Thyroid disorders can affect the whole body´s metabolism and energy level, heart rate, body weight, mood, and also skeletal growth and bone health1. Although hyperthyroidism, a condition with high circulating concentrations of thyroid hormones 3,3´,5-triiodo-L-thyronine (T3) and L-thyroxine (T4), is an established risk factor for secondary osteoporosis, the molecular mechanisms for thyroid hormone excess driven bone loss still remain incompletely understood2. In patients and preclinical models, hyperthyroidism leads to an increase of both, bone formation and bone resorption, overall shortening the bone turnover cycle2,3. Given the prevalent bone resorption and a net bone loss by 10% per cycle, hyperthyroid patients may suffer from various facets of osteoporosis such as low bone mass, impaired bone quality and higher skeletal fragility leading to an increased susceptibility to bone fractures2,4,5,6,7,8.

At the cellular level, thyroid hormones are reported to stimulate osteoblast differentiation and activity9,10,11,12,13,14,15. A few studies confirm that bone-resorbing osteoclasts express thyroid hormone receptors and take up thyroid hormones by specific transporters and thus, might respond to thyroid hormones16,17,18,19,20. Nevertheless, cytokines and paracrine factors secreted by thyroid hormone stimulated cells of the osteoblast lineage are postulated to have the greater impact on osteoclastogenesis2,21,22,23,24,25,26.

One of the most important osteogenic pathways that controls not only osteoblastogenesis but also osteoblast-osteoclast coupling is the bone morphogenetic protein (BMP) pathway27,28. BMP ligands exert their effects via serine threonine transmembrane receptors, BMP type I and type II receptors. BMP type I receptors such as BMPR1A (also ALK3) can bind BMP ligands and subsequently form a heterotetrameric receptor complex with a type 2 receptor such as BMPR2. Those receptor complexes can activate two subsequent signaling pathways: the canonical, SMAD-dependent and the non-canonical pathway involving protein kinases such p38, PI3K/AKT or MAPK27,28. Activated SMAD1/5/9 proteins can dimerize with SMAD4 to form transcription factor complexes that induce the expression of BMP target genes such as Runx2 or Id127,28.

The skeletal phenotypes of several transgenic mouse models targeting BMP ligands, BMP receptors, SMAD proteins and BMP inhibitors have been studied to decipher their role in bone growth and metabolism. As such, osteoblast- and osteocyte-specific Bmpr1a knockout mice present a striking osteosclerotic bone phenotype due to low bone turnover, with a drastic reduction in bone resorption exceeding the low bone formation29,30,31,32. This suppression of osteoclasts and thus, bone resorption is driven by a decreased ratio of receptor activator of NF-κB ligand to osteoprotegerin (RANKL/OPG), a prominent stimulator and inhibitor of osteoclastogenesis, respectively, secreted by Bmpr1a-deficient osteogenic cells29. Recent studies also confirm the importance of BMP signaling in osteoclasts, especially during maturation and in promoting their resorptive activity33,34,35,36,37.

In our previous study, we demonstrated that thyroid hormones activate SMAD-dependent BMP signaling in osteoblasts26. Pharmacological blockade of BMP signaling either at the receptor or ligand level impaired thyroid hormone response of osteoblasts26. Importantly, administration of BMPR1A-Fc, a potent BMPR1A-specific BMP ligand scavenger, prevented osteoporosis in hyperthyroid mice by normalizing both bone formation and bone resorption26. However, this systemic approach did not reveal which bone cell type mainly accounts for thyroid hormone actions in bone. In this study, we used male and female conditional knockout mice targeting Bmpr1a expression in either osteoclast precursors (LysM-Cre) or osteoprogenitors (Osx-Cre) and rendered them hyperthyroid to assess whether BMP signaling in osteoblasts or osteoclasts primarily drives the pathogenesis of hyperthyroidism-induced bone loss.


Bmpr1a deletion in osteoclast precursors does not prevent hyperthyroidism-induced osteoporosis

Given the predominant bone resorption in hyperthyroid mice, we first investigated whether Bmpr1a in osteoclast precursors might contribute to thyroid hormone excess driven bone loss.

Conditional Bmpr1a knockout in LysM-expressing cells led to a significant downregulation of Bmpr1a expression by 69.1% in untreated mature osteoclasts in vitro (Fig. S1a). Nevertheless, we did not observe any alterations of the body weight, femur length and bone phenotype in male mice due to the loss of Bmpr1a in osteoclast precursors (Fig. 1, Table 1). After 4 weeks of L-thyroxine treatment, serum concentrations of total T4 (Cre-negative: 4.1-fold; Cre-positive: 3.7-fold vs. respective untreated controls) and total T3 (Cre-negative: 2.2-fold; Cre-positive: 1.9-fold vs. respective untreated controls) increased in both Cre-negative and Cre-positive male Bmpr1afl/fl;LysM-Cre mice (Fig. S2a, b). Neither body weight nor femur length were affected by hyperthyroidism (Table 1), however, we observed trabecular bone loss at the spine and femur in hyperthyroid male mice regardless of their Bmpr1a expression in osteoclast precursors (Fig. 1a−f). In Cre-negative hyperthyroid mice, trabecular bone volume of the fourth lumbar vertebra was reduced by 43.2% with a decreased trabecular number (−26.3%) and a 1.4-fold increased trabecular separation while Cre-positive mice displayed lowered bone volume by 34.9% and decreased trabecular thickness (−11.2%) with hyperthyroidism as compared to non-treated respective littermate controls (Fig. 1a−e). In addition, all hyperthyroid male mice exhibited trabecular (BV/TV Femur: Cre-negative: −48.1%, Cre-positive: −43.8% vs. respective untreated controls) and cortical bone loss at the femur (Ct.BV/TV: Cre-negative: −1.5%, Cre-positive: −1.2% vs. respective untreated controls) as compared with respective euthyroid groups (Fig. 1f, g, Table 1). Bone strength of L5 vertebrae was impaired in all hyperthyroid males (Fmax: Cre-negative: −44.8%, Cre-positive: -50.9% vs. respective untreated controls), while cortical bone strength tested at the femur was only significantly reduced in Cre-positive male mice (Fmax: Cre-negative: −13.4%, n.s., Cre-positive: −24.1% vs. respective untreated controls) (Fig. 1h, i).

Fig. 1: Osteoclast progenitor-specific Bmpr1a deletion does not prevent hyperthyroidism-induced bone loss.
figure 1

Twelve-week old male Cre-negative (Cre-) and Cre-positive (Cre +) Bmpr1afl/fl;LysM-Cre mice remained euthyroid (CO) or were rendered hyperthyroid (T4) by adding 1.2 µg/mL L-thyroxine into their drinking water over 4 weeks. Using microCT analysis, (a) bone volume per total volume (BV/TV), (b) trabecular number (Tb.N), (c) trabecular separation (Tb.Sp), and (d) trabecular thickness (Tb.Th) were determined at the fourth lumbar vertebra (L4). e Representative 3D reconstructions of the trabecular compartment of L4. Additionally, (f) trabecular BV/TV and (g) cortical bone volume over total volume (Ct.BV/TV) were assessed at the femur. The maximal force (Fmax) as an indicator of bone strength (h) at the L5 vertebra and (i) femur was tested by compression test and 3-point bending test, respectively. Serum concentrations of (j) bone formation marker P1NP and (k) bone resorption marker TRAP were analyzed by ELISA. Furthermore, (l) the bone formation rate per bone surface (BFR) and (m) osteoclast surface per bone surface (Oc.S/BS) were assessed at the L3 and L4 vertebra, respectively. In vitro, (n) expression of acid phosphatase 5, tartrate resistant (Acp5), encoding for TRAP, in primary osteoclasts derived from Bmpr1afl/fl;LysM-Cre mice at day 7 of differentiation, with or w/o treatment with 100 nM T3 (T3) over 48 h, was quantified using quantitative real-time PCR. Each dot indicates an individual mouse. In vivo: Cre-/CO: N = 9; Cre-/T4: N = 8; Cre +/CO: N = 9; Cre + /T4: N = 9; In vitro: Cre-/CO: N = 5; Cre-/T3: N = 6; Cre +/CO: N = 6; Cre +/T3: N = 6. The horizontal lines represent the mean +/- SD. Statistical analysis was performed by Two-way ANOVA and selected p values are shown within the graph.

Table 1 Body weight, femur length as well as selected bone and bone turnover parameters of male Bmpr1afl/fl;LysM-Cre mice

In female cohorts, L-thyroxine treatment led to a significant increase of serum total T4 in both Cre-negative (6-fold) and Cre-positive (3.8-fold) mice, while total T3 concentrations were significantly elevated in Cre-negative (4.1-fold) but not Cre-positive mice (2.9-fold, p = 0.14) as compared with respective untreated controls (Fig. S2c, d). Further, female Cre-positive mice as well as hyperthyroid Cre-negative mice presented higher body weight as compared with euthyroid Cre-negative mice, however, femur length was not different between the groups (Fig. S1b, c). In line with findings in male cohorts, both Cre-negative and Cre-positive female mice showed trabecular and cortical bone loss with hyperthyroidism (Fig. S1d−k).

Bone turnover parameters P1NP and TRAP increased with hyperthyroidism in both Cre-negative and Cre-positive male and female mice, although changes in TRAP levels of euthyroid versus hyperthyroid Cre-positive mice did not reach statistical significance in the male (+31.7%, p = 0.056), but only female cohort (+82.3%) (Fig. 1j, k; Fig. S1l, m). Histomorphometric and histological osteoblast parameters (MS/BS, MAR, BFR/BS, Ob.S/BS, Ob.N/B.Pm) did not show significant changes in hyperthyroid males despite elevated P1NP serum levels (Fig. 1l, Table 1, Figs. S3, S4). Nevertheless, osteoclast surface and number increased by 1.6-fold and 1.4-fold, respectively, in Cre-positive hyperthyroid versus euthyroid males (Fig. 1m, Table 1, Fig. S3). Together with the raised circulating TRAP levels, these findings demonstrate a high bone resorption phenotype with hyperthyroidism for both sexes irrespective of the genotype. In contrast to the elevated TRAP serum levels with hyperthyroidism, in vitro treatment of premature osteoclasts with T3 until their maturation did not affect the expression of acid phosphatase 5, tartrate resistant (Acp5) (Fig. 1n), the gene encoding for TRAP, indicating that actions of thyroid hormones on osteoclasts might be indirectly mediated via other bone cells.

Osteoprogenitor-specific Bmpr1a knockout protects hyperthyroid mice from bone loss

Osteoblasts have been described as the primary target cells of thyroid hormones in bone2 and multiple osteoblast-derived cytokines and endocrine factors are known to control osteoclast development and activity, especially in the context of BMP signaling29,30,38,39. Therefore, we analyzed whether hyperthyroid mice with an osteoprogenitor-specific Bmpr1a deletion might be protected against bone loss. As described before32, conditional Bmpr1a knockout in osteoprogenitors per se led to an increase of trabecular bone volume (L4: males: 2.9-fold, females: 6.5-fold), trabecular number (L4: males: 1.7-fold, females: 2.2-fold), trabecular thickness (L4: males: 2.6-fold, females: 3.4-fold), a reduction of trabecular separation (L4: males: −47.2%, females: −58.7%) and at the spine (Fig. 2a−d, f) and femur in male (Fig. 2g, Table 2) and female mice (Fig. S5) as compared to respective Cre-negative littermates. Furthermore, Cre-positive Bmpr1afl/fl;Osx-Cre males presented improved bone strength (L5: 2.1-fold; femur: 1.4-fold versus untreated respective controls, Fig. 2e, k). Femur length, body weight and cortical bone volume were not affected by conditional Bmpr1a knockout, however, we observed thickened cortices (males: +28.6%, females: +29.7% vs. untreated Cre-negative mice) and decreased cortical bone mineral density (males: −2.9%, females: −4.0% vs. untreated Cre-negative mice), as previously described40, in male and female mice lacking Bmpr1a in osteoprogenitors (Table 2, Fig. 2h−j, Fig. S3n−p).

Fig. 2: Loss of Bmpr1a in osteoblasts protects hyperthyroid mice from osteoporosis.
figure 2

Twelve-week old male Cre-negative (Cre-) and Cre-positive (Cre +) Bmpr1afl/fl;Osx-Cre mice remained euthyroid (CO) or were rendered hyperthyroid (T4) by adding 1.2 µg/mL L-thyroxine into their drinking water over 4 weeks. Based on microCT analysis, (a) bone volume over total volume (BV/TV), (b) trabecular number (Tb.N), (c) trabecular thickness (Tb.Th), and (d) trabecular separation (Tb.Sp) were measured at the fourth lumbar vertebra (L4). e The maximal force (Fmax) was assessed by compression test at the L5 vertebra. f Representative 3D reconstructions of the trabecular compartment of L4. In addition, (g) trabecular BV/TV and (h) cortical bone volume over total volume (Ct.BV/TV), (i) Ct. thickness (Ct.Th) and (j) cortical bone mineral density (Ct.BMD) were determined at the femur. k The maximal force (Fmax) at the femur was tested by 3-point bending test. l Serum concentrations of bone formation marker P1NP was quantified using ELISA. Further, (m) osteoblast surface per bone surface (Ob.S/BS) and (n) bone formation rate per bone surface (BFR) were assessed at the spine. With regards to bone resorption, serum concentrations of (o) bone resorption marker TRAP and (p) osteoclast surface per bone surface (Oc.S/BS) were determined. Each dot indicates an individual mouse. MicroCT/Histology: Cre-/CO: N = 9; Cre-/T4: N = 9; Cre +/CO: N = 8; Cre +/T4: N = 9; ELISAs: Cre-/CO: N = 7; Cre-/T4: N = 8; Cre +/CO: N = 7; Cre +/T4: N = 8. The horizontal lines represent the mean +/- SD. Statistical analysis was performed by Two-way ANOVA and selected p values are shown within the graph.

Table 2 Body weight, femur length as well as selected bone and bone turnover parameters of male Bmpr1afl/fl;Osx-Cre mice

Serum concentrations of total T4 (Male Cre-negative: 4.8-fold; male Cre-positive: 4.7-fold; female Cre-negative: 5.1-fold; female Cre-positive: 3.1-fold vs respective untreated controls) and total T3 (Male Cre-negative: 2.4-fold; male Cre-positive: 2.1-fold; female Cre-negative: 3.9-fold; female Cre-positive: 1.9-fold vs respective untreated controls) increased with L-thyroxine treatment in all groups (Fig. S2e−h). With hyperthyroidism, Cre-negative but not Cre-positive male mice developed an osteoporotic phenotype as shown by lowered trabecular bone volume (−56.8%), abated trabecular number (−13.5%) and increased trabecular separation (+11.8%) at the L4 as compared with untreated respective controls (Fig. 2a−e). Bone strength showed a tendency to decrease by −47.3% (P = 0.072) only in spines of Cre-negative hyperthyroid males as compared to untreated controls (Fig. 2e). In femurs, we found decreased trabecular bone volume (−59.4% vs untreated, Cre-negative mice), low bone mineral density and enhanced trabecular separation with hyperthyroidism in Cre-negative mice (Fig. 2g, Table 2) and an even increased bone mineral density in hyperthyroid conditional knockout mice as compared with respective, untreated controls (Table 2). Cortical bone volume, thickness, bone mineral density and bone strength were not altered by hyperthyroidism in either wildtype mice or mice lacking Bmpr1a in osteoprogenitors (Fig. 2h−k).

With regards to bone formation, circulating levels of P1NP, osteoblast surface, and osteoblast numbers as well as MAR and BFR showed a tendency towards increase in Cre-negative male mice with L-thyroxine treatment as compared with respective untreated mice (Fig. 2l−n, Table 2, Fig. S6). In Cre-positive males osteoblast surface increased with hyperthyroidism by 1.3-fold (Fig. 2m; Fig. S7), while again other osteoblast parameters demonstrated a tendency towards increase (Fig. 2l, n, Table 2). Of note, calcein labels in bones from Cre-positive Bmpr1afl/fl;Osx-Cre males showed an overall low fluorescence signal, while T4 treatment enhanced the fluorescence intensity resulting in wide, but often not clearly separable fronts of bone formation as shown by the representative pictures in the Supplementary Figs. (Fig. S6).

Nevertheless, only hyperthyroid Cre-negative mice presented increased serum levels of bone resorption marker TRAP (+1.5-fold vs. untreated Cre-negative mice), enlarged osteoclast surfaces (1.5-fold vs. untreated Cre-negative mice) and an elevated osteoblast number, while conditional knockout mice presented low TRAP serum concentrations in general (Fig. 2o, p, Table 2, Fig. S7).

In female mice, we observed weight gain with hyperthyroidism in Cre-negative (+13.9% vs. untreated, Cre-negative mice) and Cre-positive cohorts (+10.3% vs. untreated, Cre-positive mice), while femur length was not different between the four groups (Fig. S5a, b). At the spine, no alterations of bone parameters due to thyroid hormone excess were detected in both wildtypes and conditional knockouts (Fig. S5c−h). Nevertheless, decreased trabecular number (−36.8% vs. untreated, Cre-negative mice, Fig S5k) and augmented trabecular separation (+44.3% vs. untreated, Cre-negative mice, Fig. S5m) at th femur and upregulated serum levels of both P1NP (+60.5% vs. untreated, Cre-negative mice, Fig. S5q) and TRAP (+65.9% vs. untreated, Cre-negative mice, Fig. S5r) indicate an early stage of hyperthyroidism-driven bone loss in female hyperthyroid Cre-negative mice, but not mice lacking Bmpr1a in osteoprogenitors.

Thus, we identified the osteoblastic BMP receptor Bmpr1a as a critical contributor to hyperthyroidism-driven high bone resorption in male and female mice.

BMP receptor BMPR1A mediates thyroid hormone actions in osteoblasts and regulates osteoblast-osteoclast interactions

To investigate how the thyroid hormone response in osteoblasts is altered by Bmpr1a knockout, we cultured osteoblasts derived from Cre-negative and Cre-positive Bmpr1afl/fl;Osx-Cre mice and treated them with pharmacological doses of T3 that is known to have a 10-fold higher binding affinity to thyroid hormone receptors than T42. Osterix-promotor driven Bmpr1a deletion led to a 59.9% decrease of Bmpr1a expression in mature osteoblasts as compared to wildtype osteoblasts (Fig. 3a). T3 treatment increased the expression of prominent osteoblast marker genes Osx, alkaline phosphatase (Alpl) and osteocalcin (Bglap) (1.7-fold, 1.7-fold, 4.3-fold vs. untreated, respective control) as well as improved mineralization capacity and ALP activity (both 1.6-fold vs. untreated, respective control) in wildtype osteoblasts (Fig. 3b−g). Bmpr1a-lacking osteoblasts presented downregulated Osx expression (−75.2% vs. untreated, respective control), lowered mineralization capacity (−45.6% vs. untreated, respective control) and decreased ALP activity (−86.3% vs. untreated, respective control) and were not affected by T3 treatment (Fig. 3b−g).

Fig. 3: BMPR1A mediates the response to T3 in osteoblasts in vitro.
figure 3

Using quantitative real-time PCR analysis, mRNA expression of (a) Bmpr1a, (b) osterix (Sp7), (c) alkaline phosphatase (Alpl) and (d) osteocalcin (Bglap) was measured in primary murine osteoblasts derived from Bmpr1afl/fl;Osx-Cre mice after 48 h treatment with 100 nM T3 (T3). Further, (e) mineralization capacity was evaluated after differentiation and T3-treatment over 10 days. f Representative image of Alizarin Red stained bone matrix. g ALP activity was assessed after 72 h treatment with T3. In addition, transcript levels of BMP target genes (h) runt related transcription factor 2 (Runx2) and (i) inhibitor of DNA binding 1 (Id1), thyroid hormone target gene (j) Krueppel-like factor 9 (Klf9), glucose transporters (k) Slc2a4 (l) Slc2a1 and (m) Slc2a3, (n) receptor activator of NF-κB ligand (Rankl), and (o) osteoprotegerin (Opg) were quantified and the (p) Rankl/Opg ratio was calculated. Each dot indicates an individual mouse. Real-time PCR: Cre-/CO: N = 8-11; Cre-/T3: N = 8-14; Cre +/CO: N = 11-13; Cre +/T3: N = 8-13; Mineralization: Cre-/CO: N = 7; Cre-/T3: N = 6; Cre +/CO: N = 8;Cre +/T3: N = 8; ALP assay: Cre-/CO: N = 8; Cre-/T3: N = 5; Cre +/CO: N = 9;Cre +/T3: N = 5. The horizontal lines represent the mean +/- SD. Statistical analysis was performed by Two-way ANOVA and selected p values are shown within the graph.

While expression of BMP target genes Runx2 and Id1 and thyroid hormone target gene Klf9 were not altered with T3 in Bmpr1a knockout osteoblasts, their T3-dependent regulation in wildtype osteoblasts indicates that the transcriptional activation of both pathways relies on Bmpr1a expression (Fig. 3h−j). As a link to increased energy demands with hyperthyroidism, we checked the expression of Slc2a1, Slc2a3 and Slc2a4 encoding for class I glucose transporters GLUT1, GLUT3 and GLUT4, respectively, that facilitate glucose transport across the cell membrane. Interestingly, we found 2.2-fold elevated levels of Slc2a4, the gene coding for the primary glucose transporter GLUT4 in osteoblasts41, in T3-treated wildtype but not Bmpr1a knockout osteoblasts as compared to respective untreated controls (Fig. 3k). Expression of Slc2a1 and Slc2a3 tended to increase in both T3-treated groups, but did not reach statistical significance (Slc2a1: Cre-/T3: 1.3-fold, P = 0.27; Cre +/T3: 1.4-fold, P = 0.14; Slc2a3: Cre-/T3: 1.5-fold, P = 0.18; Cre +/T3: 1.5-fold, P = 0.29 vs. respective untreated control) (Fig. 3l, m).

Further, we showed that Rankl was upregulated in T3-treated wildtype osteoblasts (2.4-fold vs. untreated respective control) and expression of Opg was downregulated in both untreated and T3-treated knockout osteoblasts (CO: −30.7%, T3: −41.3%) as compared with non-treated wildtype cells, resulting in 2.3-fold increased Rankl/Opg ratio only in thyrotoxic wildtype but not Bmpr1a-deficient osteoblasts (Fig. 3n−p). In bone tissue of Bmpr1afl/fl;Osx-Cre mice, Bmpr1a expression tended to decrease in mice with active Cre-recombinase (−59.4%, P = 0.07) (Fig. 4a) and Klf9 expression was upregulated by 78.4% with hyperthyroidism (Fig. 4b) as compared with Cre-negative, euthyroid mice. While transcript levels of Rankl and Opg were not different between the four groups, Rankl/Opg ratio again was 2.6-fold increased with hyperthyroidism in bone tissue obtained from wildtype mice only, supporting our in vitro findings (Fig. 4c−e).

Fig. 4: BMPR1A controls the Rankl/Opg ratio in response to T4 in vivo.
figure 4

Twelve-week old male Cre-negative (Cre-) and Cre-positive (Cre +) Bmpr1afl/fl;Osx-Cre mice remained euthyroid (CO) or were rendered hyperthyroid (T4) by adding 1.2 µg/mL L-thyroxine into their drinking water over 4 weeks. Total RNA was isolated from bone tissue and the expression of (a) Bmpr1a, (b) Klf9, (c) Opg and (d) Rankl was evaluated using quantitative real-time PCR. Further, (e) the ratio of Rankl and Opg was calculated. Each dot indicates an individual mouse. Cre-/CO: N = 8; Cre-/T4: N = 7; Cre +/CO: N = 8; Cre +/T4: N = 8. The horizontal lines represent the mean +/- SD. Statistical analysis was performed by Two-way ANOVA and selected p values are shown within the graph.

Given that RANKL and OPG are soluble factors secreted by osteoblasts, we used osteoblast supernatants for the treatment of wildtype osteoclasts to investigate whether thyroid hormones can indirectly affect osteoclastogenesis and whether this effect might be BMPR1A-dependent. A direct treatment of early osteoclasts with T3 did not alter the number of TRAP-positive osteoclasts or the expression of common osteoclast marker genes (Fig. 5a−d). Interestingly, thyroid hormone target genes Klf9 and deiodinase 3 (Dio3) were upregulated 1.5-fold and 1.6-fold, respectively, suggesting that osteoclasts do respond to thyroid hormone treatment, however, this did not translate into phenotypic changes (Fig. 5e, f). Despite osteoclast numbers not being affected by supernatant treatment, expression of osteoclast marker genes Acp5 (1.7-fold), cathepsin K (Ctsk, 2.1-fold), and dendrocyte expressed seven transmembrane protein (Dcstamp, 1.4-fold) as well as Klf9 (1.4-fold) and Dio3 (1.4-fold) increased when osteoclasts were challenged with supernatants derived from T3-treated Cre-negative osteoblasts as compared to untreated Cre-negative cells (Fig. 5g−l). In contrast comparing the treatment with supernatants derived from untreated versus T3-treated Cre-positive osteoblasts, only the expression of Dio3 was significantly upregulated by 1.8-fold indicating that BMPR1A is involved in osteoblast-osteoclast interactions under thyrotoxic conditions (Fig. 5l).

Fig. 5: Direct versus indirect actions of T3 on primary wildtype osteoclasts.
figure 5

Early wildtype osteoclasts (d4) were treated directly with 100 nM T3 over 48 h (af) and (a) number of TRAP-positive multinucleated osteoclasts (≥3 nuclei) was counted after TRAP staining. In addition, expression of osteoclast marker genes (b) acid phosphatase 5, tartrate resistant (Acp5), (c) cathepsin K (Ctsk), and (d) dendrocyte expressed seven transmembrane protein (Dcstamp) as well as thyroid hormone target genes (e) krueppel-like factor 9 (Klf9) and (f) deiodinase 3 (Dio3) was quantified. Furthermore, early wildtype osteoclasts (d4) were treated over 48 h with supernatants from untreated (CO) or T3-treated (T3) Cre-negative (Cre-) and Cre-positive (Cre +) Bmpr1afl/fl;Osx-Cre osteoblasts (gl). Subsequently, (g) number of TRAP-positive multinucleated osteoclasts was evaluated as well as the expression of (h) Acp5, (i) Ctsk, (j) Dcstamp, (k) Klf9, and (l) Dio3. Each dot indicates an individual mouse. TRAP staining (a, g): n = 8 per group. RNA expression analysis: direct treatment (bf): n = 5 per group. Indirect coculture (hl): Cre-/CO: N = 8; Cre-/T3: N = 8; Cre +/CO: N = 8; Cre +/T3: N = 7. The horizontal lines represent the mean +/- SD. Statistical analysis was performed by a two-sided student´s t test comparing two groups or a two-way ANOVA for the comparison of four groups. Significant p values are shown within the graph.

In conclusion, we show that the BMP receptor BMPR1A is an important mediator of thyroid hormone actions in osteoblasts and regulates the Rankl/Opg ratio in osteoblasts and bone tissue in response to thyroid hormones, thus potentially promoting osteoclastogenesis and bone resorption.


Hyperthyroidism is an established cause of osteoporosis, still, the underlying molecular mechanisms are not fully understood yet. In our previous study, we showed that thyroid hormones activate BMP signaling in osteoblasts and that, in vivo, blocking BMP signaling using a BMPR1A-specific ligand scavenger prevented bone loss in hyperthyroid mice26. Nevertheless, this systemic approach did not reveal which bone cell type might be the main driver of hyperthyroidism-induced osteoporosis.

Given the high bone resorption observed with thyroid hormone excess, we first analyzed the bone phenotype of mice lacking Bmpr1a in osteoclast progenitors. In contrast to a histomorphometry-based study showing mild trabecular bone gain in femurs of 8 to 10-week-old Bmpr1afl/fl;LysM-Cre-positive versus control mice42, we did not observe any significant skeletal alterations using microCT analysis or changes in serum bone turnover markers at an age of 16 weeks in male and female conditional knockout mice. Cathepsin K (Ctsk)-promotor-driven Bmpr1a knockout targeting mature osteoclasts, however, leads to marked trabecular bone gain due to enhanced bone formation at a young age implying that BMPR1A mediates osteoclast-to-osteoblast coupling through mature osteoclasts mainly36. In vitro analysis of osteoclasts lacking Bmpr1a based on LysM-Cre versus Ctsk-Cre driven deletion showed again contradictory results with reduced number of TRAP-positive osteoclasts and low Acp5 expression42 or no effect on Acp5 expression in Bmpr1afl/flLysM-Cre-positive cells, while osteoclastogenesis increased with Ctsk-Cre-based Bmpr1a knockout. Thus, the timing of Bmpr1a deletion during osteoclast development seems to be critical and primarily differentiated osteoclasts are affected by disrupted BMP signaling.

With hyperthyroidism, Bmpr1afl/fl;LysM-Cre mice presented osteoporotic bones with impaired bone strength independent of osteoclast progenitor-specific Bmpr1a deletion, indicating that either BMPR1A plays only a minor role mediating effects of thyroid hormones on osteoclastogenesis and osteoclast-osteoblast coupling or that osteoclasts per se are not directly responsive to thyroid hormone treatment. In line, we could not observe any changes in Acp5 expression in either knockout or wildtype osteoclasts when treated with T3. Several studies show that thyroid hormone actions on osteoclasts and bone resorption are indirectly mediated via cells of the osteoblast lineage2,21,22,23,24,25,26 corroborating our findings of direct versus indirect thyroid hormone treatment of wildtype osteoclasts. In contrast to our previous studies26,43,44 and the results from the Bmpr1afl/fl;Osx-Cre cohorts, only circulating P1NP levels but not histomorphometric osteoblast parameters increased with L-thyroxine treatment in Bmpr1afl/fl;LysM-Cre mice. Still, osteoclast numbers showed a tendency to increase or were significantly enhanced in hyperthyroid Cre-negative and Cre-positive Bmpr1afl/fl;LysM-Cre mice, respectively, suggesting together with elevated TRAP serum levels high bone resorptive activities under thyroid hormone excess regardless of Bmpr1a expression in osteoclast progenitors.

Given that loss of Bmpr1a expression in osteoclasts did not prevent hyperthyroidism-driven bone loss, we next focused on osteoblasts. While osteoblast-specific Bmpr1a deletion led to a osteosclerotic bone phenotype as described before32, thyroid hormone treatment did not promote bone resorption or bone loss in male and female Bmpr1afl/fl;Osx-Cre-positive mice revealing an important role of osteoblastic BMPR1A in the pathogenesis of hyperthyroidism-induced bone loss and fragility.

Consistently, besides more general approaches such as blocking BMP type I receptors or scavenging BMP ligands using the inhibitor LDN193189 and anti-BMP ligand antibodies, respectively26, also targeted genetic Bmpr1a knockout hampered the response of osteoblasts to T3 treatment in vitro. In particular, mineralization capacity, ALP activity and expression of the early osteoblast marker osterix were impaired with Bmpr1a loss, in line with findings in osteoblasts derived from Bmpr1afl/fl;Og2-Cre45, Bmpr1afl/fl;Prx1-CreERT46 and Bmpr1afl/fl;Col1-ERTTM-Cre47 mice, and were not enhanced with T3 in contrast to osteoblasts expressing Bmpr1a. Further, we show that not only the expression of BMP target genes (Runx2, Id1), but also the thyroid hormone target gene Klf9 was not induced in T3-treated Bmpr1a-deficient osteoblasts and bone tissue from hyperthyroid Bmpr1afl/fl;Osx-Cre-positive mice, respectively, allowing for speculations on close mutual interactions between components of thyroid hormone signaling and the BMP signaling pathway.

Dynamic histomorphometric analysis revealed overall low fluorescence signals from calcein labels in untreated Cre-positive Bmpr1afl/fl;Osx-Cre animals suggesting a slow bone formation in these mice (Fig. S6) as reported before in mouse models with Col1- or Og2-driven Bmpr1a deletion29,45. T4 treatment increased the overall fluorescence intensity of the labels, however, a detection of clearly separated double labels was difficult (Fig. S6). In addition, enlarged osteoblast surface and a tendency to increase of other osteoblast parameters (Ob.N, P1NP) indicate that, in contrast to our in vitro findings, Bmpr1a-lacking osteoblasts are still stimulated by thyroid hormone treatment in vivo. However, within the bone marrow and next to osteoblasts, several other cell types such as hematopoietic cells and marrow adipocytes reside and were shown to also be affected by thyroid hormones48,49,50,51,52. Thus, we cannot exclude indirect stimulation of the Bmrp1a-depleted osteoblasts by other surrounding cells and further investigations are needed to estimate the extent of these cell-cell interaction-based effects in our hyperthyroidism mouse model.

High osteoblast activity due to thyroid hormone excess requires high amounts of energy, likely in the form of glucose53,54. Osteoblasts utilize glucose primarily via aerobic glycolysis, a tightly regulated process that can generate adenosine triphosphate quickly53,54. Glucose uptake is mainly performed by class I glucose transporters, GLUT1, -3 and -441. GLUT2 expression was not detected in osteoblasts41. Based on in vitro Slc2a knockout studies, especially Slc2a4 expression was shown to be crucial for osteoblast proliferation and differentiation, albeit also other GLUTs might contribute to a balanced glucose uptake and thus, be necessary to cover the energy needed for new bone formation41. A previous study confirmed that T3 stimulates glucose uptake via glucose transporters GLUT1 and GLUT3 into osteoblasts in vitro, however, GLUT4 was not investigated in that study55. Vice versa, we have previously shown that hypothyroidism leads to reduced glucose uptake and correspondingly low Slc2a4 expression in bone tissue in mice56. Linking glucose metabolism to BMP signaling, BMPR1A was reported to regulate the expression of glucose transporter Slc2a1 in chondrocytes in vivo57 and its BMP ligand BMP2 was shown to upregulate Slc2a4 expression in adipocytes58. In this study, we observed increased expression of Slc2a4 in osteoblasts with hyperthyroidism in vitro indicating possibly elevated glucose uptake due to higher energy demands in thyroid hormone-stimulated osteoblasts, an effect that was blocked in Bmpr1a-deficient cells. Thus, Bmpr1a loss could hamper the glucose uptake needed to meet high energy demands and might disrupt osteogenic thyroid hormone response not only at a transcriptional but also metabolic level. Further in-depth metabolic analyses are needed to unravel the effects of thyroid hormones on metabolic processes and energy levels in osteoblasts in detail, also considering the different states of osteogenic differentiation with distinct needs for energy and the role of the BMP signaling pathway therein.

As shown before30,39,46,59, BMPR1A is not only critical for osteogenic differentiation, but also regulates the expression of Rankl and Opg in osteoblasts to promote osteoclastogenesis and thus, couples bone formation and bone resorption via RANKL forward signaling. In response to thyroid hormones, wildtype, but not Bmpr1a-deficient osteoblasts upregulated Rankl expression in vitro and the Rankl/Opg ratio increased in both Cre-negative osteoblasts and bone tissue. Loss of osteoblastic Bmpr1a inhibited the thyroid hormone-driven increase of the Rankl/Opg ratio and thus, potentially prevented high bone resorption and bone loss in hyperthyroid mice. Corroborating our hypothesis, osteoblast-mediated but not direct treatment of wildtype osteoclasts with T3 promoted the expression of osteoclast marker genes in a BMPR1A-dependent manner. In contrast to the in vivo setting, osteoclast numbers were not affected suggesting that additional factors such as cell-cell-interactions might play an important role. As such, RANKL reverse signaling via vesicular RANK secreted by maturing or apoptotic osteoclasts binding to membranous RANKL on osteoblast surfaces was shown to upregulate Runx2 expression and promote osteogenic differentiation and bone formation60,61. Thus, BMPR1A-induced Rankl expression could be an important coupling factor between bone formation and bone resorption in both ways and loss of Bmpr1a might disrupt their fine-tuned balance in bone remodeling. Furthermore, thyroid hormone-induced Rankl expression in osteoblasts might start an amplification loop between osteoblasts and osteoclasts resulting in high bone turnover-induced bone loss that can be prevented by Bmpr1a deletion in osteoblasts.

Besides its strength, our study may have potential limitations. Aside from the great value of the Cre/LoxP system to investigate genes of interest in a cell/tissue-specific manner, possible off-targeting need to be considered. As such, LysM-Cre can target not only monocytes and macrophages, but also neutrophils62 and Osx-Cre can target also additional cell populations such as osteocytes, hypertrophic chondrocytes, stromal cells, adipocytes, and perivascular cells of bone marrow63. Beside osteoblasts, also osteocytes are crucial for orchestrating bone remodeling by secreting RANKL64, can import thyroid hormones16 and have been shown to exhibit resorptive TRAP activity themselves in response to hyperthyroidism in vivo65. Osteocytes are not the main target cells for the Osx-Cre-driven recombination66, but are the dominant cells type (90−95%) within bone tissue67. In line, we observed only a tendency to decreased Bmpr1a expression in bone tissue derived from Bmpr1afl/fl;Osx-Cre positive mice (Fig. 4a) and thus, results based on total RNA from bone tissue should be interpreted carefully. Whether osteocytes directly contribute to thyroid hormone excess-driven osteoclastic bone resorption in a BMPR1A-dependent manner needs to be investigated in further studies using osteocyte-targeting Cre-lines. Furthermore, other signaling pathway besides BMP signaling might be activated by thyroid hormones in osteoblasts and thus, unbiased approaches such as RNA sequencing or proteomic profiling are necessary to identify further molecular changes that can drive thyroid hormone responses in bone cells.

In conclusion, this study is the first to reveal the importance of the osteoblastic BMP receptor BMPR1A in the pathogenesis of hyperthyroidism-induced osteoporosis by coupling thyroid hormone-stimulated osteoblast activity to high bone resorption potentially via the RANKL/OPG system.


Animal models

Floxed Bmpr1a mice on C57BL/6 background that carry loxP sites flanking exon 4 of the gene Bmpr1a (Bmpr1afl/fl) were a kind gift from Prof. Yuji Mishina (University of Michigan, MI, USA)68. For generating osteoblast-specific Bmpr1a-deficient animals, Bmpr1afl/fl mice were crossed with doxycycline-repressible Osterix-Cre (Osx-Cre) mice on a C57BL/6 background69. Due to skeletal alterations seen in young Osx-Cre mice even without floxed alleles70, it is advised to repress Cre-recombinase activity prenatally up to the first 4−5 weeks of postnatal development71,72. Therefore, breeding pairs of the generated Bmpr1afl/fl;Osx-Cre mice as well as their offspring received a 10 mg/mL doxycycline dissolved in 3% sucrose solution in their drinking water ad libitum up to 5 weeks of age. In addition, Bmpr1afl/fl;LysM-Cre were generated to inactivate Bmpr1a specifically in osteoclast progenitors by crossing Bmpr1afl/fl mice with LysM-Cre mice on a C57BL/6 background73. Efficient Bmpr1a deletion and recombination in Cre-targeted cells and organs of Cre-positive animals was confirmed by PCR analysis68. Respective age-matched Cre-negative littermates were used as controls. All mice were on a C57BL/6 background. Mice were maintained in groups up to four animals in a light-dark cycle of 12/12 h at room temperature in filter-top cages with cardboard houses for enrichment purposes and had ad libitum access to their respective drinking water and standard chow diet.

To induce hyperthyroidism, 12-week-old Cre-negative (Cre-) and Cre-positive (Cre +) male and female mice were treated over 4 weeks with 1.2 µg/ml L-thyroxine (T4, Sigma-Aldrich, Munich, Germany) in their drinking water. Control littermates received regular tap water (CO). Group sizes were as follows: Bmpr1afl/fl;Osx-Cre mice: Males: Cre-/CO: N = 8; Cre-/T4: N = 9; Cre +/CO: N = 9;Cre +/T4: N = 9; Females: Cre-/CO: N = 9; Cre-/T4: N = 7; Cre +/CO: N = 9;Cre +/T4: N = 8; Bmpr1afl/fl;LysM-Cre mice: Males: Cre-/CO: N = 9; Cre-/T4: N = 9; Cre +/CO: N = 8;Cre +/T4: N = 9; Females: Cre-/CO: N = 7; Cre-/T4: N = 6; Cre +/CO: N = 8;Cre +/T4: N = 8.

At an age of 16 weeks, all mice were euthanized using CO2. Blood was collected via heart puncture and serum was obtained by centrifugation. The fourth lumbar vertebrae (L4) and femurs were collected postmortem, fixed in 4% PBS-buffered paraformaldehyde for 48 h and stored in 50% ethanol. For RNA isolation of bone tissue, humeri were flushed with PBS, immediately frozen with liquid nitrogen and stored at −80 °C until further analysis. All subsequent analyses (ELISA, RIA, microCT, bone biomechanics, histology/dynamic histomorphometry, RNA analysis of bone tissue) were performed in a blinded manner.

Animal procedures were approved by the institutional animal care committee of the Technische Universität Dresden and the Landesdirektion Sachsen (TVV 09/2021) and were performed according to the ARRIVE guidelines. We have complied with all relevant ethical regulations for animal use.

Serum analysis

Serum concentrations of bone formation marker procollagen type 1 amino-terminal propeptide (P1NP) and bone resorption marker tartrate-resistant acid phosphatase (TRAP) were quantified using ELISAs according to the manufacturers´ protocols (IDS, Frankfurt/Main, Germany). Total T4 and total T3 serum concentrations were assessed using radioimmunoassays (IM1447 and IM1699 respectively; Beckman Coulter/DRG Instruments, Marburg, Germany) as previously described3,44. The lower limit of analysis was 10 nM and 0.22 nM for T4 and T3, respectively. Intra-assay coefficients of variations were below 7.5%.

Analysis of femur length, bone mass and microarchitecture

The length of the femurs was measured postmortem using a caliper. Also, femurs and L4 were analyzed using micro-computed tomography (microCT) (vivaCT 40, Scanco Medical, Brüttisellen, Switzerland) with an X-ray energy of 70 kVp and isotropic voxel size of 10.5 µm (114 mA, 200 msec integration time). Following standard protocols from Scanco medical, trabecular (Tb.) and cortical (Ct.) bone parameters including bone volume/total volume (BV/TV), trabecular number (Tb.N), trabecular separation (Tb.Sp) and thickness (Tb.Th) were evaluated based on calculations including 100 scan slices. Trabecular bone analysis of L4 was conducted at the center contouring 50 slices above and 50 slices below the middle of the vertebral body. Femur trabecular bone parameters were assessed in the metaphyseal region starting 20 slices below the growth plate, while cortical bone was tested within the diaphyseal region midway between femoral head and distal condyles.

Static and dynamic histomorphometry

After fixation in 4% PBS-buffered paraformaldehyde for 48 h, L4 vertebrae were decalcified in Osteosoft (Merck, Germany) for 7 days and dehydrated using ascending ethanol series. Subsequently, decalcified bones were embedded in paraffin and cut into 2 µm thick sections. Tartrate-resistant acid phosphatase (TRAP) staining was performed to quantify osteoclasts in an area of 0.90 mm² in the center of vertebrae using the Microscope Axio Imager M1 (Carl Zeiss Jena, Jena, Germany) and Osteomeasure software (OsteoMetrics, Atlanta, GA, USA). In addition, osteoblasts were identified by their morphology and localization along the bone surface and quantified in an area of 0.90 mm² in the center of vertebrae. Representative photos were taken using CellSens Entry Software Version 1.5 (OLYMPUS Cooperation, Shinjuku, Japan).

Dynamic bone histomorphometry was performed as described previously26,43. Five and two days before sacrifice, mice received intraperitoneal injections with the fluorochrome calcein (20 mg/kg BW; Sigma-Aldrich, Munich, Germany) that incorporates into newly formed bone. The third lumbar vertebrae were fixated in 4% PBS-buffered paraformaldehyde for 48 h and dehydrated via ascending ethanol series. Then, bones were embedded in methyl methacrylate (Technovit 9100, Heraeus Kulzer, Hanau, Germany) and cut into 7 µm sagittal sections for calcein label quantification of the trabecular bone. The mineralized surface/bone surface (MS/BS), the mineral apposition rate (MAR) and the bone formation rate/bone surface (BFR/BS) were quantified in an area of 1.44 mm² in the center of vertebrae according to established protocols16,17,26,65,74 using fluorescence microscopy (Microscope Axio Imager M1) and the Osteomeasure software. Representative photos were taken using Axio Vision 3.1 Software (Carl Zeiss Jena, Germany). Terminology and quantification procedures were conducted according to guidelines of the Nomenclature Committee of the ASBMR75.

Bone biomechanics

Femurs were used for 3-point bend testing to determine cortical bone strength while L5 vertebrae were used for a compression test (Zwick Roell, Ulm, Germany). Beforehand, femurs and L5 vertebrae were rehydrated in PBS overnight. Femurs were then placed onto two supports with an intermediate distance of 6 mm and a mechanical force was applied vertically onto the middle of the femoral midshaft. Vertebrae were placed onto the center of the lower plate and pressure was applied via the upper platen. After reaching a preload of 1 N, the measurement started and continued with a load rate of 0.05 mm/s until failure. The maximal load (Fmax) was evaluated as an indicator of bone strength using testXpert II—V3.7 software (Zwick Roell, Ulm, Germany).

Cell culture of primary murine osteoblasts and osteoclasts

Hind legs of Cre-negative and Cre-positive Bmpr1afl/fl;Osx-Cre mice were collected and mesenchymal stromal cells (MSC) were obtained by flushing the bone marrow. To acquire primary murine osteoblasts, MSC were first cultured in growth medium (DMEM, 10% FCS, 1% Penicillin/Streptavidin (P/S)) until 80% confluence in a humidified atmosphere of 95% air and 5% CO2 at 37 °C. Subsequently, MSC were differentiated by adding an osteogenic cocktail (100 μM ascorbate phosphate, 5 mM β-glycerol phosphate, both from Sigma-Aldrich, Munich, Germany) to the growth medium over 7 days. For RNA and alkaline phosphatase (ALP) activity analysis, osteoblasts were then starved overnight in DMEM with 1% FCS and 1%P/S and treated the next day with 100 nM T3 (Sigma-Aldrich, Munich, Germany) over 48 h and 72 h, respectively. ALP activity was assessed as a marker of osteoblastic differentiation. After the 72 h treatment, cells were washed twice in PBS and then scraped in ALP lysis buffer (10 mM Tris-HCl pH 8.0, 1 mM MgCl2, and 0.5% Triton X-100) for isolation of total proteins. Subsequently, cell lysates were processed through a 24-gauge needle and centrifuged at 25,000 × g for 30 min at 4 ˚C. 10 µL of diluted supernatants (1:5 in distilled water) were incubated with 90 µl ALP substrate buffer (100 mM diethanolamine, 150 mM NaCl, 2 mM MgCl2, and 2.5 g/ml p-nitrophenylphosphate) for 30 min at 37 ˚C. Color change occurred due to hydrolysis of p-nitrophenylphosphate by enzymatic activity of ALP. Absorbance was measured at 410 nm using FluoStar Omega (BMG Labtech, Offenburg, Germany). Results were normalized to the total protein content that was quantified using Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocol.

To determine the mineralization capacity, MSC were differentiated and simultaneously treated with 100 nM T3 over 10 days, then fixed in 10% paraformaldehyde in PBS for 15 min and stained with 1% Alizarin Red S solution (pH 5.5, Sigma-Aldrich, Munich, Germany) for 30 min at room temperature. Repeated washing steps with distilled water were performed to remove excess stain. 100 mM cetylpyridinium chloride solution (Sigma-Aldrich, Munich, Germany) was used to dissolve incorporated calcium that was then quantified by photometric measurement at a wavelength of 540 nm using FluoStar Omega (BMG Labtech, Offenburg, Germany).

For osteoclast culture, long bones from Bmpr1afl/fl;LysM-Cre mice were harvested and flushed with 1xPBS containing 5% FCS to isolate bone marrow. Following erythrocyte lysis with ACK lysis buffer (Thermo Fisher Scientific, Waltham, MA, USA), monocytes were isolated by magnetic cell separation (MACS) using DynabeadsTM Biotin Binder and Anti-Mouse CD11b Clone M1/70 (both from Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocols. Isolated monocytes were first cultured in a-MEM with stable glutamine, 10% FCS, 1% P/S and 25 ng/µl recombinant mouse macrophage colony-stimulating factor (M-CSF; R&D systems, Minneapolis, MN, USA). After 72 h, 50 ng/µl recombinant mouse RANKL (R&D systems, Minneapolis, MN, USA) was added to the medium to induce osteoclastogenesis. As soon as small osteoclast precursors were spotted (3−4 days later), cells were treated with 100 nM T3 over 48 h for RNA expression analysis.

Indirect coculture experiments

To conduct indirect coculture experiments, supernatants were collected from over 10 days differentiated Cre-negative and Cre-positive Bmpr1afl/fl;Osx-Cre osteoblasts that were cultured beforehand with or w/o 100 nM T3 over 48 h in a-MEM with 1% FCS and 1% P/S. To obtain osteoclasts as reported before76. bone marrow cells from long bones of 12-week-old male C57BL/6JRj were cultured in a-MEM supplemented with 10% FCS, 1% P/S, 2 mM L-glutamine and 10 ng/ml macrophage colony-stimulating factor (MCSF, from R&D Systems, Minneapolis, USA) over 24 h in a cell culture flask. Only the floating cells (mostly myeloid precursors) were collected and grown again in medium supplemented with 25 ng/ml M-CSF over 48 h. Osteoclastogenesis was induced by additionally adding 50 ng/ml receptor activator of nuclear factor kappa-B ligand (RANKL, from R&D Systems, Minneapolis, USA) over 4 days. Early osteoclasts were then treated with fresh a-MEM supplemented with 1% FCS, 1%P/S, 2 mM L-glutamine and osteoblast supernatants (1:1 ratio) supplemented with 25 ng/ml M-CSF and 50 ng/ml RANKL over 48 h and samples were used for subsequent RNA isolation and TRAP staining. As controls, we used d4 osteoclasts treated directly with 100 nM T3 (T3) over 48 h in a-MEM supplemented with 1% FCS, 1%P/S, 2 mM L-glutamine, 25 ng/ml M-CSF and 50 ng/ml RANKL.

TRAP-positive osteoclasts were quantified by TRAP staining at day 6 after starting RANKL treatment. At first, osteoclasts were fixated for 10 min in acetone/citrate buffer containing 37% paraformaldehyde at room temperature. Following two washing steps with tap water, TRAP staining was performed in the dark over 20 min using TRAP-Kit according to the manufacturer´s protocol (from Sigma-Aldrich, St. Louis, USA). TRAP-positive cells with three or more nuclei were counted as osteoclasts.

RNA isolation, RT-PCR and quantitative real-time PCR

Total RNA from primary cells and bone tissue was extracted using ReliaPrepTM RNA Tissue Miniprep System (Promega, Madison, WI, USA) and TRIzol reagent (Invitrogen, Darmstadt, Germany), respectively, following the manufacturer’s protocol and quantified using a Nanodrop spectrophotometer (Peqlab, Erlangen, Germany).

Five hundred nanograms of RNA were reverse transcribed using M-MLV Reverse Transcriptase (Promega, Madison, WI, USA) followed by GoTaq® qPCR Master Mix-based quantitative real-time PCR (Promega, Madison, WI, USA) according to established protocols (ABI7500 Fast, Applied Biosystems, Carlsbad, CA, USA). Primer sequences for mice are listed in supplemental Table S1. PCR conditions were: 50 °C for 5 min and 95 °C for 10 min followed by 40 cycles with 95 °C for 15 s and 60 °C for 1 min. Melting curves were evaluated using the following scheme: 95 °C for 15 s, 60 °C for 1 min and 95 °C for 30 s. Results were calculated based on the ∆∆CT method and are represented as x-fold increase normalized to β-actin and GAPDH mRNA levels.

Statistical analysis and reproducibility

Data are presented as mean ± standard deviation. Statistical analysis comparing four groups are based on a Two-way ANOVA followed by Bonferroni’s multiple comparison post-hoc test using GraphPad Prism 9.0 (GraphPad, La Jolla, CA, USA). Values of p < 0.05 were considered statistically significant. Significant outliers were excluded based on Grubbs’ test provided in GraphPad by Dotmatics (

Each in vitro experiment was performed at least twice with a number of >3 animals per group. All subsequent analyses (RNA analysis, ALP activity evaluation, Alizarin Red staining, TRAP staining) were conducted in a blinded fashion.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.