Osteoblast-Like Cells From Estrogen Receptor α Knockout Mice Have Deficient Responses to Mechanical Strain


  • The authors have no conflict of interest


Osteoblast-like cells in primary cultures derived from ERα−/− mice do not proliferate in response to mechanical strain, unlike those from their ERα+/+ littermates. ERα−/− cells also lack strain-related NO production and responsiveness to IGFs. Proliferative responsiveness to strain is rescued by transfection with functional ERα. ERα number or function in bone cells may limit bones' adaptability to mechanical loading.

Introduction: In vivo, bones' osteogenic response to mechanical loading involves proliferation of surface osteoblasts. This response is replicated in vitro and involves ERK-mediated activation of the estrogen receptor (ER) α and upregulation of estrogen response element activity. This proliferative response can be blocked by selective estrogen receptor modulators and increased by transfection of additional ERα.

Materials and Methods: We have now investigated the mechanisms of ER involvement in osteoblast-like cells' early responses to strain by comparing the responses of primary cultures of these cells derived from homozygous ERα knockout (ERKO) mice (ERα−/−) with those from their wildtype (ERα+/+) and heterozygous (ERα+/−) littermates and from ERβ knockout (BERKO) mice (ERβ+/+, ERβ+/−, and ERβ−/−).

Results: Whereas ERα+/+, ERα+/−, ERβ+/+, and ERβ−/− cells proliferate in response to a single 10-minute period of cyclic strain, ERα−/− cells do not. Transfection of fully functional, but not mutant, ERα rescues the proliferative response to strain in these cells. The strain-related response of ERα−/− cells is also deficient in that they show no increased activity of an AP-1 driven reporter vector and no strain-related increases in NO production. Their strain-related increase in prostacyclin production is retained. They proliferate in response to fibroblast growth factor-2 but not insulin-like growth factor (IGF)-I or IGF-II, showing the importance of ERα in the IGF axis and the ability of ERα−/− cells to proliferate normally in response to a mitogenic stimulus that does not require functional ERα.

Conclusions: These data indicate ERα's obligatory involvement in a number of early responses to mechanical strain in osteoblast-like cells, including those that result in proliferation. They support the hypothesis that reduction in ERα expression or activity after estrogen withdrawal results in a less osteogenic response to loading. This could be important in the etiology of postmenopausal osteoporosis.


THE ABILITY OF the skeleton to withstand load-bearing, without damage, is dependent on bone cell populations being able to establish and subsequently maintain structurally appropriate bone architecture. Resident bone cells are commonly believed to do this by adjusting their modeling and remodeling activity in response to the magnitude and distribution of strains that load-bearing engenders within the bone tissue.(1,2) The ability of bone cells to continue to maintain an appropriate match between structure and load-bearing seems to become less effective with estrogen withdrawal at the menopause. The result is reduced bone mass, less structurally effective bone architecture, and increased incidence of fracture. Despite this suggestive link between estrogen and loading-related remodeling, there are no data, to our knowledge, identifying a mechanism by which estrogen itself directly affects bone cells' adaptive response to strain.

In vivo bones' mechanically adaptive response is presumed to involve some or all of the range of osteoblast and osteocyte responses reported to be stimulated by mechanical strain and/or fluid flow in vitro. These include stretch-activated ion channels,(2–5) G-proteins,(6) release of Ca2+ from intracellular stores,(7,8) increases in G6PD activity,(9) increased cyclic adenosine monophosphate (cAMP)(10) and phosphoinositide 3-kinase levels,(11) increased protein kinase C (PKC) activity,(12) MAPK activation,(13–15) ERα phosphorylation,(16) release of NO and prostaglandins,(5,17–19) upregulation of insulin-like growth factor (IGF)-II mRNA,(20) and Cbfa1 expression(21) and proliferation.(22,23)

Previous in vitro work in our laboratory has implicated a role for estrogen receptor α (ERα) in some of these responses to mechanical strain. For example, the strain-related proliferation of primary rat long bone-derived osteoblasts from both males and females is inhibited by tamoxifen, a partial ER antagonist, and by ICI 182,780, a full ER antagonist.(24,25) Exposure of osteoblast-like cells to strain also increases phosphorylation of ERα, in an extracellular regulated kinase (ERK)-dependent manner,(16) and is accompanied by upregulation of estrogen response element (ERE) activity.(26) Moreover, we have previously shown that the proliferative response of ROS 17/2.8 cells to a single period of strain is augmented by transfection of additional ERα.(26) This involvement of ERα in bone cells' adaptive responses to their mechanical environment has led us to hypothesize(27) that the bone loss associated with the menopause is caused by a reduction in bone cell responsiveness to mechanical strain after the reduction in ERα number and/or function associated with estrogen deficiency. Such reductions have been observed previously.(28–30)

ERα is a ligand-inducible transcription factor responsible for mediating the actions of estrogen in target cells.(31,32) It has a modular structure consisting of distinct functional domains including a DNA binding domain, a ligand-binding domain, and two transcriptional activation functions.(33,34) Activation-function 1 (AF-1) is located in the amino-terminal domain and AF-2 in the ligand-binding domain (C terminus). The traditional interpretation of ERα activation involves ligand binding, receptor dimerization, phosphorylation, and interaction with ERE sequences in the promoter region of target genes. There are now, however, numerous examples of both ligand-independent ERα activation(16,32,35–37) and important nongenomic actions.(38–42) ERα has also been shown to influence gene expression through alternative response elements(43–45) and by protein-protein interactions rather than DNA binding.(45–47) Ligand-independent activation of ERα has been shown to occur through both AF-1 and AF-2 domains.(35–37,48)

To address the role of ER, mice with disruption of ERα (ERα knockouts [ERKO]) and ERβ (BERKO) have been generated.(49,50) Longitudinal bone growth is reduced in the female ERKO mice but increased in female BERKO mice, in which there are corresponding changes in serum IGF-I levels.(51–53) Trabecular and cortical BMC are higher in female ERKO mice. However at 12 weeks of age, levels of bone formation and resorption are reduced relative to wildtype (WT) littermates.(53,54) The male ERKO mouse also has shortened long bones and decreased IGF-1 levels in serum.(55,56) Contrary to the situation in females, male ERKO mice have lower BMC compared with WT littermates.(55,56) The mechanical strength of femurs from 17-week-old ERKO males, determined at mid-diaphysis, is lower than in their WT littermates.(56)

If ERα is obligatorily involved in bone's mechanically adaptive responses, ERα knockout animals should have a “disuse” phenotype and should show a defective response to artificial loading. The phenotype of the ERKO mouse is not unambiguously evocative of “disuse,” a fact that may be related to factors such as the high endogenous levels of estrogen. Nevertheless, the ulna in ERKO mice is straighter than in WT littermates (a “disuse” phenomenon),(57) and these bones exhibit a substantially lower osteogenic response to loading in vivo.(58)

This study investigates the mechanisms of ER's involvement in bone cells' adaptive responses to mechanical strain. We therefore compared the response of primary cultures of osteoblast-like cells from ERKO homozygous (ERα−/−), heterozygous (ERα+/−), WT (ERα+/+), and BERKO (ERβ+/+ and ERβ−/−) mice to mechanical strain and IGFs. The responses examined included proliferation, ERK activation, activation of an AP-1 reporter vector derived from the promoter region of the IGF-II gene, and the production of NO and prostaglandins. We also examined the ability of ERα mutants to rescue the deficient responses of ERα−/− cells.


Colony details

The genetic background of both the ERKO and BERKO is the mouse strain C57B16J. The mice were housed in polycarbonate or polypropylene cages with wood chip and paper bedding. Weaners up to 8 weeks of age were fed a standard rodent breeding diet and thereafter a standard rodent maintenance diet (Special Diet Services, South Witham, UK). Postweaning mice were housed in single-sex littermate groups of two or three. They were not segregated by genotype, and no differences in physical activity among the different genotypes were observed. The bone density of these animals was not measured, but this information has been published previously.(52,53,56,59) Ex vivo calibration experiments confirmed that loading engendered the same strain levels in the ulna of knockout and WT mice.(60)

Cell isolation

Osteoblast-like cells were isolated from cortical explants of the long bones of 6- to 10-week-old male and female ERKO and BERKO homozygotes, heterozygotes, and WT littermates by a method described previously.(20) Briefly, this involved the removal of the attendant soft tissue from the long bone and flushing out the bone marrow with sterile PBS. The bones were cut longitudinally, and each half was cut into small (∼1 mm) fragments. The bone fragments were cultured in tissue culture petri dishes (Corning) in DMEM (Invitrogen, Paisley, UK) containing heat-inactivated 10% FCS (Invitrogen), supplemented with 2 mM L-glutamine (Invitrogen), 100 U/ml penicillin (Invitrogen), and 100 μg/ml streptomycin (Invitrogen) and incubated at 37°C in a humidified 5% CO2 incubator. The fragments were incubated for 21 days to allow cell outgrowth, with media replenished weekly.

Exposure to mitogens

Cells were seeded onto sterile, tissue culture-treated, plastic strips (Nunc, Dossel, Germany) at a density of 80,000 cells/strip, and three strips were incubated together in 150-cm2 plastic sterile dishes (VWR, Poole, UK) with DMEM and 10% charcoal-dextran stripped FCS throughout the treatment period. The cells (three strips of cells/condition) were exposed to mechanical strain. Cells were seeded at a density of 30,000 cells/well (three wells/condition) in 8-well plates (Nunc) and exposed to 2 ng/ml fibroblast growth factor (FGF)-2 or vehicle (media), 40 ng/ml (des 1-3) IGF-I (GroPep, Adelaide, Australia), 40 ng/ml (des 1-6) IGF-II (GroPep), or a vehicle (final concentration, 4 μM HCl) control for 24 h. The IGFs used were truncated to prevent interaction with IGF-binding proteins. The cells were washed in PBS and fixed for 10 minutes in neutral-buffered formalin (BDH), washed twice in PBS, and stored at 4°C.

Mechanical straining

Cells seeded onto plastic strips were subjected to mechanical stimulation in a custom-designed loading apparatus. Each well of the loading apparatus contained one strip and 10 ml of fresh DMEM containing 10% charcoal dextran-stripped media. The strips were subjected to a single 10-minute stimulus of four-point bending consisting of 600 1-Hz cycles of strain generating a peak strain of 3400 με, which is within the physiological range engendered by locomotion in vivo.(61) After the 10-minute period of strain, strips were replaced in the plastic dish and incubated for a further 24 h with media transferred from the wells of the loading apparatus. Static (undisturbed) controls were removed from the incubator, provided with fresh media, and replaced in the incubator for the same 24 h. The cells were washed in PBS and either fixed, for cell counting, lysed for protein, or freeze-thawed with reporter lysis buffer (Promega) for reporter gene analysis, and conditioned media were collected and stored at −20°C.

Cell transfection

Cells for transfection were seeded at a density of 50,000 cells/strip and maintained in DMEM containing 10% FCS throughout the transfection. The cells were transiently transfected with a 3-h incubation with Effectene (Qiagen, Crawley, UK) following the manufacturer's recommendations. The plasmids were used at a concentration of 1 μg DNA/strip. To determine the transfection efficiency, cells were transfected with a lac z expression plasmid, at a concentration of 1 μg DNA/strip. β-galactosidase activity was detected by histochemical staining (lac z staining kit; Promega), and the transfection efficiency was determined by cell counting with phase contrast microscopy.

Plasmids for transfection

pRST7-ER is a human WT ERα expression vector, and pRST7-AF-2 is a human ERα mutant lacking a functional AF-2 domain.(33,62) Both were from D McDonnell (Duke University, Durham, NC, USA). L-543A/L-544A is a leucine to alanine mutation of the full-length mouse ERα sequence and was a gift from R White (Imperial College, London, UK). In this mutant, helix 12 of ERα is permanently in the antagonist-induced conformation that prevents interaction with co-activators.(63) The sequence of the AP-1 Luc reporter vector was derived from the IGF-II promoter and was a gift from A Ward (University of Oxford, Oxford, UK).

Nuclei counting

To determine the number of cells after exposure to strain or mitogens, the nuclei of fixed cells were labeled with a flurochrome (Hoechst 33258; Sigma).(64) The fixed strips were washed with distilled water, blotted dry on tissue, and exposed to 2 ml/strip of 1 mg/ml Hoechst 33258 in TNE (10 mM Tris, 1 mM EDTA, 2M NaCl, pH 7.4) buffer or 1 ml/well and incubated in the dark for 30 minutes. Cells were washed three times in TNE buffer. Excess liquid was blotted off on tissue before adding glycerol and a coverslip (VWR). The stained nuclei were visualized (magnification, ×4) on an ultraviolet microscope (Olympus BH-2), and 16 frames were individually counted per strip. The results are expressed as the average number of nuclei per frame for three strips (or wells) per condition in each experiment.

β-Galactosidase assay

After treatment, the cells were washed in PBS and lysed using 1 ml Reporter Lysis Buffer (Promega, Southampton, UK) per strip. The strips were stored overnight at −20°C and thawed before scraping the lysate into a sterile microcentrifuge tube and vortexed for 15 s. The lysates were stored at −70°C. The β-galactosidase assay (Promega) was performed in 96-well plates (Nunc) using cell lysates and measured on a fluorescence plate reader (Bio-Tek Instruments, Winooski, VT, USA). The results were used to normalize the AP-1 Luc reporter gene activity.

Luciferase assay

Luciferase activity was measured in 20-μl cell lysates in opaque 96-well plates using the Luciferase Assay System from Promega and carried out following the manufacturer's instructions. The samples were measured on a Wallac Victor 1420 multilabel Counter using Wallac 1420 Workstation software version 2 (Wallac Oy, Turku, Finland).

Nitrite assay in conditioned media

Because NO is labile, its production could not be measured directly but was assessed from the concentration of nitrite in the medium. Media samples were stored at −20°C before determination of nitrite concentration by chemiluminescence as described previously.(17,65)

Measurement of prostacyclin concentration in conditioned medium

Once the cells were harvested, conditioned medium was collected and stored (−20°C) before determination of 6-keto-PGF1α (the stable hydrolytic product of PGI2) levels by enzyme immunoassay (Assay Designs, Ann Arbor, MI, USA). None of the compounds that were added to the culture medium interfered with the immunoassays, and the assay was carried out following the manufacturer's instructions.

Statistical analyses

The data were analyzed using the (unpaired) Student t-test and SPSS software (version 10.0). Comparisons were made between the static and strain results within the same genotype. In the transfection experiments, statistical analysis was restricted to the static and strain results for the same vector. Comparisons between genotypes or between different expression vectors were not carried out because of the possibility of difference in expression levels. Significance was determined as p = 0.05 or lower.


Cells derived from both male and female animals were used in this study. No apparent differences were observed in their behavior in vitro. Therefore, the results described represent the combined results of both male and female cells.

ERα−/− osteoblast-like cells do not proliferate in response to strain

Osteoblast-like cells from ERKO WT mice (ERα+/+) as well as those from ERKO heterozygous (ERα+/−) and BERKO WT (ERβ+/+), heterozygous (ERβ+/−), and homozygous (ERβ−/−) mice exhibited significantly increased cell number in response to strain. Cells derived from ERα−/− mice showed no such strain-related response (Fig. 1) instead there was a reduction in cell number compared with the static controls. Cells from ERβ−/− mice showed a greater increase in cell number in response to strain than cells derived from ERβ+/− or ERβ+/+ mice. This suggests that ERβ may exert a negative influence on strain-related responses in osteoblast-like cells.

Figure FIG. 1..

Proliferation of primary murine osteoblasts in response to mechanical strain. ERα+/+, ERα+/−, and ERα−/− osteoblasts and ERβ+/+, ERβ+/−, and ERβ−/− osteoblasts in DMEM + 10% charcoal dextran-stripped FCS were exposed to a single 10-minute period of strain. The cells were fixed 24 h after the end of the strain stimulus, and the nuclei were stained and counted as described in the Materials and Methods section. The percentage change in cell number after exposure to strain compared with static controls is plotted. Error bars represent SEM, with the following p valuesa0.05,b0.03,c0.05, andd0.02. All genotypes from both ERKO and BERKO animals significantly increased in cell number after exposure to strain compared with static controls, except ERα−/− osteoblasts, which did not increase in cell number after strain (n = 4; all ERKO cells were female; two BERKO experiments were with male cells).

ERα−/− osteoblast-like cells exhibit increased proliferation in response to FGF-2, but not IGFs

IGFs have been suggested to mediate bone cell proliferation in response to strain.(23,66) Consistent with this, osteoblast-like cells derived from ERα+/+ mice showed increased cell number when incubated for 24 h with either 40 ng/ml truncated IGF-I or IGF-II (Fig. 2). In contrast, osteoblast-like cells derived from ERα−/− mice showed no increase in number in response to exogenous IGF-II and only a small, but not statistically significant, increase in response to IGF-I.

Figure FIG. 2..

Proliferation of primary ERKO-derived osteoblasts in response to exogenous IGF-I, IGF-II, and FGF-2. ERα+/+and ERα−/− osteoblasts in DMEM + 10% charcoal dextran-stripped FCS were exposed to 40 ng/ml IGF-I, IGF-II, or vehicle (HCl) control or 2 ng/ml FGF-2 or vehicle for 24 h. The cells were fixed, and the nuclei were stained and counted as described in the Materials and Methods section. The percentage change in cell number after exposure to the mitogens compared with vehicle is plotted. Error bars represent SEM, with the following p values:a0.01,b0.001, andc0001. ERα−/− osteoblasts failed to proliferate in response to exogenous IGFs but were capable of proliferation similar to WT controls when exposed to FGF-2 (n = 3; two females and one male).

Despite this relative insensitivity of ERα−/− osteoblast-like cells to IGFs, both ERα−/− and ERα+/+ osteoblast-like cells showed a robust and equivalent increase in cell number in response to exogenous FGF-2 (Fig. 2).

All cells derived from BERKO mice, regardless of genotype (ERβ+/+, ERβ+/−, ERβ−/−), showed significant proliferative responses to 40 ng/ml IGF-I and 40 ng/ml IGF-II (data not shown).

ERα−/− osteoblast-like cells produce prostaglandins, but not NO, in response to strain

Osteoblast-like cells release both NO(17,19) and prostaglandins(5,6,18–20) after exposure to strain. Blockade of the production of either NO(67) or prostaglandins(68–71) in vivo have been shown to abrogate the osteogenic response to loading.

In this study, media was collected 10 minutes after the end of exposure to mechanical strain and assayed for nitrite as an indicator of NO release. While osteoblast-like cells derived from ERα+/+ mice increased their medium concentration of nitrite in response to strain, there was no such increase in the case of osteoblast-like cells derived from ERα−/− mice (Fig. 3A). Interestingly, the medium conditioned by ERα−/− osteoblast-like cells contained higher basal levels of nitrite than conditioned media from ERα+/+ cells. Exposure to strain failed to increase this further.

Figure FIG. 3..

The concentration of nitrite and prostacyclin in media from ERKO-derived cells exposed to mechanical strain. Confluent cells in DMEM + 10% charcoal dextran-stripped FCS were exposed to a single 10-minute period of mechanical strain. (A) Media was collected 10 minutes after the end of the strain stimulus, and the nitrite concentration was determined as described in the Materials and Methods section. The mean concentration from three independent experiments (all female) was plotted, and the SEM is represented by error bars. The nitrite concentration is indicative of the amount of NO released by the cells.aThe concentration of nitrite was increased significantly in media derived from ERα+/+ cells; p = 0.03. In ERα−/− cells, the basal levels of nitrite were higher but were reduced after exposure to strain. (B) Media was collected 24 h after the end of the strain stimulus, and the concentration of 6-keto-PGF1α (the stable hydrolytic product of PGI2) was determined as described in the Materials and Methods section. The mean concentration from three independent experiments (all female) was plotted, and the SEM is represented by error bars. The 6-keto-PGF1α concentration is indicative of the amount of prostacyclin released by the cells. These increases were not statistically significant.

The concentration of 6-keto-PGF1α, the stable metabolite of PGI2, was measured in conditioned media from cells 24 h after exposure to a single period of mechanical strain. We found that the medium concentration of 6-keto-PGF1α was increased in response to the application of strain in media conditioned by either ERα+/+ or ERα−/− osteoblast-like cells (Fig. 3B).

ERα−/− osteoblast-like cells do not exhibit activation of an AP-1 reporter construct in response to strain

The activity of an AP-1 driven reporter vector (measured as luciferase activity normalized for β-galactosidase activity) was increased 8 h after exposure to mechanical strain in osteoblast-like cells derived from ERα+/+ mice (Fig. 4). This increase was not observed in osteoblast-like cells derived from ERα−/− mice. However, the basal level of activation of the AP-1 Luc reporter vector was similar in both genotypes.

Figure FIG. 4..

Strain-related AP-1 activity in ERKO-derived osteoblasts. ERα+/+ and ERα−/− osteoblasts were transfected with 1 μg AP-1 reporter vector and 0.5 μg lac Z control vector per strip, as described in the Materials and Methods section. Cells were exposed to a single 10-minute period of strain 24 h after transfection and harvested 8 h after the end of the strain stimulus. Activity of the luciferase reporter gene was normalized for β-galactosidase activity. Strain-related increases in AP-1 activity were reduced in ERα−/− compared with ERα+/+ cells (n = 3, all female), but basal levels of activation were similar. These strain-related increases in luciferase activity were not statistically significant.

Transfection with fully functional, but not mutant, ERα confers strain-related proliferation on ERα−/− cells

ERα−/− osteoblast-like cells transfected with fully functional ERα (pRST7-ERα) or with mutant ERα dysfunctional in one or more domains (pRST7-AF-2, L-543A/L-544A) were strained, and cell nuclei counted as described previously. Transfection of fully functional ERα (pRST7-ER) into these cells conferred a proliferative response to strain (Fig. 5), whereas transfection of the AF-2 mutant ERα did not. Transfection of the L-543A/L-544A mutant (in which the helix 12 region is misaligned) produced a small, but not statistically significant increase in cell number following strain compared with the ERα−/− controls. The transfection efficiency is these experiments was measured by histochemical detection of β-galactosidase activity and determined to be 20 ± 2%.

Figure FIG. 5..

The effect of ERα transfection on strain-related proliferation in ERKO-derived homozygous osteoblasts. ERα−/− osteoblasts in DMEM + 10% FCS were transfected with 1 μg ERα expression vector per strip, as described in the Materials and Methods section. The transfection efficiency was determined to be 20 ± 2%. Cells were exposed to a single 10-minute period of strain in DMEM + 10% charcoal dextran-stripped FCS 24 h after transfection. The cells were fixed 24 h after the end of the strain stimulus, and the nuclei were stained and counted as described in the Materials and Methods section. The percentage change in cell number after exposure to strain compared with static controls is plotted. Error bars represent SEM, with p values as follows:a0.05,b0.004, andc0.02. Transfection of the full-length ERα rescued the strain-related proliferation in ERα−/− cells (n = 5, all female).


The aim of this study was to determine which aspects of ER activity are involved in bone cells' responses to mechanical stimulation. To do this, we examined the responses to mechanical strain and IGFs of primary cultures of long bone-derived osteoblast-like cells from mice with different complements of functional and dysfunctional ERs.

In vivo a consistent and repeatable response to mechanical loading is periosteal and endosteal new bone formation.(66–70) An early feature of this response is proliferation of the cells in the osteogenic layers of both periosteum and endosteum.(69) Without such proliferation, any osteogenic response to load-bearing would be confined to the new bone formation produced by the osteoblasts already active on the bone surface. An inference from the observation that ER antagonists block strain-related proliferation in vitro(24,25) is that strain-related proliferation of osteoblast-like cells requires a functional ER. The data presented here support that inference.

Our current experiments confirm that, when exposed to a short period of dynamic strain, within the physiological strain range, osteoblast-like cells that possess functional ERα (ERα+/+ or ERα+/−) proliferate. In cells where ERα is dysfunctional (ERα−/−) exposure to strain does not stimulate proliferation. This lack of responsiveness can be rescued by transfection of the gene for intact ERα. Because ERα−/− cells contain functional ERβ and do not proliferate with strain, and ERβ−/− cells contain no ERβ but do proliferate with strain, it seems that ERβ is neither normally involved in the proliferative response to strain nor can it substitute for absent or dysfunctional ERα. The finding that strain-related proliferation in cells lacking ERβ is greater than cells with a full complement of ERβ perhaps suggests a negative influence of ERβ on this response. However, we cannot exclude the possibility that the expression of ERα is altered in these cells, which may also account for the increased response.

In addition to their lack of proliferation in response to strain, osteoblast-like cells derived from ERα−/− mice also fail to proliferate in response to exogenous IGF-I and IGF-II. This contrasts with the response of cells from ERα+/+ and ERα+/− littermates. We have shown previously that ROS 17/2.8 osteoblast-like cells show an increase in proliferation in response to estradiol, which can be blocked by neutralizing antibodies to IGF-I, and strain, which can blocked by neutralizing antibodies to IGF-II.(23,72) That osteoblast-like cells from ERα−/− do not proliferate in response to either strain or exogenous IGFs implies that ERα, and not ERβ, is involved in the mitogenic response to both these stimuli. The cell counting assay used in this study does not exclude the possibility that some of the effects we have attributed to proliferation may be contributed to by changes in cell viability.

A number of ERα isoforms have been identified recently, including a 46-kDa truncated ERα isoform expressed in a number of cell types(73,74) and a 55-kDa truncated isoform, resulting specifically from the mutated ERα gene in ERKO-derived cells.(75,76) Both of these isoforms have a truncated A/B domain lacking a functional AF-1 region, but still containing AF-2 and ligand binding domains. The importance of having a full-length, fully functional ERα to mediate bone cells' proliferative response to strain is not only shown by the inability of the truncated forms present in ERKO homozygotes to do so but also by the failure of mutated ERα to rescue this response after transfection. Introduction of the AF-2 mutant, which has a functional AF-1 domain, had no effect on strain-related proliferation, suggesting that a functional AF-1 domain is not sufficient to mediate strain-related proliferation any more than the endogenous AF-2 region of the truncated isoforms. Thus, it is likely that ERα requires both AF-1 and AF-2 regions to mediate a proliferative response to strain.

The L-543A/L-544A mutant has both functional AF-1 and AF-2 regions. Its introduction supported a small increase in strain-related proliferation in ERα−/− osteoblast-like cells, but this was not statistically significant. This mutant contains a misaligned helix 12 region that is known to be involved in receptor interaction with co-activators.(77) Consequently, it seems that strain-related proliferation of osteoblast-like cells requires the presence of both functional AF-1 and AF-2 regions of ERα and, to a lesser extent, an appropriately aligned helix 12 region. However, caution should be taken in comparing the effects of different mutants in case of variation in expression levels. We have shown previously that tamoxifen blocks strain-related proliferation in primary rat long bone-derived osteoblasts.(24) This inhibition would be consistent with tamoxifen blocking AF-2 activity by disruption of helix 12 conformation.(31,33,77,78) This raises an important issue regarding the possible involvement of a ligand in the use of ERα by strain. Tamoxifen induces a receptor conformation that is distinct from that of estrogen.(79) Preliminary results comparing the peptide recruitment of ERα in response to strain and estrogen suggest a distinct ERα conformation induced by strain (D. Ong, personal communication, February 6, 2003). It is likely that the inhibitory effects of tamoxifen observed previously(24) relate to the production of a receptor conformation that inhibits its use by strain rather than inhibition by competition for binding with an endogenous ligand.

It has been reported that female ERKO-derived endothelial cells from the aorta still produce NO in response to estrogen.(80) However, in cerebral arteries of female ERKO mice estrogen-related increases in eNOS and COX-1 were reported to be absent.(75) In this study, ERKO-derived bone cells failed to produce NO in response to mechanical strain, but spontaneous NO release from these cells was enhanced. It is possible that the failure to increase NO production in response to strain is a direct consequence of these high basal levels. The high basal levels may reflect aberrant downstream signaling in ERα−/− osteoblast-like cells, resulting in the loss of negative feedback regulation of NOS activity. The inhibition of NOS in vivo blocks loading-related bone formation(81,82) and the application of NO donors can prevent ovariectomy-induced bone loss,(83) but the signaling cascades that mediate responsiveness to NO are not well-defined in bone. With respect to this study, the significance of the loss of strain-related increases in NO production to the absence of strain-related and IGF-related proliferation has yet to be determined. This is complicated by the observation that, because of the high basal level of NO in ERα−/− osteoblast-like cells, the poststrain levels of nitrite in conditioned media from ERα−/− and ERα+/+ osteoblast-like cells are actually similar although the downstream events differ. This suggests ERα−/− osteoblast-like cells may be relatively insensitive to NO, whether basal or induced.

A number of in vitro studies have shown that bone cells produce prostaglandins in response to mechanical stimulation by strain and/or fluid flow.(5,6,18–20) In vivo blockade of prostaglandin production prevents the osteogenic response to load-bearing.(68–70) Exogenous prostaglandins have been shown to increase the activity of an ERE-driven reporter in osteoblast-like cells,(26) suggesting that the strain-related production of prostaglandins is upstream of ERα-dependent transcription. The current observation that ERα−/− osteoblast-like cells retain strain-related increases in prostaglandin production supports this.

The lack of responsiveness, in terms of strain-related proliferation, may relate to the reduced AP-1 response observed in ERα−/− osteoblast-like cells. ERα is known to function at AP-1 binding sites in the promoter of some target genes, including IGFs,(45,84) and functional ERα AF-1 and AF-2 regions seem to mediate this interaction.(85) The mutated ERα isoforms present in the nucleus of ERα−/− osteoblast-like cells may be unable to carry out this function effectively. This may result in a significant reduction in the production of IGF-II in response to strain. Loss of IGF-II production may explain the lack of strain-related cell proliferation observed in ERα−/− osteoblast-like cells.

The inability of ERα−/− cells to proliferate in response to IGFs is not confined to osteoblast-like cells because uterine cells are similarly unresponsive.(86) Kahlert et al.(87) have demonstrated a physical interaction between ERα and the IGF-I receptor in response to estrogen, resulting in IGF-IR phosphorylation. This interaction is dependent on ERα phosphorylation mediated by ERK. This ERK-mediated phosphorylation occurs in the N-terminal region of ERα, which is truncated in the isoforms expressed in ERKO-derived homozygous cells. Absence of the physical interaction described by Kahlert et al.(87) in ERα−/− cells offers a possible explanation for the lack of proliferation in response to exogenous IGFs.

In summary, our findings support the implication of previous studies that a reduction in ERα expression or activity, or even an alteration in the ERα isoforms present, can abrogate many of osteoblast-like cells' responses to mechanical strain. The responses affected include proliferation, responsiveness to IGFs, NO production, and the activation of ERK and AP-1. Prostaglandin production seems unaffected. Taking into consideration the high levels of estrogen produced by the ERKO mouse, it is evident that it is the activity of the ER and not the presence of estrogen itself that is the determining factor in the ability of bone cells to respond to strain. This inference is supported by the phenotype of a man with an ER null mutation and elevated estrogen levels(88) and the findings of Braidman et al.,(89) who have demonstrated a reduction in ERα expression in young men with idiopathic osteoporosis who are nonetheless estrogen replete. The loss of estrogen after menopause (or ovariectomy) has been shown to result in a decrease in protein and mRNA levels of ERα(29,30) and in ERα activity.(28) This situation, in vivo, would lead to reduced responsiveness to mechanical loading, which, for cells downstream of this response, would be the same as disuse. We hypothesize that the characteristic, functionally inappropriate, bone loss at the menopause reflects the skeleton's remodeling response to apparent disuse that arises at least in part from less effective processing of strain-related information by ERα.


We are grateful to Ken Korach for permission to use the ERKO mouse, to Donald McDonnell and Roger White for providing ER mutant expression vectors and to A. Ward for the AP-1 reporter vector. We also thank the Wellcome Trust & the BBSRC for funding the work.