A circRNA-miRNA-mRNA network plays a role in the protective effect of diosgenin on alveolar bone loss in ovariectomized rats

Six-month-old female rats undergoing ovariectomy have been commonly used as a model to study osteoporosis in postmenopausal women [18, 19]. This study included 48 female Wistar rats (six months old). We obtained the rats (average weight 300 ± 20.0 g) from the National Institutes for Food and Drug Control of China. All rats were kept under a regular 12 h/12 h light/dark cycle and at constant room temperature (22 ± 1 °C). The rats underwent sham operation (SHAM, n = 12) or bilateral OVX (n = 36). The ventral approach was used in bilateral ovariectomy [20]. Briefly, at first, the anesthetized rats were fixed ventrally and the abdominal skin in the mid-lower back was preparated and disinfected. A 0.5 cm incision closed the spine under the costal margin was made and skin was separated. Secondly, dorsal muscle was cut and then ovary was exposed adequately. Thirdly, the fat tissue coating the ovary was separated by blunt dissection and the blood vessels and oviduct were ligated. Fourthly, ovariectomy was completed and the preserved tissue or organs were put back into the abdominal cavity, followed by suturing incision. Ovariectomy on the other side was in the same way. The SHAM group rats underwent same operation except that the ovaries were preserved.

The model rats subjected to OVX were classified into 3 categories randomly, namely, OVX group (n = 12), estradiol valerate (EV) group (n = 12) and DIO group (n = 12). According to studies published previously [21‐23], EV group rats were administered oral gavage of EV (0.1 mg/kg body weight, Bayer Health Care, Guangzhou, China) every day, while DIO group rats were treated by oral gavage with DIO (100 mg/kg body weight, purity≥93%, Sigma-Aldrich, Saint Louis, MO, USA) every day. For the SHAM or OVX group, rats were administered distilled water (of equal volume) by oral gavage. A standard chow was used to feed all rats during our present experiment. The treatments started one week after the operation and lasted 12 weeks. No rats died during the whole treatment period.

At the end of treatment, all experimental rats were anesthetized via intraperitoneal injection by ketamine (80 mg/kg) and xylazine (12 mg/kg) and were subsequently exsanguinated for sacrifice. Under anesthesia, we punctured the abdominal aortae to collect blood (8–10 mL) which was subsequently transferred into tubes added with heparin. The rats subjected to haemospasia were palpated for 5 min to ensure asystole and were considered as dead after confirming asystole, respiration cease and corectasis. Next, the plasma was separated from blood by centrifugation (12 min, 2500 g, 4 °C) and stored (− 80 °C) for the following experiments. To observe and determine the alveolar bone microstructure of the rats, the right mandibles were first excised and kept under the temperature of − 20 °C and then scanned using microcomputed tomography (micro-CT). Thereafter, the right mandibles also served as the specimens for histological examination. For microarray assays and reverse transcription-quantitative polymerase chain reaction (RT-qPCR), we used osseous tissue between the first molar and incisor in left mandibles.

We used ELISA to determine plasma concentrations of osteocalcin (OCN) and tumor necrosis factor-alpha (TNF-alpha). A rat OCN ELISA kit (Novus Biologicals, Littleton, CO, USA) and a rat TNF-alpha ELISA kit (Abcam, Cambridge, MA, USA) were used to determine the rat plasma levels of TNF-αand OCN from all groups. The absorption value at 450 nm was detected by a BioTek ELISA reader (BioTek Instruments., Winooski, VT, USA).

The micro-CT scanning method has been previously described [24]. The right mandibles of the rats were scanned by using a high-resolution micro-CT system (Skyscan 1174, Bruker Corporation, Ettlingen, Germany) without any sample processing. The resolution of the scan was 9.8 μm, and analuminum filter (0.5 mm) was used to remove image noise. A global threshold (upper gray threshold: 255, lower gray threshold: 55) was used for the quantity of the parameters about trabecular bone.

The image capture conditions were 800 μA and 50 keV. A cube (0.5 mm × 0.5 mm × 0.5 mm) termed as “volume of interest” (VOI) was reconstructed starting from 0.5 mm beneath the base of crown in the first molar. The trabecular bone morphological characteristics within the VOI were measured applying the Skyscan software package. Furthermore, a 3-D analysis was performed to determine eight key parameters, namely, the trabecular bone volume fraction (BV/TV), the bone surface (BS), the trabecular separation (Tb.Sp), the trabecular number (Tb.N), the trabecular thickness (Tb.Th), the trabecular pattern factor (Tb.Pf), the degree of anisotropy (DA) and the structural model index (SMI) of the identical VOL [25].

The prepared solution containing formalin (10%) was used to fix the right mandibles. Next, the mandibles were decalcified with ethylenediaminetetraacetic acid (EDTA, 14%) and embedded in paraffin. After a regular microtome cutting process, the glass slides were used to affix sections, which then were stained with hematoxylin and eosin.

Twelve samples of alveolar bones in the DIO group (n = 6) and OVX group (n = 6) were chosen for microarray studies randomly. The microarray assay was completed by KangChen Biotech Inc. (Shanghai, China).

For our circRNA study, we used an Arraystar Rat circRNA Array (Arraystar, Rockville, MD, USA). We used RNase R (Epicentre, Madison, WI, USA) to digested total RNA, and further remove linear RNA and concentrate circular RNA. Subsequently, we amplified and transcribed the concentrated circRNA into fluorescent RNA. This step was performed using an Arraystar Super RNA Labeling Kit (Arraystar, Rockville, MD, USA), by a random priming method. Thereafter, the labeled cRNA was subjected to hybridization, washing and scanning on Arraystar mouse circRNA Array v1.0 using DNA Microarray Scanner (Agilent Technologies, Santa Clara, CA, USA).

For mRNAs, we used a rat 4 × 44 K Gene Expression Array (Agilent Technology, Santa Clara, CA, USA). In line with the protocol, samples were labeled and arrays were hybridized. Then, the obtained array images were evaluated by the Agilent Feature Extraction software (version 11.0.1.1). The data was normalized and further processed using an R software package.

All DE mRNAs obtained from our array analyses were input into the IPA system, a system that predicts canonical pathways, global functions and biological networks of a particular gene dataset and is referred to as the Ingenuity Pathways Knowledge Base (IPKB). In IPKB, “Role of Osteoblasts, Osteoclasts and Chondrocytes in Rheumatoid Arthritis (ROOCRA)” pathway is a exclusive pathway because this pathway includes nearly all known pivotal molecules and signaling pathways that are closely related to the development, differentiation, degradation, mineralization and apoptosis of osteoblasts and osteoclasts. As this pathway is essential in bone metabolism, the DE mRNAs that belong to the ROOCRA pathway were therefore regarded as key DE mRNAs.

Potential target miRNAs of circRNAs were predicted by miRanda (v3.3a) [26] and TargetScan (version 7.2) [27]. Software developed by Arraystar was applied to predict the interactive relationship between circRNAs and miRNAs, while the predicted interactions of mRNAs and miRNAs were obtained using miRanda (version 3.3a). After Pearson correlation analyses of the predicted mRNA-miRNA and circRNA-miRNA coexpression networks, the circRNA-miRNA-mRNA coexpression network were constructed. Cytoscape (version 2.8.2) [28] was applied to visualize the whole circRNA-miRNA-mRNA network.

qRT-PCR assays used for detecting the expression levels of the predicted miRNAs were all performed in an ABI 7500 system (Applied Biosystems, Foster City, CA, USA) using SYBR RT-PCR kits (Takara, Dalian, Liaoning, China). We used the method of 2^-ΔΔCt cycle threshold, and the expression of U6 served as the internal normalization control. For circRNAs/mRNAs expression data, the internal control was Gapdh expression.

The mean ± standard deviation was used to express all values. We utilized SPSS 19.0 (SPSS Inc., Chicago, IL, USA) for analyzing data statistically. The difference of the assessed parameters between rats from two groups or four groups was tested using the the t-test or the one-way analysis of variance (ANOVA) and subsequently the least significant difference (LSD) test, respectively. Kolmogorov-Smirnov statistics was used to test the normality of data from all groups. Statistical significance was set at p < 0.05.

Figure 1A and B showed the plasmic concentrations of OCN and TNF-alpha in the rats that were subjected to 12-week long treatment. After the treatment, the OCN and TNF-alpha contents of rats in the OVX group were remarkably superior to those of rats in the SHAM group (p < 0.01). Contrastively, the DIO and EV group rats exhibited significantly lower plasmic OCN and TNF-alpha levels than the OVX group rats (p < 0.05).

Our micro-CT results suggested that in the OVX group, morphological parameters of bone, such as BV/TV, BS, Tb.Th, and Tb.N were substantially reduced, yet Tb.Pf, Tb.Sp and SMI were raised in comparison with the SHAM group (Fig. 2A-H). As shown in Fig. 2, significant changes in the bone morphological parameters were found when the rats were treated with EV or DIO. Furthermore, trabecular impairment due to ovariectomy was lessened by DIO or EV treatment (Fig. 3A-D).

To understand the effect of DIO from a histological perspective, we observed the alveolar bone beneath the first molar. Representative histomorphological images from four group rats are shown in Fig. 4. It is clear that the SHAM group rats had thicker trabeculae in alveolar bone and scant and smaller medullary cavity (Fig. 4A) in comparison to the OVX group rats. It appears that OVX decreases alveolar bone volume and increases the medullary cavity size (Fig. 4B). Furthermore, bone volume beneath the first molar was significantly enlarged after DIO or EV treatment (Fig. 4C and D), and EV appeared to have a stronger effect in this regard. With these data, our results from histological observations and micro-CT were consistent with each other.

Our microarray results revealed that there were, in total, 10 DE exonic circRNAs (Table 1) and 614 DE mRNAs (Table S4) in the alveolar bone between DIO and OVX group rats. Four and 6 circRNAs were up-and downregulated, respectively, while almost half of the mRNAs were increased and half were decreased (300 were upregulated and 314 downregulated).

Eight key DE mRNAs detected from the groups of OVX and DIO were assigned to the pathway of ROOCRA in IPA (Table 2). Figures 5 and 6 illustrate the potential effects of these key mRNAs on osteoclasts and osteoblasts, respectively, on the basis of the ROOCRA pathway.

The 10 circRNA expressions are listed in Tables 1 and 8 key DE mRNA expressions listed in Table 2 were measured. As a result, rno_circRNA_016717 was the only circRNA whose RT-qPCR data were in agreement with the microarray results (Fig. 7). As for mRNAs, RT-qPCR data of all mRNAs except Birc3 agreed with the microarray results (Fig. 8). Tables 3-4 listed the primers of DE circRNAs and key DE mRNAs for qRT-PCR experiments.

Based on the validated DE circRNA (rno_circRNA_016717) and 7 DE key mRNAs (Sfrp1, Pik3c2g, Wnt9b, Csf1, Il1rl1, Nfatc4, and Tnfrsf1a), 60 potential miRNA targets were predicted using miRanda and TargetScan. Thereafter, a circRNA-miRNA-mRNA coexpression network was established that included 380 nodes and 2173 edges (Fig. S1).

According to the interactions among circRNA, miRNAs and mRNAs, we searched for overlapping miRNAs that were downstream molecules of the DE circRNA (rno_circRNA_016717) and were upstream molecules of one DE upregulated mRNA (Sfrp1). Finally, only one circRNA-miRNA-mRNA axis, circRNA_016717/miR-501-5p/Sfrp1 (Fig. 9), was explored. The roles of this axis in the regulatory effect of DIO require further intensive study.

MiR-501-5p, the only predicted miRNA, was validated by RT-qPCR. As predicted, miR-501-5p expression in alveolar bone of DIO group rats was decreased compared to that in OVX rats (Fig. 10). Tables 5 listed the primers of predicted miRNA for qRT-PCR experiments.