Bisphosphonates and Estrogens Inhibit Osteocyte Apoptosis via Distinct Molecular Mechanisms Downstream of Extracellular Signal-regulated Kinase Activation*

Both estrogens and bisphosphonates attenuate osteocyte apoptosis by activating the extracellular signal-regulated kinases (ERKs). However, whereas estrogens activate ERKs via an extranuclear function of the estrogen receptor, bisphosphonates do so by opening connexin 43 hemichannels. Here, we demonstrated that the signaling events downstream of ERKs induced by these two stimuli are also distinct. Inhibition of osteocyte apoptosis by estrogens requires nuclear accumulation of ERKs and activation of downstream transcription factors. On the other hand, anti-apoptosis induced by bisphosphonates requires neither transcription nor ERK-dependent transcription factors. Instead, the effect of bisphosphonates is abolished when ERKs are restricted to the nucleus by blocking CRM1/exportin1-mediated nuclear protein export or by expressing nuclear-anchored ERKs, but it is unaffected in cells expressing cytoplasmic-anchored ERKs. Connexin 43/ERK-mediated anti-apoptosis induced by bisphosphonates requires the kinase activity of the cytoplasmic target of ERKs, p90RSK, which in turn phosphorylates the pro-apoptotic protein BAD and C/EBPβ. Phosphorylation of BAD renders it inactive, whereas phosphorylation of C/EBPβ leads to binding of pro-caspases, thus inhibiting apoptosis independently of the transcriptional activity of this transcription factor. Consistent with the evidence that estrogens and bisphosphonates phosphorylate diverse targets of ERKs, probably resulting from activation of spatially distinct pools of these kinases, the two agents had additive effects on osteocyte survival.

Previous work from our group (1,2) demonstrates that bisphosphonates prevent apoptosis of osteocytes and other cells of the osteoblastic lineage via the opening of hemichannels formed by connexin (Cx) 1 43 leading to activation of Src and the extracellular signal-regulated kinases (ERKs). Remarkably, Cx43 (but not other connexins) confers de novo responsiveness to these agents to connexin-naïve cells. The ability of Cx43 to transduce pro-survival signals requires both the pore-forming domain of Cx43 and the C-terminal portion of the protein, which is physically associated with the kinase Src (3,4). Src activation and its interaction with Cx43 are indispensable for the anti-apoptotic effects of bisphosphonates (2).
Similar to bisphosphonates, ligand-induced activation of sex steroid receptors also attenuates osteocyte and osteoblast apoptosis by a mechanism that requires activation of ERKs and Src. However, the molecular events that lead to activation of the Src/ERK pathway by estrogens or androgens are different from the ones involved in bisphosphonate action. Thus, either class of sex steroids activate ERKs via an extranuclear function of their receptors that can be mediated by the ligand binding domain of the proteins (5). In addition, kinase-dependent activation of transcription factors and gene transcription is required for the anti-apoptotic effect of sex steroids on osteoblastic cells (6).
Heretofore, the mechanism by which bisphosphonates prevent apoptosis downstream of ERKs, however, has remained unknown. We report that, unlike the anti-apoptotic effect of ERK activation induced by sex steroids, anti-apoptotic effects triggered by the Cx43/Src/ERK pathway do not require gene transcription or ERK-dependent transcription factors. Instead, they depend on the cytoplasmic localization of ERKs and the phosphorylation of the ERK-activated kinase p90 RSK and its cytoplasmic targets BAD and C/EBP␤. Moreover, and consistent with the evidence that estrogens and bisphosphonates phosphorylate diverse targets of ERKs, the two agents had additive effects on osteocyte survival, supporting the notion that they activate spatially distinct pools of these kinases.
Cell Culture-MLO-Y4 osteocytic cells derived from murine long bones were cultured as described previously (1,20). Embryonic fibroblasts isolated from wild type (Cx43 ϩ/ϩ ) or from Cx43-deficient mice (Cx43 Ϫ/Ϫ ) were provided by Dr. A. F. Lau (University of Hawaii at Manoa, HI) and cultured as described previously (21). Subcellular Localization of ERK-MLO-Y4 cells were transiently transfected with wtGFP-ERK2 to allow the visualization of ERK and wtMEK to anchor inactive ERK2 in the cytoplasm along with nRFP to allow the localization of cell nuclei. Following transfection, the cells were cultured in growth medium for 40 h followed by incubation with ␣-MEM containing 2% bovine serum albumin for 3.5 h. Subsequently, ethanol or 1 ng/ml leptomycin were added together with 10 Ϫ7 M alendronate or 10 Ϫ7 M 17␤-estradiol for 30 min. For the time-course experiment, cells were fixed in neutral buffer formalin after incubation with the stimuli for 2 min to 4 h. The percentage of cells showing nuclear accumulation of GFP-ERK2 was quantified by enumerating cells exhibiting increased GFP in the nucleus compared with the cytoplasm, using a fluorescence microscope. At least 250 cells from fields selected by systematic random sampling were examined for each experimental condition.
Quantification of Apoptotic Cells-Apoptotic cells were quantified either by trypan blue uptake or by evaluating the nuclear morphology of cells transfected with nGFP or nRFP, as described previously (1). Cells exhibiting chromatin condensation and/or nuclear fragmentation The increase in apoptotic cells induced by etoposide in the absence of any anti-apoptotic stimuli was designated as 100%. Actinomycin D and cycloheximide reduced the incorporation of [ 3 H]uridine into RNA to 9.6 Ϯ 5.1% and the incorporation of [ 3 H]leucine into protein to 21.4 Ϯ 0.8% of control cultures, respectively. Bars represent mean Ϯ S.D. of triplicate determinations. *, p Ͻ 0.05 versus vehicle-treated cultures. b, cells were transiently transfected with dominant negative forms of the indicated transcription factors together with nGFP. Forty hours after transfection, cells were treated as in a, and apoptosis was quantified by evaluating the nuclear morphology of transfected (fluorescent) cells, as described under "Experimental Procedures." *, p Ͻ 0.05 versus vehicle-treated cultures for each construct. c-e, nuclear accumulation of GFP-ERK2 was evaluated as described under "Experimental Procedures." Cells were treated with alendronate or E 2 for the indicated times (c), with alendronate or E 2 for 5 min or leptomycin B for 30 min (d), or with alendronate or E 2 alone or in combination with leptomycin B for 30 min (e). *, p Ͻ 0.05 versus vehicle-treated cultures for each time point (c). f, cells transiently transfected with Elk-1/GAL4-Luc/Renilla luciferase constructs were treated with vehicle (Ϫ), alendronate or E 2 , alone or in combination with leptomycin B for 24 h. * and # indicate p Ͻ 0.05 versus vehicle-or leptomycin B-treated cultures, respectively (e and f).
were considered apoptotic. Data are presented as percentage of etoposideinduced apoptosis in the absence of alendronate or 17␤-estradiol. The percentage of apoptosis was calculated using the formula (% DC eϩs Ϫ % DC s )/(% DC e Ϫ % DC v ) ϫ 100, where DC ϭ dead cells, e ϭ etoposidetreated cultures, s ϭ alendronate-or 17␤-estradiol-treated cultures, and v ϭ vehicle-treated cultures, as published previously (2,5).
Inhibition of RNA or Protein Synthesis-MLO-Y4 cells were incubated with 1 Ci/ml [ 3 H]uridine or [ 3 H]leucine for 2.5 h followed by the addition of 2 ϫ 10 Ϫ6 M actinomycin D or 10 Ϫ6 M cycloheximide and incubation for 7.5 h. The cells were scraped off the plates, protein and RNA were precipitated with ice-cold 5% trichloroacetic acid, and the radioactivity was quantified. Protein concentration in the precipitates was determined using a detergent-compatible Bio-Rad kit. [ 3 H]Uridine or [ 3 H]leucine incorporation was expressed as counts/min/g protein.
Reporter Assay-Cells were transiently transfected with a construct containing amino acids 307-428 of the C-terminal domain of Elk-1 and the DNA binding domain of GAL4 fused to the firefly luciferase reporter (GAL4-Luc) together with the Renilla luciferase plasmid pRL-SV40. Forty hours after transfection, the cells were treated with vehicle or 1 ng/ml leptomycin B alone or together with 10 Ϫ7 M alendronate or 10 Ϫ7 M 17␤-estradiol for 24 h in ␣-MEM containing 0.2% serum. Cell lysates were obtained and luciferase activity was measured using a dual luciferase kit (Promega, Madison, WI) according to the manufacturer's instructions. Elk-1 promoter activity was normalized to Renilla activity, as published previously (6).
Western Blot Analysis-Cells maintained in the absence of serum were treated with vehicle, 10 Ϫ7 M alendronate or 10 Ϫ7 M 17␤-estradiol for the indicated times. Proteins in the cell lysates were separated on 10% SDS-polyacrylamide gels, electrotransferred to polyvinylidene difluoride membranes, and immunoblotted with rabbit anti-phospho-Thr 217 -C/EBP␤, anti-total C/EBP␤, anti-HA-tagged antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) or anti-phospho-Ser 112 -BAD antibody (Cell Signaling Technology, Beverly, MA), or mouse monoclonal anti-␤-actin antibody (Sigma). Blots were developed by chemiluminescence (Pierce), and the intensity of the bands was quantified using the Versadoc Imaging system (Bio-Rad).
In-gel Kinase Assay-p90 RSK in-gel kinase activity was determined using an assay described previously (22), with modifications. Briefly, MLO-Y4 cells cultured in 2% bovine serum albumin ␣-MEM for 4 h were treated with 10 Ϫ7 M alendronate or 10 Ϫ7 17␤-estradiol for the last 5 or 30 min of the incubation, respectively. Fifty g of protein were separated on 7.5% SDS-polyacrylamide gels containing 0.05 mg/ml of the p90 RSK synthetic substrate peptide (Arg-Arg-Arg-Leu-Ser-Ser-Leu-Arg-Ala) corresponding to amino acids 231-239 of human 40 S ribosomal protein S6 (Upstate Biotechnology, Lake Placid, NY) (23). The gel was then incubated with 40 mM HEPES, pH 8, 2 mM dithiothreitol, 0.5 mM EGTA, 10 mM Mg 2 Cl, 40 M ATP, and 25 Ci [␥-32 P]ATP for 1 h at 25°C and then washed and dried. Radioactivity was detected using the Versadoc Imaging system (Bio-Rad). A parallel 10% SDS-polyacrylamide gel was run, and the levels of ␤-actin were determined by West-FIG. 2. Restriction of ERKs to the nucleus abolishes the anti-apoptotic effect of alendronate. a, MLO-Y4 cells transiently transfected with a GFP-ERK2 fusion protein along with nRFP were treated with vehicle or 1 ng/ml leptomycin B for 30 min followed by the addition of 10 Ϫ7 M alendronate (A) for 30 min. Subsequently, vehicle or 50 M etoposide (etop) were added. Apoptosis was quantified 6 h later by examining nuclear morphology. Arrows point to apoptotic cells. A similar experiment was performed in untransfected cells, and apoptosis was quantified by trypan blue uptake. *, p Ͻ 0.05 versus vehicle-treated cultures. b, MLO-Y4 cells were transiently transfected with wild type GFP-ERK2 or with GFP-ERK2 mutants that localize in the nucleus (nuc) or in the cytoplasm (cytop), independent of their activation status, along with nRFP. Apoptosis was assessed by examining nuclear morphology. ern blotting. The levels of phosphorylated p90 RSK substrate were normalized to the levels of ␤-actin in the same sample.
Image Acquisition-Fluorescent images were collected on an inverted microscope (Axiovert 200, Carl Zeiss Light Microscopy, Gottingen, Germany) with a LD A-Plan, 32ϫ/0.40 lens and a low light camera (Polaroid DMC Ie, Polaroid Corporation, Cambridge, MA) using a filter set for GFP and the Image-Pro Plus acquisition software (Media Cybernetics, Silver Spring, MD).
ERK Phosphorylation-ERK activation was determined using a fast activated cell enzyme-linked immunosorbent assay kit (FACE, Active Motif, Carlsbad, CA). Briefly, cells cultured in ␣-MEM containing 2% serum for 24 h were incubated in ␣-MEM containing 2% bovine serum albumin for 4 h and then treated with alendronate, 17␤-estradiol, or both at 10 Ϫ9 or 10 Ϫ7 M for the last 2 min. The cells were then fixed and incubated with specific anti-phosphorylated or anti-total ERK antibodies followed by incubation with a secondary horseradish peroxidaseconjugated antibody. The levels of phosphorylated or total ERKs were quantified using a colorimetric readout and are expressed as the ratio between phosphorylated and total ERKs.
Statistical Analysis-Data were analyzed by one-way analysis of variance, and the Student-Neuman-Keuls method was used to estimate the level of significance of differences between means.

Inhibition of Apoptosis by Alendronate Does Not Require
Nuclear Functions of ERKs-ERKs promote anti-apoptosis by inducing changes in gene transcription or by modifying the functional activity of proteins, independent of any transcriptional effects (6,24,25). We, therefore, examined here whether transcription, at large, or ERK-dependent transcription factors, in particular, were a requirement for the anti-apoptotic effect of the bisphosphonate alendronate on MLO-Y4 osteocytic cells. We found that inhibition of RNA or protein synthesis with actinomycin D or cycloheximide, respectively, did not influence the protective effect of alendronate on etoposide-induced apoptosis (Fig.  1a). Likewise, dominant negative forms of Elk-1, CREB, Runx2/ Cbfa1, C/EBP␤/NF-IL6, or STAT3, transcription factors activated by ERKs (26 -30), did not alter alendronate-induced antiapoptosis (Fig. 1b). In contrast and consistent with earlier studies (6), the anti-apoptotic effect of 17␤-estradiol in the same cells required gene transcription and was abolished by dominant negative forms of Elk-1, CREB, and C/EBP␤ (Fig. 1, a and b). 3. The anti-apoptotic effect of alendronate requires the cytoplasmic substrate of ERKs p90 RSK and inactivation of BAD. a, p90 RSK activity of cells treated with alendronate (A) or 17␤-estradiol (E 2 ) for 5 or 30 min, respectively, was assessed by in-gel kinase assay, as detailed under "Experimental Procedures." An autoradiograph representative of three independent experiments is shown. Cells were transiently transfected with empty vector or wild type or a kinase-dead p90 RSK mutant along with nGFP. Apoptosis was assessed by examining nuclear morphology. b, BAD phosphorylation in Ser 112 was determined in lysates of cells transfected with wild type HA-BAD and treated with 10 Ϫ7 M alendronate or E 2 for 5 and 30 min, respectively. A Western blot representative of three independent experiments is shown. Cells were transiently transfected with expression vectors for wild type HA-BAD or HA-BAD mutants lacking the phosphorylation site in Ser 112 , Ser 136 , or Ser 155 along with nGFP. Apoptosis was assessed by examining nuclear morphology. *, p Ͻ 0.05 versus vehicle-treated cultures for each construct; 3S/A, S112A/S136A/S155A.
Unexpectedly for an agent that induces ERK phosphorylation, exposure of MLO-Y4 cells to alendronate for up to 4 h did not result in the accumulation of ERKs into the nucleus, as it was assessed using a GFP-ERK2 fusion protein (Fig. 1, c and  d). Once again, this biologic profile was different to that of 17␤-estradiol, which caused nuclear accumulation of GFP-ERK2. Indeed, this effect of 17␤-estradiol was manifested as early as 2 min, reached a plateau after 30 min, and returned to base line by 4 h. Consistent with the distinct pattern of ERK translocation into the nucleus in response to the two agents, alendronate failed to induce a significant increase in Elk-1mediated transcription (Fig. 1f). On the other hand, 17␤-estradiol did have a potent effect on Elk-1-driven transcription.
To determine whether the inability of alendronate to induce nuclear ERK accumulation or ERK-dependent transcription results from either lack of nuclear translocation of the active kinases or from an accelerated nuclear export of ERKs, we investigated the effect of leptomycin B, a specific inhibitor of the nuclear protein exporter CRM1/exportin1 (31) in this phenomenon. Leptomycin B did not affect GFP-ERK2 nuclear accumulation or Elk-1 reporter activity in response to alen-dronate (as well as 17␤-estradiol), suggesting that the failure of alendronate to induce ERK nuclear accumulation does not result from increased nuclear export of ERKs (Fig. 1, e and f).
That nuclear accumulation of ERKs was not a requirement for the anti-apoptotic effect of bisphosphonates was subsequently documented by restricting proteins, in general, into the nuclear compartment, using leptomycin B, or by specific restriction of ERKs into the nuclear compartment, using a GFP-ERK2 mutant with impaired association to MEK (nuclear GFP-ERK2) (17). Either experimental manipulation completely abolished the anti-apoptotic effect of alendronate (Fig. 2).
The Anti-apoptotic Effect of Alendronate Requires p90 RSK and Its Target, BAD-Activated p90 RSK is a downstream effector of nuclear or cytoplasmic events initiated by ERK phosphorylation (32,33). However, p90 RSK is not a universal requirement for the pro-survival effects of ERKs (24,34). To establish whether p90 RSK is involved in the anti-apoptotic effect of alendronate, we searched for and found that alendronate and also 17␤-estradiol increased p90 RSK kinase activity (Fig. 3a). Consistent with this finding, cells expressing a catalytically inactive dominant negative p90 RSK (kinase-dead) failed to respond FIG. 4. Alendronate-induced activation of ERK/p90 RSK /C/EBP␤ is mediated via connexin43 hemichannels. a, the levels of C/EBP␤ phosphorylation in Thr 217 were determined in lysates of cells treated with 10 Ϫ7 M alendronate (A) or 17␤-estradiol (E 2 ) for 5 and 10 min, respectively. A Western blot representative of three independent experiments is shown. Cells were transiently transfected with expression vectors for wild type C/EBP␤, a C/EBP␤ mutant lacking transcriptional activity, or a C/EBP␤ mutant lacking the phosphorylation site in Thr 217 along with nGFP. Apoptosis was assessed by examining nuclear morphology. b, MLO-Y4 cells were transiently transfected with empty vector or with Cx43 mutants with impaired permeability (Cx43⌬130) or lacking the C-terminal tail (Cx43⌬245), or with dominant negative kinase-dead (K Ϫ ) MEK or p90 RSK . Cells were treated with alendronate for 5 min and lysates were prepared. C/EBP␤ phosphorylation in Thr 217 was analyzed by Western blotting. Results shown are representative of three independent experiments. c, MLO-Y4 cells were transiently transfected with empty vector or with expression vectors for Cx43⌬130 or Cx43⌬245 along with nGFP. Apoptosis was assessed by examining nuclear morphology. d, embryonic fibroblasts derived from wild type or Cx43-deficient mice were cultured in serum-free ␣-MEM for 1 h and treated with vehicle or 10 Ϫ7 M alendronate for the last 5 min. Cell lysates were prepared, and C/EBP␤ phosphorylation in Thr 217 was analyzed. A Western blot representative of three independent experiments is shown. *, p Ͻ 0.05 versus vehicle-treated cultures.
to alendronate. As shown earlier in osteoblastic cells (6), the anti-apoptotic effect of 17␤-estradiol in MLO-Y4 osteocytic cells also required p90 RSK catalytic activity.
The pro-apoptotic protein BAD is one of several cytoplasmic substrates of the ERK/p90 RSK pathway. Specifically, phosphorylation of BAD by p90 RSK in Ser 112 renders BAD inactive and thereby promotes cell survival (24). As shown in Fig. 3b, alendronate induced BAD phosphorylation in Ser 112 . Moreover, cells expressing dominant negative BAD mutants in which Ser 112 is substituted by Ala (S112A/S136A/S155A or S112A) were refractory to the anti-apoptotic effect of alendronate (Fig.  3b). On the other hand, cells expressing BAD mutants, in which serines not involved in p90 RSK phosphorylation were mutated (S136A or S155A), were protected from etoposideinduced apoptosis by alendronate as effectively as cells expressing the wild type BAD. Remarkably, even though 17␤estradiol also induced BAD phosphorylation in Ser 112 , phosphorylation at this site was not essential for anti-apoptosis in response to the hormone, as cells expressing S112A BAD were still protected from etoposide-induced apoptosis by 17␤estradiol. In contrast, 17␤-estradiol was unable to protect from apoptosis cells expressing S136A or S115A mutants, indicating that phosphorylation of BAD in Ser 136 and Ser 155 was required for its effect (Fig. 3b). Hence, both alendronate and 17␤-estradiol inactivate BAD by phosphorylating this pro-apoptotic pro-tein. Although phosphorylation of Ser 112 (but not Ser 136 and Ser 155 ) is critical for the effect of alendronate, phosphorylation of Ser 136 and Ser 155 (but not Ser 112 ) are indispensable for the effect of 17␤-estradiol.
An Extranuclear, Transcription-independent Function of C/EBP␤ Is Required for the Anti-apoptotic Effect of the Cx43/ Src/ERK/p90 RSK Pathway-It has been demonstrated that p90 RSK -mediated phosphorylation of C/EBP␤ in Thr 217 promotes its binding to pro-caspases 1 and 8, thereby leading to cell survival. Caspase inactivation represents a C/EBP␤ action not predicted by its well known function as transcription factor, as it is indeed independent of transcription (14). Both alendronate and 17␤-estradiol increased phosphorylation of C/EBP␤ in Thr 217 (Fig. 4a). However, this phosphorylation event was only required for the anti-apoptotic effect of alendronate but not for that of 17␤-estradiol. Thus, cells expressing a C/EBP␤ mutant in which Thr 217 is replaced by Ala (rendering a caspase binding-deficient protein) were not protected from etoposide-induced apoptosis by alendronate. 17␤-Estradiol, however, was as effective in these cells as in cells expressing wild type C/EBP␤. Conversely, cells expressing a C/EBP␤ mutant lacking transcriptional activity due to its inability to bind DNA (8) were protected from apoptosis by alendronate (Figs. 1b  and 4a). On the other hand, 17␤-estradiol failed to induce survival in cells expressing this mutant (Fig. 4a).  Fig. 1c. b, cells transiently transfected with Elk-1/GAL4-Luc/Renilla luciferase constructs were treated with vehicle (Ϫ), 10 Ϫ7 M alendronate, E 2 , or both. The levels of Elk-1-driven luciferase and Renilla activity in the same cell lysates were measured as indicated in the legend to Fig. 1f. c, cells were treated with vehicle (Ϫ) or the indicated concentrations of alendronate, E 2 , or both for 2 min. Cells were fixed and the levels of phosphorylated and total ERKs were determined as indicated under "Experimental Procedures." *, p Ͻ 0.05 versus vehicle-treated cultures; #, p Ͻ 0.05 for 10 Ϫ7 M alendronate ϩ E 2 versus 10 Ϫ7 M alendronate or E 2 alone. d, MLO-Y4 cells were incubated for 1 h with the indicated concentrations of alendronate, E 2 , or both, followed by the addition of vehicle or 50 M etoposide and incubation for 6 h. Dead cells were enumerated by trypan blue uptake. *, p Ͻ 0.05 versus vehicle-treated cultures.
We next examined the ability of alendronate to phosphorylate C/EBP␤ in cells in which hemichannels or the ERK/p90 RSK pathway had been disrupted. Similar to the findings with untransfected MLO-Y4 cells shown in Fig. 4a, alendronate induced C/EBP␤ phosphorylation in Thr 217 in MLO-Y4 cells transfected with empty vector. However, alendronate was ineffective in cells transfected with a dominant negative form of Cx43 with impaired permeability (Cx43⌬130) (19) or in cells transfected with a Cx43 mutant that lacks the C-terminal domain and therefore lacks the ability to bind Src and activate the ERK pathway (Cx43⌬245) (18) (Fig. 4b). Transfection with these mutants also abolished the anti-apoptotic effect of alendronate (Fig. 4c). Moreover, alendronate did not increase C/EBP␤ phosphorylation in cells transfected with dominant negative forms of MEK or p90 RSK that lack kinase activity (Fig.  4b). Further, alendronate induced C/EBP␤ phosphorylation in embryonic fibroblasts derived from wild type mice, but it failed to do so in embryonic fibroblasts derived from Cx43-deficient mice (Fig. 4d). Altogether, these results confirm that C/EBP␤ phosphorylation in Thr 217 results from alendronate-induced opening of Cx43 hemichannels and the subsequent activation of the Src/ERK/p90 RSK pathway leading to cell survival.
Bisphosphonates and Estrogens Exhibit Additive Effects on ERK Activation and Anti-apoptosis-Because the lack of nuclear ERK accumulation by bisphosphonates is probably due to cytoplasmic retention of ERKs, we next investigated whether bisphosphonates interfere with estrogen-induced nuclear ERK accumulation and ERK-mediated activation of transcription factors. We found that the simultaneous addition of both alendronate and 17␤-estradiol induced similar increase in GFP-ERK2 nuclear accumulation or Elk-1-mediated transcription as compared with 17␤-estradiol alone (Fig. 5, a and b). Moreover, the combination of both agents at optimal concentrations (10 Ϫ7 M) resulted in higher levels of ERK phosphorylation than individual treatments (Fig. 5c). In addition, alendronate, 17␤estradiol, or a combination of both agents prevented etoposideinduced apoptosis (Fig. 5d). Furthermore, whereas treatment with suboptimal concentrations of either agent (10 Ϫ9 M) neither induced ERK phosphorylation nor prevented etoposide-induced apoptosis, simultaneous addition of both agents did (Fig. 5, c  and d). Taken together, these findings indicate not only that alendronate and 17␤-estradiol do not interfere with the antiapoptotic action of each other, but that, in fact, these agents exhibit additive effects on ERK activation and anti-apoptosis. DISCUSSION The studies reported herein demonstrate that, unlike most ERK-activating survival stimuli that induce changes in gene transcription, the ERK/p90 RSK anti-apoptotic signal cascade initiated by the opening of Cx43 hemichannels by bisphosphonates in osteocytic cells does not. Instead, in this case, ERK/ p90 RSK activation causes inactivation of the pro-apoptotic protein BAD and the creation of a functional caspase-inhibitory domain in C/EBP␤ (Fig. 6). Most remarkably, all of these events result from actions of ERKs exerted in the cytoplasm, and in fact, they are abolished when ERKs are experimentally restricted to the nucleus. In addition, we showed that bisphosphonate-induced ERK/p90 RSK activation modifies C/EBP␤ post-translationally. This modification leads to a novel conformation and the appearance of a domain known as "caspaseinhibitory box," which allows C/EBP␤ to bind to and inhibit the activation of pro-caspases (14). The cytoplasmic restricted actions of ERKs in the case of bisphosphonate-initiated activation are sharply different from the actions of ERKs required for the prevention of apoptosis, in response to another anti-apoptotic stimulus for osteocytes that also requires ERKs, namely the steroid hormone 17␤-estradiol. Indeed, whereas the creation of the "caspase-inhibitory box" of C/EBP␤ is dispensable for the anti-apoptotic effect of 17␤-estradiol, the function of C/EBP␤ as a transcription factor is essential. The elucidation of these mechanistic differences has evidently important pharmacotherapeutic implications, as we also show here that the antiapoptotic effects of these two agents on osteocytes are indeed additive.
Cytoplasmic Versus Nuclear ERK Functions-A variety of extracellular stimuli converge on the ERK cascade to elicit specific biological responses. Upon their activation in the cytoplasm by the kinase MEK, ERKs phosphorylate proteins on serine or threonine residues that are followed by proline residues (35). The p90 ribosomal S6 kinase (p90 RSK ) is one of the proteins phosphorylated by ERKs. p90 RSK in turn phosphorylates other cytoplasmic molecules or undergoes nuclear translocation resulting in transcription factor activation (32,36). Consistent with previous evidence implicating p90 RSK in ERKmediated cell survival (24,37), we found that ERK-dependent anti-apoptosis induced by bisphosphonates requires p90 RSK . Direct phosphorylation of the immediate early gene X-1 by ERKs, independently of p90 RSK , is required for the inhibition of apoptosis in other cell types (34). However, we found that alendronate still inhibits apoptosis in MLO-Y4 cells expressing an unphosphorylatable immediate early gene X-1 mutant (not shown). Therefore, taken together, our findings are consistent with an action of ERKs mediated by p90 RSK in the anti-apoptotic effect of bisphosphonates.
Several examples in the literature indicate that the subcellular localization of ERKs dictates the outcome of their activation. Thus, whereas nuclear accumulation of ERKs induces differentiation, cytoplasmic retention of the kinases leads to apoptosis of erythromyeloblast D2 cells (38). Likewise, constitutively active ERK2 localized in the nucleus prevents apoptosis induced by inhibition of the Bcr-Abl chimeric protein kinase FIG. 6. Proposed model for the survival effect of bisphosphonates in osteocytes. Based on the present findings, we proposed that the transient opening of Cx43 hemichannel and Src activation induced by bisphosphonates results in ERK activation without nuclear accumulation, followed by activation of the kinase p90 RSK and phosphorylation of the cytoplasmic substrates BAD and C/EBP␤, which are required for cell survival. TF, transcription factor. For simplicity, only one of the six molecules of Cx43 that form the hemichannel is depicted.
but not by serum starvation of chronic myelogenous leukemia cells; yet, constitutively active ERK2 localized in the cytoplasm inhibits apoptosis induced by the latter, but not the former, stimulus (39). Further, CD40 activation induces ERK nuclear accumulation and inhibition of apoptosis; on the other hand, B cell receptor activation results in ERK cytoplasmic accumulation and induction of apoptosis (40,41). In distinction from these paradigms, the findings of the present report demonstrate that two different ERK activators lead to the same outcome (inhibition of osteocyte apoptosis) by triggering cytoplasmic actions of ERKs (in the case of bisphosphonates) or nuclear functions of the kinases (in the case of 17␤-estradiol). Moreover, we show that the cytoplasmic retention of ERKs induced by bisphosphonates does not interfere with ERK nuclear accumulation induced by 17␤-estradiol. In fact, simultaneous treatment with both agents induces additive effects on ERK phosphorylation as well as anti-apoptosis (Fig. 5). This evidence strongly suggests that different subpopulations of ERKs are activated by each stimulus and that, when concurrently activated, they cooperate with each other.
It is unclear why ERKs phosphorylated by bisphosphonates do not accumulate in the nucleus and do not activate ERK-dependent gene transcription. The present results indicate that the lack of nuclear accumulation is not due to translocation into the nucleus followed by rapid export but rather to cytoplasmic retention of ERKs. This phenomenon could result from tethering of the kinases to the cytoskeleton (42,43) or from their sequestration by scaffolding proteins, such as ␤-arrestin (44,45) or Sef (46). All these mechanisms inhibit nuclear functions of ERKs without affecting their ability to phosphorylate cytoplasmic substrates. Future studies will be required to determine whether any of these anchoring mechanisms are responsible for the cytoplasmic retention of ERKs activated by the Cx43/Src pathway.
Conclusions and Significance-In conclusion, the findings reported in this study, together with our previous observations (1,2), reveal that bisphosphonates activate a previously unknown signaling pathway that is triggered by the opening of Cx43 hemichannels, followed by the activation of ERKs, exclusive phosphorylation of cytoplasmic substrates of ERKs (such as BAD and C/EBP␤), and BAD-and C/EBP␤-mediated osteocyte survival, independent of the transcriptional activity of the latter (Fig. 6). Recent evidence indicates that osteocyte viability might contribute to the maintenance of the mechanical competence of the skeleton, independent of bone mineral density (47,48), and that the effectiveness of bisphosphonates as well as other therapeutic agents employed in the treatment of metabolic bone diseases (e.g. estrogen, and even daily parathyroid hormone injections) may result, in part, from their ability to prevent apoptosis of osteocytes and osteoblasts (1, 2, 5, 49 -52). The demonstration in the present report of a sharp difference in the anti-apoptotic actions of bisphosphonates and 17␤-estradiol on osteocytes and that the effects of these two agents on apoptosis are additive in vitro provides a mechanistic explanation for the clinical observations that bisphosphonates and estrogens in combination may be more effective in protecting the skeleton than either treatment alone (53)(54)(55)(56).