Radix Rehmanniae and Cornus Officinalis raw herbs were purchased from Haichang Chinese Medicine (Jiangsu, China, Batch No. 190101) and identified by a professor from Nanjing University of Chinese Medicine; the voucher numbers of CO (No. 20170505) and RR (No. 20181020) were deposited. The herbs were cut into small pieces, soaked in 10 volumes of distilled water for 30 min, and boiled twice. To prepare RR-CO, the two original herbs were mixed at a ratio of 2:1, which is a common dose in clinical practice. The final doses of RR, CO, and RR-CO were 200, 100, and 300 mg/mL, respectively. Antibodies against IL-6 (Batch No. sc-57315) and TNF-α (Batch No. sc-52B83) were obtained from Santa Cruz Biotechnology (United States). Antibodies against NF-κB (Batch No. 10745-1-AP) were purchased from Wuhan Proteintech Group (China). Antibodies against Glucagon-like peptide-1 receptor (GLP-1R) (Batch No. bs-1559R) were purchased from Beijing Bioss Biotechnics Co., Ltd. (China).

Ten-week-old male C57BL/6J and KK-Ay mice (license number: 2014-0004) were purchased from Beijing Huafukang Biotechnology Co., Ltd. The mice were bred adaptively for 2 weeks, and the temperature and humidity conditions maintained in the animal house were 25 ± 5°C and 55 ± 5%, respectively. All the animal experiments performed were approved by the Animal Ethics Committee of the Nanjing University of Traditional Chinese Medicine (code: 201903A019). C57BL/6J mice were fed a normal diet, while KK-Ay mice were fed a high-fat diet (458 kcal/100 g, containing 10% fat). Then, the KK-Ay mice were randomly divided into RR-CO, RR, and CO groups. The three groups were given RR-CO (3,000 mg/kg/d), RR (2,000 mg/kg/d), and CO (1,000 mg/kg/d), respectively, for 8 weeks. The KK-Ay mice in the model group were administrated saline. Seven C57BL/6J mice that were fed with a normal diet were also given saline. At the end of the experiment, the mice were sacrificed using the cervical dislocation technique, and blood was drawn from the orbit and serum was extracted. Testes were resected and weighed to calculate the ratio of the bilateral testis to body weight. The flow chart of the animal study is shown in Figure 1A.

The tail of the epididymis was homogenized in PBS and incubated at 37°C for 5 min to release the sperms. The suspension was then mixed with 1% trypan blue of the same volume and observed using an optical microscope. Live sperm eliminated the dye, while the dead sperm accumulated it and presented a blue head. The serum and testicular testosterone level was detected using enzyme-linked immunosorbent assay (ELISA) kits (Anhui Joyee Biotechnics Co., Ltd., China), the FSH and LH were measured using ELISA kit from Shanghai Jianglai Biotechnics Co., Ltd., China.

The testicular tissue was fixed with 4% paraformaldehyde, embedded in paraffin, sliced, dewaxed with xylene and ethanol, then stained with hematoxylin and eosin (H&E), and sealed with neutral gum. Use a microscope to photograph randomly selected areas. The semi-quantitative scoring of the H&E image was conducted according to the grading arrangement and structure of spermatogenic cells, interstitial cell proliferation, interstitial vascular dilatation, congestion, bleeding, inflammatory cell infiltration, etc. For transmission electron microscopy (TEM) assay, the testicular tissue was fixed with 2.5% glutaraldehyde solution at 4°C for 24 h, followed by gradient dehydration and imaging of the seminiferous tubule at random fields. Immunofluorescence staining was performed as follows: paraffin sections of mouse testes were dewaxed, rehydrated in xylene and ethanol, and then incubated for 20 min at 37°C in proteinase K working solution, after PBS rinsing, the cells were incubated with primary antibody against GLP-1R (dilution 1:200) for 2 h at room temperature, followed by secondary antibody with FITC fluorescence, nuclei were stained with DAPI for 5 min and a Zeiss LSM 900 confocal microscope Images were taken and immunofluorescence intensity was quantified using ImageJ software.

The major SCFAs in mice feces of all groups, such as acetic acid, propionic acid, butyric acid, and valeric acid, were detected using gas chromatography–mass spectrometry (GC-MS, Thermo Scientific TSQ 8000, United States). About 50 mg of feces of each mouse were collected, and 1,000 μL 0.005 mol/L NaOH aqueous solution was added to each sample separately. After vortexing for 10 min, the feces were centrifuged at 4°C at 10,000 rpm for 15 min. About 500 μL of the supernatant was taken and added to a 10-mL glass centrifuge tube. Then, 300 μL pure water, 500 μL propanol pyridine solution, and 100 μL n-propyl chloroformate were added to the above extract. The mixture was vortexed for 10 s and subjected to ultrasound sonication for 1 min. After the derivatization, n-hexane was used to extract the derivatized substances. First, 300 μL n-hexane was added to the reaction system, and the sample was centrifuged at 3,000 rpm for 5 min. About 200 μL of the supernatant was taken in an injection bottle, and 200 μL n-hexane was added to the reaction system again, followed by vortexing for 1 min and centrifugation for 5 min at 3,000 rpm. Another 200 μL of the supernatant was taken into the injection bottle and 60 μL of the supernatant was used for GC-MS analysis.

According to the characteristics of the amplified 16S ribosomal DNA (rDNA) region using the Illumina Miseq technology sequencing platform, a small fragment library was constructed using paired-end sequencing. Complete standard information analysis and advanced information analysis were performed to analyze the microbial diversity in the sample. Genomic DNA of the gut microbiome (GM) was extracted from fecal samples and the bacterial V3–V4 region of the 16S rDNA gene from each sample was amplified. Sequencing was performed using an Illumina Miseq PE250 platform (Shanghai Bioprofile Technology Co., Ltd.). Raw reads were stored in the NCBI database.

Testis and colon tissue was lysed with RIPA buffer (Solarbio Life Sciences, Beijing, China) to produce lysate. Protein sample (15–30 μg per sample) was separated using 10% SDS-PAGE polyacrylamide gel, and then transferred to the PVDF membrane. Then the membrane is blocked with 5% skimmed milk for 2 h, incubated with the primary antibody, corresponding horseradish peroxidase-conjugated secondary antibodies were then followed. At the end, protein bands were detected using a chemiluminescence kit (Millipore, United States) with ECL chemiluminescence system (Tanon-5200Multi, Shanghai, China). Data was qualified by ImageJ software. The experiments were repeated three times.

A total of 100 mg of fresh stool samples from the Model and RC groups were collected immediately and resuspended in 1 mL of saline, vortexed for 5 min, and filtered using a sterile nylon net. Then, 20% glycerin was added to the bacterial suspension and the suspension was stored at −80°C. Twelve C57BL/6J recipient mice were fed a high-fat diet (FB-D12492, 60% fat, Wuxi Fanbo Biotechnology Co., Ltd) for 1 month and injected with 60 mg/kg Streptozotocin twice a week to establish the type 2 diabetes model. Furthermore, six C57BL/6J mice on a normal diet without STZ injection were used as normal control. The recipient mice were then given water containing antibiotics (1 g/L ampicillin, 1 g/L metronidazole, 0.5 g/L vancomycin, and 0.5 g/L neomycin) for 2 weeks and then gavaged with 200 μl of fecal sample supernatant every day for another 3 weeks. The flow chart of the FMT animal study was shown in Figure 5A.

The data were expressed as mean ± standard deviation (SD) and analyzed by one-way ANOVA with Spss19.0 software, and then the differences between groups were analyzed using LSD t-test, the correlations were statistically evaluated using Pearson correlation tests with Spss19.0 software, P < 0.05 showed statistical difference.

Hematoxylin and eosin results demonstrated that all stages of spermatogenic cells were abundant and orderly with filamentous sperm cells in the lumen in the normal group, while in the model group, the spermatogenic layer was destroyed and the cells were disorganized. Also, the mature sperm cells in the lumen decreased in the model group, suggesting that the spermatogenic function of diabetic animals was reduced (Figures 1B,D). Compared to the normal group, the TEM results showed an uneven distribution of nuclear chromatin, the chromatin particles concentrated to the heterochromatin region, the tight junctions between cells were damaged, and the vacuolization of cytoplasm increased in the model group. The study of the testicular ultrastructure in the RR-CO group showed that the tight junctions between the cells were clear, vacuolization was ameliorated, and the nuclear chromatin was well-distributed (Figure 1C). The RR-CO, RR, and CO groups showed reduced fasting blood glucose (FBG) level and an increase in the live sperm rate, serum and testicular testosterone level, and testis/body weight ratio to varying degrees compared to the model group (Figures 1E–I). The serum FSH level in the model group was also dramatically decreased compared with the normal group, while RR-CO administration can increase the level. Nonetheless, serum LH levels did not show significant differences between the groups (Figures 1J,K).

Western blotting was performed to evaluate the inflammatory status of the testicular and colon tissues. The results revealed that, compared to the normal group, the expression of pro-inflammatory factors IL-6 and TNF-α significantly increased in the model group, while treatments with RR-CO, RR, and CO down-regulated their expression to varying degrees (Figures 2A–F).

Chao1 and ACE indices were used to estimate the total number of microbial species in the sample, showing no significant differences across the groups (Figures 3A,B). Shannon and Simpson’s indices represent the diversity of the bacterial community. The Shannon index in the RR-CO and CO group was significantly higher than that in the model group, whereas the Simpson’s index was lower than that in the model group (P < 0.05) (Figures 3C,D). The rarefaction curve already reached stable values, suggesting that the amount of sequencing data is reasonable (Figure 3E), while the Shannon curve indicated that the sequencing data reflected majority of GM information in the sample (Figure 3F). The species accumulation curves also reached a plateau, indicating that the species in this environment will not increase significantly with the increase in sample size, implying that the sampling was sufficient (Figure 3G).

Partial Least Squares Discriminant Analysis (PLS-DA) was used to study the differences in sample community composition. PLS-DA results indicated that there were significant structural differences in the GM across the normal, model, and RR-CO groups (Figure 3H).

To further explore the differences in the structure of GM across groups, the bacteria from each sample were analyzed at the phylum, family, and genus levels. The histogram shows the overall community structure of GM and the top nine relatively abundant phyla of GM in different groups (Figure 4A). The relative abundance of Actinobacteria was lower in the model group compared to the normal group, while RR-CO groups showed an increase in the abundance compared to the model group (Figure 4B). At the family level, 17 predominant families in the five groups are shown in Figure 4C. The relative abundance of Clostridiaceae_1 was dramatically lower in the model group compared to the normal group, whereas RR-CO increased the abundance of Clostridiaceae_1 significantly (Figure 4D). Meanwhile, the relative abundance of certain pathogenic microorganisms, such as Catabacteriaceae and Helicobacteraceae, increased significantly in the model group and decreased in the normal, RR-CO, RR, and CO groups to varying degrees (Figures 4E,F). At the genus level, 26 intestinal bacteria were found to be significantly differentially abundant across the five groups, and the abundance level of four genera reversed after the treatments (Figure 4G). Catabacter, Marvinbryantia, and Helicobacter, which increased in the model group, were down-regulated with RR-CO, RR, or CO treatment to a varying degree. On the contrary, the decreased numbers of Clostridium_sensu_stricto in the model group were restored in the RR-CO group. However, RR and CO alone did not significantly upregulate the abundance of Clostridium_sensu_stricto (Figures 4H–K).

To further verify the role of GM dysbiosis in diabetes-induced reproductive damage, we conducted FMT experiments, transplanting feces of DM (model group) and RR-CO-treated DM mice into the recipient pseudo-germfree C57BL/6J mice with or without diabetic symptoms (Figure 5A). Feces of the model group (M-FMT) resulted in the aggravation in the testicular damage, manifested as an abnormal structure of the spermatogenic cells and a reduction in the number of mature spermatogenic cells in the lumen. Also, the live sperm rate in the M-FMT group showed a significant reduction compared to the Normal FMT (N-FMT) group. Conversely, RR-CO FMT (RC-FMT) group mice demonstrated remarkable attenuation in the testicular pathological damage and an up-regulation in the live sperm rate (Figures 5B–D). However, there was no significant difference in testis/body weight ratio across the three groups (Figure 5E).

Principal Component Analysis (PCA) revealed a clear distinction in gut microbial ecosystem across the N-FMT, M-FMT, and RC-FMT groups, and the altered GM in M-FMT was restored by RC-FMT, close to that in the N-FMT mice (Figure 6A). A total of 23 genera were found to be significantly differentially abundant across the three groups, among which RC-FMT treatment increased the abundance of butyric acid-producing Clostridium_sensu_stricto and Lactobacillus genera, decreased the amount of Helicobacter genera (Figures 6B–E).

Several studies have reported that the occurrence and development of T2DM are closely related to changes in GM composition and function. GM produces SCFAs, such as acetic acid, propionic acid, and butyric acid. SCFAs promote the secretion of Glucagon-like peptide-1 (GLP-1) by the intestinal L-cells (Wang et al., 2020). Therefore, we evaluated the effect of 8 weeks of administration of RR-CO on the SCFA/GLP-1 axis. The results of GS-MS demonstrated that the content of butyric acid in colon feces in the model group significantly decreased compared to that in the normal group; nevertheless, RR-CO and CO significantly increased the content of butyric acid in the model group. There were no significant differences between the model group and the RR-CO, RR, and CO groups in the levels of acetic acid, propionic acid, and isovaleric acid (Figures 7A–D). To further explore the relationship between GM-mediated regulation of SCFAs/GLP-1 axis and testicular inflammation in diabetes, we detected GLP-1 levels in serum and GLP-1R expression in testicular tissue. Compared to the normal group, the serum GLP-1 level and GLP-1R expression in the model group markedly diminished. Reciprocally, administration of RR-CO, RR, and CO for 8 weeks up-regulated the levels of GLP-1 and GLP-1R expression compared to the model group (Figures 7E–H).

A correlation heatmap was used to assess the association between GM and FBG, sperm damage indicators, butyric acid, and GLP. Based on the heatmap, at the family level, the abundance of Moraxellaceae, Desulfovibrionaceae, Erysipelotrichaceae, Enterococcaceae, and Helicobacteraceae showed a positive correlation with FBG and pathological testicular score, while Clostridiaceae_1, Coriobacteriaceae, Bifidobacteriaceae, Rikenellaceae, and Peptostreptococcaceae abundance exhibited a negative correlation with FBG and pathological testicular score. However, Rikenellaceae, Clostridiaceae_1, and Bifidobacteriaceae abundance demonstrated a positive correlation with live sperm rate and testosterone level in the blood. The abundance of Rikenellaceae, Clostridiaceae_1, and Deferribacteraceae demonstrated a positive correlation with butyric acid and GLP content (Figure 8A). At the genus level, butyric acid producing related bacteria Clostridium_sensu_stricto exhibited significantly positive correlation with live sperm rate, testosterone level, butyric acid and GLP level, but a negative correlation with FBG and pathological testicular score. Some conditionally pathogenic bacteria such as Helicobacter, Enterococcus, and Marvinbryantia were positively correlated with FBG and pathological testicular score, but negatively correlated with live sperm rate and testosterone level (Figure 8B).

Since RR-CO significantly increased the content of butyric acid in feces of diabetic mice, we conducted correlation analysis of butyric acid and sperm damage parameters, and the results revealed that the content of butyric acid in feces was positively correlated with testosterone, live sperm rate, and testis/body weight ratio (Figures 9A–C). To further confirm the role of butyric acid in alleviating reproductive damage in diabetes, sodium butyrate (SB) was intraperitoneally administered to diabetic mice to observe its alleviating effects on diabetic symptoms, testicular damage, and inflammation. After 8 weeks of administration, SB significantly improved the pathological damage to the testis and reduced the expression of inflammatory factors IL-6 and TNF-α in the model group (Figures 9D–G). In addition, compared to the model group, the content of butyric acid in colon feces and GLP-1 in the blood of the mice in the SB group increased significantly. Also, SB significantly up-regulated the expression of GLP-1R in the testicular tissue (Figures 9H–J).