Distinct effects of loss of classical estrogen receptor signaling versus complete deletion of estrogen receptor alpha on bone

Abstract

Estrogen receptor (ER)α is a major regulator of bone metabolism which can modulate gene expression via a “classical” pathway involving direct DNA binding to estrogen-response elements (EREs) or via “non-classical” pathways involving protein-protein interactions. While the skeletal consequences of loss of ERE binding by ERα have been described, a significant unresolved question is how loss of ERE binding differs from complete loss of ERα. Thus, we compared the skeletal phenotype of wild-type (ERα^+/+) and ERα knock out (ERα^−/−) mice with that of mice in which the only ERα present had a knock-in mutation abolishing ERE binding (non-classical ERα knock-in [NERKI], ERα^−/NERKI). All three groups were in the same genetic background (C57BL/6). As compared to both ERα^+/+ and ERα^−/− mice, ERα^−/NERKI mice had significantly reduced cortical volumetric bone mineral density and thickness at the tibial diaphysis; this was accompanied by significant decreases in periosteal and endocortical mineral apposition rates. Colony forming unit (CFU)-fibroblast, CFU-alkaline phosphatase, and CFU-osteoblast numbers were all increased in ERα^−/− compared to ERα^+/+ mice, but reduced in ERα^−/NERKI mice compared to the two other groups. Thus, using mice in identical genetic backgrounds, our data indicate that the presence of an ERα that cannot bind DNA but can function through protein-protein interactions may have more deleterious skeletal effects than complete loss of ERα. These findings suggest that shifting the balance of classical versus non-classical ERα signaling triggers pathways that impair bone formation. Further studies defining these pathways may lead to novel approaches to selectively modulate ER signaling for beneficial skeletal effects.

INTRODUCTION

Estrogen plays a critical role in regulating bone metabolism in both sexes in humans and in mice [1]. The effects of estrogen in various tissues, including bone, are mediated by two related receptors, estrogen receptor (ER)α and β [2], and the importance of receptor-mediated estrogen signaling in bone has been well characterized using mice with deletions in ERα (ERα knock out [ERαKO]), ERβ (ERβKO), or both receptors (double ER KO, DERKO). These studies have found that ERα is the major ER mediating the effects of estrogen on bone in females [3] and in males [4]; however, ERβ may play a role in estrogen action in trabecular bone [3, 5] and can, under certain circumstances, substitute for [5] or enhance [3] ERα action.

Estrogen can regulate gene expression through either ERα or β using a number of signaling pathways (Figure 1). The “classical” pathway involves direct DNA binding of the liganded ER to estrogen response elements (EREs) [6], as in the case of the prolactin [7], progesterone receptor [8], and c-fos [9] genes. In addition, however, the ER can also regulate gene expression via a number of “non-classical” pathways which do not involve direct DNA binding by the ER but rather are due to specific protein-protein interactions. Thus, the ER can modulate the transcription of genes containing AP-1 sites, as in the case of the collagenase [10] and IGF-I [11] genes. Similarly, suppression of IL-6 gene expression by estrogen occurs via interactions of the liganded ER with the NF-κB complex [12]. Finally, estrogen can also regulate gene expression through membrane pathways which involve alterations in MAP kinase activity [13]; these effects appear to be particularly important for the anti-apoptotic effects of estrogen on osteoblasts.

Figure 1.

ER signaling pathways. The classical pathway of ERα action (left) requires direct binding to EREs found in the control regions of estradiol (E₂)-regulated genes. The non-classical pathway can involve either nuclear (middle) or membrane (right) signaling. In non-classical nuclear signaling, the ER interacts with other transcription factors, such as the jun/fos complex which binds to AP-1 sites, whereas in the membrane pathway, ER signaling is via alterations in kinase activity.

In order to better define non-classical estrogen signaling and the balance between the classical and non-classical estrogen signaling pathways of ERα in vivo, Jakacka and colleagues generated mice which have two substitution mutations (E207A/G208A) in the first zinc finger of the DNA binding domain in one of the ERα alleles (non-classical ER knock-in mice, NERKI) [14]. This mutant receptor fails to activate ERE reporter activity in vitro, but has intact transcriptional activity for AP-1 reporter constructs and can bind c-jun normally in a mammalian two-hybrid cell assay [15]. Heterozygote female NERKI mice are infertile and display pronounced reproductive defects [14], suggesting that the balance between classical and non-classical estrogen signaling has important biological consequences in vivo. We subsequently generated female ERα^−/NERKI mice which completely lack classical ERα signaling but have intact signaling via the non-classical ERα pathway and characterized the skeletal phenotype of these mice [16]. Relative to wild type (ERα^+/+) mice, female ERα^−/NERKI mice had reduced cortical volumetric BMD (vBMD) and thickness at the tibial diaphysis, preserved trabecular vBMD at the tibial metaphysis, and a trend for reduced bone volume/total volume (BV/TV) in trabecular bone at the lumbar spine [16]. Similar to ERα^−/− females [5, 16], female ERα^−/NERKI mice also had significant increases in circulating estradiol levels, which may account for the relative preservation of trabecular bone in these mice through actions on ERβ [5, 16].

While our previous studies indicated that the NERKI receptor had negative effects on bone, we were unable to address the important question of whether the presence of the NERKI receptor was, in fact, more deleterious for bone than simply the absence of ERα. This was because our ERα^−/NERKI mice were in a 50:50 C57BL/6:129/SvJ background, whereas the ERα^−/− were in a C57BL/6 background, making it impossible to directly compare the skeletal phenotypes of ERα^−/− and ERα^−/NERKI mice [16]. Thus, in this study, we assessed bone parameters in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice, all of who were in a C57BL/6 background for > 6 generations. Overall, our findings indicate that loss of classical estrogen receptor signaling may have distinct skeletal consequences as compared to complete deletion of ERα.

METHODS

Generation, breeding, and care of animals

Heterozygote ERα^+/NERKI male mice from our earlier study [16] in a 50:50 C57BL/6:129/SvJ background were backcrossed with heterozygote ERα^+/− female mice [17] in a C57BL/6 background for more than 6 generations. The female ERα^+/+, ERα^−/− and ERα^−/NERKI littermates were then studied at 3– 5 months of age. To assess dynamic measurements of bone formation, the mice received two calcein injections intraperitoneally (10 mg/kg) a week apart, with the last injection given two days prior to sacrifice. The animals were housed in a temperature controlled room (22 ± 2°C) with a daily 12-h light/12-h dark schedule. During the experiment, animals had free access to water and were pair fed a casein based diet (AIN 93M; Dyets Inc). Pups were genotyped at 4–5 weeks of age by PCR as described previously [18]. The Institutional Animal Care and Use Committee approved all animal procedures.

Sex steroid measurements

Serum estradiol (E₂) was measured using an ultra-sensitive quantitative radioimmnoassay (Beckman Coulter, Brea, CA; inter-assay CV<12%). Testosterone (T) was also measured by radioimmunoassay (Siemens, Los Angles, CA; inter-assay CV 11%).

Bone Phenotypic Characterization

Three-month old female ERα^+/+, ERα^−/− and ERα^−/NERKI mice were scanned by dual energy x-ray absorptiometry (DXA) to obtain whole body, femur and spine areal BMD (aBMD) measurements as well as whole body % fat mass. The tibial metaphysis and diaphysis were scanned by peripheral quantitative computed tomography (pQCT) to obtain total, trabecular and cortical volumetric BMD (vBMDs) and other bone structural parameters. For the in vivo BMD measurements, the mice were anesthetized with Avertin (2,2,2 tribromoethanol, 720mg/kg i.p.). DXA measurements were carried out using a Lunar PIXImus densitometer (software version 1.44; Lunar Corp.). Mice were placed on an animal tray in a prone position and whole body scans were carried out. After scanning, regions of interest (whole body, right femur and lumbar spine) were analyzed. The coefficients of variation (CVs) for the total body, lumbar and femoral areal BMD (aBMD) were 4.9%, 2.7%, and 4.3%, respectively. pQCT measurements were performed with the mice placed in a supine position on a gantry using the Stratec XCT Research SA Plus using software version 5.40 (Norland Medical Systems Inc.). Slice images were measured at 1.9 mm (corresponding to the proximal tibial metaphysis) and at 9 mm (corresponding to the diaphysis of the tibia) from the proximal end of the tibia. For trabecular bone, the threshold was set at 480 mg/cm³ and for cortical bone at 710 mg/cm³. The CV was 4.4% for the total tibial vBMD at the metaphysis and 1.4% for the cortical vBMD at the diaphysis.

Histology and histomorphometry

The lumbar vertebrae and femurs were dehydrated in alcohol, embedded in methyl methacrylate, sectioned and stained using Goldner’s stain. The bone volume/total volume (BV/TV), trabecular number (TbN, /mm), trabecular thickness (TbTh, µm) and trabecular separation (TbSp, mm) were obtained using standard methods. The measurements excluded the primary spongiosa. All measurements were done at a 10× magnification using the OsteoMeasure analysis system (Osteometrics). The dynamic measurements were carried out as described previously [18, 19].

MicroCT Measurements

The quantitative analysis of the metaphyseal cancellous bone of the proximal tibia was done by µCT (µCT 35, Scanco Medical AG). Using two-dimensional data from scanned slices, three-dimensional analysis was conducted to calculate morphometric parameters defining trabecular bone mass and micro-architecture, including BV/TV, TbN, TbTh, and ThSp.

Osteogenic and adipogenic assays

Three-month old ERα^+/+, ERα^−/−, and ERα^−/NERKI mice were sacrificed by CO₂ inhalation and the tibiae and femora were excised under aseptic conditions. The metaphyseal ends of the bones were removed in order to expose the marrow cavity. The marrow was flushed using a 27 gauge syringe into maintenance medium (phenol-free αMEM containing 10% heat inactivated fetal bovine serum and 1% antibiotic/antimycotic, GIBCO/BRL). The cells were then plated after counting on a hemocytometer using trypan blue exclusion of dead cells. These freshly isolated cells derived from 5–7 individual mice of each genotype were then plated at a density of 1 ×10⁷ cells/well for the colony-forming unit-fibroblast (CFU-F), CFU-alkaline phosphatase (AP), CFU-osteoblast (OB), and adipocyte assays. Osteogenic conditions (for the CFU-AP and CFU-OB assays) included the addition of 0.01M β glycerophosphate (Sigma), 10⁻⁸M dexamethasone (DAKO), and 50 µg/ml ascorbic acid (Sigma). Adipogenic conditions included the addition of 10⁻⁶M rosiglitazone. Half media changes were carried out on day 3. On day 15, the cells were stained with methylene blue for enumerating CFU-Fs, with an AP stain (Sigma) for CFU-APs, and with an Oil Red O stain for adipocytes. The CFU-OBs were enumerated on day 24 by von Kossa staining. To identify CFU-Fs, methylene blue positive colonies (>20 cells) were counted; similar criteria were used to identify CFU-AP and CFU-OB colonies. For the adipocyte assay, we counted the number of Oil Red O positive cells per plate.

In vitro effects of estrogen on osteoblast versus adipogenic differentiation

The following experimental groups used in order to assess the osteogenic potential of bone marrow stromal cells derived from the different genotypes and their modulation by estrogen : (i) osteogenic conditions without estradiol (E₂) (αMEM +10% triple charcoal stripped fetal bovine serum + 1% antibiotic/antimycotic solution [GIBCO/BRL] containing 0.01M β glycerophosphate [Sigma], 10⁻⁸M dexamethasone [DAKO] and 50 ug/ml ascorbic acid [Sigma]); and (ii) osteogenic conditions with E₂ (αMEM +10% triple charcoal stripped fetal bovine serum + 1% antibiotic/antimycotic solution [GIBCO/BRL] containing 0.01M β glycerophosphate (Sigma), 10⁻⁸M dexamethasone [DAKO], 50 ug/ml ascorbic acid [Sigma] and 10⁻⁸M E₂ [Sigma]). Cells were replenished every 4 days with fresh media and on day 24 cells were washed with PBS and fixed with cold 70% ethanol and stained with 6% silver nitrate solution. The excess stain was bleached by briefly incubating with 5% sodium thiosulfate solution. The von Kossa-positive colonies were counted and expressed as a percent of the osteogenic control. To evaluate estrogen effects on adipogenesis in the various genotypes, we used the following groups: (i) adipogenic Treatment without E₂ (αMEM +10% triple charcoal stripped fetal bovine serum + 1% antibiotic/antimycotic solution [GIBCO/BRL] containing 10⁻⁶ M rosiglitazone) and (ii) adipogenic treatment with E₂ (αMEM +10% triple charcoal stripped fetal bovine serum + 1% antibiotic/antimycotic solution [GIBCO/BRL] containing 10⁻⁶ M rosiglitazone and 10^{− 8}M E₂ [Sigma]). Cells were replenished every 4 days with fresh media and on day 15 cells were stained with Oil Red O. The number of Oil Red O Positive cells were counted and the results expressed as percentage of the adipogenic control.

Statistical Methods

For comparisons between groups, we used an ANOVA; where this was significant, the Fisher Protected Least Significant Difference test was used for pairwise comparisons between groups. All data are presented as mean ± SEM, and P < 0.05 was considered significant.

RESULTS

Body weight and composition

Figure 2 shows the body weights (Figure 2A) and percent fat mass (Figure 2B), assessed by DXA, in the ERα^+/+, ERα^−/−, and ERα^−/NERKI mice. As is evident, either complete deletion of ERα or loss of ERE signaling led to increased body weight and in percent body fat, consistent with a role for ERα and specifically, ERE signaling, in regulating body composition in mice.

Figure 2.

(A) Body weight (n = 48–49 per group) and (B) percent body fat by DXA in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 14–25 per group). **P < 0.01 and ***P < 0.001 versus ERα^+/+; ^†††P < 0.001 versus ERα^−/−.

Serum sex steroid levels

We previously found that ERα^−/NERKI mice in a 50:50 C57BL/6:129/SvJ background had significantly increased serum E₂ levels as compared to ERα^+/+ mice in the same background [16]. We now reassessed serum E₂ and T levels in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice who were all in the C57BL/6 background (Table 1). As is evident, in this genetic background, while serum E₂ and T levels were significantly higher in the ERα^−/− mice as compared to the ERα^+/+ mice, serum E₂ levels in the ERα^{− /NERKI} mice were similar to those in the ERα^+/+ mice However, the ERα^−/NERKI mice did have significant increases in serum T levels as compared to the ERα^+/+ mice.

Table 1.

Serum E₂ and T levels in the ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 7–11 per group).

	ERα+/+	ERα−/−	ERα−/NERKI
E2, pg/mL	9 ± 2	21 ± 4*	11 ± 3
T, ng/dL	40 ± 9	111 ± 23*	72 ± 11*

^*P < 0.05 versus ERα^+/+

Areal and volumetric BMD

Figure 3 shows total body aBMD (Figure 3A), spine aBMD (Figure 3B), and femur aBMD (Figure 3C) in the 3 groups of mice. In the ERα^−/− mice, the increases in E₂ and T levels noted above likely activate ERβ and the androgen receptor, respectively [5], leading to the observed preservation of aBMD at all 3 sites in these mice; by contrast, despite the observed increased T levels, the ERα^−/NERKI mice had reduced total body, spine, and femur aBMD values as compared to both the ERα^+/+ and ERα^−/− mice (Figure 3 A – C).

Figure 3.

(A) Total body aBMD, (B) spine aBMD, and (C) femur aBMD by DXA in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 14–25 per group). ***P < 0.001 versus ERα^+/+; ^†††P < 0.001 versus ERα^−/−.

Total vBMD at the tibial metaphysis was reduced in the ERα^−/− and in the ERα^−/NERKI mice (Figure 4A), with greater reductions in the ERα^−/NERKI mice. However, trabecular vBMD was increased in both the ERα^−/− and ERα^−/NERKI mice (albeit to a lesser extent in the ERα^−/NERKI as compared to the ERα^−/− mice) (Figure 4B). By contrast, the reduction in total vBMD in both the ERα^−/− and ERα^−/NERKI mice was largely due to reductions in cortical vBMD (Figure 4C).

Figure 4.

(A) Total vBMD, (B) trabecular vBMD, and (C) cortical vBMD at the tibial metaphysis by QCT in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 18–31 per group). ***P < 0.001 versus ERα^+/+; ^†††P < 0.001 versus ERα^−/−.

Since the tibial diaphysis is a more robust site for assessment of cortical bone, we also assessed cortical parameters at this site. Cortical vBMD (Figure 5A) and thickness (Figure 5B) were slightly higher in ERα^−/− as compared to ERα^+/+ mice, but both parameters were significantly reduced in the ERα ^−/NERKI as compared to both the ERα^+/+ and ERα^−/− mice. Periosteal circumference was increased in the ERα^−/− , but not the ERα^{− /NERKI} mice (Figure 5C), whereas the ERα^−/NERKI mice had significant increases in endosteal circumference at this site as compared to both the ERα^+/+ and ERα^−/− mice (Figure 5D).

Figure 5.

(A) Cortical vBMD, (B) cortical thickness, (C) periosteal circumference, and (D) endosteal circumference at the tibial diaphysis in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 18–31 per group). ^aP = 0.051, *P < 0.05, **P < 0.01 and ***P < 0.001 versus ERα^+/+; ^††P < 0.01 and ^†††P < 0.001 versus ERα^−/−.

µCT and histomorphometry

We also assessed trabecular bone parameters in more detail at the proximal tibial metaphysis using µCT (Table 2) and at the lumbar spine using histomorphometry (Table 3). This did reveal site-specificity in trabecular bone changes in the ERα^−/− and ERα^−/NERKI mice as compared to the ERα^+/+ mice. Thus, BV/TV at the tibial metaphysis was increased in both the ERα^−/− and ERα^−/NERKI mice (albeit to a lesser extent in the ERα^−/NERKI mice) as compared to the ERα^+/+ mice (Table 2). TbTh was reduced in the ERα^−/NERKI as compared to both the ERα^+/+ and ERα^−/− mice. Both the ERα^{− /−} and ERα^−/NERKI mice had increases in TbN and decreases in TbSp as compared to the ERα^+/+ mice (Table 2). Figure 6A shows representative µCT images from the tibial metaphysis in the 3 groups of mice.

Table 2.

Trabecular bone parameters at the proximal tibial metaphysis by µCT in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 3–5 per group). BV/TV, bone volume/tissue volume; TbTh, trabecular thickness; TbN, trabecular number; TbSp, trabecular separation.

	ERα+/+	ERα−/−	ERα−/NERKI	ANOVA
BV/TV, %	4.8 ± 0.28	12.7 ± 1.1***	8.4 ± 1.1*,†	<0.001
TbTh, mm	0.036 ± 0.001	0.041 ± 0.001*	0.031 ± 0.001*,†††	<0.001
TbN, mm	3.18 ± 0.12	4.71 ± 0.11***	5.00 ± 0.14***	<0.001
TbSp, mm	0.322 ± 0.011	0.221 ± 0.007***	0.202 ± 0.006***	<0.001

^*P < 0.05,

^**P < 0.01 and

^***P < 0.001 versus ERα^+/+;

^†P < 0.05 and

^†††P < 0.001 versus ERα^−/−.

Table 3.

Static and dynamic histomorphometric parameters at the lumbar spine in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 9–12 per group). BV/TV, bone volume/tissue volume; TbTh, trabecular thickness; TbN, trabecular number; TbSp, trabecular separation; OS/BS, osteoid surface/bone surface; Ob.S/BS, osteoblast surface/bone surface; N.Ob/B.Pm, number of osteoblasts/bone perimeter; ES/BS, eroded surface/bone surface; Oc.S/BS, osteoclast surface/bone surface; N.Oc/B.Pm, number of osteoclasts/bone perimeter; MAR, mineral apposition rate; BFR/BS, bone formation rate, surface referent.

	ERα+/+	ERα−/−	ERα−/NERKI	ANOVA
Static parameters
BV/TV, %	23.0 ± 1.1	21.9.0 ± 1.4	21.0 ± 1.1	0.492
TbTh, mm	0.050 ± 0.002	0.044 ± 0.003a	0.051 ± 0.002†	0.095
TbN, /mm	4.6 ± 0.2	5.0 ± 0.1	4.1 ± 0.1*,†††	0.0014
TbSp, mm	0.170 ± 008	0.158 ± 0.004	0.193 ± 0.008*,††	0.007
OS/BS, %	6.25 ± 0.72	2.92 ± 0.70**	3.92 ± 0.59*	0.01
Ob.S/BS, %	17.2 ± 1.4	9.3 ± 1.3***	13.9 ± 0.9a, †	0.002
N.Ob/B.Pm, /mm	20.0 ± 1.7	12.4 ± 1.5**	17.4 ± 1.3†	0.009
ES/BS, %	1.27 ± 0.22	1.15 ± 0.19	0.85 ± 0.09	0.273
Oc.S/BS, %	4.43 ± 0.39	2.81 ± 0.55*	3.06 ± 0.24*	0.032
N.Oc/B.Pm, /mm	1.68 ± 0.14	1.37 ± 0.24	1.41 ± 0.10	0.415
Dynamic parameters
MAR, µm/d	0.936 ± 0.059	0.805 ± 0.016*	0.805 ± 0.041*	0.060
BFR/BS, µm3/µm2/d	0.192 ± 0.020	0.152 ± 0.015	0.188 ± 0.026	0.308

^aP = 0.075,

^*P < 0.05,

^**P < 0.01 and

^***P < 0.001 versus ERα^+/+;

^†P < 0.05 versus ERα^−/−.

Figure 6.

(A) Representative µCT images of the tibial metaphysis and (B) representative histological sections of the lumbar spine in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice.

In contrast to the increase in BV/TV at the tibial metaphysis in the ERα^−/− and ERα^−/NERKI mice as compared to the ERα^+/+ mice, all 3 groups of mice had similar values for BV/TV at the lumbar spine (Table 3). However, while trabecular microarchitecture was relatively well preserved in the ERα^−/− mice, the ERα^−/NERKI mice had significant reductions in TbN and increases in TbSp at the lumbar spine; this did result in a small but non-significant decrease in BV/TV in the ERα ^−/NERKI mice at this site. Overall, ERα^−/NERKI mice clearly had more severe deficits in trabecular structure at the lumbar spine as compared to the ERα^−/− mice. Figure 6B shows representative histological sections from the lumbar spine in the 3 groups of mice.

Table 3 shows additional trabecular static and dynamic histomorphometric parameters at the lumbar vertebrae. Osteoid surface/bone surface (OS/BS) and osteoblast surface/bone surface (Ob.S/BS) were reduced in both the ERα^−/− and ERα^−/NERKI as compared to the ERα^+/+ mice, with a similar trend for number of osteoblasts/bone perimeter (N.Ob/B.Pm). Eroded surface/bone surface (ES/BS) or number of osteoclasts/bone perimeter (N.Oc/B.Pm) did not differ between the groups, while osteoclast surface/bone surface (Oc.S/BS) was reduced in both the ERα^−/− and ERα^−/NERKI as compared to the ERα^+/+ mice. Mineral apposition rate (MAR), but not the surface referent bone formation rate (BFR/BS), was also reduced in both the ERα^−/− and ERα^−/NERKI as compared to the ERα^+/+ mice.

Table 4 shows the cortical static and dynamic histomorphometric parameters at the femur diaphysis in the 3 groups of mice. Consistent with the pQCT data at the tibial diaphysis (Figure 5B), histologically assessed cortical thickness was reduced at the femur diaphysis in the ERα^−/NERKI, but not the ERα^−/−, mice. Periosteal MAR and BFR were reduced in both the ERα^−/− and ERα^−/NERKI as compared to the ERα^+/+ mice. By contrast, endocortical MAR (but not BFR) was significantly reduced in the ERα^−/NERKI mice, but not in the ERα^−/− mice, consistent with the more severe reductions in cortical thickness present in the ERα^−/NERKI mice.

Table 4.

Static and dynamic histomorphometric parameters at the femur diaphysis in ERα^+/+, ERα^−/− , and ERα^−/NERKI mice (n = 6 per group).

	ERα+/+	ERα−/−	ERα−/NERKI	ANOVA
Static parameters
Cortical thickness, mm	0.185 ± 0.008	0.190 ± 0.010	0.127 ± 0.012***,†††	<0.001
Dynamic paramters
Periosteal surface
MAR, µm/d	0.925 ± 0.038	0.620 ± 0.167*	0.366 ± 0.081***	0.001
BFR, µm2/µm/d	0.140 ± 0.025	0.032 ± 0.016***	0.026 ± 0.012***	<0.001
Endocortical surface
MAR, µm/d	0.796 ± 0.063	0.692 ± 0.082	0.508 ± 0.050**,a	0.013
BFR, µm2/µm/d	0.122 ± 0.013	0.091 ± 0.046	0.073 ± 0.022	0.430

^*P < 0.05,

^**P < 0.01 and

^***P < 0.001 versus ERα^+/+;

^aP = 0.068

^†††P < 0.001 versus ERα^−/−.

Collectively, these data demonstrate that while trabecular bone volume at the tibial metaphysis is increased in both the ERα^−/− and ERα^−/NERKI mice as compared to the ERα^+/+ mice, this is not the case for trabecular bone volume at the spine. However, at the spine, ERα^−/NERKI mice do have microstructural changes (reduced TbN and increased TbSp, with a trend for reduction in BV/TV) not present in the ERα^−/− mice. In addition, cortical vBMD and thickness are reduced in the ERα^−/NERKI mice as compared to either the ERα^+/+ or ERα^−/− mice. MAR, which represents the work done by a team of osteoblasts, is reduced in trabecular and in cortical bone in the ERα^−/− and ERα^−/NERKI mice, with the deficit in MAR in cortical bone tending to be more severe in the ERα ^−/NERKI as compared to the ERα^−/− mice.

In vitro studies

To better understand the cellular basis for the alterations in skeletal structure and dynamics in the ERα^−/− and ERα^−/NERKI mice, we performed CFU-F, CFU-AP, and CFU-OB assays in the 3 groups of mice. As shown in Figure 7, all 3 parameters were increased in the ERα^−/−, but reduced in the ERα^−/NERKI, as compared to the ERα^+/+ mice. These data indicate that in contrast to loss of ERα, the presence of the NERKI receptor leads to a reduction in osteoblast progenitors in the ERα^−/NERKI mice. Figure 8 shows the number of adipocytic (Oil Red O+) cells in marrow cultures from the 3 genotypes. As is evident, these were similar in the ERα^+/+ and ERα^−/− cultures, but reduced in the ERα^−/NERKI cultures, indicating that the presence of the NERKI receptor is suppressive also of adipocytic precursors.

Figure 7.

(A) CFU-F, (B) CFU-AP, and (C) CFU-OB colonies (per 10⁷ cells plated) in ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 4 per group). *P < 0.05, **P < 0.01 and ***P < 0.001 versus ERα^+/+; ^†††P < 0.001 versus ERα^−/−.

Figure 8.

Oil-red O+ (adipocyte) cells in bone marrow cultures from ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 5–10 per group). *P < 0.05 versus ERα^+/+ mice.

Since both osteoblasts and adipocytes share a common progenitor [20], in further studies we also examined the effects of in vitro estrogen treatment of bone marrow stromal cells from the 3 groups of mice on osteoblast versus adipocyte colonies. As shown in Figure 9A, while estrogen treatment significantly increased the number of osteoblastic colonies in cells from the ERα^+/+ mice, estrogen failed to have this effect on cells from the ERα^−/− or the ERα^−/NERKI mice. By contrast, estrogen treatment reduced the numbers of adipocytes in vitro in cells from ERα^+/+ as well as ERα^−/NERKI mice, but failed to have this effect in cells from the ERα^−/− mice (Figure 9B). These findings thus demonstrate that while stimulation of mineralization by estrogen requires both ERα and intact ERE activity, the suppression of adipogenesis by estrogen, although requiring ERα, can also occur in cells from ERα^−/NERKI mice.

Figure 9.

(A) Osteoblast colonies and (B) Oil-red O+ (adipocyte) cells (open bars, vehicle control; solid bars, 10⁻⁸ M E₂) in bone marrow cultures from ERα^+/+, ERα^−/−, and ERα^−/NERKI mice (n = 5–10 per group). *P < 0.05 and **P < 0.01 versus the vehicle control.

DISCUSSION

The effects of estrogen on bone are complex and mediated by multiple signaling pathways [18]. In the present study, we compared the skeletal phenotype of wild type (ERα^+/+) mice with those having complete deletion of ERα (ERα^−/−) versus mice in whom the only active ERα was the NERKI receptor, which is unable to bind DNA (ERα^−/NERKI) [14]. All genotypes were in the same genetic background, allowing for a direct comparison between groups. In cortical bone at the tibial diaphysis, we found that ERα ^−/NERKI mice had significantly reduced values for cortical vBMD and thickness as compared to either the ERα^+/+ or ERα^−/− mice. This was accompanied by significant decreases in periosteal and endocortical MARs in the ERα^−/NERKI mice. These data thus suggest that for cortical bone, the presence of the NERKI receptor has effects that are distinct from simple loss of the receptor. Similar to our findings, Almeida et al. [21] also found reduced cortical thickness at the tibial diaphysis in ERα^−/NERKI as compared to ERα^+/+ mice, but these investigators did not compare values for cortical bone parameters in the ERα^−/NERKI mice versus the ERα^−/− mice.

In our previous study [16], we found that ERα^−/NERKI mice in a 50:50 C57BL/6:129/SvJ background had significantly increased serum E₂ levels as compared to ERα^+/+ mice in the same background. Interestingly, when we compared ERα^−/NERKI mice in a complete C57BL/6 background to ERα^+/+ mice, serum E₂ levels did not differ between the two groups, but serum T levels were significantly higher in the ERα^−/NERKI mice. While we did not previously assess serum T levels in the ERα^−/NERKI mice in a 50:50 C57BL/6:129/SvJ background, these findings do indicate that there may be significant strain effects on changes in serum sex steroid levels in mice with ERα mutations.

The trabecular bone phenotype of the ERα^−/NERKI mice was more complex than the cortical bone phenotype and differed somewhat for trabecular bone at the tibial metaphysis versus the spine. As compared to the ERα^+/+ mice, trabecular vBMD and BV/TV at the tibial metaphysis was increased in both the ERα^−/− and ERα^−/NERKI mice, although lower in the ERα^−/NERKI as compared to the ERα^−/− mice. Similar to our findings, Sims et al. [5] reported that ERα^−/− had increased trabecular BV/TV at the femoral metaphysis; the authors postulated that this was likely secondary to the elevated E₂ levels in the female ERα^−/− mice activating ERβ signaling [5].

In contrast to the tibia, trabecular BV/TV was not increased in either the ERα^−/− or ERα^−/NERKI mice at the lumbar spine; instead, as compared to both the ERα^+/+ and ERα^−/− mice, ERα^−/NERKI mice had significantly reduced trabecular numbers and increased trabecular separation. Trabecular BV/TV was somewhat reduced (by 9%) in the ERα ^−/NERKI as compared to the ERα^+/+ mice, but this difference was not statistically significant. While Almeida et al. [21] did not compare the ERα^−/NERKI to the ERα^−/− mice, the ERα^−/NERKI mice in their study did have significant reductions in spine BV/TV as compared to the ERα^+/+ mice. These reductions were of a greater magnitude than those observed in our study, perhaps due to the fact that the mice used by Almeida et al. [21] were in a mixed C57BL/6–129Sv/J background, rather than in a complete C57BL/6 background. This difference notwithstanding, our findings in trabecular bone at the spine do demonstrate that the ERα^−/NERKI mice have significant structural deficits as compared to the ERα^−/− mice. It should also be noted that, as compared to the ERα^+/+ mice, osteoclast numbers or eroded surfaces were not increased in either the ERα^−/− or ERα^−/NERKI mice. Thus, while osteoclast-specific deletion of ERα clearly results in increased osteoclast numbers and enhanced bone resorption [22, 23], global deletion of ERα or global knock-in of the NERKI receptor does not increase either parameter, perhaps due to the increase in sex steroid levels in these mice activating ERβ or the androgen receptor in osteoclasts. Further studies are needed, however, to address this important issue.

Our in vitro studies provide further support for differential effects of deletion of ERα versus the NERKI mutation on CFU-F, -AP, and OB numbers. All three parameters were increased in bone marrow cells from the ERα^−/− mice, consistent with previous work from Di Gregario et al. [24] showing that estrogen suppresses the proliferation of early mesenchymal progenitors. By contrast, all three parameters were reduced in bone marrow cells from ERα^−/NERKI as compared to either the ERα^+/+ or ERα^−/− mice. Finally, our studies with estrogen treatment of bone marrow stromal cells derived from mice of the three genotypes indicated that the presence of ERα and ERE signaling was necessary for the stimulation of mineralization by these cells. However, estrogen was able to suppress adipocyte colony formation in cells from ERα^+/+ and ERα^−/NERKI, but not ERα^−/−, mice, indicating that ERE signaling was not required for this effect. The molecular mechanisms mediating these differential effects of ERα on mineralization versus adipocyte formation warrant further study, but these data do suggest that ERα (and in particular, the classical function of ERα) plays a role in osteoblastic versus adipocytic differentiation of mesenchymal progenitor cells. Additional studies are also needed to examine changes in marrow fat with aging in ERα^−/− and ERα^−/NERKI as compared to wild type mice.

It is also important to address the apparent discrepancy between our data in Figure 7, where complete loss of ERα leads to an increase in CFU-F, CFU-AP, and CFU-OB (implying a suppressive effect of ERα on these parameters, which also is present in the ERα^−/NERKI mice), whereas in Figure 9, estrogen treatment facilitates osteoblast differentiation in wild type mice. We believe this is due to the fact that the CFU assays measure progenitor cells for CFU-F, CFU-AP, and CFU-OB in vivo in the absence of estrogen treatment in vitro (which are increased in the absence of ERα and reduced in the presence of the NERKI receptor), which is distinct from the effects of estrogen in enhancing differentiation (which is what is assessed in Figure 9). Thus, estrogen appears to suppress the pool of progenitor cells (as demonstrated, for example, by Di Gregorio et al. [24]), leading to the increase in CFU-F, CFU-AP, and CFU-OB populations in the ERα^−/− mice, but when progenitor cells are exposed to estrogen, they do show enhanced differentiation. In short, effects of estrogen on the progenitor pool seem to be different from the effects on estrogen on osteoblast differentiation. In this context, it is also important to note that the MAR reflects the activity of a team of osteoblasts in the basic multicellular unit, whereas the CFU assays reflect the numbers of progenitor cells in vivo. Thus, changes in these two parameters may differ, depending on the effects of estrogen on the progenitor pool versus osteoblast differentiation and/or activity

Since our 3 genotypes were wild type (ERα^+/+), complete ERα deletion (ERα^−/−), or mice where the only receptor is the NERKI receptor (ERα^−/NERKI), and we did not rely on or use heterozygous mice (i.e., ERα^+/NERKI) mice, the issue of the NERKI receptor being a dominant negative is not a concern with any of the data in this paper. Moreover, previous work by Jackaka et al. [14] has provided evidence that the NERKI receptor does not function as a dominant negative receptor for wild type ERα.

Previous studies by Jackaka and colleagues [15] have shown that the NERKI receptor fails to activate ERE reporter activity in vitro, but has intact transcriptional activity for AP-1 reporter constructs and can bind c-jun normally in a mammalian two-hybrid cell assay. Of note, recent studies have provided evidence for a requirement for “pioneer” factors, such as FoxA1, in mediating direct ER binding to DNA, at least in breast cancer cells [25]. Based on what is known regarding preserved protein-protein interactions of the NERKI receptor [14,15], this receptor should be able to bind factors such as FoxA1, although further studies are needed to address this possibility.

In summary, our study provides a detailed skeletal characterization of ERα^−/NERKI compared to ERα^−/− mice and indicates that for a number of bone parameters, the presence of the NERKI receptor results in greater skeletal deficits than complete loss of the receptor. We recognize that our models are limited by the fact that the deletion of ERα or the NERKI mutation was present in all tissues, and not specifically restricted to osteoblasts or osteoclasts. However, generating mice with replacement of the wild-type ERα by the NERKI receptor in specific cell types is not a simple exercise, although the construction of such mice has been initiated in our laboratory. This limitation notwithstanding, our findings point to the need for further studies aimed at defining the cellular and molecular mechanisms whereby shifting the balance of ERα signaling from the classical (ERE) to the non-classical (non-ERE) pathway leads to deleterious skeletal effects. In addition to better understanding the fundamental biology of ER signaling, such studies may identify novel approaches to selectively modulate ER signaling pathways for beneficial skeletal effects.