m²) to exclude the effects of HTG and obesity. Enzyme-linked immunosorbent assay (ELISA) detection exhibited higher RTN3 levels in part patients (Figure 1A and Table 1).

The GSE165601 database including the RNA-seq data of ventral aortas from wild-type (WT) mice (n = 4) and angiotensin II (Ang II)-treated mice (n = 4) showed that the mRNA levels of RTN3 in Ang II-treated mice were higher than those in WT mice (Figure 1B). To confirm this discovery, hypertension mouse models were constructed by Ang II osmosis pumps with a permeability of 500 ng/kg/min for 7 days, which is a classic hypertension modeling (Figure 1C). Both systolic blood pressure (SBP) and diastolic blood pressure (DBP) were obviously increased in Ang II-treated mice compared with saline-treated mice (Figure 1D). Real-time PCR and western blot (WB) authenticated the increased RTN3 in thoracic aortas of Ang II-treated mice (Figures 1E and F). These discoveries suggested the association between high expression of RTN3 and hypertension and that this linkage may be independent of obesity and HTG.

2.2 RTN3-null mice present hypotensive conditions

We have reported that RTN3 overexpression is associated with HTG and obesity, which are important factors for secondary hypertension. We then generated RTN3-null mice as we previously described to continue the study.¹² We identified that RTN3 was expressed in thoracic aortas but silenced in RTN3-null mice (Figure 2A). As expected, compared with WT mice, RTN3-null mice were obviously hypotensive, and their SBP, DBP, and mean arterial pressure were significantly lower (Figures 2B–D). Hematoxylin–eosin (H&E) and Masson staining showed the incompact elastin fibers of aortas in the RTN3-null mice (Figure 2E).

Next, we applied Ang II (500 ng/kg/min) to WT and RTN3-null mice for 7 days. The RTN3-null mice presented lower SBP and DBP than WT mice after Ang II treatment at all observation times (Figures 2G and H). Simultaneously, the increase in SBP on the 7th day after Ang II treatment in RTN3-null mice was also less than that in WT mice (Figures 2I and J). Our findings revealed that RTN3 triggered hypotension in mice and may attenuate Ang II-induced elevated blood pressure.

2.3 RTN3 reduction promotes the migration and tube formation of HUVECs

VSMCs and endothelial cells are the two most dominant classes of cells in thoracic aortas.¹⁸ First, we analyzed the effects of RTN3 reduction on VSMCs. We stripped the mouse aortic media in RTN3-null mice or transfected RTN3 siRNA into human aortic VSMCs (HA-VSMCs) and then tested vinculin, actin alpha 2, smooth muscle (α-SMA), calponin 1 (CNN1), and calponin 2 (CNN2), which are associated with blood pressure regulation or phenotypic switch of VSMCs.^19-21 However, we did not observe prominent alterations in the expression of these proteins (Figures S1A and B). Transwell results indicated that reduced RTN3 did not disturb HA-VSMC migration (Figure S1C). This result suggested that the lack of RTN3 had very little effect on VSMCs.

Hence, we then focused on the effect of RTN3 reduction on endothelial cells. Immunofluorescence staining found that RTN3 colocalized with the endothelial cell marker CD31 (platelet and endothelial cell adhesion molecule 1) in WT mouse thoracic aortas, which suggested that RTN3 was expressed in endothelial cells (Figure 3A). To assess the effect of RTN3 reduction in endothelial cells, we transfected RTN3 siRNA into human umbilical vein endothelial cells (HUVECs; Figure 3B). The scraping line method and transwell studies revealed that knocking down RTN3 expression in HUVECs may promote cell migration ability (Figures 3C and D). Cell tube formation manifested that silencing the expression in HUVECs may increase the angiogenesis of cells (Figure 3E). In addition, siRTN3 suppressed the transition from G1 to S phase but did not induce cell apoptosis to reduce cell viability in HUVECs (Figure S2). Together, these findings suggested that downregulated RTN3 can promote the migration and tube formation of HUVECs.

2.4 Lack of RTN3 activates S1P the signaling pathway

Previous studies have claimed that S1P plays a significant role in blood pressure regulation, cell migration, and tube formation of endothelial cells.²² We detected the S1P levels in the mouse models. S1P from the plasma of RTN3-null mice was markedly increased (Figure 4A). However, the S1P levels in the aortas of RTN3-null mice and WT mice were not significantly different (Figure S3A). We speculated that the majority in the arteries were tunica media and that S1P from tunica media/VSMCs covered the difference in S1P from the endothelium.

Then, we assessed S1P levels in HUVECs. The cell supernatant of HUVECs transfected with siRTN3 contained more S1P than that of the controls (Figure 4B). Simultaneously, HUVECs lacking RTN3 were accompanied by the increase in extracellular NO and intracellular cGMP, which may be induced by S1P signaling (Figures 4C and D). Moreover, WB analysis found that the sphingosine kinase isozyme SPHK2, but not SPHK1, which catalyzed the phosphorylation of sphingosine into S1P, was increased in HUVECs with RTN3 knockdown (Figures 4E and S3B).⁸ However, similar to S1P in aortas, both SPHK1 and SPHK2 were also not discriminative in RTN3-null mouse aortas and WT mouse aortas (Figures S3C and D). Conversely, an overabundance of RTN3 in HUVECs caused diminished S1P, NO, cGMP, and SPHK2 levels (Figures 4F–I). Thus, we reasoned that lack of RTN3 can activate the S1P signaling pathway to lower blood pressure and stimulate the migration and tube formation of endothelial cells.

3-(4-Chlorophenyl)-adamantane-1-carboxylic acid (pyridin-4-ylmethyl) amide, named ABC294640, is a selective inhibitor of SPHK2 that can decrease S1P synthesis and rescue Ang II-induced hypertension in mice.^{23, 24} Here, ABC294640 and/or RTN3 siRNA were applied to treat HUVECs, and the results showed that the levels of S1P and NO after ABC294640 treatment were lower than those in the controls, while compared with the siN+ABC294640 group, they were obviously increased in the siRTN3+ABC294640 group (Figures 4J–L). These studies suggested that silencing RTN3 expression in HUVECs can attenuate the inhibitory effects of ABC294640 on S1P biosynthesis. All these findings suggested that reducing RTN3 activated the S1P signaling pathway in HUVECs.

2.5 RTN3 interacts with CERS2 to play a negative role in ceramide synthesis

We found that RTN3 can regulate the S1P pathway in HUVECs, which plays a crucial role in blood pressure regulation, but the underlying molecular mechanisms are unclear. Mass spectrometry was applied to detect the candidate RTN3-interacting proteins in HUVECs. Interestingly, CERS2, a ceramide synthase and a core enzyme in S1P synthesis pathways, was found to be a possible RTN3-interacting protein (data not shown).⁷ Coimmunoprecipitation (Co-IP) analysis further validated that CERS2 can interact with RTN3 in human embryo kidney 293 (HEK293) cells and HUVECs (Figures 5A and B). CERS2 predominated in mouse aortas, as detected by unique identifier-mRNA sequencing (Figure S4A). In addition, CERS1, which was slightly expressed in mouse aortas, also interacted with RTN3 in HEK293 cells and HUVECs (Figures 4B and C). WB analysis revealed that overexpressed RTN3 impeded CERS1 in HEK293 and CERS2 in HUVECs (Figures 5C and S4D). Correspondingly, the downstream total ceramide (intracellular) level was also reduced (Figures 5D and S4E). In contrary, the expression of CERS2 and the concentration of intracellular ceramides increased when HUVECs were transfected with siRTN3 (Figures 5E and F).

However, why can RTN3 regulate the expression of CERS2? By reviewing literatures, we found that RTN3 has been reported to be involved in the formation of autophagic vesicles.¹³ WB validated that autophagy markers, autophagy-related protein LC3 A and B (LC3A/B II) and autophagy receptor P62 (p62), were more highly expressed in HUVECs with increased RTN3 after starvation for four hours (Figure 5G). Confocal microscopy found that only a little LC3A/B did not colocalized with CERS2 in HUVECs transfected with RTN3 overexpression plasmids, less than cells transfected with blank vector (Figure 5H), which indicated that autophagic vesicles were easier to degrade the CERS2 when RTN3 increased in HUVECs. In fact, RTN3 is one of binding sites of LC3 in ER, to induce the formation of autophagic vesicles.²⁵ CERS2, as an interacting protein of RTN3, may be more liable to be parceled into autophagic vesicles.

Collectively, our data suggested that RTN3 can interact with CERS2 in endothelial cells. As a marker of cell autophagy, RTN3 may promote the formation of autophagic vesicles, and more CERS2 was bound by RTN3 into autophagic vesicles to be degraded. As a ceramide synthase, reducing CERS2 may decrease ceramide to further inhibit S1P produce. However, reduced RTN3 correspondingly retained more CERS2 to stimulate the synthesis of ceramide and S1P. The imbalance of S1P metabolism would lead to blood pressure variation (Figure 6).

2.6 An RTN3 variant (c.116C>T, p.T39M) is identified in a patient with hypertension

Additionally, 48 hypertensive families were enrolled to perform mutation detection by whole-exome sequencing (WES). An RTN3 variant (c.116C>T, p.T39M) was identified in Family 31 (Figures 7A and B). The proband (II:3) was a 42-year-old male with SBD of 150 mmHg and PBD of 110 mmHg. He was diagnosed with hypertension 6 years ago, and his mother (I:2) also had hypertension with the onset of her forties. He and his daughter (III:1) harbored the RTN3 variant, while the girl was unaffected thus far (Table 2). Bioinformatics analysis manifested that p.T39 in RTN3 was highly conserved in mammals, especially primates (Figure 7C). Three-dimensional protein modeling revealed that the variant p.T39M sharply reduced the distance between the 39th amino acid and the 66th (p.E66), possibly changing the spatial structure of RTN3, and that the substitution of positively charged threonine by noncharged methionine moderately altered the surface charge of RTN3 (Figure 7D).

To verify the pathogenicity of the RTN3 variant, we constructed WT and RTN3 mutated (p.T39M) plasmids. We transfected WT and mutant plasmids into HUVECs, and after treatment with cyclohexane (CHX), RTN3 mutant protein presented higher related expression than WT RTN3, suggesting that the variant p.T39M made RTN3 more stabilized and resistant to degradation (Figure 7E). Ceramide and S1P levels were also reduced in the mutated RTN3 group (Figures 7F and G). Thus, we reasoned that the RTN3 variant (c.116C>T, p.T39M) is one of the etiological factors in the proband.

5.1 Subject recruitment

A total of 18 controls and 56 hypertension patients were recruited. All subjects were male with serum triglycerides < 1.7 mmol/L, 18.5 kg/m² < BMI < 24 kg/m², and they ranged in ages from 20 to 50. All subjects or their guardians consented to publication of the clinical details.

5.2 Mouse strains and modeling

RTN3-null mice were generated, and their genotypes were identified as described previously.¹² C57BL/6J mice were purchased from the Chinese Academy of Sciences (Shanghai, China) and bred in the Department of Zoology, Central South University.

Ang II (Solarbio, Beijing, China) was induced in male mice at 10 weeks of age. Ang II powder was diluted to an appropriate concentration using sterile normal saline and injected into Alzet implantable osmotic pressure capsules (Durect, Cupertino, USA) with a special needle, and the penetration rate was 500 ng/kg/min⁶. Alzet capsules were inserted into the mouse dermis, and the wound was sutured. A Coda Non-Invasive Blood Pressure System (Kent Scientific, Torrington, USA) was used to detect the noninvasive blood pressure in the tail vein of mice.

5.3 GSE165601 database analysis

The RNA-seq datasets of GSE165601 were obtained from the GEO database (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE165601). R (version 4.0.4) and RStudio (version 1.2.5033) were used to address all the data in this study. GSE185051 contained the expressed genes in the abdominal aorta of five C57BL/6 WT mice and five Ang II-treated mice.

5.4 H&E staining and Masson staining

Paraformaldehyde-fixed aortas were embedded in paraffin and sliced into 5 μm sections. The sections were dried, dewaxed with xylene and rehydrated with decreasing concentrations of alcohol. For H&E staining, slides were stained with hematoxylin (Beyotime, Shanghai, China) for 15 min, and stained with 1% eosin (Beyotime) for 2 min. Masson's trichrome reagent (Solarbio) was used for Masson staining.³⁸

5.5 Cell culture and transfection

The pcDNA3.1-flag-RTN3 plasmid and pcDNA3.1-myc-RTN3 plasmid were purchased from Sangon Biotech Company (Shanghai, China). Mutagenesis of RTN3 c.116C>T (pcDNA3.1-flag-mutRTN3-T39M) was performed using Fast Mutagenesis Kit V2 (Vazyme, Nanjing, China).

HUVECs and HEK293 cell lines were obtained from the Cell Bank of Shanghai Institutes for Biological Sciences (Shanghai, China). HUVECs were cultured in Roswell Park Memorial Institute (RPMI-1640) medium, and HEK293 cells were cultured in Dulbecco's modified Eagle's medium. HUVECs were transfected with the pcDNA3.1-flag-RTN3 plasmid. Cells were seeded in 6-well plates and transfected with RTN3 siRNA (RiboBio, Guangzhou, China) or plasmids for 24 h to operate the scraping line method, transwell, cell tube formation, and 60 μM ABC294640 (APExBIO, Houston, USA) treatment with test after 24 h, or for 48 h to operate WB, ELISA, Co-IP, 100 μM CHX (Solarbio) treatment, and flow cytometry.

5.6 WB, Co-IP, and confocal

Proteins were extracted by ristocetin-induced platelet agglutination buffer (Solarbio), and WB, Co-IP, and confocal microscopy were performed as previously described.¹² Anti-RTN3 primary antibody (1:2000) was self-produced. Anti-GAPDH primary antibody (10494-1-AP, 1:5000), anti-flag primary antibody (20543-1-AP, 1:5000), and anti-p62 primary antibody (18420-1-AP, 1:1000) were purchased from Proteintech Company (Wuhan, China). Anti-SPHK2 primary antibody (YT4383, 1:1000) and anti-CYP27A1 primary antibody (YT1202, 1:1000) were purchased from ImmunoWay Biotechnology Company (Suzhou, China). Anti-myc primary antibody (sc-40, 1:100), anti-CERS2 primary antibody (sc-390745, 1:100), and anti-LC3A/B primary antibody (sc-398822, 1:100) were purchased from Santa Cruz Biotechnology Incorporated (Santa Cruz, USA). Anti-myc magnetic beads (Beyotime) and anti-flag magnetic beads (Beyotime) were used for Co-IP, and self-produced anti-RTN3 primary antibody (1:400) and anti-CD31 primary antibody (66065-2-Ig, 1:200) were purchased from Proteintech Company; anti-LC3 primary antibody (PM036, 1:200) were purchased from MBL International Corporation (Kyushu, Japan); anti-CERS2 primary antibody (sc-390745, 1:100) and DAPI (Solarbio) were used for confocal microscopy.

5.7 ELISA and NO assay

Human plasma was isolated from blood samples collected by coagulation promoting. Plasma RTN3 levels were tested by self-produced RTN3 ELISA embedding assay. Eight microliters of plasma and 92 μL embedding buffer (2.93 g NaHCO₂ and 1.59 g Na₂CO₂ dissolved in ultrapure water) were embedded in 96-well plates, washed, blocked, and incubated with primary antibody and secondary antibody, and ultimately, 3,3′,5,5′-tetramethylbenzidine substrate (Solarbio) and stop buffer were added to measure in 450 nm absorbance.

S1P was tested in mouse plasma isolated from blood sampled by eyeball extirpating, in the aortic tissue homogenate of mice dissolved in PBS, and in the supernatant from cells cultured in complete medium for 48 h using S1P ELISA assay kits (Mlbio, Shanghai, China). A NO assay kit (Beyotime) was used to detect the NO content in the culture supernatant. The cell number was counted by a cell counter. Cells were collected, dissolved in PBS, and lysed by freeze–thaw cycles. The lysates were used to detect ceramide and cGMP using ceramide and cGMP ELISA assay kits (Mlbio). Details are provided in the Supplementary Materials and Methods.

5.8 Scraping line method, transwell, and cell tube formation assay

After transfection of siRTN3 into HUVECs for 24 h, a straight line was drawn in each well of 12-well plates, where the change in width was recorded every 24 h. Five thousand to ten thousand cells were seeded in transwell (Costar, Suzhou, China) and cultured for 24 h to check cell migration. Fifteen thousand cells were seeded in Matrigel (Corning, Corning, USA) for 3 h to check cell tube formation.

5.9 Real-time PCR

Total RNA was extracted using an RNA Extraction Kit (Qiagen, Dusseldorf, Germany) and stored at −80°C. Mouse RTN3 and GAPDH primer pairs (mouse RTN3 real-time PCR f: 5′-GGTAGAAGACTTGGTTGACTCC-3′; mouse RTN3 real-time PCR r: 5′-GGCGAGAATCAGAAGGGTAAT-3′; mouse GAPDH real-time PCR f: 5′-CCCTTCATTGACCTCAACTACA-3′; mouse GAPDH real-time PCR r: 5′-CCCTTCATTGACCTCAACTACA-3′) were designed. Total RNA was reverse transcribed using a RevertAid First Strand cDNA Synthesis Kit (Thermo, Waltham, USA) and then subjected to real-time PCR with 2xSYBR Green qPCR Mix (Thermo).

5.10 Genetic screening

Genomic DNA was extracted using the DNeasy Blood & Tissue Kit (Qiagen, Valencia, USA). WES was performed by Berry Genomics Company (Chengdu, China) using the Illumina HiSeq4000 platform (Illumina, San Diego, USA), and WES data were filtered as described previously.³⁹ The identified RTN3 variant was verified by Sanger sequencing. RTN3 primer pairs (RTN3 f: 5′-GCCAGTTGCCGGATTATTCTA-3′; RTN3 r: 5′-TGCACAACCTTACGTTCCC-3′) were designed by IDT browser.

RTN3 amino acid sequences of multiple species were obtained from NCBI (https://www.ncbi.nlm.nih.gov/protein/?term=RTN3). The structure of the human RTN3 protein fragment (Q7RTN4) was downloaded from the AlphaFold database (https://alphafold.ebi.ac.uk/entry/Q7RTN4), and the mutant RTN3 model was constructed using PyMol.

5.11 Statistical analysis

Data were subjected to statistical analysis using Graph Pad Prism 8. The mean ± SEM were calculated based on at least three independent experiments. Two-tailed Student's t-tests and ANOVA were used for two-group comparisons. Differences were considered statistically significant at p ≤ 0.05.