All animal manipulations were approved by the Institutional Animal Care and Use Committee of Central South University, Changsha, China (#2019sydw0021). All efforts were made to minimize mouse suffering. Eight-week‐old male C57BL/6J mice (wildtype, WT) and phosphofurin acidic cluster sorting protein 2 (PACS2) knockout mice (PACS2^−/−) were obtained from the Nanjing Biomedical Research Institute of Nanjing University and housed under standard pathogen-free conditions at the Xiangya Medical School of Central South University.

After 1 week of acclimatization, the mice were randomly assigned to the standard chow diet (ND) group and the STZ plus HFD (STZ/HFD) group, with five mice each. The latter was fed HFD (D12109C, Research Diets, New Brunswick, NJ, USA) and intraperitoneally injected 35 mg/kg STZ (S0130, Sigma‐Aldrich, San Louis, MO, USA) twice between 9 and 10 weeks. STZ was dissolved in sodium citrate-hydrochloric acid buffer solution, pH 4.5 (SSC) (Feng et al., 2019; Zhang et al., 2019; Kim et al., 2020). The ND group was injected with the same volume of SSC. Blood glucose was measured at week 11, and mice with blood glucose >11.1 mM were considered to have diabetes. The flow chart of establishing the HFD diabetic mouse model is shown in Supplementary Figure S1A.

The results of population studies have shown that patients with diabetes mellitus complicated with hyperlipemia have a higher incidence of rapid progression of diabetic vascular impairment (Haile and Timerga, 2020; Gebreegziabiher et al., 2021). For these patients, glucose control alone cannot reduce the final incidence of ESRD (Coca et al., 2012; Peltier et al., 2014). We believe that glucose and lipid overload are jointly involved in the pathogenesis of high-risk diabetes patients. Therefore, HGHF conditions were created in this study to simulate the disease state (Sun et al., 2021; Wu et al., 2021).

Primary human umbilical vein endothelial cells (HUVECs) were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured in an Endothelial Cell Medium (Invitrogen, Carlsbad, CA, USA) containing 5% fetal bovine serum at 37°C in a 5% CO₂ incubator. Cells were treated with 30 mM glucose and 0.1 mM palmitic acid (P0500, Sigma-Aldrich, USA) dissolved in 0.5% bovine serum albumin (BSA) for 48 h to simulate HGHF treatment. The Akt inhibitor MK-2206 2HCl was purchased from Selleck Chemicals (S1078, Houston, TX, USA).

Blood was drawn through the submandibular vein at 11, 14, 17, and 20 weeks to measure serum biochemistry parameters, including total cholesterol (CHO), triglyceride (TG), blood glucose, creatinine (CRE), and blood urea nitrogen (BUN). The 24-h urine microalbuminuria (u-mALB) was measured once a week from the 8th week, but no urine was collected 1 week after each blood collection to avoid the effect of transient blood loss on renal function. All mice were sacrificed at week 20, and kidney tissues were fixed in 4% paraformaldehyde and stained with hematoxylin-eosin (HE) and Masson trichrome.

To verify the function of PACS2 in HUVECs, PACS2 siRNA was delivered into cells according to the manufacturer’s instructions. Briefly, cells were transfected with 50 nM PACS2 siRNA or negative control (Ribobio, Guangzhou, Guangdong, China) using Lipofectaminutese 3,000 reagent (Invitrogen) in Opti-MEM I reduced serum medium (ThermoFisher Scientific, Grand Island, NY, USA) for 12 h, and then stimulated with HGHF for another 48 h.

After deparaffinization and rehydration through 100, 95, and 70% ethanol, kidney tissue slides were serially stained in Weigert’s iron hematoxylin working solution for 10 min and Biebrich scarlet-acid fuchsin solution for another 10–15 min. To differentiate nuclei, slides were submerged into the phosphomolybdic-phosphotungstic acid solution for 10–15 min or until the collagen was not red. Slides were then directly transferred into blue aniline solution and stained for 5–10 min. Following differentiation in 1% acetic acid solution for 2–5 min and rapid dehydration through 95% ethyl alcohol and absolute ethyl alcohol, slides were dipped into xylene and finally mounted with a resinous mounting medium.

As an indicator of vascular permeability, albumin extravasation was evaluated in the Miles assay by measuring the extravasation of albumin-bound Evans blue. In short, mice were anaesthetized with isoflurane inhalation. Five min later, 150 µL of Evans Blue (30 mg/ml; Sigma-Aldrich) was anaesthetized via the tail vein and allowed to circulate for 30 min, then extravasated Evans blue dye was evaluated (Shan et al., 2017). Following photographing the extravasation of Evans blue in the coronal section of kidneys. The kidneys were dried on foil at 150°C for 48 h. To extract Evans blue from the tissue, 500 µL formamides was added and incubated at 55°C for 72 h. Plasma extravasation was quantified by measuring absorbance at 620 nm and calculated per gram of dry weight.

Confluent HUVECs were treated with or without HGHF for 48 h on type I collagen-coated culture inserts with 0.4-µm pores (BD Bioscience, Franklin Lakes, NJ, USA). After that, 15 µL of 5 mg/ml FITC-dextran 40 (Sigma-Aldrich) was added to the upper chamber and incubated at 37°C for 15, 30, and 60 min. The fluorescence intensity of FITC-dextran diffused into the lower chamber was measured at Ex/Em = 485/590 nm.

Endothelial cell permeability was detected by using an In Vitro Vascular Permeability Assay kit following the manufacturer’s protocol (Millipore, Burlington, MA, USA). After forming a tighter monolayer and exposure to HGHF, the cells were stained with FITC-dextran 40 for 20 min. The amount of FITC-dextran 40 (Aman et al., 2012) diffused across the endothelial monolayer was observed under a fluorescent microscope with a FITC filter and quantified on a fluorescence plate reader.

Leaked ratio = Transwell lower chamber fluorescence intensity (Ex₄₈₀/Em₅₉₀)/upper chamber fluorescence intensity (Ex₄₈₀/Em₅₉₀).

The kinetic changes of extracellular oxygen levels, an indicator of aerobic metabolism in living cells, were measured using the MitoXpressXtra oxygen consumption assay kit (Luxel Bioscience, Santa Clara, CA, USA) (Du et al., 2019). Antimycin A, a potent electron transport chain complex III inhibitor, was used to shut down mitochondrial respiration (zero oxygen consumption control). The uncoupling agent carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP) that disrupts the mitochondrial proton gradient was employed to drive its maximal mitochondrial respiration rate. In principle, the phosphorescent signal of MitoXpressXtra is quenched by oxygen and produces a signal that is inversely proportional to the amount of oxygen present. Time-resolved fluorescence (TR-F) was measured at Ex/Em = 380/650 nm and the recommended delay time. The parameters of oxygen consumption included PA-based maximal oxygen consumption rate (PA-OCR) obtained by adding 2.5 μmol/L FCCP plus PA-only-based substrate and negative control (NegOCR) obtained by adding 1 μmol/L antimycin A in the presence of PA-only-based substrate.

Cells were seeded in triplicate in 96-well plates and treated with firefly luciferase for the indicated time. Cellular ATP levels were evaluated using an ATPlite assay (Perkin-Elmer, Waltham, MA, USA) (Liu et al., 2019a).

Intracellular NADPH/NADP ratio was assayed using an NADP/NADPH quantification colorimetric kit (K347‐100,BioVision, Milpitas, CA, USA) as per the manufacturer’s instructions.

Kidney tissues were fixed in 4% glutaraldehyde at 4°C for 48 h, and then post-fixed in 1% osmium tetroxide at room temperature for 2 h. After pre-staining with barbitone acetate for 10 min, histological samples were dehydrated with acetone and embedded in paraffin wax. Immunohistochemical staining was performed to detect PACS2 (PA5-100167, ThermoFisher) and CD31 (ab28364, Abcam, Cambridge, UK) in endothelial cells and visualized using fluorochrome-conjugated goat anti-mouse (ab150115, Abcam) or anti-rabbit IgG H&L (ab150077, Abcam). PACS2-positive cells were counted and normalized to total endothelial cells in the same field using computer-assisted morphometric analysis.

Chemiluminescent based on horseradish peroxidase (HRP) were used to detect and visualize western blots. The following antibodies were used: anti-CPT1α (12252S, Cell Signaling Technology, Beverly, MA, USA), PACS2 (GTX17244, GeneTex, Irvine, CA, USA), VE-cadherin (ab33168, Abcam), β‐actin (A5441, Sigma‐Aldrich), AKT (2938S, Cell Signaling Technology), p-AKT (9018S, Cell Signaling Technology), Smad2 (5339T, Cell Signaling Technology), p-Smad2 (ab280888, Abcam, UK), PPARα (ab215270, Abcam, UK), and HRP conjugated goat anti-rabbit IgG secondary antibody (AS09 602, Agrisera, Vannas, Sweden).

Cells were isolated on nickel fitters, stained with 2% uranyl acetate for 10 min, and then stained with Reynold’s lead citrate for 5 min. The endoplasmic reticulum (ER)-mitochondria contacts of endothelial cells were evaluated by transmission electron microscopy (TEM; Hitachi-7650, Tokyo, Japan) at 60 kV. The ER-mitochondrial contacts were quantified as described previously (Wu et al., 2019). The images were analyzed using ImageJ (National Institutes of Health, Bethesda, MD, USA). The mitochondrial and ER membranes were delineated using the freehand tool. The selected areas were converted to a mask, and the perimeters of ER were calculated. Two independent investigators blindly quantified the images. To quantify MAMs, the total ER perimeter connected to mitochondria to the total ER perimeter was normalized.

HUVECs were labelled with ER-Tracker Blue-White DPX dye (E12353, ThermoFisher Scientific) and Mito Tracker Deep Red FM (M22426, ThermoFisher Scientific) at 37°C for 30 min and observed under a confocal microscope (LSM800, Carl Zeiss Microscopy, Cambridge, MA, USA). The Pearson correlation coefficient mode, a well-defined and generally accepted means of describing the overlap between image pairs, was applied to quantify the degree of co-localization between the fluorophores representing ER-Tracker Blue and Mito Tracker Deep Red. The Pearson correlation coefficient was analyzed using the built-in Carl Zeiss co-localization analysis module from the ZEN software and the threshold obtained from single-label control samples.

Total proteins were extracted from HUVECs in the presence or absence of HGHF. Peptide samples (2 μg) were separated and analyzed by Kangchen Biotech (Shanghai, China) using the EASY-nLC1200 system and Q Exactive mass spectrometer (120 min/sample), respectively (Zhu et al., 2020). Differentially expressed peptides were identified at p < 0.05 and fold-change > 2, followed by gene ontology (GO) enrichment analysis by using Blast2Go 4.0.7 software.

Each cell sample is slowly thawed at 4°C, mixed with 1 ml of cold methanol/acetonitrile/H₂O (2:2:1, v/v/v) and vortexed adequately. The homogenates were sonicated for two cycles at low temperature, 30 min/cycle, incubated at -20°C for 60 min to precipitate proteins and centrifuged at 13,000 rpm at 4°C for 15 min. Supernatants were collected, dried under vacuum, and stored at -80°C. The dried samples were re-dissolved in 100 µL acetonitrile/water (1:1, v/v), vortexed, and centrifuged. The resulting supernatants were used for LC-MS/MS analysis, performed at Applied Protein Technology (Shanghai, China). The metabolites were blasted against the online Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://geneontology.org/) to retrieve their COs and were subsequently mapped to pathways in KEGG. The corresponding KEGG pathways were finally extracted.

The SPSS 22.0 software (SPSS, Chicago, IL, USA) was used for statistical analysis. Data are expressed as mean ± standard deviation (SD). Statistical differences were assessed by Student t-test of the means between two groups or by one-way analysis of variance of the means between multiple groups. p < 0.05 was considered a significant difference.

In order to observe the pathological changes of kidney vascular vessels, we created a diabetic mouse model. Hyperglycemia and hyperlipidemia developed after 14 weeks of HFD feeding plus low-doze STZ injection (Figure 1A). Blood CRE, BUN, and u-mALB levels increased significantly at week 20, indicating that the kidney of the HFD group was dysfunctional. Consistent with the serum and urinary biochemical analysis results, HE and Masson trichrome staining showed pathological changes in renal tissues, manifested as atrophy of glomerular vascular loops (Figure 1B) and renal microvascular fibrosis (Figures 1C,D). Moreover, Evans blue extravasation demonstrated an increased renal tissue extravasation index in the HFD group (Figures 1E,F). TEM observation of renal endothelial cells showed a significant increase in MAMs of renal cortical glomerular endothelial cells, accompanied by mitochondrial swelling and crista degeneration (Figure 1G). These results indicate that mitochondria are involved in endothelial cell dysfunction induced by HGHF.

To explore the underlying mechanism behind endothelial cell dysfunction in diabetic mice, we next examined the barrier function changes of endothelial cells. As shown in Figure 2A, FITC-dextran permeability was significantly increased in the HGHF group from 15, 30 until 60 min. VE-cadherin is the central adhesion molecule of endothelial cell-cell junctions (Lampugnani et al., 2018). The dissociation of p120-catenin and β-catenin from VE-cadherin complexes will increase VE-cadherin internalization and weaken the endothelial barrier (Yang et al., 2015; Juettner et al., 2019). The immunofluorescence of VE-cadherin showed a significant decrease from the cell membrane in Figure 2B, indicating that the adherent contact between endothelial cells was damaged.

By mining the MS reference database, protein identification was performed on data-independent acquisition (DIA) data results (identification criteria: precursor threshold, 1.0% false discovery rate (FDR) and protein threshold, 1.0% FDR). We identified 26,201 peptides and 3,678 proteins in the two sets of samples. The identified proteins were screened for differentially expressed ones between the ND and HGHF groups. The relative fold change of expression >1.2 and the Student t-test q-value < 0.05 were used as filter criteria to process DIA quantitative data. On this basis, 320 differentially expressed proteins were selected out (Supplementary Figure S2A). Of them, 138 were upregulated and 182 downregulated after HGHF treatment. GO annotation analysis showed that they were mainly enriched in the unfolded protein response in the ER, and the mitochondrial translational elongation and termination processes (Figure 2C) indicate mitochondrial damage or ER stress. Proteomics also supported that the proteins were enriched in the mitochondria and ER (Supplementary Figure S2B). Several proteins related to renal damage and inflammation increased, consistent with the chronic renal inflammation state (Supplementary Figure S2C). We also found that the expression of the integrin-related proteins (like ITGB4, ITGA3) and cell-cell adhesions protein MLLT4 had down-regulated significantly (Figure 2C), which indicates damage to membrane integrity (Tremblay et al., 2014). Further, we verified the potential changes in mitochondria-ER contact in vivo. The same cell samples were observed under TEM, showing enrichment of ER and MAMs (Figure 2D) under HGHF. Double staining of mitochondria and ER revealed their more significant co-localization in HGHF-treated HUVECs (Figures 2E,F).

Based on the TEM and immunofluorescence observations of renal endothelial cells in vivo and HGHF-treated HUVECs in vitro, we further explored the expression changes of three MAMs regulatory proteins, acyl-CoA synthetase 4 (FACL4) (Lewin et al., 2002), PACS2 (Yang et al., 2021), and glucose-regulated protein 75 (GRP75) (Liu et al., 2019b), in HUVECs under HGHF exposure. Among these regulatory proteins, PACS2 was significantly increased in HGHF-induced HUVECs (Figures 3A,B), and it also had an inevitable increase in diabetic mouse renal vascular tissues (Figure 3C). To further understand the role of PACS2, we silenced PACS2 in HUVECs using siRNA (Figures 3D,E). We found knocking down PACS2 could attenuate HGHF-enhanced co-localization of mitochondria and ER (Figures 3F,G). These results indicate that PACS2 is an active regulator of MAMs and may be involved in endothelial cell dysfunction in HGHF-treated endothelial cells.

PACS2 knockout mice (PACS2^−/−) were used to confirm the effect of PACS2 on vascular endothelium under HGHF in vivo. Diabetic PACS2^−/− mouse modeling was created the same as depicted in Supplementary Figure S1. Microscopic immunofluorescence observations verified the gene knockout efficiency (Figure 4A). We observed that the HFD treatment altered the biochemical parameters of renal function in WT mice. However, PACS2 knockout eliminated the increase in blood glucose, TG, CRE, and BUN caused by HFD treatment. However, it did not affect CHO (Figure 4B), indicating that PACS2 is implicated in HFD-induced renal dysfunction.

Compared to HFD-treated WT mice, PACS2-deficient mice showed a significant reduction in HFD-induced glomerular fibrosis in Masson staining (Figures 4C,D). In addition, we found that the glomerular cortex endothelial cells and vascular endothelial cell MAMs in the kidney tissues of PACS2^−/− + HFD mice were significantly reduced (Figure 4E), as was the glomerular Evans blue leakage index (Figures 4F,G), suggesting that PACS2 plays a role in HFD-induced kidney injury by regulating MAMs.

In order to explore how HGHF-induced endothelial PACS2 and MAMs vary, the upstream regulators of PACS2 were then examined. Recent reports noted that AKT could mediate vascular barrier leakage and inhibit renal fibrosis (Lewin et al., 2002; Yang et al., 2021) and controlling MAMs integrity (Peltier et al., 2014). Therefore, we evaluated whether AKT works upstream of PACS2 in HUVECs treated with HGHF.

Figures 5A,B show that HGHF induced endothelial cell fibrosis, indicated by increased FN expression. At the same time, the phosphorylation levels of AKT and Smad2 were significantly increased (Figures 5A,B), suggesting that HGHF activates both AKT and Smad2 signaling pathways. Extracellular matrix (ECM) is secretable from endothelial cells, and its accumulation is the pathological basis of DN. However, the transcription factor Smad2 signaling mediates endothelial-related ECM deposition and renal fibrosis (Weng et al., 2020). To understand the potential role of PACS2 in the activation of AKT and Smad2, we silenced PACS2 and then treated cells with HGHF. We found that knocking down PACS2 reduced FN expression and Smad2 phosphorylation but did not inhibit AKT phosphorylation, indicating that PACS2 regulates Smad2 signaling during HGHF-induced endothelial cell fibrosis. Interestingly, MK2206, a selective AKT inhibitor, decreased HGHF-induced PACS2 expression and Smad2 phosphorylation, indicating that AKT signaling works upstream of PACS2 regulation under HGHF treatment (Figures 5C,D). These results indicate that HGHF-induced PACS2 upregulation in endothelial cell dysfunction is partly through the AKT signaling.

Recent research reported that PACS2 could regulate cellular energy metabolism, in which the free fatty metabolism was also related to cell-cell junction (Xiong et al., 2018). In order to understand how AKT/PACS2 signaling impacts endothelial barrier function, we explored the metabolic index in endothelial cells. Non-targeted metabolomics were tested under HGHF induced HUVECs and FAO’s key enzyme, carnitine palmitoyl-transferase 1α (CPT1α), was examined in diabetic kidney tissue. KEGG pathway analysis results revealed that most differential metabolites were enriched in the metabolism pathways (Figure 6A). Significant changes were found in adenosine monophosphate, adenosine 5′-diphosphate, glycerol 3-phosphate (phosphoglycerol 3), N-acetylglucosamine-1-phosphate, and phosphorylcholine (Supplementary Figure S2), all of which are related to fatty acid metabolism, indicating that fatty acid metabolism is easily affected by HGHF and has a role in HGHF-induced HUVEC damage.

Next, we measured FAO ability through mitochondrial respiration using PA as an energy substrate. We evaluated FAO-dependent OCR (PA-OCR) by adding FCCP or etomoxir (ETO) into cultured HUVECs. The results showed that PA-OCR was significantly reduced after HGHF exposure. However, silencing PACS2 reversed the inhibition of PA-OCR by HGHF (Figure 6B). Quantitative NADPH/NADP⁺ (Figure 6C) and ATP (Figure 6D) analyses showed that the NADPH/NADP⁺ ratio and ATP production in HUVECs was significantly reduced after HGHF treatment. However, knocking down PACS2 prevented the effect of HGHF on the NADPH/NADP⁺ ratio and showed a tendency to block the effect of HGHF on ATP production. These results suggest that HGHF interferes with mitochondrial aerobic respiration, where PACS2 is a positive regulator.

We also found that, contrary to the changes in PACS2 expression, CPT1α expression was significantly reduced under HGHF stress but was rescued by PACS2 knockdown (Figures 6E–G). This result was consistent with the CPT1α expression in kidney tissues of diabetic mice fed with HFD and found that its protein expression was reduced (Figure 6H) in Oil red-positive tissues (Figure 6I). These results further suggest that PACS2 regulates fatty acid metabolism, possibly by disturbing the balance of endothelial redox homeostasis through NADPH production.

As observed above, HGHF affects endothelial FAO. However, it is unclear if PACS2 regulation of barrier function under HGHF is reliable for the FAO change. Therefore, the FAO inhibitor ETO were used to block FAO in endothelial cells to test this. We found that ETO increased VE-cadherin internalization (Figure 7A) and FITC leakage (Figure 7B) under non-HGHF conditions. Moreover, PACS2 knockdown attenuated the effect of HGHF on FITC leakage and VE-cadherin internalization (Figure 7B). However, this PACS2 effect was not seen when ETO was present. These results indicate that FAO is downstream of PACS2 in regulating endothelial barrier function under HGHF.