C/EBPβ dictates postmenopausal FSHβ transcription and blockade of AEP/C/EBPβ pathway alleviates osteoporosis

TL;DR

This study shows that C/EBPβ controls FSHβ production in the pituitary gland and blocking the AEP/C/EBPβ pathway can treat osteoporosis in postmenopausal women.

Contribution

The study reveals C/EBPβ as a transcription factor for FSHβ and demonstrates that inhibiting AEP/C/EBPβ pathway can alleviate osteoporosis.

Findings

C/EBPβ directly binds to and regulates the fshb gene promoter in the pituitary gland.

Inhibiting AEP or knocking out C/EBPβ reduces FSHβ levels and prevents osteoporosis in mice.

An AEP inhibitor (#11a) performs as well as teriparatide in treating OVX-induced osteoporosis.

Abstract

Follicle-stimulating hormone (FSH), a gonadotropin that rises in post-menopausal females, activates its receptor FSHR to trigger bone loss via increasing bone resorption by osteoclasts. FSH stimulates CCAAT/enhancer binding protein beta (C/EBPβ) /asparagine endopeptidase (AEP) pathway, facilitating neural degeneration in the brain of mouse models with Alzheimer’s disease (AD). However, whether C/EBPβ/AEP pathway feeds back and modulates FSHβ bone resorption action remains elusive. Here we show that C/EBPβ acts as a transcription factor for fshb gene and directly binds its promoter, mediating its mRNA transcription in the pituitary gland. Knocking down C/EBPβ in primary pituitary cells significantly blunts GnRH (gonadotropin-releasing hormone)-induced FSHβ expression. Knockout of C/EBPβ also robustly diminishes FSHβ levels in mice. Inactivation of AEP, either by knockout of AEP or its small molecular inhibitor, antagonizes C/EBPβ and suppresses FSHβ levels, attenuating ovariectomy (OVX)-elicited osteoporosis. Markedly, a specific AEP inhibitor (#11a) displays comparable therapeutic effect as an FDA-approved drug teriparatide in OVX-induced osteoporosis. Hence, these findings support that C/EBPβ dictates FSHβ transcription and blocking AEP by its inhibitor represses C/EBPβ-mediated FSHβ levels, exerting prominent therapeutic efficacy toward osteoporosis.

Figures

C/EBPβ regulates the FSHβ expression in vitro and in vivo. a–c The Knockout of C/EBPβ significantly reduces FSHβ rather LHβ expression in OVX mice. Wild-type or C/EBPβ ± mice (3–4 months old) were subjected to the operation of sham (a: left) or OVX (a: right) and their pituitary glands were isolated to monitor the FSHβ protein (a). Quantification of FSHβ proteins in western blotting (b) and mRNA (c) levels in pituitary glands by RT-qPCR (*n* = 3) after sham or OVX. d, e Over-expression of C/EBPβ via lentivirus in primary pituitary gland cells induced FSHβ expression under vehicle treatment. Th

CF3CN and #11a diminish OVX-triggered osteoporosis by inhibiting the bone turnover in female mice. Femoral bone structures were assessed by in vitro μCT: a Images of the femoral indices of trabecular bone structure measured by in vitro μCT scan. Scale bar= 200 μm; b μCT scanning measurements of trabecular bone volume fraction (BV/TV), Conn.D., Structure model index (SMI), Trabecular number (Tb.N), Trabecular spacing (Tb.Sp), trabecular thickness (Tb.Th). (*n* = 7, one-way ANOVA). c DXA measurements of bone mineral density (BMD) and bone mineral content (BMC). (*n* = 6, one-way ANOVA). d Hemato

CF3CN and #11a promote MC3T3-E1 cell differentiation and mineralization. a ALP staining in MC3T3-E1 cells treated with BDNF, 7,8-DHF, CF3CN or #11a for 14 days. b Alizarin Red S mediated calcium staining in MC3T3-E1 cells treated with BDNF, 7,8-DHF, CF3CN or #11a for 21 days. c Western blotting of MC3T3-E1 cells treated with CF3CN at different time points. d, e Images and quantification of Western blotting of MC3T3-E1 cells treated with BDNF, 7,8-DHF, CF3CN or #11a for 4 days. (*n* = 3, one-way ANOVA). f AEP enzymatic activity assay of MC3T3-E1 cells. (*n* = 4, one-way ANOVA). Data represented

CF3CN and #11a inhibit osteoclastogenesis. a Representative images (left) and quantification (right) of Tartrate-resistant acid phosphatase-stained (TRAP-stained) sections of the distal femur bone. Scale bar = 500 μm (upper panel), Scale bar = 100 μm (lower panel). Areas of interest used in quantification were randomly selected around the circle-marked regions. Area 1 is located near the distal growth plate, area 2 represents cortical or trabecular bone. *n* = 3 slice from three mice. b The in-vitro bone resorption assay of RAW 264.7 cells induced by RANK-L with or without treatment of BDNF, 7

The model revealing AEP inhibitor and TrkB agonist distinctive mechanisms for anti-osteoporosis. The schematic diagram of AEP inhibitor (#11a) and TrkB agonist (CF3CN) treatment on osteoporosis. In menopause, elevated inflammation or oxidative stress activates C/EBPβ, which upregulates FSHβ transcription in the pituitary gland. AEP inhibitor #11a blocks AEP, leading to repression of C/EBPβ and FSH in the pituitary gland. It also blocks AEP proteolytic cleavage of fibronectin, which is crucial for osteoblast differentiation in the bone. Hence, #11a represses FSH elevation under OVX in the pitui

Full text

Introduction

FSH is a heterodimeric glycoprotein expressed by gonadotrophs in the anterior pituitary. The hormone-specific FSHβ-subunit is non-covalently associated with the common α-subunit that is also present in the luteinizing hormone (LH), another gonadotrophic hormone secreted by gonadotrophs and thyroid-stimulating hormone (TSH) secreted by thyrotrophs.^1^ Produced within gonadotrope cells in the anterior pituitary, FSH synthesis is rate-limited by transcription of the fshb gene.^2,3^ Transcription of fshb, the gene encoding the beta subunit, is regulated by both GnRH (gonadotropin-releasing hormone) and activin. FSH synthesis and release are driven by GnRH from the hypothalamus and activins and their inhibitors produced by the ovaries and within the pituitary.^4,5^ Knockout of GnRH in mice reduces FSH levels in serum by 60%–90%.^6^ The primary role of FSH is implicated in procreation via mediating both estrogen production in females and spermatogenesis in males.^7^ FSH concentration in serum sharply increases in the early menopausal transition, though estrogen levels remain within normal limits.^8^ Serum FSH level begins to escalate approximately 2–3 years before menopause. In addition to the classical FSH actions in gonads, FSH also stimulates bone loss via increasing bone resorption by osteoclasts^9^ and regulates body fat.^10^ Blockade of FSH from binding to FSHR not only elevates bone mass^11^ but also decreases body fat. Clearly, FSH acts as an important hormone in peri-menopause, exhibiting alternative mechanisms in parallel with the role of estrogen.

Transcription factor CCAAT/enhancer binding protein beta (C/EBPβ) mediates steroidogenic acute regulatory protein (StAR) and prostaglandin endoperoxide synthase 2 genes in ovarian granulosa cells upon FSH stimulation.^12–14^ On the other hand, FSH induces various C/EBP isoforms and activates C/EBPβ transcriptional activities via cAMP in Sertoli cells.^15^ In addition to these effects, C/EBPβ affects both osteoblast activity in bone formation and osteoclast activity in bone resorption. C/EBPβ usually expresses a long transactivator isoform LAP (active) and a truncated repressor LIP (inhibitory).^16^ The distinct C/EBPβ isoforms differentially regulate expression of the downstream transcription factor MafB that acts like a brake to restrict osteoclastic differentiation and osteolytic activity.^17^ C/EBPβ acts as a scaffold in the assembly of osteogenic TFs,^18^ including Runx2 or ATF4, stimulating osteoblast differentiation.^19,20^

AEP (asparagine endopeptidase, also called legumain) is a ubiquitously expressed lysosomal cysteine protease that cleaves its substrates after asparagine (N) under acidosis.^21^ We have shown it is activated in the brain and cuts SET, a DNase inhibitor, triggering neuronal cell death under stroke.^22^ AEP also acts as δ-secretase that simultaneously shreds both APP at N585 and Tau at N368 residues, respectively, promoting Aβ production, senile plaques and neurofibrillary tangles (NFT) formation in Alzheimer’s disease (AD).^23,24^ Furthermore, C/EBPβ regulates AEP expression in an age-dependent manner in the brain, mediating AD pathogenesis.^25^ In addition to AEP, C/EBPβ also mediates APP, MAPT and ApoE4 mRNA transcription in the brain, driving AD pathologies.^26^ Remarkably, FSH but not LH activates C/EBPβ/AEP signaling after menopause, facilitating AD pathogenesis, shedding light on the clinical observations of why women are more susceptible to AD than men.^27^ Women with one ApoE4 allele display greater risk and earlier onset of AD compared with men. FSH and ApoE4 additively contribute to AD pathogenesis via activating C/EBPβ/AEP pathway.^28^

Osteoporosis is a systemic skeletal disease characterized by low bone mass and micro-architectural deterioration of bone tissue, with a resultant increase in bone fragility and vulnerability to fracture.^29,30^ Bone homeostasis is determined by the balance between bone resorption and bone formation. Bone resorption is mediated by osteoclasts from hematopoietic stem cells, whereas formation of new bone is regulated by osteoblasts from bone marrow mesenchymal stromal cells (BMSCs), through several pathways, such as OPG/RANK-L/RANK. AEP modulates the differentiation and functions of both osteoclasts and osteoblasts. AEP is secreted as inactive pro-AEP (56 kD) and processed into enzymatically mature 36 kDa forms, as well as a 17 kDa C-terminal fragment, which inhibits osteoclast differentiation.^31^ AEP-deficient mice exhibit higher bone formation and lower bone resorption after OVX. Knockout of AEP improves trabecular bone density in OVX female mice.^32^ BDNF (brain-derived neurotrophic factor)/TrkB signaling via its downstream effector Akt phosphorylation on T322 residue inhibits AEP proteolytic activation.^33^ Activation of BDNF/TrkB signaling pathway by R13, a prodrug of small molecular TrkB agonist 7,8-dihydroxyflavone (7,8-DHF), prevents OVX-induced bone loss through Akt phosphorylation and inhibition of AEP, and increases osteoprotegerin (OPG) levels.^32^ In the current study, we explored whether C/EBPβ fed back and mediated FSH mRNA transcription in OVX and found that knockout of C/EBPβ or AEP substantially suppressed FSHβ expression. Inhibition of AEP with its specific inhibitor #11a robustly alleviated OVX-induced osteoporosis with reduced FSH levels.

Results

C/EBPβ regulates the FSHβ expression in vitro and in vivo

To assess whether C/EBPβ/AEP signaling mediates FSHβ expression, we conducted sham or OVX surgery in female wild-type (WT) and C/EBPβ^±^ mice (Fig. S1a). Immunoblotting (IB) with pituitary gland lysates revealed a prominent reduction of FSHβ proteins in OVX-treated C/EBPβ^±^ samples compared to WT tissues. By contrast, FSHβ levels were increased in sham-operated C/EBPβ^±^ mice versus WT mice (Fig. 1a and b). These observations suggest that C/EBPβ positively regulates FSHβ expression under pathological OVX condition. To investigate the mechanism underlying this phenotype, we assessed GnRHR expression. Both GnRHR protein and mRNA levels were significantly higher in these mice (Fig. S1b). Given that C/EBPβ is a transcriptional regulator, this inverse correlation suggests that it may function as a transcriptional suppressor of GnRHR (as depicted in the model in Fig. S1c). FSHR was consistently augmented in both sham and OVX C/EBPβ^±^ mice compared to WT mice. As a positive control, C/EBPβ downstream target AEP was also reduced (Fig. 1a and b). ERK signaling in the pituitary is required for female but not male fertility.^34^ GnRH-induced activation of MAPKs, including ERK1/2, JNK and p38.^35–37^ The p-Erk_1/2_ and its total levels were clearly elevated in C/EBPβ^+/−^ mice versus WT mice after both of OVX and sham operation (Fig. 1a and b), consistent with the proposed role of C/EBPβ as a negative regulator of GnRHR (Fig. S1c). Quantitative RT-PCR (qRT-PCR) analysis of FSHβ and LH revealed that depletion of C/EBPβ selectively repressed FSHβ but not LH mRNA levels in the pituitary gland (Fig. 1c). We also performed sham and OVX surgery on AEP^-/-^ mice (Fig. S1a) and found that mRNA and protein levels of C/EBPβ and FSHβ were significantly decreased in the pituitary glands from OVX-operated AEP^-/-^ mice in comparison to WT mice (Fig. S1d–f). Conversely, in the sham-operated groups, AEP knockout led to an increase in FSHβ. Again, FSHR and p-Erk_1/2_ and total Erk_1/2_ levels were noticeably augmented, when AEP was deleted (Fig. S1d–f).Fig. 1C/EBPβ regulates the FSHβ expression in vitro and in vivo. a–c The Knockout of C/EBPβ significantly reduces FSHβ rather LHβ expression in OVX mice. Wild-type or C/EBPβ ± mice (3–4 months old) were subjected to the operation of sham (a: left) or OVX (a: right) and their pituitary glands were isolated to monitor the FSHβ protein (a). Quantification of FSHβ proteins in western blotting (b) and mRNA (c) levels in pituitary glands by RT-qPCR (n = 3) after sham or OVX. d, e Over-expression of C/EBPβ via lentivirus in primary pituitary gland cells induced FSHβ expression under vehicle treatment. The primary cultures were either treated by vehicle or GnRH (10 nmol/L, 6 h). The protein and mRNA levels of C/EBPβ and FSHβ were detected by western blotting analysis (d and e: upper) and RT-qPCR (e: lower). pERK and ERK levels were detected by western blotting as well (d). f, g Knockdown of C/EBPβ via lentivirus-mediated shRNA expression in primary pituitary gland cells decreased FSHβ expression under GnRH treatment. The protein and mRNA levels of FSHβ were detected by western blotting analysis (f, g: upper) and RT-qPCR (g: lower). pERK and ERK levels were detected by western blotting as well (f). RT-qPCR data in (e, g) represent three independent experiments. Data represented as mean ± SEM, Student’s t test, ns, no significance, *P < 0.05, ** P < 0.01, ***P < 0.001, ****P < 0.000 1

To explore in-depth the effect of C/EBPβ on FSHβ expression, we prepared rat primary pituitary gland cells. Overexpression of C/EBPβ significantly augmented FSHβ under vehicle treatment (Fig. 1d and e). Of note, although GnRH alone potently increased FSHβ protein and mRNA levels, overexpression of C/EBPβ partially attenuated this enhancement (Fig. 1d and e), suggesting that C/EBPβ may function as a negative regulator of GnRH signaling. These findings were consistent with the observations in C/EBPβ^+/−^ mice (Fig. S1b and c). Knocking down C/EBPβ clearly decreased FSHβ levels under vehicle and GnRH treatments, revealed by IB and qRT-PCR analysis (Fig. 1f, g). Taken together, these results indicate that C/EBPβ acts as a key downstream transcription factor in GnRH signaling to stimulate FSHβ expression and provides a feedback mechanism to repress GnRHR expression. Hence, C/EBPβ/AEP signaling selectively regulates FSHβ expression in the pituitary glands.

To investigate whether C/EBPβ could act as a transcription factor and directly modulate FSHβ mRNA expression, we analyzed the mouse fshb gene promoter and identified 4 potential C/EBPβ binding elements with site 4 located in the 5’ UTR region from +1 to +43 (Fig. S2a). Chromatin immunoprecipitation (CHIP) assay using pituitary tissue of OVX WT mice showed that site 1, 2 and 4 demonstrated detectable binding activities with site 1 the most robust (Fig. S2b and c). Luciferase promoter activity assay showed mutation of site 1 substantially abolished C/EBPβ-mediated transcriptive activities (Fig. S2d and e). EMSA (electrophoretic mobility shift assay) showed that transfected C/EBPβ in the nuclear fraction strongly bound to the labeled fshb promoter (Site 1), which was stripped by addition of a large quantity of cold probe (Fig. S2f). Mutation of the binding motif in site 1 cold probe abrogated the competitive effect (Fig. S2f), indicating that this domain is indeed implicated in association with C/EBPβ. Therefore, C/EBPβ binds to fshb promoter and regulates its activity.

Based on previous findings showing significantly elevated C/EBPβ levels in the pituitary of OVX mice (Fig. 1) and prominent enrichment of C/EBPβ protein on the FSHβ promoter (Fig. S2a–f) and considering that OVX leads to a sharp decline in estrogen (E2) levels, we sought to investigate whether the activation of FSHβ by C/EBPβ after OVX is modulated by E2. Primary pituitary cells were cultured in charcoal-stripped FBS and phenol red-free medium. Under conditions with or without C/EBPβ overexpression, cells were treated with either vehicle or E2, followed by RT-qPCR and ChIP-qPCR analyses. GREB1, one of the most sensitive E2-responsive genes,^38,39^ exhibited a marked increase in mRNA levels upon E2 treatment (Fig. S2g). The finding that E2 treatment partially attenuated both the C/EBPβ-induced FSHβ mRNA upregulation and C/EBPβ promoter binding (Fig. S2g and h) suggests that E2 withdrawal following OVX may contribute to the FSHβ up-regulation by C/EBPβ.

#11a but not CF3CN suppresses the elevation of FSH induced by OVX

Recently, we showed that knockout of AEP improves trabecular bone density in OVX-treated female mice.^32^ Since both C/EBPβ and FSH are strongly reduced in AEP knockout mice, and FSH triggers bone loss, we wondered if AEP inhibitor #11a could exert any therapeutic efficacy toward osteoporosis. Accordingly, we subjected WT female knockout mice to OVX at the age of 12 weeks, followed by daily oral administration of AEP inhibitor #11a (7.5 mg/kg) or TrkB agonist CF3CN (5 mg/kg) consecutively for 3 months. CF3CN is a synthetic TrkB receptor agonistic derivative, which potently activates TrkB signaling after oral administration.^40^ The rationale to include a TrkB receptor agonist is for a head-to-head comparison of TrkB agonist that indirectly inhibits AEP via Akt-mediated phosphorylation and a direct AEP inhibitor in pharmacologically treating osteoporosis. The doses were chosen based on our previous studies that #11a specifically targets AEP and CF3CN selectively activates TrkB receptors.^40,41^ As expected, the shrunken uterine morphology and reduced uterus weight revealed that OVX surgery was successful. Drug treatment did not affect shrunken uterus (Fig. 2a). OVX strongly escalated FSH levels in the serum. Interestingly, serum FSH levels were selectively suppressed by #11a but not CF3CN after 3 months of drug treatment (Fig. 2b). Another TrkB agonist R13 was also unable to reduce FSH concentrations in serum in either WT or BDNF^±^ mice (Fig. S3a). Though OVX potently augmented LH levels in both the serum and pituitary glands, they were not affected by #11a or CF3CN treatment (Fig. S3a and b). Therefore, these data strongly support that AEP inhibitor #11a selectively antagonizes FSH but not LH expression.Fig. 2#11a but not CF3CN suppresses elevation of FSH expression induced by OVX. a Ovariectomized (OVX) mice fed with vehicle, #11a or CF3CN all displayed hypoplastic thread-like uteri and atrophic ovaries, which were normal in sham mice. Uterine weight was quantified and is presented on the right. b The serum FSH levels of sham and OVX mice fed with vehicle, #11a or CF3CN (left); the serum FSH levels of ovariectomized WT and BDNF^-/+^ mice treated with vehicle or TrkB agonist R13 (right). n  = 4–6 mice per group. c Immunoblotting of the pituitary glands from sham, OVX+ vehicle, OVX + #11a and OVX + CF3CN mice showed that #11a but not CF3CN inhibited FSHβ, pC/EBPβ, C/EBPβ and AEP levels induced by OVX. Data are representatives of three independent experiments. d Quantification of western blotting in (c), n = 3 per group. e The FSH levels in pituitary gland from sham and OVX mice treated with vehicle, #11a or CF3CN, detected by ELISA. n  = 6 mice per group. f The RT-qPCR results of FSHβ and C/EBPβ mRNA levels in the pituitary glands from sham, OVX+ vehicle, OVX + #11a and OVX + CF3CN mice, n = 3 per group. GAPDH was employed as the internal control. Data represented as mean ± SEM, one-way ANOVA, ns no significance, *P < 0.05, ** P < 0.01, ***P < 0.001

We wonder whether #11a influences FSHβ levels through repressing its expression in the pituitary gland. We performed IB analysis and found that OVX markedly augmented FSHβ, which were selectively blunted by #11a but not CF3CN. Noticeably, C/EBPβ levels correlated with FSHβ expression pattern. As expected, AEP auto-activation (mature active 36 kD form) via its own proteolytic cleavage was manifestly antagonized by #11a but not CF3CN (Fig. 2c and d). The FSH levels in the pituitary gland measured by ELISA were consistent with WB results, showing that #11a suppressed the elevation of FSH in OVX mice (Fig. 2e). qRT-PCR of fshb and cebpb corroborated IB findings (Fig. 2f). Our previous study shows that BDNF/TrkB signaling antagonizes C/EBPβ/AEP pathway.^33,42,43^ The observation that neither R13 nor CF3CN exerts anti-osteoporosis via repressing FSH levels allows us to hypothesize that this effect might be due to a lack of TrkB receptor expression in the pituitary gland. IB and qRT-PCR analysis validated that TrkB receptors were barely detectable in the pituitary gland, whereas they are abundantly distributed in the cortex, hippocampus and hypothalamus regions in the brain (Fig. S3c and d). The distinctive TrkB gene distributions in different tissues and deficiency from the pituitary gland were also confirmed in microarray and SAGE (serial analysis of gene expression) database (Fig. S3e). Hence, the TrkB agonists do not attenuate FSH levels in the serum due to TrkB receptors deficiency in the pituitary gland.

Recently, we reported FSHR expression in the brain, mediating FSH-triggered AD pathogenesis.^27^ Given #11a is brain permeable,^44^ in addition to suppression of FSHβ, #11a treatment also pronouncedly eradicated its receptor FSHR expression in the dentate gyrus and entorhinal cortex in the brain, which appeared to be escalated by OVX; These receptor levels appeared elevated following ovariectomy (OVX) (Fig. S4a). Moreover, immunofluorescent staining on the bone also displayed augmented C/EBPβ/AEP signals, coupled with FSHR elevation upon OVX, which were robustly blunted by #11a (Fig. S4b). Hence, these results indicate that OVX might activate C/EBPβ/AEP pathway via augmented FSH, resulting in FSHR escalation. AEP inhibitor #11a potently blocks its expression in the bone and brain.

CF3CN and #11a diminish OVX-triggered osteoporosis in female mice

To compare the therapeutic efficacy between #11a and CF3CN in treating osteoporosis, we performed microcomputed tomography (μCT) analysis of femurs harvested at sacrifice after chronic drug treatment and found a higher trabecular bone volume fraction (BV/TV), connectivity density (Conn. D) and a lower structure model index (SMI) in #11a or CF3CN-treated mice compared with vehicle-treated mice after OVX. As expected, OVX diminished trabecular number (Tb.N) and elevated trabecular separation (Tb.Sp), whereas trabecular thickness (Tb.Th) indices remained comparable among the groups. These indicators were similar between two pharmacological agents (Fig. 3a and b). Bone mineral density (BMD) and bone mineral content (MBC) analysis by dual-energy X-ray absorptiometry (DXA) measurements showed that OVX strongly decreased BMD than sham group, which were significantly attenuated by #11a or CF3CN (Fig. 3c).Fig. 3CF3CN and #11a diminish OVX-triggered osteoporosis by inhibiting the bone turnover in female mice. Femoral bone structures were assessed by in vitro μCT: a Images of the femoral indices of trabecular bone structure measured by in vitro μCT scan. Scale bar= 200 μm; b μCT scanning measurements of trabecular bone volume fraction (BV/TV), Conn.D., Structure model index (SMI), Trabecular number (Tb.N), Trabecular spacing (Tb.Sp), trabecular thickness (Tb.Th). (n = 7, one-way ANOVA). c DXA measurements of bone mineral density (BMD) and bone mineral content (BMC). (n = 6, one-way ANOVA). d Hematoxylin and eosin (H&E) staining of the distal femur bone. Scale bar = 500 μm. e, f Trabecular calcein double-fluorescence labeling images of the representative sections and histo-morphometric statistics. Mineral Apposition Rate (MAR), Bone Formation Rate per Bone Surface (BFR/BS), Osteoblast Surface per Bone Surface (Ob.S/BS), Osteoblast Number per Bone Surface (N.Ob/BS), Osteoclast Surface per Bone Surface (Oc.S/BS), Osteoclast Number per Bone Surface (N.Oc/BS). N.Oc/BS and Oc.S/BS are indices of bone resorption. N.Ob/BS, Ob.S/BS, MAR, BFR/BS, and MS/BS are indices of bone formation. (n = 4, one-way ANOVA). Data represented as mean ± SEM, one-way ANOVA, *P < 0.05, ** P < 0.01, ***P < 0.001

To further characterize the therapeutic effect of #11a or CF3CN in OVX-induced osteoporosis, we conducted the H&E staining and compared the bone morphology and white adipocytes after OVX surgery. Significant white adipocyte hyperplasia was observed in the bone marrow of OVX mice. Furthermore, trabecular bone volume was decreased in these mice compared to the sham-operated controls (Fig. 3d). Calcein double-fluorescence labeling allows the determination of the onset time and location of mineralization and the direction and speed of bone formation (Fig. 3e). Based on dynamic indices of femur trabecular bone formation, mineral apposition rate (MAR) and bone formation rate (BFR) were significantly decreased after OVX versus sham, and #11a or CF3CN treatment substantially reversed the reduction. Analysis of static indices of bone formation and resorption revealed that both number of osteoclasts (N. Oc/BS) and the percentage of surfaces covered by osteoclasts (OcS/BS) and number of osteoblasts (N. Ob/BS) were comparable among the groups (Fig. 3e and f). Together, these data suggest that CF3CN and #11a inhibit the bone turnover induced by OVX in female mice.

CF3CN and #11a promote MC3T3-E1 cells differentiation and mineralization

To compare the molecular mechanisms between CF3CN and #11a in promoting bone density increase in mice, we tested their effects in MC3T3-E1 cells in the presence of OIM (osteogenic induction medium). ALP (alkaline phosphatase) staining demonstrated that OIM treatment clearly enhanced osteoblast cell differentiation at 14 days, which were further elevated by BDNF or its agonists CF3CN and 7,8-DHF (Fig. 4a). Similar findings occurred in the presence of AEP inhibitor #11a. Alizarin Red staining also confirmed these observations at 21 days (Fig. 4b), indicating that CF3CN and #11a strongly facilitate MC3T3-E1 differentiation and calcium deposition. To assess whether CF3CN mimics BDNF in triggering TrkB-mediated signaling cascade in OIM-treated MC3T3-E1 cells, we conducted a temporal experiment and found that CF3CN triggered p-TrkB activation at 5 min and climaxed at 15 min, then the activities started to decay, whereas the downstream effectors including p-Akt and p-Erk1/2 and p-CREB signals began to escalate and reached the peak at 15 min and plateaued till 60 min (Fig. 4c), supporting that CF3CN indeed mimics BDNF by activating TrkB neurotrophic signaling. By contrast, p-C/EBPβ and its downstream AEP were inversely coupled with p-Akt and p-Erk1/2 signals (Fig. 4c). Immunoblotting showed that OIM notably activated C/EBPβ/AEP pathway, which was conspicuously blunted by BDNF or its agonists. As expected, blockade of AEP with #11a antagonized C/EBPβ/AEP pathway (Fig. 4d and e). Remarkably, OPG (osteoprotegerin), an inhibitory binding partner for RANK-L, was induced by BDNF and its agonists and #11a, while OIM-elicited RANK-L remained comparable among the groups (Fig. 4d and e). Numerous transcription factors including c-Jun and CREB are implicated in OPG mRNA transcription.^45,46^ Our previous study revealed that 7,8-DHF via activating CREB, a well-characterized downstream transcription factor of BDNF/TrkB pathway, stimulates OPG expression levels.^32^ Noticeably, p-JNK and p-CREB signals were time-dependently increased by CF3CN, suggestive of activation of these transcription factors. Consistently, CF3CN increased OPG levels (Fig. 4c and d). Osterix, a key early gene in the bone formation cascade, is usually used as a predictive measure of bone formation. Runx2 is a transcription factor that is essential for osteoblast differentiation. Immunoblotting showed that OIM prominently elevated both osterix and RUNX2 levels as compared to vehicle, which was further elevated by BDNF or its agonists CF3CN and 7,8-DHF (Fig. 4d). Similar findings occurred in the presence of AEP inhibitor #11a. As expected, BDNF and its agonists robustly activated p-TrkB/p-Akt/p-Erk1/2, whereas #11a displayed no effect (Fig. 4d and e). Consistent with active AEP reduction by BDNF or its agonists and #11a, the enzymatic assay confirmed that AEP protease activities were highly antagonized (Fig. 4f).Fig. 4CF3CN and #11a promote MC3T3-E1 cell differentiation and mineralization. a ALP staining in MC3T3-E1 cells treated with BDNF, 7,8-DHF, CF3CN or #11a for 14 days. b Alizarin Red S mediated calcium staining in MC3T3-E1 cells treated with BDNF, 7,8-DHF, CF3CN or #11a for 21 days. c Western blotting of MC3T3-E1 cells treated with CF3CN at different time points. d, e Images and quantification of Western blotting of MC3T3-E1 cells treated with BDNF, 7,8-DHF, CF3CN or #11a for 4 days. (n = 3, one-way ANOVA). f AEP enzymatic activity assay of MC3T3-E1 cells. (n = 4, one-way ANOVA). Data represented as mean ± SEM, one-way ANOVA, *P < 0.05, ** P < 0.01, ***P < 0.001

CF3CN and #11a inhibit osteoclastogenesis

Tartrate-resistant acid phosphatase (TRAP) staining revealed that OVX induced more osteoclast cells than sham control, which were manifestly repressed by CF3CN or #11a (Fig. 5a). Although the initial software-based general quantification of osteoclast number (N. oc/BS) and osteoclast surface/bone surface (Oc.S/BS) (Fig. 3f) showed no significant differences among groups, subsequent region-specific re-analysis revealed a significant increase of both parameters in cortical and trabecular bone following OVX, and this elevation was effectively suppressed by treatment with CF3CN and #11a (Fig. 5a, left panel). To investigate whether CF3CN and #11a alleviate osteoporosis via repressing osteoclastogenesis, we tested the effects of two compounds in RAW264.7 cells, which is a well-characterized cellular model for osteoclastic differentiation and has been widely used in the bone study. RANK-L elicits RAW264.7 cell osteoclastic differentiation, which proficiently generates osteoclasts in vitro.^47^ To assess osteoclastogenesis and osteoclast bone resorption activity, we cultured RAW264.7 cells with bone slices in the presence of RANK-L (30 ng/mL). Ten days after RANK-L treatment, the number of multinucleated osteoclastic cells and resorption pits on the bone slice were significantly elevated, and this increase was attenuated by BDNF, 7,8-DHF, or CF3CN. We observed a similar effect with #11a. These findings suggest that TrkB agonists or AEP inhibitor suppress RANK-L stimulatory effect on osteoclastogenesis and suppress bone resorption activity (Fig. 5b). Besides, the bone resorption marker CTX1 was measured by ELISA in serum samples from sham-operated or OVX mice administered vehicle, CF3CN, or #11a, and in culture media from undifferentiated and differentiated RAW 264.7 cells treated with vehicle, BDNF, 7,8-DHF, CF3CN, or #11a (Fig. 5c). Immunoblotting showed that RANK-L treatment strongly activated C/EBPβ/AEP signaling, which was robustly inhibited by BDNF and its agonists and #11a (Fig. 5d and e). Consequently, RANK-L-triggered AEP enzymatic activities were potently suppressed by these pharmacological agents (Fig. 5f). Immunohistochemistry (IHC) staining on the bone demonstrated that OVX treatment strongly increased C/EBPβ and AEP signals, which were abolished by CF3CN or #11a (Fig. 5g). Hence, CF3CN and #11a antagonize osteoclastogenesis accompanied with C/EBPβ/AEP signaling inhibition.Fig. 5CF3CN and #11a inhibit osteoclastogenesis. a Representative images (left) and quantification (right) of Tartrate-resistant acid phosphatase-stained (TRAP-stained) sections of the distal femur bone. Scale bar = 500 μm (upper panel), Scale bar = 100 μm (lower panel). Areas of interest used in quantification were randomly selected around the circle-marked regions. Area 1 is located near the distal growth plate, area 2 represents cortical or trabecular bone. n = 3 slice from three mice. b The in-vitro bone resorption assay of RAW 264.7 cells induced by RANK-L with or without treatment of BDNF, 7,8-DHF, CF3CN or #11a for 10 days. Scale bar = 100 μm for toluidine blue staining images (upper). Scale bar = 50 μm for TRAP staining images (lower). c ELISA assays detecting CTX1, a bone resorption marker. Serum levels of CTX1 were measured in sham-operated or OVX mice treated with vehicle, CF3CN, or #11a. CTX1 levels were also assessed in the culture medium of undifferentiated RAW 264.7 cells, as well as in differentiated RAW 264.7 cells treated with vehicle, BDNF, 7,8-DHF, CF3CN, or #11a. d, e Images and quantification of Western blotting of RAW 264.7 cells induced by RANK-L with or without treatment of BDNF, 7,8-DHF, CF3CN or #11a for 4 days. (n = 3, one-way ANOVA). f AEP enzymatic activity assay of RAW 264.7 cells induced by RANK-L with or without treatment of BDNF, 7,8-DHF, CF3CN or #11a for 4 days. (n = 4, one-way ANOVA). g Representative images of immunohistochemistry staining of C/EBPβ and AEP of the distal femur bone. Data represented as mean ± SEM, ns no significance, one-way ANOVA, *P < 0.05, **P < 0.01, ***P < 0.001

#11a displays the similar effect to Teriparatide on blocking trabecular bone loss induced by OVX in female mice

Since #11a suppresses FSH elevation in OVX mice, we compared the therapeutic efficacy of this AEP inhibitor with that of teriparatide (PTH 1-34 peptide fragment) for osteoporosis treatment. Teriparatide is an anabolic bone-forming agent and an FDA-approved drug. Teriparatide improves bone mineral density (BMD) and alters the levels of bone formation and resorption markers. Histo-morphometric studies showed teriparatide-induced effects on bone structure, strength and quality.^48^ Four weeks after OVX, the mice were treated with #11a or teriparatide, respectively, consecutively for 8 weeks as previously reported.^49^ Remarkably, micro CT (μCT) analysis revealed comparable anti-osteoporosis efficacy by both #11a and teriparatide (Fig. 6a). Assessment of femoral bone structure by in vitro μCT revealed elevated trabecular bone volume fraction (BV/TV), Conn.D and a reduced Structure model index (SMI) in #11a or teriparatide-treated mice compared with vehicle control after OVX. Moreover, OVX increased trabecular separation (Tb.Sp), while trabecular number (Tb.N) and trabecular thickness (Tb.Th) indices remained comparable among the groups. Noticeably, #11a demonstrated the similar therapeutic efficacy in the bone mineral density (BMD) and bone mineral content (BMC) to teriparatide treatment (Fig. 6b). H&E staining disclosed that both #11a and teriparatide increased trabecular bone marrow density and decreased the white adipocyte contents in the bone after OVX (Fig. 6c). TRAP staining indicated that OVX-induced osteoclast cells were decreased by both drugs (Fig. 6d). Therefore, these results suggest that #11a exhibits equivelent therapeutic efficacy as teriparatide in decreasing OVX-induced bone loss.Fig. 6#11a displays the similar effect with Teriparatide on blocking trabecular bone loss induced by OVX in WT female mice. a Representative images of the femoral indices of trabecular bone structure measured by in vitro μCT scan. Femoral bone structures were assessed by in vitro μCT in wild-type mice which were obtained OVX or sham operation at 12 weeks old, and some of which were treated with vehicle, #11a (7.5 mg/kg, feeding for 8 weeks) or Teriparatide (40 μg/kg, daily subcutaneous injection for 8 weeks). b μCT scanning measurements of BV/TV, Conn.D., SMI, Tb.Pf, Tb.N, Tb.Sp, Tb.Th, BMD and BMC. Data are shown as mean ± SEM, n = 3 mice per group, one-way ANOVA. c Hematoxylin and eosin (H&E) staining of the distal femur bone. Scale bar = 500 μm. d Tartrate-resistant acid phosphatase-stained (TRAP-stained) sections of the distal femur bone. Scale bar = 300 μm. The significant TRAP-positive signals are indicated by arrow heads. Data represented as mean ± SEM, ns no significance, *P < 0.05, **P < 0.01, ***P < 0.001

Discussion

In the current work, we show that C/EBPβ physically binds to the promoter of Fshb gene and mediates its mRNA transcription in the pituitary gland. Genetic knock-down of C/EBPβ strongly reduces FSHβ levels in the OVX-treated mice, and knockout of AEP feeds back and diminishes C/EBPβ, resulting in FSHβ reduction as well (Fig. 1, Fig. S1). These findings demonstrate that inactivation of downstream target AEP attenuates upstream C/EBPβ activation, leading to decreased FSHβ. In alignment with these results, inhibition of AEP with #11a potently reduces FSH but not LH levels in OVX-treated mice (Fig. 2, Fig. S3). We have shown that BDNF/TrkB signaling reciprocally antagonizes C/EBPβ/AEP pathway.^28,42,50^ Nonetheless, neither CF3CN nor R13, two agonists for activating TrkB receptors, fails to decrease FSH levels in the serum. To delineate the molecular mechanism why TrkB agonists fail to manipulate FSHβ levels, we conducted immunoblotting and qRT-PCR with different brain tissues, and found that TrkB receptors are barely expressed in the pituitary gland, whereas they are abundantly expressed in other brain regions including cortex, hippocampus and hypothalamus (Fig. S3). Transcription of fshb gene is rate-limiting in FSH production and is regulated by both GnRH and activin. Activin and GnRH synergistically induce fshb transcription via a distal enhancer.^51–53^ Notably, FSHβ levels are higher in the pituitary gland of C/EBPβ^±^ mice than WT mice under sham condition (Fig. 1a). In addition, whereas genetic ablation of C/EBPβ in primary pituitary cells abrogates GnRH-induced FSHβ expression, the converse manipulation—overexpression of C/EBPβ—elevates FSHβ under basal conditions but exhibits a mild suppressive effect under GnRH treatment (Fig. 1d and e). Based on the elevated GnRHR levels in C/EBPβ^+/−^ mice (Fig. S1b), we posit that C/EBPβ functions downstream of GnRH to repress its receptor’s expression. Collectively, our data reveal that C/EBPβ mediates a dual role: it is required for GnRH to escalate FSHβ expression, yet it also acts downstream of GnRH to suppress GnRHR expression (Figs. 1 and S1). We propose that this latter function constitutes a negative feedback mechanism (Fig. S1c).

In addition to the central dogma of classical FSH actions in gonads, FSH is directly involved in bone physiology.^9^ In contrast to declining estrogen levels in postmenopausal women, FSH levels increase sharply with osteoporosis. Interestingly, FSH is required for hypogonadal bone loss. Neither FSHβ nor FSH receptor (FSHR) null mice have bone loss despite severe hypogonadism. Bone mass is increased and osteoclastic resorption is decreased in haploinsufficient FSHβ^±^ mice with normal ovarian function, suggesting that the skeletal action of FSH is estrogen-independent.^9^ FSH regulates the reproductive axis via binding and activating FSHR expressed in granulosa cells in ovaries and Sertoli cells in testes.^54^ Inactivating either Fshb or Fshr leads to consistent reproductive defects.^55–57^ The canonical Gs/cAMP/PKA pathway has been considered as the major mechanism by which FSH exerts its actions within target cells.^58^ In neuronal cells, we found that FSH/FSHR via small G-protein Gαi-mediated Akt and Erk1/2 signaling to activate downstream C/EBPβ/AEP pathway to elicit AD pathologies.^27^ Remarkably, FSHR are upregulated in the hippocampus and entorhinal cortex after OVX, which are diminished by #11a. FSHR levels are augmented in bone upon OVX, coupled with activated C/EBPβ/AEP signals (Fig. S4). It is worth noting that FSHR was upregulated when either C/EBPβ or AEP was inactivated in the pituitary gland and ovary, and were downregulated in the cortex and hippocampus of C/EBPβ^±^ mice (Fig. S4C). On the contrary, FSH was highly reduced in the pituitary glands in C/EBPβ^+/−^ or AEP^-/-^ mice when compared to their WT littermates (Fig. 1, Fig. S1). These observations indicate that FSHR displays distinct expression patterns in different cell types by C/EBPβ/AEP signaling. FSH is solely expressed in the anterior pituitary, and FSHRs are expressed in numerous tissues, including the CNS and peripheral tissues.^59,60^ Conceivably, FSHR elevation in the pituitary gland compensates for FSH level repression, when C/EBPβ/AEP signaling is inactivated. Recently, we reported that the deletion of FSHR from the hippocampus of 3xTg AD mouse model abolishes OVX-induced AD pathogenesis. Neutralizing OVX-induced FSH with its specific anti-FSHβ antibody also diminishes AD pathologies in 3xTg female mice without altering LH levels.^32^ Chronic treatment with #11a results in FSHR reduction in the brain and bone is in agreement with the findings that AEP inhibition exerts pronounced therapeutic efficacy in AD^44^ and osteoporosis (Fig. 3), because antagonizing FSH with anti-FSHβ also demonstrates pronounced therapeutic effects toward both AD^27^ and osteoporosis.^61^

Mounting evidence shows that C/EBPβ/AEP signaling mediates not only neural degeneration^28^ but also bone degeneration.^32^ The C/EBPβ LAP isoform enhances expression of MafB that subsequently blunts osteoclastogenesis. MafB binds to and inactivates the osteoclastic TF including c-Fos, Mitf and NFATc1. Inactivation of these key TFs prevents osteoclast differentiation by inhibition of osteoclast target gene expression of OSCAR and NFATc1. On the other hand, the LIP isoform enhances osteoclastogenesis by decreasing expression of MafB, and lack of MafB allows access of the osteoclast TF (In conjunction with NFATc1 as a major osteoclast TF) to activate target genes and osteoclast differentiation. Moreover, C/EBPβ mediates osteoclast recruitment by regulating endothelial progenitor cell expression of SDF-1α.^62^ Nevertheless, C/EBPβ LIP also enhances osteoblast differentiation and function,^63^ possibly by acting as a coactivator for Runx2.^20^ Fitting with these observations, C/EBPβ^−/−^ mice demonstrate that bone mass is decreased^64^ due to reduced osteoblast differentiation and function,^19^ indicative of its pivotal roles in osteoblast differentiation and bone formation. Based on the observation that CF3CN and #11a suppressed C/EBPβ levels during bone cell differentiation (Figs. 4 and 5), we hypothesized that C/EBPβ heterozygosity might directly affect the differentiation of primary osteoblasts and osteoclasts. Ex vivo differentiation assays revealed that C/EBPβ heterozygosity promoted osteoblast differentiation and mineralization under vehicle treatment, an effect that was exacerbated by CF3CN but not by #11a (Fig. S5a–e). This supports that #11a promotes osteoblast differentiation by suppressing the C/EBPβ/AEP pathway. Furthermore, the differentiation of monocytes into osteoclasts was inhibited by C/EBPβ heterozygosity, and this inhibition was further enhanced by CF3CN treatment (Fig. S5f and g).

Conceivably, different C/EBPβ isoforms LAP and LIP may play distinctive roles in bone formation and resorption. In different contexts, LIP isoform may reveal different functions in promoting either osteoblast or osteoclast differentiation. Generally, a higher LAP/LIP ratio in bone is known to promote osteoblast differentiation and bone formation through upregulation of Runx2 and Osterix expression,^65^ whereas a lower ratio favors osteoclastogenesis and bone resorption by increasing the RANKL/OPG ratio.^66^ In the pituitary, while both of LAP and LIP levels were significantly up-regulated following OVX, the LAP/LIP ratio was down-regulated. Treatment with #11a or CF3CN effectively suppressed both isoforms in this context and increased the LAP/LIP ratio at the same time.

AEP contributes to decreased bone mass in postmenopausal osteoporosis.^32,67^ AEP regulates lineage commitment of human bone marrow stromal cells and its expression level and cellular localization are altered in postmenopausal osteoporotic patients. AEP inhibits osteoblast differentiation and in vivo bone formation through degradation of the bone matrix protein fibronectin. Inactivation of AEP leads to precocious osteoblast differentiation and increases vertebral mineralization in zebrafish. Noticeably, localized increased expression of AEP in bone marrow adipocytes is inversely correlated with adjacent trabecular bone mass in a cohort of patients with postmenopausal osteoporosis.^67^ On the other hand, AEP inhibits the formation of osteoclast-like multinucleated cells in human or mouse bone marrow cell cultures in the presence of 1,25-dihydroxyvitamin D3 [1,25-(OH)2D3] and parathyroid hormone-related protein (PTHrP). Implantation of AEP-overexpressing HEK293 cells in immune-deficient mice decreases PTHrP-induced hypercalcemia, associated with decreased osteoclast surface and numbers.^31^ RANK-L reduces AEP expression in RAW264.7 macrophages during differentiation to osteoclast-like cells. Inhibition of AEP enhances formation of osteoclast-like cells in MDSC (Myeloid-derived suppressor cells) cultures, possibly through altering the cleavage and activation of cathepsin L.^68^ Relative to these reports, we show that both TrkB agonists (7,8-DHF and CF3CN) and AEP inhibitor (#11a) potently repress RNAK-L-induced C/EBPβ/AEP signaling, blunting osteoclastogenesis (Fig. 5). Noticeably, both CF3CN and #11a strongly accelerate OIM-induced osteoblast cell differentiation, associated with suppressed C/EBPβ/AEP signaling and escalated OPG expression (Fig. 4d and e). Though TrkB agonist CF3CN substantially attenuates C/EBPβ/AEP signaling in cultured MC3T3-E1 and RAW264.7 cells (Figs. 4 and 5), it fails to repress C/EBPβ levels in OVX-treated animals due to deficiency of TrkB receptor distribution on the pituitary gland (Fig. S3). Consequently, FSH levels are not significantly reduced upon CF3CN treatment (Fig. 2).

Bone homeostasis depends on the resorption of bones by osteoclasts and the formation of bones by osteoblasts. Osteoblasts can also affect osteoclast formation, differentiation, or apoptosis through several pathways, such as OPG/RANK-L/RANK. In the current study, we demonstrate that CF3CN and R13 do not interfere with FSH levels, though both of them display robust anti-osteoporosis therapeutic efficacy (Figs. 3 and S3a). In contrast, AEP inhibitor #11a pronouncedly inactivated C/EBPβ, resulting in repression of FSH levels in the OVX-treated mice (Fig. 2). These results are in agreement with the observations that pronounced C/EBPβ reduction takes place in AEP^-/-^ mice (Fig. S1d). It is noteworthy that #11a demonstrates comparable therapeutic effects as FDA-approved drug teriparatide in treating osteoporosis (Fig. 6). Menopause elicits augmented inflammation and oxidative stress in female,^69–71^ which drive C/EBPβ activation.^72,73^ Subsequently, active C/EBPβ dictates FSHβ transcription in the pituitary gland. Inactivation of AEP via gene knockout or pharmacological inhibitor #11a represses C/EBPβ, resulting in FSH reduction. Together, our study demonstrates that both AEP inhibitor #11a and TrkB agonists are effective anti-osteoporosis therapeutic agents with different emphases on the interference mechanisms. AEP inhibitor blocks both FSH production and osteoclast differentiation, in addition to stimulating osteoblast differentiation, whereas TrkB agonist primarily promotes osteoblast differentiation and inhibits osteoclast differentiation without a prominent effect in lowering FSH production (Fig. 7).Fig. 7. The model revealing AEP inhibitor and TrkB agonist distinctive mechanisms for anti-osteoporosis. The schematic diagram of AEP inhibitor (#11a) and TrkB agonist (CF3CN) treatment on osteoporosis. In menopause, elevated inflammation or oxidative stress activates C/EBPβ, which upregulates FSHβ transcription in the pituitary gland. AEP inhibitor #11a blocks AEP, leading to repression of C/EBPβ and FSH in the pituitary gland. It also blocks AEP proteolytic cleavage of fibronectin, which is crucial for osteoblast differentiation in the bone. Hence, #11a represses FSH elevation under OVX in the pituitary gland and increases OPG and full-length Fibronectin levels in osteoblasts, while CF3CN primarily affects bone via elevating OPG rather than the pituitary gland to restore the bone homeostasis

Materials and methods

Animals

Female C57BL6/J wild-type mice (#000664), C/EBPβ^+/−^ mice (#006873) and BDNF^+/−^ mice (#002267) were obtained from Jackson Laboratory, then held and underwent breeding at Shenzhen Institute of Advanced Technology (SIAT), Chinese Academy of Sciences (CAS). The AEP knockout mice on a mixed C57BL/6 and 129/Ola background were generated as reported.^74^ All in vivo experiments were carried out in female mice. All mice were kept under specific pathogen-free conditions in an environmentally controlled clean room with the humidity ranging 40%–60% and housed at 22 °C on a 12-h/12-h light/dark cycle. Food and water were provided ad-lib. All procedures performed in studies involving animals were in accordance with the ethical standards of the Institutional Animal Care and Use Committee of Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences. BDNF^+/−^ mice and WT littermates, C/EBPβ^±^ mice and WT littermates, AEP^-/-^ mice and WT littermates were bilaterally ovariectomized or sham-operated at 12 weeks of age. Four days after OVX, the WT and BDNF^+/−^ mice received vehicle or R13 dissolved in 5% DMSO/0.5% methylcellulose at a dose of 21.8 mg/kg/d, six days per week, for 8 weeks by gavage. In another group, 4 weeks after OVX, WT mice were either treated with AEP inhibitor (compound #11a) or TrkB agonist (CF3CN) dissolved in 5% DMSO/0.5% methylcellulose by oral gavage at doses of 7.5 mg/kg/d (#11a) and 5 mg/kg/d (CF3CN), six days per week, for 8 or 12 weeks. As the positive drug to evaluate the therapeutic efficacy, Teriparatide (40 µg/kg) was subcutaneously injection daily as reported^75^ for 8 weeks. Mice were sacrificed, after which the uteri were carefully dissected out. In the sham group, the uteri were dissected along with the ovaries, rinsed with PBS, and then photographed. The ovaries were then carefully removed, and the uteri were lightly blotted on absorbent paper to remove surface moisture before final weighing.

Cell line culture

Murine MC3T3-E1 (subclone 4) cells and RAW 264.7 cells were obtained from American Type Culture Collection (ATCC, Manassas, VA, USA, catalog#: CRL-2593 and catalog#: TIB-71). The MC3T3-E1 cells were cultured in alpha-MEM with 10% FBS and 0.1% penicillin/streptomycin, but without ascorbic acid. The RAW 264.7 cells were cultured in DMEM supplemented with 10% FBS and 0.1% penicillin–streptomycin. The cells were maintained at 37 °C in a humidified atmosphere of 95% air and 5% CO_2_.

Primary rat pituitary cell culture and treatment

Primary culture and treatment of rat pituitary cell was performed by a modification of the method reported elsewhere.^76,77^ Briefly, the anterior pituitary glands from male Sprague-Dawley rats were excised and were rinsed with ice-cold Dulbecco’s A PBS (PBS), and then minced into small pieces. Pituitary pieces were placed in Hank’s balanced salt solution with 25 mmol/L HEPES, pH 7.2, containing 0.25% collagenase and 0.25% trypsin 15 min at 37 °C in a shaking water bath. Tissue pieces were then incubated sequentially with pancreatin (2.5 mg/mL; Sigma, St. Louis, MO) for 15 min. The same volume of DMEM with 10% fetal bovine serum (FBS) was added to stop the reaction and then deoxyribonuclease I (DNase-I; Sigma) was added to a final concentration of 20 μg/mL and incubated for another 5 min at 37 °C. Tissue was pelleted between steps by centrifugation at 400 × g for 10 min. After enzyme treatment, tissue fragments were resuspended in 2 mL DMEM with FBS and pipette to mechanically dissociate the cells, and then filtered through a nylon mesh (Becton Dickinson Labware, Franklin Lakes, NJ) to remove tissue debris. Cells were re-suspended by DMEM with 10% FBS and plated at 10^6^ cells/well density in 6-well plates. Cells are cultured under a saturated atmosphere of 5% CO_2_,-95% air in a CO_2_, incubator at 37 °C, the medium was exchanged for a fresh one every 2 days. The charcoal-stripped FBS (Moocow, cat # CSF15-0212) and phenol red-free DMEM medium (Gibco, cat# 11054020) were used here for cellular experiments investigating the E2 effect (Fig. S2g, S2h). After 7 days, cells were infected by lentivirus. Virus solution (1 µL) and polybrene (5 µg/mL) were added to 1.5 mL culture medium and applied to primary cultures in each well for 24 h. The knockdown virus LV-U6-shC/EBPβ and LV-U6-shRNA(scramble)-CMV-EGFP-WPRE were ordered from OriGene Technologies, Inc. (Rockville, MD). The overexpression virus LV-CMV-C/EBPβ-WPRE and LV-CMV-EGFP-WPRE are packaged by OBiO Tech, Inc. Virus titers are 3 × 10^9^ IU/mL–5 × 10^9^ IU/mL. Three days post-infection, the medium was changed to serum-free DMEM with 0.1% bovine serum albumin (BSA) 16 h before reagent treatment. Cells were then treated with GnRH (10 nmol/L, Sigma Chemical Co., St. Louis, MO) or E2 (10 nmol/L, MedChemExpress, HY-B0141) for 4 h, after which the cells were lysed for WB, ChIP-PCR or RT-qPCR assay.

Antibody and reagents

Antibody to FSHβ (sc-374452, C12, 1:500 dilution for western blotting), C/EBPβ (HT-7) (catalog#: sc-7962, 1:1 000 for western blotting, 1:200 for immunofluorescence and immunohistochemistry), RANK-L (catalog#: sc-377079, 1:1 000 for western blotting), OPG (catalog#: sc-390518, 1:1 000 for western blotting), osterix (catalog#: sc-393325, 1:1 000 for western blotting) and RUNX2 (catalog#: sc-101145, 1:1 000 for western blotting) were from Santa Cruz; Antibodies to Legumain (D6S4H) (catalog#: 93627, 1:2 000 for western blotting, 1:500 for immunohistochemistry), p-C/EBPβ(catalog#: 3084 s, 1:1 000 for western blotting), Akt (catalog#:4691 s, 1:2 000 for western blotting), p-Akt(S473) (catalog#: 9271 s, 1:1 000 for western blotting), SAPK/JNK (catalog#: 9252 s, 1:1 000 for western blotting), Phospho-SAPK/JNK (Thr183/Tyr185) (catalog#: 9251 s, 1:1 000 for western blotting), ERK1/2 (9102 s, 1:1 000 dilution for western blotting), pERK1/2 (9106 s, 1:2 000 dilution for western blotting), CREB (catalog#: 9197 T, 1:1 000 for western blotting) and p-CREB (catalog#: 9198 T, 1:1 000 for western blotting) were purchased from Cell Signaling Technology; Antibody to TrkB (catalog#: MAB397, 1:1 000 for western blotting) was from R&D System; Antibodies to β-actin (catalog#: A5316, 1:3 000 for western blotting) and Fibronectin (catalog#: F3648, 1:1 000 for western blotting) were from Sigma-Aldrich; Antibody against Legumain (catalog#: AF2058-SP, 1:400 for immunofluorescence) was purchased from R&D Systems; Antibody against FSHR (PA5-50963, 1:1 000 dilution for western blotting and 1:200 for immunofluorescence) was purchased from Thermo Fisher Scientific; Antibody to MAP2 (catalog#:17490-1, 1:5 000 for immunofluorescence) was from Proteintech. Antibody to p-TrkB (Tyr816) (1:1 000 for western blotting) was developed in the Ye lab; Anti-mouse IgG-HRP (catalog#: 70765, 1:5 000 for western blotting) and anti-rabbit IgG-HRP (catalog#: 70745, 1:5 000 for western blotting) were from Cell Signaling Technology.

Alpha-MEM (catalog#: A1049001) was obtained from Gibco. Lipo3000 transfection reagent (catalog #: L3000008) was obtained from Invitrogen. K252a (catalog#: ab120419) were obtained from Abcam. TRACP&ALP double-staining kit (catalog#: MK300) was from TakaRa Bio. Alkaline Phosphatase Assay kit (catalog#: ab83369) was from Abcam. The AEP substrate Z-Ala-Ala-Asn-AMC (catalog#: 4033201) was from Bachem, and EZ-Link Sulfo-NHS-LC-Biotinylation Kit (catalog #: 21435) was obtained from Thermo Fisher. Serum mouse follicle-stimulating hormone (FSH) ELISA Kit (catalog#: CSB-E06871m-96T) and mouse luteinizing hormone (LH) ELISA kit (catalog#: CSB-E12770m-96T) were from CUSABIO. Mouse CTX1 ELISA Kit (catalog#: E-EL-M3023) were from Elabscience (Wuhan, China), Teriparatide (catalog#: HY-P0059) for mice treatment was obtained from MCE (MedChemExpress).

Osteogenic differentiation of primary monocyte and MC3T3-E1 cell line

Primary osteoblasts were isolated from C/EBPβ mice and cultured according to a well-established protocol.^78^ Two days after seeding, when the osteoprogenitors reached approximately 70%–80% confluence, the culture medium was replaced with osteoblast induction medium (containing α-MEM, 10% FBS, 10 mmol/L β-glycerophosphate, and 50 μg/mL ascorbic acid). The cells were maintained in this induction medium for two days. Subsequently, they were treated with either DMSO (vehicle control), 10 nmol/L CF3CN, or 10 nmol/L #11a for 7 days. During this period, the medium was replaced every three days. Alkaline phosphatase (ALP) staining was performed at the end of the 7-day treatment period.

MC3T3-E1 cells were seeded into plates in complete medium and cultured for 24 days until the cells reached 70% confluence for initiation of the differentiation. Cells were incubated in an osteogenic induction medium (OIM) containing α-MEM, 10% FBS, dexamethasone (10^−7^ mol/L), β-glycerophosphate (10 mmol/L) and ascorbic acid (50 μg/mL). The differentiation medium was replaced every 3 days, with DMSO, BDNF (50 ng/mL), 7,8 DHF (0.5 μmol/L), CF3CN (10 nmol/L), or #11a (10 nmol/L) added into the medium.

Osteoclast differentiation of primary osteoblast and RAW264.7 cell line

Primary osteoclasts were isolated from C/EBPβ mice and cultured according to a well-established protocol.^78^ Specifically, monocytes were induced with osteoclast stimulation medium containing M-CSF (30 ng/mL) and RANKL (60 ng/mL). During the 7-day induction period, the cells were treated with either DMSO (vehicle control), 10 nmol/L CF3CN, or 10 nmol/L #11a. The medium was replaced every three days.

RAW264.7 cells were seeded in 24 wells plates and cultured for 24 h in DMEM with 10% FBS and 0.1 penicillin/streptomycin. The medium was changed to α-MEM with 5% FBS, 0.1% penicillin/streptomycin. The receptor activator of NF-κB ligand (RANK-L, 30 ng/mL) was added to induce osteoclast differentiation. The medium was replaced every 3 days, accompanied with DMSO, BDNF (50 ng/mL), 7,8 DHF (0.5 μmol/L), CF3CN (10 nmol/L), or #11a (10 nmol/L) added into the medium.

ALP staining

MC3T3-E1 cells were plated in 24-well plates, cultured in complete medium or OIM, and treated with DMSO, BNDF (50 ng/mL), 7,8 DHF (0.5 μmol/L), CF3CN (10 nmol/L), or #11a (10 nmol/L) for 14 days. The Cells were washed in PBS twice, and fixed for 10 min with fixing buffer at room temperature, stained the ALP staining with the TRACP&ALP double-staining kit.

Alizarin red S staining

MC3T3-E1 cells were plated in 24-well plates, cultured in complete medium or OIM, with treatment of DMSO, BNDF (50 ng/mL), 7,8 DHF (0.5 μmol/L), CF3CN (10 nmol/L), or #11a (10 nmol/L), Cells were washed in distilled water twice and fixed in 70% ice-cold ethanol. Then the cells were stained with 2% Alizarin Red S solution to detect calcification.

TRAP and H&E staining

RAW 264.7 cells were cultured in α-MEM with or without RANK-L, in the presence or absence of DMSO, BNDF (50 ng/mL), 7,8 DHF (0.5 μmol/L), CF3CN (10 nmol/L), or #11a (10 nmol/L) for 5 days. The cells were washed in PBS twice, fixed in fixing solution for 10 min at room temperature, and then stained the TRAP activity with the TRACP&ALP double-staining kit according to the supplied protocols.

The sections of right femurs were subjected to hematoxylin and eosin (H&E) staining and TRAP staining. After being freed from soft tissues, the right femurs were fixed in 4% paraformaldehyde for 24 h at 4 °C and/or decalcified and then dehydrated. Subsequently, the specimen was embedded in wax, made into 4 mm paraffin sections, and baked at 60 °C. The sections were deparaffinated before being subjected to H&E staining and TRAP staining. After staining, the sections were sealed in neutral gum and analysis was performed with NIKON Eclipse Ci microscope equipped with a digital camera.

Western blotting

Tissues and cells were washed with ice-cold PBS and lysed in (50 mmol/L Tris, pH 7.4, 40 mmol/L NaCl, 1 mmol/L EDTA, 0.5% Triton X-100, 1.5 mmol/L Na3VO4, 50 mmol/L NaF, 10 mmol/L sodium pyrophosphate, 10 mmol/L sodium β-glycerophosphate, supplemented with protease inhibitors cocktail) on ice for 0.5 h, and centrifuged for 25 min at 15 000 r/min. The supernatant was boiled in SDS loading buffer. After SDS-PAGE, the samples were transferred to a nitrocellulose membrane. The membrane was blocked with TBS containing 5% nonfat milk and 0.1% Tween 20 (TBST) at room temperature for 2 h, followed by the incubation with primary antibody at 4 °C overnight, and with the secondary antibody at room temperature for 2 h. After washing with TBST, the membrane was developed using the enhanced chemiluminescent detection system.

AEP activity assay

Cell lysates (10 μg) were incubated in 200 μL assay buffer (20 mmol/L citric acid, 60 mmol/L Na_2_HPO_4_, 1 mmol/L EDTA, 0.1% CHAPS, and 1 mmol/L DTT, pH 6.0) containing 20 μmol/L δ-secretase substrate Z-Ala-Ala-Asn-AMC (Bachem). AMC released by substrate cleavage was quantified by measuring at 460 nm in a fluorescence plate reader at 37 °C for 2 h in kinetic mode. Each sample has three technical repeats.

Quantitative real-time PCR analysis

Total RNA was isolated by TRIzol (Life Technologies). Reverse transcription was performed with PrimeScript™ RT reagent Kit (TAKARA, #RR037Q). All real-time PCR reactions were performed using the ABI 7500-Fast Real-Time PCR System and the SYBR™ Green PCR Mix Kit (Applied Biosystems, #4309155). The relative quantification of gene expression was calculated using the ΔΔCt method. The real-time qPCR primers were designed as below:

mFSHβ-F: AGGGATCTGGTGTATAAGGACC

mFSHβ-R: CACACTTGCCACAGTGACA

rFSHβ-F: AGGGATCTGGTGTATAAGGACC

rFSHβ-R: GCATTCAGTGGCTACTGGATATG

LHβ-F: GCTGCTGCTGTGGCTGCTG

LHβ-R: GGGCAGAACTCATTCTCTGC

CGA(α-subunit)-F: GCTGTCATTCTGGTCATGCTG

CGA(α-subunit)-R: AAGCAACAGCCCATACACTG

mFSHR-F: ATAGAGATCTCTCAGAATGATGTC

mFSHR-R: GGAAGGCCTCAGGGTTGAT

mC/EBPβ-F: GCCGCGACAAGGCCAAGAT

mC/EBPβ-R: GGCTCGGGCAGCTGCTTGA

GAPDH-F: CATCACCATCTTCCAGGAGC

GAPDH-R: CCTTCTCCATGGTGGTGAAGA

For each data point, at least three technical duplicated wells were used. GAPDH was used for housekeeping gene control.

μCT measurements

μCT scan and analysis were performed in femurs ex vivo using a μCT- 40 scanner, as previously reported.^79,80^ Voxel sizes were 12 μm^3^ for the in vitro measurements of femurs. For the femoral trabecular region, we analyzed 140 slices, beginning with 50 slices below the distal growth plate. X-ray tube potential was 70 kVp, and integration time was 200 ms. Representative samples were reconstructed in 3D to generate visual representations of trabecular structure.

Quantitative bone histomorphometry

The measurements, indices, and units for histo-morphometric analysis were recommended by the Nomenclature Committee of the American Society of Bone and Mineral Research.^81^ Mice were injected with calcein (25 μg/g) subcutaneously at day 10 and day 3 before sacrifice. Bone histo-morphometric analysis was performed at the University of Alabama at Birmingham Center for Metabolic Bone Disease Histomorphometry and Molecular Analysis Core Laboratory. The Goldner’s trichrome-stained plastic-embedded sections of calcein double labeled femora of the mice were analyzed by an operator blinded as to the nature of the samples.

Prediction of C/EBPβ binding motif

The DNA sequence of the human FSHβ was obtained from the UCSC genome browser (https://genome.ucsc.edu). The binding motifs of C/EBPβ on human FSHβ promoters and 5’ UTR were analyzed by JASPAR online tools (http://jaspar.genereg.net). The Relative profile score thresholds were set >80%.

Plasmid construction

The luciferase reporter plasmid of human FSHβ promoter with 5’ UTR luciferase reporter plasmids was constructed by inserting the FSHβ gDNA fragment (−1 086 to +899) into a pGL3-basic plasmid (Promega, #U47295). To generate the construction of FSHβ promoters with 5’ UTR using the In-Fusion HD cloning Kit (TAKARA, #639649), two forward primers and two reverse primers were used as follows:

pLv6-Luc linearization-F: ATGGAAGACGCCAAAAACATAAAG

pLv6-Luc linearization-R: AAGGGCGAATTCGAAGCTTGAGCTCG

FSHβ promoter insert-F: cttcgaattcgcccttatactcaacatcaatccagcaga

FSHβ promoter insert-R: tttggcgtcttccatCCTGTGCAGTCAGCTGTCTC

Mutations of C/EBPβ binding motif were generated with the Hieff Mut™ Site-Directed Mutagenesis Kit (Yeasen, #11003ES10). The primers for the mutagenesis of promoter luciferase were:

Site1-F:CtctagCCtcGaGGtatatccagagtaatagcatgactcat

Site1-R:aCCtCgaGGctagaGagaacatttctgctggattgatgt

Luciferase assay

Primary pituitary cells were seeded in six-well plates and cultured for 7 days. Cells were then infected with lentivirus carrying FSHβ promoter luciferase reporters, as well as C/EBPβ or control overexpressing plasmids. Cells were harvested in passive lysis buffer at 3 days after infection and analyzed using a luciferase reporter assay system according to the manufacturer’s protocol (Promega) on a microplate reader. Relative light units of FSHβ promoter luciferase were normalized to luciferase light units without promoter drive. The experiments were performed in triplicate.

Electrophoretic mobility shift assay (EMSA)

Nuclear proteins of HEK293 with C/EBPβ overexpression were extracted by using NEPER Nuclear and Cytoplasmic Extraction Reagents (Life Technologies). Protein concentrations were determined using a Coomassie Brilliant Blue protein assay kit (Bio-Rad). A double-stranded oligonucleotide probe: 5’-ttctagtttcaaaatatatc-3’-Biotin for site 1 of FSHβ promoters was designed with prediction results and labeled with biotin. Unlabeled probe (cold probe): 5’-ttctagtttcaaaatatatc-3’ was used as competitor. Unlabeled mutant probes, mut probe1: 5’-CtctagCCtcGaGGtatatc-3’ was used to verify the binding specificity. The C/EBPβ (Santa Cruz, sc-7962) antibody was added in super-shift group. EMSA assay was performed as described in LightShift Chemiluminescent EMSA Kit (Life Technologies).

ChIP assay

ChIP assay was performed using the ChIP kit (#ab500, Abcam). Pituitary glands from OVX wildtype C57BL/6 J mice were excised and choped into small pieces using two razor blades. Then tissue pieces were immediately fixed with 10% formaldehyde at room temperature for 20 min, stopped by adding glycine to a final concentration of 0.125 mol/L. After PBS washed, tissues were resuspended in PBS/PMSF and homogenized by 2–3 strokes in a Dounce homogenizer. After obtaining the single cell suspensions, transferred to a sonication buffer, and sonicated nine times for 10–20 s each at a 30% setting in the ice bath (VibraCell Sonicator). Sample DNA was fragmented by sonication to an average size of 400 bp, cleared of debris by centrifugation 5 min at 14 000 g and the supernatant harvested. A 50 μL aliquot of the supernatant was saved for input DNA analysis. Add 5 μg IgG or C/EBPβ antibody and 40 μL Protein-A or G Sepharose into 25 μg of DNA samples, incubate overnight at 4 °C. After washing the beads, DNA purifying slurry and Proteanase K were added to purify DNA. The input sample was processed in parallel. Eluted IP and input DNA samples were used for PCR and qPCR analysis. The primers used to evaluate the C/EBPβ enrichment on predicted binding sites of FSHβ promoter and 5’ UTR are:

Site1-F:ATCTTTAGACTCATCCTCACCTTT

Site1-R:GTGTTTGTTTGGGCCAATG

Site2-F:ACCTGTGTTACTACAAACCATCT

Site2-R:CTCTGTGTGATTTGGTCCACT

Site3-F:ATACACTTGGAGTGTTCAGTCT

Site3-R:CTGATCTTTAACTTGGCGAACTC

Site4-F:ATTGGTCATGTTAACACCCAGT

Site4-R:CAGAATAAGATGCAAAGCTGGAT

Immunostaining staining

Immunohistochemical staining was performed with an IHC staining kit (Invitrogen). Briefly, mice brain or femurs paraffin-embed sections on slides were deparaffinated and treated with 3% hydrogen peroxide at room temperature for 10 min. Manufacturer-supplied blocking buffer (Invitrogen) was used for each reaction. Then the sections were incubated with primary antibodies overnight at 4 °C. Biotin-conjugated secondary antibodies (Jackson ImmunoResearch), streptavidin-conjugated HRP (Invitrogen) were applied to enhance the signals. For double or triple immunofluorescence staining, brain or femurs sections were deparaffinated and blocked in 5% BSA and 0.3% Triton X-100 for 30 min, followed by overnight incubation with primary antibodies at 4 °C. After washing with PBST, the sections were incubated with a mixture of Alexa Fluor 488-, 555- and 647-coupled secondary antibodies (Invitrogen) for detection. DAPI (1 μg/mL) (Sigma) was used for staining nuclei. Then, coverslips were mounted on glass slides and imaged using a confocal microscope (LSM 980, 640 Zeiss). The fluorescence intensity was quantified using ZEN (blue) software.

Statistical analysis

Information on biological replicates (n) is indicated in the figure legends. All statistical analyses were performed by GraphPad Prism 7 software. All data are presented as means ± SEM unless otherwise stated. When only two groups were compared, the statistical differences were assessed with the double-sided Student’s t test. The number of samples per group (n ≥ 3) is stated in the figure legends. Comparisons among multiple groups were performed using one-way analysis of variance (ANOVA) with Tukey’s post hoc test. Two-way ANOVA was used for analysis of multiple groups with Tukey’s multiple comparison post hoc test. For all experiments, *P ≤ 0.05, **P < 0.01 or ***P < 0.001 was considered a significant difference.

Supplementary information

Original blots Figure S1 Figure S2 Figure S3 Figure S4 Figure S5 Supplementary Figure Legend