Scrophularia ningpoensis Hemsl., Lonicera japonica Thunb., Angelica sinensis (Oliv.) Diels, and Glycyrrhiza uralensis Fisch. ex DC. were purchased from Kangqiao Co., Ltd. (Shanghai, China). These four TCMs were authenticated by Professor Hong Xu, and voucher samples were kept in the Institute of Chinese Materia Medica (Shanghai University of Traditional Chinese Medicine). Chlorogenic acid (Jin-Yin-Hua), luteoloside (Jin-Yin-Hua), harpagide (Xuan-Shen), harpagoside (Xuan-Shen), ferulic acid (Dang-Gui), glycyrrhizinic acid (Gan-Cao), and liquiritin (Gan-Cao) (purity≥98%) were bought from Dalian Meilun Biotechnology Co., Ltd. (Dalian, China). The intercellular adhesion molecule 1 (ICAM1) antibody (ABclonal, Shanghai, China), the ionized calcium-binding adapter molecule 1 (Iba1) antibody (GeneTax Inc., Alton Parkway Irvine, CA), nuclear factor κB (NFκB) p65, ß-actin, Lamin B1, phosphorylated IκB kinase (p-IKK), the phosphorylated NFκ-B inhibitor (p-IκB), VEGF receptor 2 (VEGFR2), and the phosphorylated VEGFR2 antibodies (Cell Signaling Technology, Danvers, MA) were obtained. Tumor necrosis factor-α (TNFα), claudin1, claudin19, hypoxia inducible factor 1α (HIF1α), and vascular cell adhesion protein 1 (VCAM1) antibodies (Santa Cruz, Santa Cruz, CA), Alexa Fluor 568 goat anti-rabbit IgG (BD Biosciences, Franklin Lakes, NJ), and peroxidase-conjugated goat anti-rabbit immunoglobulin G (IgG) (H + L) and anti-mouse IgG (H + L) (Jackson ImmunoResearch, West Grove, PA) were obtained. BCA Protein Assay Kits and NE-PER nuclear and cytoplasmic extraction reagents were bought from Thermo Fisher Scientific (Waltham, MA). Enzyme-linked immunosorbent assay (ELISA) kits were from R and D (Minneapolis, MN). PrimeScript RT Master Mix and SYBR Premix Ex Taq kits (Takara, Shiga, Japan) were obtained. Fluorescein isothiocyanate-coupled concanavalin A (FITC-ConA) was bought from Vector Laboratories (Burlingame, CA). 4’,6-Diamidino-2-phenylindole (DAPI) and Trizol were purchased from Life Technology (Carlsbad, CA). Other reagents were bought from Sigma Chemical Co. (St. Louis, MO).

The SMYAD powder was prepared as recorded in classic literature reports. The mixed Scrophularia ningpoensis Hemsl., Lonicera japonica Thunb., Angelica sinensis (Oliv.) Diels, and Glycyrrhiza uralensis Fisch. ex DC. (The ratio of weight is 3:3:2:1) were heated under reflux for 2 h with 10 times the amount of distilled water. The residue was heated under reflux again under the same condition after the extracts were filtered, and this procedure was repeated three times. The filtered extracts were concentrated and further freeze-dried into fine powder with the final yield of 36.5%.

High-performance liquid chromatography (HPLC) was performed on a Shiseido C18 column (250 nm × 4.6 mm, 5 μm, 30°C). The wavelength was set at 260 nm. The mobile phase consisted of a mixture of acetonitrile–0.05%H₃PO₄–Et₃N (at 0–5 min, 0.2:99.8, v/v; at 5–15 min, increased from 0.2:99.8 to 3:97, v/v; at 15–30 min, 7:93, v/v; at 30–45 min, 8:92, v/v; at 45–65 min, 6:94, v/v; at 65–67 min, 10:90, v/v; at 67–70 min, 11:89, v/v; at 75–90 min, 12:88, v/v; at 90–115 min, 17:83, v/v; at 115–130 min, 20:80, v/v; at 130–135 min, 23:77, v/v; at 135–145 min, 30:70, v/v; at 145–175 min, 50:50, v/v) with a flow rate of 1.0 ml/min. Seven reference compounds including chlorogenic acid, luteoloside, harpagide, harpagoside, ferulic acid, glycyrrhizinic acid, and liquiritin were dissolved in methanol. The peak of seven main compounds in SMYAD was identified by comparing the retention time with that of reference compounds. Additionally, the contents of those above seven compounds in SMYAD were detected by using the external standard method.

Specific-pathogen-free male C57BL/6 mice (20 ± 2 g) were obtained from the Shanghai Laboratory Animal Center of Chinese Academy of Science (Shanghai, China) and fed under a controlled room temperature (22°C ± 1 °C) and humidity (65 ± 5%) with a 12:12 h light/dark cycle. Mice have free access to food and water and have received humane care according to the institutional animal care guidelines approved by the Experimental Animal Ethical Committee of Shanghai University of Traditional Chinese Medicine (approved number: PZSHUTCM200724021).

A total of 45 mice were intraperitoneally injected (ip) with 55 mg/kg BW of streptozotocin (STZ) (Sigma No. S0130) in citric acid solution and 0.1 M trisodium citrate solution pH = 4.4 (vehicle) for 5 consecutive days. Also, other 15 mice in the control group were injected with physiological saline. Seven days later, mice with a blood glucose concentration over 16.7 mmol/L were considered as diabetic mice and included in this study. These diabetic mice were randomly divided as follows: DM, DM + SMYAD (1.5 g/kg), and DM + SMYAD (4.5 g/kg). Each group has 15 mice. One month later, SMYAD (1.5, 4.5 g/kg, intragastric administration) was orally given to mice for 1 consecutive month, and then, mice were sacrificed to collect blood samples and eyes. The body weight was monitored and the blood glucose concentration was determined with a Accu-Check glucometer (Roche Diagnostics, Germany) every 7 days during the whole experimental process.

Blood-retina-barrier (BRB) breakdown was detected according to the published method (Mei et al., 2019; Zhang et al., 2019). Briefly, mice were injected with 2% Evan’s blue (10 μl/g, i. p.). Two hours later, retinas were dissected after the Evan’s blue dye was completely removed from retinal vessels. Formamide was used to dissolve the retinas, and the absorbance in the extracts was determined at 620 nm. The amount of Evan’s blue dye was calculated, and the results were normalized to the dried weight of the retina.

Leukostasis was detected as previously described (Chen et al., 2019). Briefly, FITC-ConA (120 μl) was intravenously injected into mice. Twenty minutes later, the isolated eyes were fixed in a 4% paraformaldehyde (PFA) solution. Retinas were dissected and flat-mounted for the observation under an inverted microscope (IX81, Olympus, Japan). FITC-ConA + cells adhered to retinal blood vessels were counted.

The retina was fixed in a 4% PFA solution for 24 h and gently rocked overnight after four times of cleaning with ultrapure water and then digested with 3% trypsin and stained with hematoxylin-eosin (H and E) the next day. The morphology of retinal vessels was pictured by using an inverted microscope.

Network pharmacology analysis was conducted as described in the guidance (Li et al., 2021).

Target Collection: Active targets of SMYAD were acquired from BATMAN-TCM (http://bionet.ncpsb.org.cn/batman-tcm/) and TCMSP (http://tcmspnw.com/) databases. According to the suggested criterion, compounds were selected if the drug-likeness was ≥0.18 and the oral bioavailability was ≥30%. Additionally, the indicator chemical compounds recorded in Chinese Pharmacopoeia from four Chinese botanical drugs included in SMYAD (Chinese Pharmacopoeia Commission, 2020) with low drug-likeness and oral bioavailability values were also retained in this study.

Protein–Protein Interaction (PPI) Network Construction: Diabetic Retinopathy (ICD-11, 9B71.0Z) therapeutic targets were collected by using GeneCards (www.genecards.org/). Cytoscape 3.6.1 software was used to construct a drugs–ingredients–targets network. A target-to-target and function-related PPI network was established via the STRING database.

KEGG and GO Analysis: DAVID 6.8 was used to analyze the KEGG pathway enrichment and GO function of the proteins involved in the PPI network. The R package of ggplot2 was used for enrichment analysis, and the cutoff value was set as p < 0.05.

Retinal Iba1 immunofluorescence staining was conducted according to the published method (Mei et al., 2019; Zhang et al., 2019), but in this study, retinas were incubated with the Alexa fluor 568 goat anti-rabbit IgG (H + L) antibody after incubating with the Iba1 antibody overnight.

Retinal RNA was extracted, cDNA was synthesized, and real-time RT-PCR was conducted as described in kits. The relative expression of target genes was normalized to actin, analyzed by the 2^−ΔΔCt method, and calculated as the ratio compared with the control. The primer sequences are shown in Table 1.

Proteins were extracted, and the protein concentration was determined. The denatured samples were run on SDS-PAGE gels and electrophoretically transferred to PVDF membranes that were further incubated with primary and secondary antibodies. The protein expression was visualized by using a chemiluminescent reagent and quantified by calculating gray densities of the target protein blots with internal controls.

Serum contents of TNFα, monocyte chemoattractant protein-1 (MCP-1), interleukin-6 (IL-6), and VEGF were detected by using commercial ELISA kits.

Data are shown as the mean ± standard error of the mean (SEM). The statistical significance of differences between groups is obtained through one-way ANOVA with the LSD post hoc test. p < 0.05 is considered to be the significant difference.

Data in Figures 1A–G demonstrated the HPLC chromatogram of chlorogenic acid, luteoloside, harpagide, harpagoside, ferulic acid, glycyrrhizinic acid, and liquiritin. Figures 1H–J demonstrate the HPLC chromatogram of SMYAD, and the above seven compounds were pointed out in SMYAD. Next, the content of these above seven compounds in SMYAD was determined. After calculations, the content of chlorogenic acid, luteoloside, harpagide, harpagoside, ferulic acid, glycyrrhizinic acid, and liquiritin in SMYAD was 0.882, 0.034, 0.284, 0.103, 0.034, 0.258, and 0.201%, respectively.

The increased blood glucose concentration and the reduced body weight of diabetic mice did not change after the administration of SMYAD (1.5, 4.5 g/kg) (data not shown). The elevated Evan’s blue dye leakage in retinas from diabetic mice was reduced after the administration of SMYAD (1.5, 4.5 g/kg) (Figure 2A). As shown in Figure 2B, the decreased expression of retinal claudin1 and claudin19 (two typical tight junctions, TJs) in diabetic mice was enhanced when diabetic mice received SMYAD (4.5 g/kg). Moreover, the decreased number of acellular capillaries marked with black arrows in retinas from diabetic mice was enhanced after the administration of SMYAD (1.5, 4.5 g/kg) (Figure 2C).

A total of 109 active compounds and 398 compound-related targets were acquired from TCMSP and BATMAN-TCM. A DR target network was established to find the relationship between compounds in SMYAD and common targets of DR. As shown in Figure 3A, the yellow diamond-shape node in the central area represents the total 398 compounds-related targets, the green circular-shape node on the upper-left corner represents 14 active compounds from Jin-Yin-Hua, the red circular-shape node on the upper-right corner represents five compounds from Xuan-Shen, the purple circular-shape node on the bottom-left corner represents one compound from Dang-Gui, the pink circular-shape node on the bottom-right corner represents 84 compounds from Gan-Cao, and the blue circular-shape node represents five co-owned compounds from Jin-Yin-Hua, Xuan-Shen, Dang-Gui, and Gan-Cao. A total of 3,497 targets involved in DR development were obtained from the GeneCards database. After integrating the targets of compounds from SMYAD and the targets involved in DR development, 198 common matched-targets were found, and the Venn diagram is shown in Figure 3B. Next, those above 198 common matched-targets were imported into the STRING database to construct a PPI network. After removing 60 disconnected nodes, there were a total of 138 nodes (representing 138 common matched targets) and 817 edges (representing the interaction between two targets) (Figure 3C). After screening according to the degree of these nodes, the core targets are shown in Figure 3D. The node became larger and its color changed from yellow to red with the increased degree of the targets. Among these core targets, the first six molecular targets with higher degrees were protein kinase B (AKT1), IL-6, tumor protein p53 (TP53), TNF, mitogen-activated protein kinase 8 (MAPK8), and VEGFA.

These above 198 common matched-targets were further imported into the DAVID database, and a total of 960 GO items and 133 KEGG pathways were obtained. As shown in Figure 3E, inflammation (GO: 0006954), immune responses (GO: 0006955), and angiogenesis (GO: 0001525) marked with a red square accounted for a large proportion of biological processes. Moreover, the results of KEGG pathways showed that the TNF signaling pathway (hsa04668) (critically involved in inflammation and immune responses) and the HIF1 signaling pathway (hsa04066) (critically involved in angiogenesis) marked with a red square were the key signal pathways involved in the SMYAD-provided protection against DR.

Next, we consulted some transcriptome chips conducted in DR research included in GEO Databases (http://www.ncbi.nlm.nih.gov/gds/). The transcriptome chip results from mice at about 6 weeks after STZ injection (GSE111465) showed that the TNF signaling pathway marked with a red square was the most important signaling pathway involved in DR development in mice at 6 weeks after STZ treatment (Supplementary Figure S1A). Additionally, the transcriptome chip results from mice at about 12 weeks after STZ injection (GSE19122) showed that the VEGF signaling pathway marked with a red square was the most important signaling pathway involved in DR development in mice at 12 weeks after STZ treatment (Supplementary Figure S1B).

Studies have reported that Iba1 is a microglia/macrophage specific marker, which is widely used to label microglia (Kanazawa et al., 2002; Ibrahim et al., 2011). The enhanced number of Iba1-staining microglia (marked with white arrows) in retinas from diabetic mice was decreased after the administration of SMYAD (Figure 4A). The elevated retinal Iba1 expression in diabetic mice was decreased when diabetic mice received SMYAD (4.5 g/kg) (Figure 4B). The increased number of leukocytes adhered to the retinal capillary (marked with white arrows) from diabetic mice was also decreased after the administration of SMYAD (4.5 g/kg) (Figure 4C). Next results showed that the enhanced serum IL-6 and MCP-1 contents in diabetic mice were decreased when diabetic mice received SMYAD (1.5, 4.5 g/kg) (Figures 4D,E). The elevated ICAM1 mRNA expression in retinas from diabetic mice was decreased after the administration of SMYAD (4.5 g/kg), and the elevated retinal MCP-1 mRNA expression in diabetic mice was decreased after the administration of SMYAD (1.5, 4.5 g/kg) (Figure 4F). Moreover, the elevated protein expression of both ICAM1 and VCAM1 in retinas from diabetic mice was decreased when diabetic mice received SMYAD (4.5 g/kg) (Figure 4G).

Data in Figure 5A showed that the enhanced serum TNF-α content in diabetic mice was obviously decreased when diabetic mice received SMYAD (1.5, 4.5 g/kg). Meanwhile, the enhanced retinal TNF-α mRNA and protein expression in diabetic mice was also decreased when diabetic mice received SMYAD (4.5 g/kg) (Figures 5B,C). Next results showed that SMYAD (4.5 g/kg) abrogated the nuclear accumulation of NFκ-B subunit p65 in retinas from diabetic mice (Figure 5D). Moreover, the elevated phosphorylation of IκB and IKK in retinas from diabetic mice was decreased when diabetic mice received SMYAD (1.5, 4.5 g/kg) (Figure 5E).

SMYAD (1.5, 4.5 g/kg) abrogated the nuclear accumulation of HIF1α in retinas from diabetic mice (Figure 6A). Data in Figure 6B showed that the elevated serum VEGF content in diabetic mice was decreased after the administration of SMYAD (1.5, 4.5 g/kg). The increased retinal VEGF mRNA expression in diabetic mice was decreased when diabetic mice received SMYAD (4.5 g/kg) (Figure 6C). The enhanced retinal HIF1α-mRNA expression in diabetic mice was also obviously decreased when diabetic mice received SMYAD (1.5, 4.5 g/kg) (Figure 6C). Moreover, the enhanced VEGFR2 phosphorylation in retinas from diabetic mice was decreased when diabetic mice received SMYAD (4.5 g/kg) (Figure 6D).