Male Wistar rats with initial body weights of 200–250 g were purchased from Beijing HFK Bio-Technology Co., Ltd. All animals were housed in a room with constant temperature and a 12:12-h light–dark cycle. They were allowed free access to a standard diet and tap water. The rats were allowed to acclimatize for 1 week before the experiment began. Sham control rats underwent a sham operation consisting of an incision in the skin and muscle in the right renal region and manipulation of the renal pedicles, without damage to the kidney. To accelerate the DKD development, the other rats were subjected to right uninephrectomy at 8 weeks of age (Huang et al., 2009). One week after uninephrectomy, diabetes was induced by a single intraperitoneal injection of streptozocin (STZ; Sigma-Aldrich, St. Louis, MO, United States), diluted in a citrate buffer (0.1 mol/L, pH 4.0), at a 45 mg/kg dose. Finally, 16 rats developed hyperglycemia, with blood glucose levels over 16.7 mmol/L, at 72 h after STZ injection. After confirmation of hyperglycemia in the diabetic rats, all animals were randomized into four groups: (1) sham (n = 6), (2) uninephrectomy (unx; n = 6), (3) diabetes with uninephrectomy (DKD; n = 8), and (4) diabetes with uninephrectomy treated with SAL (DKD + SAL; n = 8). The rats in the sham and unx groups, which were not induced with STZ, were used as non-diabetic controls. SAL (purity > 98%; National Institutes for Food and Drug Control, Beijing, China) was administered daily by gavage at a dose of 70 mg/kg body weight for 8 weeks in the SAL-treated rats (Zheng et al., 2015), while the other groups received the vehicle control without SAL. Blood glucose levels were monitored at least weekly in all diabetic rats by tail-vein blood sampling. After 8 weeks, one rat died in the DKD group. The rats were housed individually in metabolic cages for urine collection. Within 1–2 days after the last urine collection, the animals were sacrificed. Blood samples were obtained, and the left kidney was immediately removed. Part of the kidney tissue was fixed in 4% paraformaldehyde, while the remaining tissue was stored at -80°C. The study was conducted in accordance with the Guiding Principles for the Care and Use of Laboratory Animals of China, and the protocol was approved by the Ethics Committee of Shandong Provincial QianFoShan Hospital, China.

Renal function was assessed by measuring the kidney index, 24-h urine protein and albumin, blood urea nitrogen (BUN), and serum creatinine (SCr) of the rats. The kidney index (in mg/g) was calculated as a ratio of the left kidney’s weight to the body weight (K/W). Urine protein was assessed by the Bradford method, while urine albumin was measured using an enzyme-linked immunosorbent assay kit (CUSABIO Engineering Co., Wuhan, China). Plasma biochemical parameters were measured using an automatic biochemical analyzer (Chemray 240; Rayto, Institute of Biotechnology, Shenzhen, China).

The removed kidney tissues were fixed in 4% paraformaldehyde and embedded in paraffin. Paraffin sections (3–4 mm) were stained with periodic acid-Schiff (PAS), periodic acid silver methenamine (PASM) and Masson’s trichrome. The sections were examined with light microscopy by two experienced pathologists. The index of mesangial expansion represented the percentage of PAS-positive area in the glomerulus. It was scored by a quantitative estimation of the width of mesangial zones at 40× power for 20 cortical fields. Injury to tubules was assessed by determining the percentage of affected tubules per 10 fields (magnification ×200) (Zhao et al., 2014). The scoring system was on a scale from 0 to 5 grades (0 = 0%, 1 = 5%, 2 = 5–10%, 3 = 10–20%, 4 = 20–30%, 5 = >30%) according to the following criteria: tubular dilation, tubular atrophy, vacuoles formation, and extracellular matrix accumulation (interstitial volume).

Cortical kidney tissue was cut into 1 mm³ cubes for standard Electron Microscopy processing. Photographs were taken with transmission electron microscope (JEM-1200EX, Japan). Five random photographs with a final magnification of 15,000× were taken from each section.

The genes related to DKD were selected from six existing databases: (1) the DrugBank database (Wishart et al., 2008), (2) the Comparative Toxico genomics database (CTD) (Davis et al., 2018), (3) the Online Mendelian Inheritance in Man (OMIM) (Amberger and Hamosh, 2017), (4) the Therapeutic Target database (TTD) (Liu et al., 2011), (5) the Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway database (Kanehisa et al., 2017), and (6) the Genetic Association database (GAD) (Becker et al., 2004). Based on the inference score computed by the CTD database, we extracted the targets scored above 60.

The targets of SAL were extracted from the Herbal Ingredients’ Targets (HIT) database (Ye et al., 2011), the Swiss Target Prediction database (Gfeller et al., 2014), the STITCH 5.0 database (Szklarczyk et al., 2016) and the ChemMapper database (Gong et al., 2013). A Canonical SMILES (C1=CC(=CC=C1CCOC2C(C(C(C(O2)CO)O)O)O)O) was recorded for SAL (PubChem CID: 159278) from the PubChem database and separately uploaded to the servers. In the ChemMapper database, if the two-dimensional similarity of a molecule was above 0.6, the targets were predicted as the targets of SAL. Noteworthy, all the species of the targets was Homo sapiens. Then, to better dissect the role of SAL in DKD treatment, those targets were mapped to DKD-related genes to obtain the candidate targets of SAL.

To normalize the gene information, inconsistent ID types were converted to UniProt ID accession numbers. The associated genes of the candidate targets were obtained from the String database (Jensen et al., 2009). Among them, only the protein with a interaction score above 0.9 were carried on for constructing the protein–protein interaction (PPI) network using Cytoscape version 3.2.1 (Shannon et al., 2003). The topological property of each node in the network was calculated by Network Analyzer (Batt et al., 2012). A node with a degree value two times greater than the average degree (5.719) was considered to an important node (Zhang Y. et al., 2016). Cytoscape plugin Reactome (Vastrik et al., 2007) was employed to perform pathway analysis of the key nodes of SAL in DKD treatment. The P-value < 0.01, FDR < 0.01 were set as the cut-off criteria. Furthermore, the top 10 signaling pathways correspond to the criteria, were displayed as bubble charts by the E Chart online analysis platform a histogram.

HK-2 cells, a proximal tubule epithelial cell line, were purchased from PriCells (Wuhan, China) and cultured as described previously (Cai et al., 2017). The cells were incubated in media containing 5.5 mM D-glucose + 24.5 mM D-mannitol (MG), 30 mM D-glucose (HG), or 30 mM D-glucose + 100 μM SAL (HG + SAL). D-glucose and D-mannitol were purchased from Sigma-Aldrich Canada Co. (Oakville, ON, Canada).

The short hairpin RNA (shRNA) targeting human BIM (5′-CGGAGACGAGTTTAACGCTTA-3′) was designed, synthesized, and cloned into the hU6-MCS-ubiquitin-EG FP-IRES-puromycin vector (BIM-shRNA) by the GeneChem Corporation (Shanghai, China). The hU6-MCS-ubiquitin-EGFP-IRES-puromycin vector (shNC) was used as a negative control. After confirming successful transfection, HK-2 cells were cultured in high glucose medium for 48 h and then harvested for RNA or protein extraction.

HK-2 cells were plated on 96-well plates at a density of 3 × 10⁴ cells/cm² in RPMI 1640 medium. Firstly, cells were incubated in 30 mM D-glucose medium for 24, 48, and 72 h, respectively. The cell viability was assessed by the Cell Counting Kit-8 assay (Dojindo Molecular Technologies, Inc., Kumamoto, Japan) according to the manufacturer’s protocol. Secondly, the cells were incubated in 5.5 mM D-glucose + 24.5 mM D-mannitol (MG), 30 mM D-glucose (HG), or 30 mM D-glucose added to SAL at various concentrations (0.1, 1, 10, 100, 1,000, and 10,000 μM). After 72 h, cell viability was assessed by the Cell Counting Kit-8 assay. Absorbance of the samples was measured at a wavelength of 450 nm.

The apoptotic cells were determined by the terminal deoxynucleotidyl transferase-mediated dUDP nick-end labeling (TUNEL) technique using an in situ cell death detection kit (Roche). Paraffin sections of the kidney were cut at 3 μm and deparaffinized for staining. The paraffin sections were assessed by the TUNEL assay according to the manufacturer’s recommendations. At least 10 renal cortex fields per slide and six slides per group were scored for apoptotic nuclei. The apoptotic cells from tubular microscopic fields (magnification ×400) were counted. Data were expressed as positive cell count, which is the mean of cells positive for apoptotic cell microscopic field (An et al., 2013; Xiao et al., 2014).

HK-2 cells were seeded on sterile glass coverslips in a 6-well plate and stimulated with 5.5 mM D-glucose + 24.5 mM D-mannitol (MG), 30 mM D-glucose (HG), or 30 mM D-glucose + 100 μM SAL (HG + SAL) for 48 h. In addition, HK-2 cells transfected with shNC and BIM shRNA were incubated in high glucose for 48 h. The HK-2 cells were assessed by the TUNEL assay according to the manufacturer’s recommendations. TUNEL-positive cells were counted under the light microscope by two independent pathologists. The rate of apoptosis in HK-2 cell was calculating by dividing the number of TUNEL positive cells to a population of 100 counted cells per microscopic field (magnification ×40 or ×100).

An Annexin V–fluorescein isothiocyanate (FITC) apoptosis detection kit (Neobioscience, Inc., Shenzhen, China) was used to assess the apoptosis. HK-2 cells were seeded in 6-well plates. The adhered cells were exposed to different media for 24 h or 72 h and then detached with trypsin. Subsequently, the cells were resuspended in a binding buffer (195 mL) and stained with Annexin V–FITC (5 μL) and propidium iodide (PI; 10 μL) at room temperature for 10 min in the dark. Apoptotic cells were determined using a flow cytometer (BD FACSAria^TM II cell sorter). Annexin V–FITC-positive and PI-negative cells were considered apoptosis.

HK-2 cells were stimulated with different media for 48 h, and then RNA isolation and cDNA synthesis were performed as previously described (Cai et al., 2017). Gene expression was measured by real-time polymerase chain reaction (PCR) using 0.2 μM gene-specific primers and the UltraSYBR One Step RT-qPCR kit (low ROX) (Takara Bio, Inc., Otsu, Japan) in a total volume of 20 μL. PCR was performed using an Applied Biosystems ViiA 7 sequence detection system. Relative expression levels of the BIM gene were determined by the ΔΔCt method, whereby the target gene expression was normalized to that of β-actin as an endogenous reference and then compared to the control. Each experiment was repeated three times. The primer sequences were as follows: BIM forward (5′-ATTACCAAGCAGCCGAAGAC-3′) and reverse (5′-TCCGCAAAGAACCTGTCAAT-3′); and β-actin forward (5′-TGACGTGGACATCCGCAAAG-3′) and reverse (5′-CTGGAAGGTGGACAGCGAGGT-3′).

Proteins were isolated from cultured HK-2 cells and animal renal cortices. Cell lysates (30–100 μg) were analyzed by immunoblotting as described previously (Zhang et al., 2017). The antibodies used were anti-BIM (monoclonal) from Abcam (1:500), anti-BAX (1:1,000, polyclonal) and anti-cleaved caspase-3 (1:500, polyclonal) from Affinity, and anti-β-actin from ZsBio (Beijing, China). Quantification was performed by measuring the signal intensity using the ImageJ software (National Institutes of Health, Bethesda, MD, United States).

For immunohistochemistry, the antibodies used in this study included a B-cell lymphoma 2 (BCL-2)-interacting mediator of cell death (BIM) rabbit antibody (1:100; Abcam, Cambridge, MA, United States), neutrophil gelatinase-associated lipocalin (NGAL) rabbit antibody (1:100, Abcam, Cambridge, MA, United States); a BCL-2-associated X (BAX) rabbit antibody (1:100; Affinity, Chicago, IL, United States), and a BCL-2 rabbit antibody (1:100; Affinity). Slides were incubated with the antibodies for 12 h at 4°C. Negative controls were incubated with phosphate-buffered saline (PBS). The bound antibodies were detected with horseradish peroxidase-conjugated anti-rabbit IgG and diaminobenzidine. Finally, the slides were counterstained with hematoxylin. All images, acquired from randomly selected microscopic fields, were measured using Image-Pro Plus 6.0 (Media Cybernetics, Silver Spring, MD, United States) as described previously (Tervaert et al., 2010).

For immunofluorescence, HK-2 cells were seeded on sterile glass coverslips in a 6-well plate and stimulated with different media for 48 h. The slides were fixed for 10 min with 4% paraformaldehyde, followed by permeabilization with 0.05% Triton X-100 in PBS for 15 min. Thereafter, the slides were incubated with the BIM rabbit antibody (1:150; Abcam) at 4°C for 24 h, followed by incubation with a secondary antibody (1:200 in PBS) for 1 h at room temperature in the dark. Finally, the slides were stained with diphenyleneiodonium. Images were examined under a fluorescence microscope (Olympus FSX100).

Data were presented as means ± standard error of the mean. SPSS 21.0 software was used for statistical analysis. Student’s t-test was used to assess significance for data between two groups. One-way ANOVA and two-way ANOVA test with subsequent post hoc Tukey’s test were used for multiple comparisons. Kruskal–Wallis test was used for non-parametric data comparison. P < 0.05 was considered indicative of statistical significance.

There are no significant differences in body weights among groups at baseline. However, after STZ injection, diabetic rats showed an obvious reduction in body weight over the 8-week study period (Figure 1A). And there were significant increases in 24-h proteinuria and albuminuria in the DKD group at the end of week 8. SAL treatment (70 mg/kg) had no obvious effect on body weight but significantly reduced 24-h proteinuria (P < 0.05 vs. DKD group) and albuminuria (approximately 66.7%, P < 0.01 vs. DKD group, Figures 1B,C).

As expected, the results summarized in Table 1 showed that DKD rats exhibited a severe polyuria (198.12 ± 18.56 mL/24 h) when compared to non-DKD rats (10.56 ± 3.31 and 9.67 ± 1.43 mL/24 h), and this parameter was reduced by approximately 16.7% in DKD rats treated with SAL. Kidney weight/body weight ratio was 5.05 ± 0.44 mg/g in uninephrectomy rats and was significantly increased in DKD rats (∼2-fold). However, treatment with SAL failed to reverse the increase. In addition, DKD rats exhibited a significant increase in SCr (∼2.3-fold) and BUN (∼2.4-fold), when compared to non-DKD rats. SAL tended to reduce the levels of SCr (P < 0.05 vs. DKD group) and BUN (P < 0.01 vs. DKD group). These results suggest that SAL could decrease urine protein and be helpful in relief of DKD.

Morphologically, representative images of histopathological slides were demonstrated in Figure 1D. There were no obvious differences in histological characteristics between the sham group and unx group. In contrast, the pathology of untreated DKD rats were characterized by mesangial matrix expansion (Figures 1D, a1–a4), glomerular hypertrophy (Figures 1D, b1–b4), the fusion and effacement of foot processes (Figures 1D, c1–c2), vacuolar formation of renal tubules (Figures 1D, d1–d4), and extracellular matrix accumulation (Figures 1D, e1–e4), which were less serious in SAL-treated rats. Specifically, treatment with SAL for 8 weeks inhibited glomerular mesangial matrix expansion (approximately 50%, P < 0.05, vs. DKD group) (Figure 1E). In the present study, we also made a quantitative analysis of renal tubular injury. As shown in Figure 1F, SAL treatment significantly decreased the tubular injury induced by diabetes (P < 0.05, vs. DKD group). Neutrophil gelatinase-associated lipocalin (NGAL) is supposed to be a marker of active renal tubular injury (Kaul et al., 2018). The expression of NGAL in tubule was independently associated with GFR decline slope (Hwang et al., 2017), which added strong evidence to the hypothesis that tubular injury plays a critical role in the progression of DKD. In the present study, immunohistochemistry assays showed a substantial increase of NGAL in the renal tubule of DKD group, while SAL treatment significant decreased the expression (P < 0.01, Figures 1D, f1–f4, 1G). The results provided evidence for the protective effects of SAL on diabetic kidney injury.

Based on the above results, we employed in silico analysis to excavate the therapeutic targets of SAL against DKD based on the existing databases (Wang T. et al., 2018) (Supplementary File S1, Websites). After deleting repeated targets, we collected 259 DKD-related genes from six databases (Supplementary Table S1) and 61 SAL-related targets from four databases (Supplementary Table S2). Then, the 61 targets were mapped to the DKD-related targets to obtain shared targets. Consequently, 15 targets were overlapped, including MMP1, CASP3 AKT1, HIF1A, IL10, MMP9, PTGS2, BCL2, BAX, MMP2, PTGS1, SLC6A2, PRKCA, LGALS3, BCHE, which were predicted as the candidate targets for SAL in treatment DKD.

System biology studies have shown that proteins and proteins are interconnected. In order to explore the mechanism of drug treatment systematically, not only the candidate targets should be screened from the database, but also the targets associated with them. PPI network has become an important tool for screening indirect targets and evaluate the role of the targets in complex diseases (Wu et al., 2014; Sneha et al., 2018). In the present study, 101 genes associated with the 15 candidate targets were acquired from the String database, which is an efficient and accurate platform that can be used for the investigation of the PPI. The PPI network of SAL against DKD was consisted of the 116 genes through 344 interactions with an average degree of 5.719 (Figure 2A). In order to elucidate the role of the 116 targets in the PPI network of SAL in treatment DKD, the topological property of each node in the network was calculated by Network Analyzer. Generally, nodes with large degree value were considered as the important nodes of the network. As shown in Figure 2A, the 13 nodes with yellow color were considered as the important targets responsible for the nephroprotective effect of SAL, including seven candidate targets with diamond shape (BCL2, CASP3, BAX, MMP2, MMP9, MMP1, AKT1) and their related-genes (BCL2L11 (protein name, BIM), TP53, APAF1, TIMP1, VEGFA, JUN) with circle shape. These genes were enriched mainly in the signaling pathway related to apoptosis, inflammation and extracellular matrix (ECM) degradation, suggesting that SAL might protect against DKD through inhibiting cell apoptosis, inflammation and fibrosis (Figure 2B).

Base on the pathway analysis, we hypothesized that the intrinsic apoptosis pathway may be an important pathway of SAL in treatment DKD. Indeed, the pathogenesis of DKD is complicated, various cells, such as podocytes, mesangial cells, tubular cells, and fibroblasts, all participate in the process. It is now increasingly recognized that the apoptosis of proximal tubular cells has an important role in the pathogenesis and progression of DKD (Gilbert, 2017; Li et al., 2017). Thus, we chose the proximal renal tubular cells as the objects to verify the anti-apoptotic activity of SAL. We validated it in vitro and in vivo.

In vivo study, no differences were observed between sham group and unx group regarding tubular cell apoptosis, yet obvious apoptosis was detected in renal tubular cells both in proximal and distal convoluted tubules in the DKD group, especially the proximal tubule (Figure 3A, red arrows). However, the amounts of cells underwent apoptosis in glomerular regions were rare and there were less than three apoptotic cells per 50 glomerulus sections (Figures 3A, a1–b3). We found the apoptotic cells in the DKD groups were higher than those in the unx group (∼10-fold), which were largely inhibited by SAL treatment (approximately 50%, P < 0.01, vs. DKD group, Figure 3B).

To further investigate the direct anti-apoptosis effect of SAL on HK-2 cells, we performed cell experiment. Firstly, we examined the cell viability stimulated by HG. As shown in Figure 3C, the cell viability began to decline at 48 h (approximately 20%). After 72 h, there was a more significant decrease (approximately 30%) in viability in medium with 30 mM glucose as compared to the cells incubated in medium with 5.5 mM glucose (P < 0.01, Figure 3C). As expected, treatment with SAL (100 μM) significantly increased the cell viability (Figure 3D). Therefore, we selected 100 μM as the concentration of SAL for all subsequent in vitro experiments. Secondly, the results of Annexin V-FITC/PI assay showed that there were 3.50% ± 0.44% and 3.90% ± 0.44% apoptotic cells in the MG group and 3.0% ± 0.30% and 12.23% ± 0.87% apoptotic cells in the HG + SAL group, as compared with 6.43% ± 0.76% and 18% ± 0.52% apoptotic cells in the HG group, after 24 h and 72 h treatment, respectively (Figure 3E). The data showed that mannitol had no effect on apoptosis, indicating that the increased apoptosis of HK-2 cell did not result from high osmolarity. And SAL treatment could significantly reduce the apoptosis induced by HG in HK-2 cells (P < 0.01, vs. HG 24 h, vs. HG 72 h, respectively, Figure 3F). The results of TUNEL assay further confirmed the anti-apoptosis activity of SAL on HK-2 cells (Figures 3G,H).

Based on the silico analysis, these apoptosis related genes (BIM, BAX, CAPS3) were thought to be the important targets in SAL treatment. BIM, one of the major members of Bcl-2 family, is a pro-apoptotic protein with only one Bcl-2 homology (BH3) domain, and is a pivotal regulator inducing intrinsic apoptosis (Kuwana et al., 2005). BIM protein has three major isoforms, among which BimEL is the main isoform involved in the regulation of apoptosis (Akiyama and Tanaka, 2011). Cumulative evidence suggested that HG can increase BIM expression in multiple tissue cells and mediate cell apoptosis (Doiron et al., 2012; Shin et al., 2014).

In vivo study, we found BIM is mainly expressed in proximal renal tubules and is hardly expressed in the glomeruli (Figure 5A), which were in line with the study of O’Reilly et al. (2000). Our previous study had revealed the obviously up-regulated expression of BIM protein induced by HG in HK-2 cells (Zhang et al., 2017). In the present study, western blotting results showed that HG induced significant time-dependent increases in the expression of BIM, BAX, and cleaved caspase-3 proteins (Figures 4A,B). The results suggested that BIM may play an irreplaceable role in apoptosis of PTECs induced by diabetes.

Our previous study found BIM silence could restrain cytochrome c releasing from mitochondria induced by HG in HK-2 cells (Zhang et al., 2017). To better understand the function of BIM in HG-induced apoptosis, we silenced the endogenous BIM protein by shRNA transfection. As shown in Figures 4C,F, the mRNA and protein expression of BIM both decreased in BIM shRNA-transfected cells, suggesting that the transfected BIM shRNA effectively inhibited BIM expression. Subsequently, the expression of BAX and cleaved caspase-3 proteins both exhibited reduce in BIM shRNA-transfected cells compared with their significant increase in the shNC-transfected cells (both P < 0.01 vs. HG + shNC, Figures 4F,G). Furthermore, the result of apoptosis determined by TUNEL revealed that BIM silence completely reversed the apoptosis of HK-2 cells induced by HG (approximately, ∼10-fold). Those results indicated that BIM was a sufficient regulator to switch apoptosis in HK2 cells (Figures 4D,E).

According to Figures 5A,B, an obvious increase expression of BIM protein was detected in DKD group when compared to sham group or unx group by immunohistochemistry (P < 0.01), while treatment with SAL markedly decreased the expression of BIM (P < 0.01, vs. DKD). The western blot confirmed the finding (P < 0.05, vs. DKD, Figures 5C,D). Furthermore, the western blot also showed that the expression of BAX and cleaved caspase-3 proteins were both significantly stronger in the DKD rats compared with that of the non-DKD rats, while treatment with SAL markedly decreased the expression (both P < 0.01 vs. DKD group, Figures 5H–J). Subsequently, we detected the expression of BAX and BCL-2 by immunohistochemistry. As illustrated in Figure 5E, treatment with SAL significantly decreased the expression of BAX (P < 0.01, vs. DKD) (Figure 5F), while enhanced the expression of BCL2 (P < 0.01, vs. DKD) (Figure 5G).

Next, we investigated this effect in vitro. The HK-2 cells were exposed to isoosmotic mannitol, or high glucose with or without SAL (100 μM) treatment for 48 h. The location and concentration of BIM were visualized using immunofluorescence microscopic analysis (Figure 6A). As illustrated in Figures 6C–F, western blotting revealed that SAL treatment markedly decreased the expression of BIM, BAX and cleaved caspase-3 proteins induced by HG (both P < 0.01 vs. HG). Meanwhile, SAL treatment completely reduced the mRNA level of BIM (P < 0.01, vs. HG, Figure 6B). Taken together, these results demonstrate that anti-apoptosis mechanism of SAL induced by HG maybe attribute to the inhibition of BIM-mediated apoptosis pathway.

As shown in Figure 7, increased ROS production was observed both in tubular cells of diabetic rats and HK-2 cells exposure to HG, while SAL significantly prevented the increase. These results indicated that the nephroprotective effect of SAL may be related to oxidative stress inhibition of renal proximal tubular cells. The inhibited effect of SAL on ROS production may participate in regulating the expression of BIM protein.