Bisphosphonates Act Directly on the Osteoclast to Induce Caspase Cleavage of Mst1 Kinase during Apoptosis

Bisphosphonates (BPs) include potent inhibitors of bone resorption used to treat osteoporosis and other bone diseases. BPs directly or indirectly induce apoptosis in osteoclasts, the bone resorbing cells, and this may play a role in inhibition of bone resorption. Little is known about downstream mediators of apoptosis in osteoclasts, which are difficult to culture. Using purified osteoclasts, we examined the effects of alendronate, risedronate, pamidronate, etidronate, and clodronate on apoptosis and signaling kinases. All BPs induce caspase-dependent formation of pyknotic nuclei and cleavage of Mammalian Sterile 20-like (Mst) kinase 1 to form the active 34-kDa species associated with apoptosis. Withdrawal of serum and of macrophage colony stimulating factor, necessary for survival of purified osteoclasts, or treatment with staurosporine also induce apoptosis and caspase cleavage of Mst1. Consistent with their inhibition of the mevalonate pathway, apoptosis and cleavage of Mst1 kinase induced by alendronate, risedronate, and lovastatin, but not clodronate, are blocked by geranylgeraniol, a precursor of geranylgeranyl diphosphate. Together these findings suggest that BPs act directly on the osteoclast to induce apoptosis and that caspase cleavage of Mst1 kinase is part of the apoptotic pathway. For alendronate and risedronate, these events seem to be downstream of inhibition of geranylgeranylation.

Bisphosphonates (BPs) 1 include potent inhibitors of bone resorption used for the treatment of osteoporosis, Paget's disease, bone metastases, and other bone diseases. It is generally accepted that BPs inhibit bone resorption by acting directly or indirectly on osteoclasts, cells of hematopoietic origin. Until recently the molecular mechanism of action of BPs was not known (1,2). Recent pharmacological studies suggest that nitrogen-containing BPs (N-BPs), such as alendronate (ALN), risedronate (RIS), ibandronate, and pamidronate (PAM), act on the cholesterol biosynthesis pathway (3,4). N-BPs were recently shown to inhibit either isopentenyl diphosphate synthase or the downstream enzyme, farnesyl diphosphate (FPP) synthase, or both (5), while clodronate (CL2) had little or no effect on these enzymes. Inhibition of either enzyme in this pathway blocks cholesterol biosynthesis, as well as farnesylation and geranylgeranylation. We recently showed that only geranylgeraniol (GGOH) prevented the effects of ALN on bone resorption (pit formation) in vitro (4), suggesting that geranylgeranylation is rate-limiting for osteoclast function. CL2 and EHDP may inhibit osteoclasts by forming toxic ATP analogs or by inhibiting protein-tyrosine phosphatases (6 -9).
The final outcome of BP action appears to be osteoclast apoptosis, which was suggested to be the means by which BPs inhibit bone resorption (10,11). It is not known if the induction of apoptosis is the result of direct action of BPs on the osteoclast, and little is known about the biochemical pathways involved. Coxon et al. (12) have shown the induction of caspases in BP-treated J774 macrophages undergoing apoptosis. However, macrophages may not fully mimic osteoclast responses to bisphosphonates, since apoptosis in these cells is blocked by addition of either farnesyl diphosphate or geranylgeranyl diphosphate (3) and only the latter is implicated as rate-limiting for N-BP effects on the osteoclast (4). Recent evidence suggests that Mammalian Sterile 20-like (Mst) kinase 1 is a substrate for caspase 3 in several hematopoietic cells and can induce apoptosis in mesenchymal cell lines (13)(14)(15). The caspase cleavage site of Mst1 (DEMD), situated between the amino-terminal kinase domain and carboxyl-terminal regulatory and dimerization domains, matches the consensus sequence for caspase 3. After Mst1 cleavage, the kinase domain, separated from its regulatory elements, has high catalytic activity.
In this study we investigated the effect of several BPs on apoptosis and activation of signaling kinases in purified osteoclasts. BPs induced osteoclast apoptosis and activation of the 34-kDa Mst1 kinase by up to 13-fold, while reducing the activity of Mst1 (59 kDa) and Mst2 (60 kDa) kinases by up to 50%. Slight 34-kDa Mst2 kinase activity was also observed, and two other responsive kinases of 50 and 36 kDa remain unidentified. Induction of apoptosis by withdrawal of serum and macrophage colony-stimulating factor (M-CSF) elicited kinase responses similar to those produced by BPs, while staurosporine activated both full-length Mst and 34-kDa Mst1 kinase activities without altering the 50-kDa kinase activity. The findings indicate that BPs act directly on osteoclasts to induce caspase cleavage of Mst1, as a signaling intermediate in the induction of apoptosis, and suggest that geranylgeranylation is the upstream target of ALN and RIS in inducing apoptosis.

EXPERIMENTAL PROCEDURES
Osteoclast (Oc) Formation-Co-cultures of murine osteoblasts and marrow cells were prepared as described by Wesolowski et al. (16) with the following modifications. Bone marrow cells were harvested from 6-week-old male Balb/C mice by flushing the marrow spaces of freshly isolated long bones (tibiae and femora) with ␣-minimal essential medium (Life Technologies, Inc.) containing penicillin/streptomycin (100 IU and 100 g/ml, respectively). Bone marrow cells were suspended in Oc medium: ␣-minimal essential medium supplemented with fetal calf serum (10% v/v; HyClone Laboratories, Logan, UT) and 10 nM 1,25-(OH) 2 vitamin D 3 (Biomol, Plymouth Meeting, PA). Bone marrow cells were then added to subconfluent monolayers of osteoblastic MB 1.8 cells and cultured for 6 -7 days at 37°C in the presence of 5% CO 2 . Cocultures were first treated with type I collagenase (Wako Pure Chemical Industries, Osaka, Japan) at a concentration of 1 mg/ml in phosphate-buffered saline (PBS) for 1 h at 37°C. Suspended osteoblasts were gently aspirated, leaving a mixture enriched in prefusion Ocs and the remaining MB1.8 osteoblasts. All cells were released with EDTA (0.2 g/liter in PBS) for 20 min at 37°C and then re-plated in Oc medium and cultured for an additional 3 days.
Oc Apoptosis Assay-Oc-forming cultures, generated as above in 24 well plates, were treated with type 1 collagenase at a concentration of 1 mg/ml (in PBS) for 1 h at 37°C. Plates were washed twice with PBS and monitored under the microscope for complete removal of osteoblastic cells. Ocs were then maintained in Oc medium supplemented with M-CSF (R&D Systems, Minneapolis, MN) at 5 ng/ml. Ocs were treated with indicated compounds in the Oc medium for 18 h and stained for tartrate-resistant acid phosphatase (TRAP) and with Hoechst nuclear (no. 33342; Sigma) stain as follows. Cells were fixed with 10% formaldehyde (in PBS) and then rinsed with PBS. Cells were then stained with Fast Red Violet LB (Sigma) dissolved in TRAP buffer (sodium acetate (50 mM), sodium tartrate (30 mM), and Triton X-100 (0.1%), naphthol AS-MX phosphate (100 mg/ml), pH 5.0) for 10 min at 37°C. After TRAP staining, nuclear staining was performed with 5 g/ml Hoechst stain in PBS for 10 min at room temperature. Cells were washed with water and stored in the dark at 4°C. Total osteoclast number (i.e. TRAP (ϩ) multinucleated cells Ն250 m in diameter) in each well and the number of osteoclasts with pyknotic nuclei were quantitated microscopically using a Nikon Diaphot microscope equipped with a 10ϫ objective and an ultraviolet light source.
Immunoprecipitation-Anti-Krs1 (Mst2) amino-terminal and anti-Krs2 (Mst1) amino-terminal antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal anti-Krs1ϩ2 carboxylterminal antibody was from Zymed Laboratories Inc. (South San Francisco, CA) and anti-Krs1 carboxyl-terminal goat polyclonal antibody was from Santa Cruz Biotechnology. Protein lysates (50 g) were diluted in 1 ml of ice-cold ␤-HBS-IP containing Tween 20 (0.05%) and then combined with primary anti-amino-terminal (3 g) or anti-carboxyl-terminal (2 g of each) Mst1 and/or Mst2 antibodies. Rabbit anti-goat secondary antibody (1 l; ICN Pharmaceuticals, Aurora, OH) was added to increase binding to protein A-conjugated agarose beads (10 l bead volume; Sigma). Lysate-antibody-bead mixtures were gently mixed overnight at 4°C. Beads were then washed three times with ice-cold ␤-HBS-IP containing Tween 20 (0.05%) and the supplemental protease inhibitor mixture described above. Antibodies and antigen were released by suspending the beads in Laemmli sample buffer (Bio-Rad) containing 5% 2-mercaptoethanol and mixed vigorously at 37°C in a Thermomixer (Brinkmann, Westbury, NY).
In-gel Kinase Assay-Kinase assays were performed in the gel using methods based on Kameshita and Fujisawa (17) and Gotoh et al. (18). Briefly, 10% SDS-polyacrylamide gels were cast with myelin basic protein (0.2 mg/ml) added prior to polymerization. Lanes were loaded with equivalent amounts of protein lysate (5-10 g/lane) or the entirety of each immunoprecipitation, and then gels were electrophoresed. Gels were washed twice at room temperature with Buffer A: 50 mM HEPES, pH 7.6; 5 mM 2-mercaptoethanol supplemented with isopropanol (20%), followed by two washes with Buffer A. Proteins were denatured with urea (6 M in Buffer A) at room temperature and renatured at 4°C with urea at 3 M, 1.5 M, and then 0.75 M (in Buffer A), followed by three washes with Buffer A containing Tween-20 (0.05%). Gels were then washed twice with Kinase Buffer (20 mM HEPES, pH 7.6, 5 mM MnCl 2 , 2 mM dithiothreitol) at 30°C. Similar, but weaker kinase activities were obtained when 20 mM MgCl 2 was used instead of 5 mM MnCl 2 . Kinase reactions lasted 30 min at 30°C using Kinase Buffer containing 0.02 mM ATP with ϳ1000 cpm/pmol [␥-32 P]ATP. Kinase reactions were stopped and unincorporated [␥-32 P]ATP was removed with six washes using 5% trichloroacetic acid/1% NaPP i (w/v). Gels were then stained, destained, and dried using standard techniques. Gels were exposed to phosphorimaging screens and scanned and analyzed using a Molecular Dynamics Storm system and associated software (Sunnyvale, CA).

Alendronate and Risedronate Induction of Osteoclast Apoptosis Is Blocked by Geranylgeraniol and Caspase Inhibitors-
N-BPs were shown to disrupt the cytoskeleton, abolish the ruffled border, and induce apoptosis in osteoclasts and macrophages (3,10,19). We found that GGOH prevents alendronate inhibition of osteoclastic bone resorption (4), suggesting that geranylgeranyl diphosphate, an isoprenylation precursor derived from the mevalonate pathway or from GGOH, is ratelimiting for osteoclast activity. Geranylgeranylated G-proteins regulate cytoskeletal organization, vesicular trafficking, and apoptosis (20,21). Focusing on the latter, we examined the effects of GGOH and of the caspase inhibitor Z-VAD-FMK on ALN-, RIS-, CL2-, and EHDP-induced apoptosis in purified Ocs. Ocs were identified by TRAP staining as described under "Experimental Procedures," and apoptosis in these cells was scored based on the presence of pyknotic nuclei. This method correlated with TUNEL staining in this system (data not shown). Base-line apoptosis in untreated Ocs maintained for 18 h was 9%. Typically, nuclei were deployed in ringlike structures (Fig. 1A).
Treatment with ALN (30 M) for 18 h increased the number of cells with nuclear condensation (apoptosis) by 3-fold ( Fig. 1, B and E). RIS (30 M) produced similar effects (Fig. 1E). Both GGOH (Fig. 1, C and E) and Z-VAD-FMK (Fig. 1, D and E) prevented ALN-or RIS-induced apoptosis, reducing the number of cells with condensed nuclei to control levels. However, while GGOH prevented both nuclear condensation and disappearance of the ringlike structure in N-BP-treated Ocs, Z-VAD-FMK-treated cells showed normal nuclear morphology without the ringlike arrangement. Instead, these nuclei were clustered in one or more regions of the osteoclast. Apoptosis induced by CL2 and EHDP was also suppressed by Z-VAD-FMK ( Fig. 1E) but was not affected by GGOH. Apoptosis induced by the withdrawal of M-CSF was also prevented by Z-VAD-FMK (Fig. 1E).
Bisphosphonates Induce Kinase Signaling during BP-induced Apoptosis in Purified Osteoclasts-Several geranylgeranylated proteins, including members of the Ras family of small GTPases, activate signaling pathways and control apoptosis through their regulation of protein kinases. We assessed kinase activity in purified murine Oc-like cells obtained as described under "Experimental Procedures." In-gel kinase assays of Oc lysates (Fig. 2) showed that treatment with ALN (panel A),  Fig. 2A). PAM had no effect at 10 or 30 M, was slightly effective at 60 M (data not shown), while 100 M PAM (Fig. 2B) elicited about the same response as 30 M ALN ( Fig. 2A) or RIS (Fig. 2C). However, at 100 M, PAM-treated cultures also showed the accumulation of debris in the medium, which was not observed with ALN or RIS. Treatment with EHDP (300 M; Fig. 2E) for 2-24 h activated the 34-kDa kinase less than 2-fold and there was no effect on kinases migrating at 36, 50, 59, or 60 kDa. In many experiments, EHDP elicited no response at all (data not shown). Tiludronate up to 300 M did not significantly alter kinase activities (data not shown).
The Oc cultures used in these experiments were 95% pure by cell count, suggesting that the BPs act directly on the osteoclast. Since Ocs in these populations are large and multinucleated, they may contribute Ն99% of the protein and kinase signals in these assays. To examine the potential contribution of kinase signals from the minor population of contaminating osteoblasts, we examined kinase responses to ALN in pure MB1.8 osteoblast cultures (Fig. 2F). Untreated MB1.8 osteoblasts show a different kinase profile than Oc controls; the 60-kDa kinase in untreated MB1.8 cells appears to migrate as a single band, and the 50-kDa kinase is a more distinct doublet (panel E, lane 1). Additional kinase activity is observed at 43 kDa, possibly corresponding to ERK1, ERK2, or p38 kinases, while no activity was observed at 34 or 36 kDa. After treatment with ALN (30 M) for 2-24 h, no changes in kinase activity were observed, indicating that the responses to BPs described above occurred in the Ocs.
Identification of Mst1 Kinase as the Target of Bisphosphonate Action-To identify the kinases that respond to BPs in the osteoclast, we used antibodies against kinases with molecular masses in the observed range and assayed immunoprecipitates using in-gel kinase assay. The kinases activated by ALN (Fig.  3) and other BPs (data not shown) were only immunoprecipitated with antibodies against Mst1 or Mst2, also know as kinase responsive to stress (Krs) 2 and 1, respectively. Antibodies directed to the amino terminus of either Mst1 (Fig. 3, lanes 3 (control) and 4 (ALN)) or Mst2 (lanes 5 (control) and 6 (ALN)) immunoprecipitated both kinases in the 59-kDa (Mst1)/ 60-kDa (Mst2) doublet in Oc lysates of control and ALN-treated cells. Mst1:Mst1 dimerization has been described and involves a carboxyl-terminal domain that is conserved in Mst2 (22). Consistent with the formation of Mst1:Mst2 heterodimers, either type of antibody precipitated both Mst1 and Mst2. Anti-Mst1 immunoprecipitated Mst1 to a greater extent than Mst2, and vice versa. This is explained by assuming random association, where anti-Mst1 precipitates Mst1:Mst1, Mst1:Mst2, and Mst2:Mst1, but not Mst2:Mst2 (yielding a 2:1 ratio), while anti-Mst2 will give a similar ratio favoring Mst2. Although bands were not separated sufficiently for quantitation, the data are consistent with this ratio. Immunodepletion with a mixture of both anti-Mst1 and Mst2 antibodies removed Ͼ95% of all kinase activity migrating at 59/60 kDa, suggesting that this activity doublet was comprised primarily of these kinases (data not shown). Several studies have recently described an apoptosis response, associated with caspase cleavage of the 59-kDa Mst1 and Mst2 kinases to form catalytically active species migrating at 34 -36 kDa (13)(14)(15). Caspase cleavage of Mst1 occurs at the DEMD sequence (residues 323-326), while Mst2 is cleaved at DELD (residues 319 -322). The catalytically active 34-kDa kinase domain is readily immunoprecipitated using antibodies directed only to the Mst amino terminus (13). Anti-Mst1 aminoterminal antibody (Fig. 3, lanes 3 (control) and 4 (ALN)) immunoprecipitated the 34-kDa kinase activated by the BPs in the osteoclast, while anti-Mst2 was much less effective (lanes 5 and 6). This suggested that the 34-kDa kinase activated by BPs was predominantly the amino-terminal fragment of Mst1. Antibodies directed to the non-catalytic carboxyl-terminal domain of Mst1 and Mst2 immunoprecipitated only the full-length Mst1 and Mst2 isoforms from both untreated (lane 7) and ALN-treated (lane 8) osteoclast lysates.

FIG. 4. Caspase inhibitors block Mst1 kinase cleavage induced by bisphosphonates.
Ocs were prepared and analyzed for in-gel kinase activity as described in Fig. 2. Osteoclasts were not treated with bisphosphonate (lanes 1-3) or treated with 30 M ALN (lanes 4 -6), 30 M RIS (lanes 7-9), and 100 M CL2 (lanes 10 -12) for 20 h in the absence or presence of Z-VAD-FMK (lanes 2, 5, 8, and 11) or Z-DEVD-FMK (lanes 3, 6, 9, and 12). Radioactive bands were visualized by phosphorimaging. Molecular mass markers (in kDa) are indicated to the left. Kinase identities are shown on the right. caspase inhibitor, relative to cells treated with ALN, RIS, or CL2, alone (lanes 4, 7, and 10, respectively). These data suggest that BP-induced activation of the 34-kDa species resulted from caspase cleavage of full-length Mst1. In the absence of BP treatment, both caspase inhibitors reduced base-line 34-kDa Mst1 activities (lanes 2 and 3) relative to control (lane 1). The findings indicate that the activity migrating at 36-kDa kinase was also caspase-dependent, although the identity of this kinase remains to be determined. Analyses of p38, c-Jun Nterminal kinase, and extracellular signal-regulated kinase using GST-ATF2 as a kinase substrate for in-gel kinase assays and using phosphospecific antibodies to probe immunoblots showed that these kinases were not activated in osteoclasts treated with ALN (data not shown).
GGOH Blocks Mst1 Cleavage Induced by ALN, RIS, or Lovastatin, but Not by CL2-Recent studies suggest that ALN and other N-BPs act on the Oc by inhibiting protein geranylgeranylation, related to their inhibition of the mevalonate pathway (4,5). To examine if inhibition of the mevalonate pathway is sufficient to induce caspase cleavage of Mst1 kinase, osteoclast cultures were treated with the hydroxymethylglutaryl-CoA reductase inhibitor, lovastatin (LOV) for 2-24 h. In-gel kinase assays (Fig. 5A) showed that, like N-BPs (Fig. 2), LOV induced 34-kDa Mst1 and 36-kDa kinases and decreased Mst1, Mst2, and 50-kDa kinase activities. The profile of these responses was similar to that of ALN, PAM, and RIS (Fig. 2, A-C, respectively) indicating that inhibition of the mevalonate pathway can induce Mst1 kinase cleavage in these cells.
Recent studies showed that induction of apoptosis in J774 macrophages by N-BPs or statins was blocked by the addition of either geranylgeranyl or farnesyl precursors (3,23). This is consistent with FPP synthase as the target for N-BP action but does not separate between geranylgeranylation and farnesylation as rate-limiting targets in N-BP action on the Oc. To distinguish between these two potential N-BP targets (4), we treated purified osteoclasts with ALN (Fig. 5B, lanes 4 -6), RIS (lanes 7-9), and LOV (lanes 10 -12) in the absence or presence of farnesol (lanes 2, 5, 8, and 11) or GGOH (lanes 3, 6, 9, and  12). Increase in 34-kDa Mst1 activity with either BP or LOV was blocked by GGOH but not by farnesol, although a partial reduction in 34-kDa kinase activity was seen in the presence of farnesol in some experiments (data not shown and Ref. 4). Other assays showed that mevalonic acid lactone, a metabolite situated between N-BP and LOV sites of action, was effective in blocking LOV, but not ALN or RIS, induction of Mst1 kinase cleavage (data not shown and Ref. 4).
To evaluate the proposed differences between the mechanism of action of N-BPs and of BPs lacking nitrogen, purified osteoclasts were treated with ALN (Fig. 5C, lanes 3 and 4), RIS (lanes 5 and 6), and CL2 (lanes 7 and 8) in the absence (lanes 3, 5, and 7) or presence of GGOH (lanes 4, 6, and 8). EHDP was not examined, since it elicited little or, occasionally, no response from 34-kDa Mst1 kinase. LOV (lanes 9 and 10) and Na 3 VO 4 (lanes 11 and 12) were used to inhibit the mevalonate pathway and protein-tyrosine phosphatases, respectively. As previously shown in Fig. 5B, ALN-, RIS-, and LOV-induced Mst1 kinase cleavage was blocked by GGOH with kinase activities similar to those observed in untreated controls (Fig. 5C,  lane 1). Similar analyses with PAM showed a significant, but incomplete, reduction in 34-kDa Mst1 activity when GGOH was added during treatment (data not shown). On the other hand, activation of 34-kDa Mst1 by CL2, which does not inhibit the mevalonate pathway, was not blocked by GGOH (lane 8). Na 3 VO 4 (lanes 9 and 10) elicited only a modest response with activity levels comparable to those observed with EHDP (Fig.  2E). Other concentrations of vanadate resulted in similar responses (data not shown).

Mst1 Kinase Is Cleaved during Osteoclast Apoptosis Induced by Withdrawal of Serum and M-CSF or by Treatment with
Staurosporine-To determine whether Mst1 kinase cleavage was unique to the BP and LOV responses or was more broadly associated with Oc apoptosis, we examined kinase activities during osteoclast apoptosis induced by treatment with staurosporine or withdrawal of serum and M-CSF. Treatment of B-cells with staurosporine, a potent inducer of apoptosis, leads to both activation of full-length Mst1 and Mst2 and caspase cleavage of Mst1 (13). Staurosporine induces apoptosis in murine macrophages but, unlike ALN, RIS, and other N-BPs, has no effect on protein prenylation (23). In purified Oc cultures, staurosporine activation of 59/60-kDa Mst kinases was detectable within 30 min of treatment (Fig. 6A, lane 2). Activity of the 34-kDa Mst1 kinase domain gradually increased, with peak activity observed at 4 and 8 h (lanes 5 and 6). In contrast to treatment with BPs or LOV, staurosporine elicited little reduction in activity of the 50-kDa kinase, although activity of the 36-kDa kinase was increased with kinetics comparable to that of the 34-kDa Mst1 kinase.
Withdrawal of M-CSF was shown to induce osteoclast apoptosis within 18 h of treatment (Fig. 1E). Withdrawal of serum  (2). The intracellular action of BPs leads to loss of cytoskeletal structure and disappearance of the ruffled border, and ultimately, apoptosis (10,11,19). Oc apoptosis was proposed to be the mechanism by which bone resorption is inhibited by BPs. However, it was not known whether BPs induce Oc apoptosis through direct or indirect action, and there has been little insight into the signaling pathways controlling this event. We show here that apoptosis is induced by BPs in purified Ocs, consistent with direct action of BPs on Ocs. Furthermore, we identified a specific kinase involved in the signal transduction pathway leading to osteoclast apoptosis.
For the N-BPs, such as ALN and RIS, apoptosis and inhibition of bone resorption seem to be initiated by the inhibition of the mevalonate pathway enzymes, isopentenyl diphosphate synthase and/or FPP synthase (5). This causes a block in biosynthesis of cholesterol, FPP, and geranylgeranyl diphosphate. The prevention of N-BP effects by the addition of GGOH indicates that geranylgeranylation is rate-limiting in this process. Geranylgeranylation is the attachment of a 20-carbon (geranylgeranyl) lipid to certain proteins, including key regulatory G-proteins such as Rac, Rho, Cdc42, and various members of the Rab family (20,21,24). Geranylgeranylation is necessary to anchor these proteins on the plasma membrane or on intracellular membranes. Absence of geranylgeranylation and intracellular targeting of G-proteins results in a block or change in signaling events that ultimately lead to the induction of caspases, possibly caspase 3, and to apoptosis.
Mst1 acts as a caspase 3 substrate and is cleaved into a highly active 34-kDa species that maintains the catalytic domain but lacks most of the carboxyl-terminal sequences (13). Included in these sequences is a domain necessary for the formation of Mst1:Mst1 dimers (22). We show here that Mst1 and Mst2 also form heterodimers that contain both isoforms, and this association is disrupted by caspase cleavage. Another domain located near the carboxyl terminus acts as a negative regulatory element, whose deletion by mutagenesis results in increased catalytic activity by up to 9-fold even when dimerization is maintained (22). Expression of Mst1 or the 34-kDa Mst1 kinase domain were shown to be sufficient to activate caspases and induce apoptosis in COS cells (13). The induction of Mst1 cleavage in the osteoclast by BPs and by the other apoptosis-inducing treatments suggests that this kinase is part of a signaling pathway leading to Oc programmed cell death.
Comparison of the potency of the tested BPs shows that the induction of Mst1 kinase cleavage by ALN and RIS in the Oc was comparable at matched concentrations, reflecting their similar efficacy in vivo (25). The mechanism for the occasional upper gastrointestinal effects produced by these agents, found to be similar for ALN and RIS in dogs and rats (26,27), could also be due to inhibition of the mevalonate pathway. PAM induces Mst1 kinase cleavage and Oc apoptosis at 10-fold higher doses, consistent with the lower anti-resorptive potency of this N-BP. BPs lacking a nitrogen also induce caspase cleavage of Mst1; however, they require higher concentrations and their action is not through the mevalonate pathway. CL2 and, to a lesser extent, EHDP and tiludronate, are converted into toxic ATP analogs in the cell (6). We observe that both CL2 and EHDP induce apoptosis in the osteoclast, but while CL2 induction of Mst1 cleavage is robust, EHDP effects are modest. Additional mechanisms may be involved in the action of these BPs. EHDP may act by inhibiting protein-tyrosine phosphatases, an effect produced by to all BPs in vitro, and tiludronate was reported to directly inhibit in vitro the osteoclast vacuolar ATPase (28).
Together, these data support the following model for BP action (Fig. 7). All BPs concentrate on bone. The BPs examined in this study can act on the Oc directly to induce apoptosis. All apoptosis-inducing treatments trigger caspase cleavage of Mst1 kinase into an active 34-kDa species. Since the 34-kDa Mst1 kinase domain is itself capable of inducing the activation of caspases that cleave the full-length kinase (13), this can create an apoptosis auto-activation loop. For ALN and RIS, inhibition of bone resorption, induction of Mst1 kinase cleavage and induction of apoptosis are prevented by GGOH, suggesting that the mevalonate pathway intermediate, geranylgeranyl diphosphate, is rate-limiting for their effects. On the other hand, CL2 does not inhibit the mevalonate pathway, is not inhibited by GGOH, and induces Mst1 kinase cleavage and osteoclast apoptosis via a different mechanism (5,6). EHDP, the least potent BP, induces apoptosis with weaker activation of Mst1 kinase cleavage, and may act via yet another mechanism. In conclusion, these findings further support the hypothesis that the mevalonate pathway is the intracellular target for the N-BPs, ALN and RIS, and identify Mst1 as a key intermediate in Oc apoptosis.