Thirty-two female Wistar albino mature rats aged 8–10 weeks, weighing between180–220 g, were used to study the induction of endometriosis. The rats were purchased from the Laboratory Animal Centre, Wenzhou Medical University. Rats were caged in a controlled environment of 22 ± 2°C with 12-h light/dark cycles. All rats were observed for 1 week to ascertain health before surgery. The Laboratory Animal Ethics Committee of Wenzhou Medical University approved this study (ethics number: wydw 2015-0458).

Endometriosis was surgically induced according to the method described previously (Oner et al., 2010). In brief, after anesthetized and abdomen opening, one uterine horn was ligated and removed. The excised horn was bisected along its macro-axis, and 5 mm × 5 mm sections dissected. These explants were then anchored onto the flank inside the abdominal wall with the endometrial surface facing the peritoneum. The development of endometriosis was determined by measuring the surface area (length × width × height in millimetres) 21 days after surgical procedure. Rats with endometriosis were randomly divided into three groups and treated by oral gavage with either vehicle (saline) or metformin (Santa Cruz, CA, United States) at two different concentrations. Group 1 (control group, n = 10) was treated with saline (4 ml/kg/day), group 2 was treated with metformin 100 mg/kg/day (ML, n = 11) and group 3 were treated with 200 mg/kg/d of metformin (MH, n = 11).

The metformin treatment dosages used in this study were similar to those described previously (Oner et al., 2010). In a clinical setting, metformin is prescribed at a dosage of 1,000 to 2,000 mg/d for women diagnosed with PCOS and insulin resistance. Thus, to correlate these dosages in a rat model, we used the conversion equation, described in (Nair and Jacob, 2016) and determined our experimental treatment dosages to be 100 and 200 mg/kg/d for rats with endometriosis.

The endometriotic implant samples from the abdominal wall and endometrium samples from the uterus were collected from rats at diestrus II stage, as described previously (Marcondes et al., 2002). Briefly, following 6 weeks of treatment, rats at diestrus II were confirmed by increased lymphocytes in vaginal smears. Rats were anaesthetized via inhalation of chloral hydrate and subsequently subjected to a laparotomy to excise the uterus endometrium as well as the endometriotic implants. The samples were immediately stored in liquid nitrogen or formalin for future analysis. For rats where the estrous cycle was not clearly identified, samples were collected for 3 days post 6 weeks of treatment and metformin treatment was continued for a further 3 days.

The key steps in the preparation of the endometriosis rat model and experimental design are summarised in Figure 1.

To evaluate the effect of metformin on the development of endometriotic implants, the obtained endometrium and endometriotic implants were measured for length, width, and height. The measurement procedure was performed by a researcher who was blinded to the study.

The endometrium and implant tissue from rats were excised and immediately fixed in 10% buffered formalin for 24 h. After de-waxing through increasing grades of ethanol and rendered transparent using Xylene, the formalin-fixed endometriotic implants and endometrium sections were embedded in formalin blocks and sectioned at 4 µm thickness. Slides were de-waxed through descending grades of ethanol to distilled water and through xylol deparaffinization. Endogenous peroxidase was blocked with 3% H₂O₂ for 10 min then washed in PBS. The slides were pretreated with citrate buffer (pH 6.0) and incubated at 90°C for 20 min and washed again with PBS. The tissues were incubated with the appropriate primary antibody overnight at 4°C. Goat anti-LIF polyclonal antibody (Santa Cruz, B2813) was used at a dilution of 1:30. A rabbit anti-HOXA-10 polyclonal antibody (Bios, YSLS25W) was used at a dilution of 1:50; A rabbit anti-MMP9 polyclonal antibody (abcam GR149385-3) was used at a dilution 1:60. A mouse anti-VEGF polyclonal antibody (abcam, GR145343-1) was used at a dilution 1:100. The secondary antibody was then incubated for 30 min and subsequently tissues were incubated with avidin and biotinylated peroxidase for a further 15 min. To visualize the staining, the tissues were incubated with 3, 3′-diaminobenzidine (DAB) for 5 min. Hematoxylin and Eosin (H&E) staining was used for counterstaining. Finally, tissues were sealed with resin. Negative controls were evaluated on additional sections by omitting the primary antibodies. All tissues were evaluated with a light microscope (Olympus BX43; Olympus Optical, Tokyo, Japan). Three random fields on each slide at ×200 magnifications were used to evaluate the samples. All tissues were evaluated by two blinded histologists. IHC results were further evaluated by a semiquantitative approach to assign H-scores for endometrium samples and endometriotic implants. Immunoreactivity was evaluated with a H-score in an area of approximately 100 cells. The staining intensity was first determined for each cell in a fixed field and the average intensity of the entire field of view, corresponding to the presence of negative, weak, intermediate, and strong staining, was allocated a score from 0 to 3. Samples were then further evaluated; the percentage of positively stained cells within the field of view was assigned 1 of 4 categories (1%–25%,26%–50%,51%–75%,76%–100%). The H-scores (H) for VEGF, MMP-9, LIF and HOXA10 were calculated with formula H = ∑ P i , where i is the intensity, and P is the percentage of positively stained cells.

Expression of HOXA10 and LIF mRNAs in the endometrium of the rat model of endometriosis were evaluated using quantitative real time polymerase chain reaction (qRT-PCR). To extract total RNA, 100 mg of each tissue sample was homogenised in TRIzol Reagent (Invitrogen, Carlsbad, CA) and incubated for 5 min. Chloroform 0.2 ml was added to the homogenate and samples incubated for a further 3 min before centrifugation at 12,000 g for 15 min at 4°C. The clear aqueous phase was transferred to a fresh tube. RNA precipitation was performed by mixing the sample with cold isopropyl alcohol, incubation for 10 min, and centrifugation at 12,000 g for a further 10 min at 4°C. Following centrifugation, RNA samples were washed twice with 75% ethanol. The RNA pellets were air-dried and resuspended with RNase-free water. Finally, RNA was stored at −80°C until RT-PCR was performed. Reverse transcription was performed using the RevertAid First Strand cDNA Synthesis kit (Fermentas Life Sciences; MA, United States) according to the manufacturer’s instructions. RT-PCR was performed using the CFX connection Real-Time PCR System with the software BIO-RAD CFX Manager. Reaction conditions included cDNA template, each primer, water, and the IQ SYBR Green Supermix for a final reaction volume of 20 μL. The sequences of all primers used are: LIF forward primer, CCC​TCT​TTA​TTT​CCT​ATT​AC; LIF reverse primer, GTA​GTC​GCA​TTG​AGT​TTG​AT; HOXA10 forward primer, CTC​CTA​CTC​CTC​CAA​CCT​GC; HOXA10 reverse primer, GTT​CCT​GCC​CAC​CGT​GCT​AT; GADPH forward primer GTG​CTG​AGT​ATG​TCG​TGG​AG; GADPH reverse primer GTC​TTC​TGA​GTG​GCA​GTG​AT. The HOXA-10 and LIF RT-PCR reactions were subjected to the following cycling parameters: 50°C for 3 min, 1 cycle; 95°C for 15 min, 1 cycle; 95°C for 10 s, 40 cycles; 59°C for 30 s, 1 cycle. Melting curve analysis was conducted to determine the specificity of the amplified products and to insure the absence of primer-dimer formation. All products obtained yielded the predicted melting temperature. Samples were run in triplicate and included negative controls. The mRNA level of each sample was normalized to that of the GADPH mRNA level. Relative gene expression data was presented as 2^−ΔΔCt method, and ΔΔCt for each sample refers to ratio between target gene and the control group. The results of quantitative RT-PCR were expressed as relative fold values.

One hundred mg of tissue was homogenized with the appropriate volume of lysis buffer using an ultrasonic homogenizer. The homogenates were centrifuged at 3,000 g for 10 min at 4°C, and the supernatant was collected. Extracted protein (30 μg) was separated using 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a methanol-activated polyvinylidene difluoride (PVDF) membrane (Millipore, Bedford, MA, United States). Membranes were blocked in 5% (w/v) skim milk/Tris buffered saline (TBS) for 1 h prior to incubation with the primary antibody. Dilution for LIF, 1:1,000; for HOXA10, 1:1,000; for MMP-9 1:1,000; for VEGF1: 1,000; for β-tubulin1:1,000 or for GADPH 1:2000. Membranes were washed three times with TBS-Tween 20 (TBST), prior to the HRP-secondary antibody (Santa Cruz, CA) incubation at 1:2500 dilutions for 1 h. Membranes were then developed using enhanced chemiluminescence (ECL) (Pierce, Illinois, United States) according to the manufacturer’s instructions and visualized using the ChemiDoc XRS + System (Bio-Rad). Statistical analysis was carried out using the ImageJ software (National Institute of Mental Health, United States) and GraphPad Prism (version 7.03) (California, United States).

The data are expressed as the Mean ± SD. Comparisons across the three groups were performed using one-way analysis of variance (ANOVA) followed by Dunn’s test to determine significant differences between the two groups using Prism version 7.03 (GraphPad Inc., San Diego, CA). p-value < 0.05 was considered statistically significant.

Endometriosis was surgically induced using a Wistar albino mature rat endometriosis model. Rats were treated with either metformin (treatment Group 2-ML and Group 3 MH) or saline (Group 1, control). At the beginning of the treatment, the mean volumes of endometriotic implants were similar in the three groups of rats (Control: 65.19 mm³ ± 19.98, ML:79.35 ± 44.76, MH:80.90 mm³ ± 30.11; p > 0.05) (Table 1). At the end of day 21 post the initial operation, endometriosis implants in the abdominal wall were formed in 32 out of the 35 rats. At the end of the treatments, implant volumes were significantly lower in rats with both dosages of metformin treatment (Control: 198.39 mm³ ± 73.75, ML: 78.60 mm³ ± 25.11, MH: 117.18 mm³ ± 55.12; p < 0.001) (Table 1) with 100 mg/kg the optimal dosage. The overall volume of endometrial implants was larger in rats treated with 200 mg/kg of metformin (117.18 mm³ ± 55.12) as opposed to those treated with 100 mg/kg of metformin (78.60 mm³ ± 25.11). However, the difference in the two treatment groups was not significant (p = 0.152). A total of 27 rats with endometriosis completed treatment and five rats died (2 in each treatment group and 1 in the control group) during the period of treatment.

To evaluate the effect of metformin on endometriotic implants in rats with endometriosis, VEGF and MMP-9 expression in implants were analyzed using IHC and western blotting.

As shown in Figure 2 left column, VEGF immunopositive staining was mainly localized in glandular epithelial cells, vascular endothelium and surrounding stromal cells, as well as in the cytoplasm of macrophage cells, shown as pale to dark brown staining. The percentage of positive stained cells and intensity of VEGF expression significantly reduced following metformin treatment. VEGF was weakly positive in macrophages and stromal cells and negative in the glands in the ML group. IHC analysis revealed that the VEGF H-scores of implants were significantly lower in metformin treated rats when compared to the control group (Control: 5.11 ± 1.76, ML:3.56 ± 10.88, MH:3.95 ± 1.35; p < 0.05) (Table 1; Figure 2).

MMP9 immunopositive staining was mainly located in the glandular epithelial cells, stromal cells and the macrophage cytoplasm, shown as pale to dark brown staining (second left column, Figure 2). After treatment, the percentage of positive stromal cells and macrophages significantly decreased, with fewer scattered brown staining in the ML group (Figure 2). Interestingly, in the MH group the intensity of MMP-9 positive cells were greater in the glandular epithelial cells compared to the ML group. The MMP-9 H-scores from the implants showed significant decrease in ML and MH groups when compared to the control group (Control: 6.44 ± 1.33, ML: 4.20 ± 1.20, MH: 4.89 ± 1.05; p < 0.01) (Table 1; Figure 2).

Western blots showed significant reduction of VEGF expression in ML treated rats when compared to control (p < 0.05, Figure 3A), consistent with the IHC H-score. Interestingly, no significant difference in VEGF was observed between the MH and control group (p > 0.05, Figure 3A). Differences in VEGF between ML and MH groups were not significant (p > 0.05, Figure 3A). MMP-9 expression was significantly decreased in the ML group when compared to the control group (p < 0.01, Figure 3B). However, MMP-9 expression was not affected in the MH group when compared to the control group (p > 0.05, Figure 3B). Similarly, no significant difference in MMP-9 was observed between the ML and MH groups (p > 0.05, Figure 3B).

Taken together, these data suggest that metformin treatment significantly decreased VEGF and MMP-9 expression and the effect was optimal at the dose of 100 mg/kg.

In this study, IHC, Western blots and qRT-PCR were applied to determine the expression of two implantation markers, LIF and HOXA10, in the rat endometrium. IHC staining demonstrated that location of LIF and HOXA10 expression was mainly in the endometrial luminal epithelial, glandular epithelium and the cytoplasm of stromal cells, presenting as pale to dark brown staining (right 2 columns, Figure 2). In the control group LIF was expressed as weakly positive in the glandular epithelia and very weak or negative in the stromal cells. After treatment, the percentage of LIF positive staining in the glandular epithelia significantly increased in both ML and MH groups. The intensity of staining in the ML group was stronger than in the control and MH groups. LIF stained negatively in stromal cells in both the control group as well as in MH groups. The H-score of LIF was significantly increased by 39.0% in the endometrium from ML rats when compared to control (Control: 4.00 ± 1.41, ML: 5.56 ± 1.26, MH: 4.55 ± 1.03; p < 0.05) (Table 1; Figure 2). Likewise, there was a significant increase (34.8%) in the H-score of HOXA10 compared to the control group (Control: 5.11 ± 1.45, ML: 6.89 ± 1.63, MH: 6.00 ± 1.73; p < 0.05) (Table 1; Figure 2).

Using Western blot analysis, LIF expression was shown to be significantly higher in the endometrium resected from rats treated with both dosages of metformin than in rats from the control group (both p < 0.01, Figure 3C). Protein expression of HOXA10 was increased in the endometrium only from the MH group (p < 0.05, Figure 3D), whereas there was no significant difference in HOXA10 expression between the ML group and the control group (p > 0.05, Figure 3D).

To advance our understanding of the molecular mechanism by which metformin influences implantation markers, mRNA for LIF and HOXA10 were further investigated in this study. The mRNA fold change of LIF significantly increased in the MH treated rats compared to the control group (p < 0.001, Figure 4A). However, the level of HOXA10 mRNA in the ML and MH groups remained similar to the control group (p > 0.05, Figure 4B). Taken together, in the endometrium resected from rats with endometriosis, metformin increased LIF protein and mRNA expression, and enhanced HOXA10 protein expression dependent on metformin dosage.