Role of lysosomal channel protein TPC2 in osteoclast differentiation and bone remodeling under normal and low-magnesium conditions

The bone is the main storage site for Ca2+ and Mg2+ ions in the mammalian body. Although investigations into Ca2+ signaling have progressed rapidly and led to better understanding of bone biology, the Mg2+ signaling pathway and associated molecules remain to be elucidated. Here, we investigated the role of a potential Mg2+ signaling-related lysosomal molecule, two-pore channel subtype 2 (TPC2), in osteoclast differentiation and bone remodeling. Previously, we found that under normal Mg2+ conditions, TPC2 promotes osteoclastogenesis. We observed that under low-Mg2+ conditions, TPC2 inhibited, rather than promoted, the osteoclast differentiation and that the phosphatidylinositol 3,5-bisphosphate (PI(3,5)P2) signaling pathway played a role in the TPC2 activation under low-Mg2+ conditions. Furthermore, PI(3,5)P2 depolarized the membrane potential by increasing the intracellular Na+ levels. To investigate how membrane depolarization affects osteoclast differentiation, we generated a light-sensitive cell line and developed a system for the light-stimulated depolarization of the membrane potential. The light-induced depolarization inhibited the osteoclast differentiation. We then tested the effect of myo-inositol supplementation, which increased the PI(3,5)P2 levels in mice fed a low-Mg2+ diet. The myo-inositol supplementation rescued the low-Mg2+ diet–induced trabecular bone loss, which was accompanied by the inhibition of osteoclastogenesis. These results indicate that low-Mg2+–induced osteoclastogenesis involves changes in the role of TPC2, which are mediated through the PI(3,5)P2 pathway. Our findings also suggest that myo-inositol consumption might provide beneficial effects in Mg2+ deficiency–induced skeletal diseases.

Ca 2ϩ is recognized as a necessary factor for bone health, and the calcium signaling pathway has been extensively studied in the bone. However, other minerals, including zinc, copper, and magnesium, are also known to be critical for bone health (1,2). In particular, bones are known to store Ͼ50% of the total Mg 2ϩ in the body (3,4), and Mg 2ϩ deficiency causes bone loss, coupled with an increase in osteoclastogenesis (5,6). However, the underlying intracellular molecular pathway and associated molecules required for osteoclastogenesis remain unidentified.
The mechanism of osteoclastogenesis has been extensively investigated (7)(8)(9)(10). Previous studies have shown that the receptor activator of nuclear factor-B (RANK)/RANK ligand (RANKL) 3 pathway induces changes in intracellular Ca 2ϩ ([Ca 2ϩ ] i ) levels. This leads to Ca 2ϩ /calcineurin-dependent dephosphorylation and activation of NFATc1 (nuclear factor of activated T cells 1), which translocates from the cytosol to the nucleus (9,10). With regard to osteoclastic molecules associated with the [Ca 2ϩ ] i pathway, we have previously reported that a lysosomal Ca 2ϩ channel, two-pore channel subtype 2 (TPC2), functions in osteoclastogenesis (10).
TPC2 is a member of a newly identified family of Ca 2ϩ -permeable channels, which is activated by nicotinic acid adenine dinucleotide phosphate (NAADP), a potent Ca 2ϩ -mobilizing cellular messenger (11,12). TPC2 has been identified as a lysosomal Ca 2ϩ channel, and thus it does not localize in the endosome, endoplasmic reticulum, Golgi apparatus, or mitochon-dria. NAADP evokes Ca 2ϩ release through TPC2 from the lysosome (11), and this leads to further Ca 2ϩ release from the endoplasmic reticulum through inositol triphosphate receptors or ryanodine receptors (11,12). In relation to bone biology, we have identified TPC2 as a previously unknown regulator of osteoclastogenesis. We have shown that in the RANK/ RANKL pathway, TPC2 promotes osteoclast differentiation through changes in [Ca 2ϩ ] i and NFATc1 localization (10). However, other studies have demonstrated that TPC2 is a phosphatidylinositol 3,5-bisphosphate (PI(3,5)P 2 )-activated sodium channel (13)(14)(15). Notably, NAADP failed to activate TPC2 in enlarged lysosomes; however, a PI(3,5)P 2 -activated Na ϩ current was recorded and shown to be inhibited in TPC2-deficient cells (13). To explain this discrepancy, a previous study has suggested that changes in Mg 2ϩ concentrations can potentially induce a functional conversion of TPC2 from a NAADP-activated channel to a PI(3,5)P 2 -stimulated channel (16).
The aforementioned finding led us to investigate whether osteoclastic TPC2 is associated with low-Mg 2ϩ -induced osteoclastogenesis and bone loss. Here, we present the results showing that TPC2 plays a different role under low-Mg 2ϩ conditions and inhibits osteoclastogenesis through PI(3,5)P 2 -dependent changes in [Na ϩ ] i but not in [Ca 2ϩ ] i . This role of the channel is functionally the opposite of that revealed by our previous findings, which showed that TPC2 normally promotes osteoclastogenesis (10). We further report here that myo-inositol supplementation, which increases the PI(3,5)P 2 levels, rescued the low-Mg 2ϩ -induced bone loss by reducing osteoclastogenesis.

TPC2 negatively regulates low-Mg 2؉ -induced osteoclastogenesis
To confirm the effects of low Mg 2ϩ on osteoclast differentiation, we used Mg 2ϩ -free DMEM as the culture medium and included only 2.5% FBS in the culture medium to eliminate as much Mg 2ϩ as possible. In this culture medium, the Mg 2ϩ concentration was considerably lower than that in the control medium (supplemental Fig. S1A). We hypothesized that osteoclastic TPC2 was functionally affected under low-Mg 2ϩ conditions (supplemental Fig. S3B). To confirm that the extracellular Mg 2ϩ concentration affects the intracellular Mg 2ϩ concentration, we performed two experiments. Based on the biochemical analysis of cell lysates, a low extracellular concentration of Mg 2ϩ decreased the concentration of Mg 2ϩ in the cell (Fig. 1A). Next, analysis of Mg 2ϩ imaging revealed that changes in extracellular Mg 2ϩ from 1 to 0 mM gradually decreased the Mg 2ϩ fluorescence intensity of RAW 264.7 cells (supplemental Fig. S1, B and C). These results suggested that the cytosolic concentration of Mg 2ϩ changed depending on the extracellular Mg 2ϩ concentration. In the low-Mg 2ϩ culture medium, osteoclast differentiation was promoted in stromal cell-free bone marrow (BM) and RAW 264.7 mouse osteoclast precursor-like cells (RAW) (Fig. 1, B versus D and G, and supplemental Fig.  S1F). To investigate the role of TPC2 in osteoclast differentiation, we used TPC2-WT and TPC2 deletion BM and RAW cells. The TPC2 protein was detected in TPC2-WT BM and RAW cells, whereas the bands were not detected upon TPC2 deletion ( Fig. 1F and supplemental Fig. S1D). To confirm the antibody specificity, the antibody was preincubated with the antigen, and no bands were detected (supplemental Fig. S1E). Thus, this antibody was concluded to bind to the TPC2 protein.
Under normal-Mg 2ϩ conditions, the deletion of TPC2 inhibited osteoclast differentiation, as determined by the activity of tartrate-resistant acid phosphatase (TRAP) (Fig. 1G and supplemental Fig. S1F), similar to a previous report on TPC2 knockdown cells (10). However, in a low-Mg 2ϩ culture, the deletion of TPC2 promoted osteoclast differentiation (Fig. 1, D versus E and G, and supplemental Fig. S1F). These effects on osteoclastogenesis were also confirmed based on the mRNA expression of osteoclast differentiation markers (calcitonin receptor, integrin ␤3, and cathepsin K; Fig. 1, H-J, and supplemental Fig. S1, G-I).
To further confirm that TPC2 plays a role in osteoclastogenesis under low-Mg 2ϩ conditions, a TPC2 inhibitor, Ned19, was added to the culture medium. Whereas inhibition of TPC2 decreased the TRAP activity and mRNA expression of osteoclast differentiation markers under normal-Mg 2ϩ conditions ( These results suggested that the role of TPC2 changed in the low-Mg 2ϩ culture medium. To investigate this change, we focused on another TPC2 ligand, PI(3,5)P 2 (13)(14)(15). The enzyme FYVEfinger-containing phosphoinositide kinase (PIKfyve) phosphorylates phosphatidylinositol 3-phosphate to generate PI(3,5) P 2 . First, we examined the effect of the PIKfyve inhibitor YM201636 on osteoclast development and found that inhibition of PIKfyve only had an effect in the low-Mg 2ϩ culture medium (Fig. 1, O-R, and supplemental Fig. S1, N-Q). Next, to investigate the direct role of PI(3,5)P 2 in osteoclastogenesis, PI(3,5)P 2 with Carrier3 (PI(3,5)P 2 ϩC3) was added to cells by using a shuttle PIP kit. The PI(3,5)P 2 ϩC3 addition affected osteoclast differentiation only under the low-Mg 2ϩ condition ( Fig. 1, S-V, and supplemental Fig. S1, R-U), and the effect was diminished in TPC2 deletion cells ( Fig. 1W and supplemental Fig. S1V). These findings indicated that the TPC2 activity was mediated via the PI(3,5)P 2 signaling pathway only under the low-Mg 2ϩ condition.

PI(3,5)P 2 increases [Na ؉ ] i and depolarizes the cell membrane potential under low-Mg 2؉ conditions in osteoclast-like cells
To investigate the signaling pathway responsible for the negative effects of PI(3,5)P 2 on osteoclast differentiation under the low-Mg 2ϩ condition, we measured changes in intracellular ion levels using fluorescent dyes. For this purpose, we used the TPC2-WT and gene-edited (TPC2 deletion) RAW cell lines. PI(3,5)P 2 is a TPC2 ligand generating a Na ϩ current (13, 15) but not a Ca 2ϩ current. In both TPC2-WT and TPC2-deletion cells, PI(3,5)P 2 ϩC3 showed no effects on relative fluorescence of the dye specific for [Ca 2ϩ ] i , under normal-and low-Mg 2ϩ conditions (Fig. 2, A-D). To reliably confirm the data of the [Ca 2ϩ ] i measurements, ATP with Carrier3 was added, which Role of TPC2 in osteoclastogenesis under low-Mg 2؉ conditions resulted in increases in the fluorescence intensity and amplitude (supplemental Fig. S2F). By contrast, the PI(3,5)P 2 ϩC3 addition increased relative fluorescence of the dye specific for [Na ϩ ] i (Fig. 2E). At 120 min after the addition of PI(3,5)P 2 ϩC3, relative fluorescence of the dye specific for [Na ϩ ] i increased under both normal-and low-Mg 2ϩ conditions (Fig. 2, E and G); however, the PI(3,5)P 2 -induced increase in [Na ϩ ] i was considerably higher under the low-Mg 2ϩ condition (Fig. 2G). Notably, this [Na ϩ ] i increase was diminished in TPC2-deletion cells (Fig.  2, F and H). To confirm that the fluorescence signal is due to Na ϩ influx, NaCl was replacement by N-methyl-D-glucamine chloride. Addition of PI(3,5)P 2 ϩC3 increased relative fluorescence; however, its increment was little in the replacement of NaCl by N-methyl-D-glucamine chloride (supplemental Fig.  S2G). This suggests that the signal is mainly due to Na ϩ influx. Next, we focused on the change in the membrane potential caused by the [Na ϩ ] i increase. Using a fluorescent dye specific for the cell membrane potential, we determined that the membrane potential gradually increased after PI(3,5)P 2 ϩC3 addition only under the low-Mg 2ϩ condition (Fig. 2, I and K). However, no increase was observed in TPC2-deletion cells (Fig. 2, J and L). These results suggested that PI(3,5)P 2 depolarized the cell membrane potential through TPC2 and depolarization occurred only under low-Mg 2ϩ conditions. After the addition of PBS with Carrier3 (PBSϩC3) in normal-and low-Mg 2ϩ culture medium, the relative fluorescence of Na ϩ gradually increased. At 120 min after the addition, the relative fluorescence changes in the [Na ϩ ] i (⌬[Na ϩ ] i ) was 10.2 Ϯ 0.7 (n ϭ 5) for normal-Mg 2ϩ culture medium and 10.5 Ϯ 0.5 (n ϭ 5) for low-Mg 2ϩ culture medium. The Mg 2ϩ concentrations did not affect the values. However, the increases by PI(3,5)P 2 ϩC3 were significantly larger than those induced by PBSϩC3 and were dependent on Mg 2ϩ concentration. These data confirmed that PI(3,5)P 2 increased [Na ϩ ] i significantly in low-Mg 2ϩ culture medium. In addition to these results, the deletion of TPC2 showed no effect on the Na ϩ levels or membrane potential under both normal and low-Mg 2ϩ conditions (Fig. 2, E-H), although it showed a functional effect on osteoclastogenesis ( Fig. 1, G-J). To elucidate this inconsistency, we investigated whether RANKL affected the level of PI(3,5)P 2 . After confirming the linearity of the signal intensity for PI(3,5)P 2 (supplemental Fig. S2, H and I), we quantified the PI(3,5)P 2 levels. At 48 h after the RANKL addition, the levels of PI(3,5)P 2 were higher than those in the control (supplemental Fig. S2J). This suggests that RANKL affects the Na ϩ levels, partly via changes in PI(3,5)P 2 levels.

Membrane depolarization inhibits osteoclast differentiation
PI(3,5)P 2 inhibited osteoclastogenesis and depolarized the membrane potential. Previously, we have confirmed that membrane hyperpolarization promotes osteoclast differentiation (2). Thus, we tested our additional hypothesis, suggesting that membrane depolarization would affect osteoclastogenesis. To allow a non-invasive control of the membrane potential, we generated a RAW cell line that stably expressed the channel rhodopsin-wide receiver-mCherry (WR), which is activated by blue light and depolarizes cells through a cation influx (17). Light-induced changes in the membrane potential were recorded under the whole-cell clamp configuration in WR-expressing and control (expressing mCherry only) RAW cells. WR-expressing cells were immediately depolarized (by 29 Ϯ 3.7 mV) upon the onset of light stimulation and returned to the prestimulus potential after termination of the stimulus (Fig. 3, A and 3B). Whereas all recorded WR-expressing cells were depolarized by the light stimulus ( Fig. 3D), control cells did not respond to the stimulation (Fig. 3, C and E). Thus, temporal patterns of membrane depolarization could be strictly controlled using this system, which we next employed to investigate how membrane depolarization affects osteoclast differentiation. Light stimulation decreased the TRAP activity in WR-expressing cells but not in control cells (Fig. 3F), which indicated that osteoclast differentiation is inhibited by membrane depolarization in the presence of RANKL. To investigate the relationship between membrane depolarization and TPC2, WR was expressed in TPC2-WT and TPC2-deletion RAW 264.7 cells. The light-induced membrane depolarization reduced the TRAP activity in TPC2-WT and TPC2-deletion cells under both normal Mg 2ϩ (Fig. 3G) and low-Mg 2ϩ conditions (Fig. 3H). Although in the latter case, the level of TRAP activity in the TPC2-deletion cells was similar to that in the TPC2-WT cells subjected to light stimulation ( Fig. 3H), these TPC2-related differences were unclear. These effects on osteoclast differentiation were identical to those observed upon the addition of a TPC2 inhibitor (Fig. 3, I and J). The results indicated that TPC2 was not involved in the inhibition of osteoclastogenesis under low-Mg 2ϩ conditions, which was caused by light-induced membrane depolarization. A, Mg 2ϩ concentrations in cell lysates from a normal-or low-Mg 2ϩ culture medium, measured 1 day after replacing the culture medium (n ϭ 4). *, p Ͻ 0.01. B-E, representative images of Mg 2ϩ -dependent osteoclast differentiation in TPC2-WT and TPC2-deletion bone marrow-derived osteoclasts (BM-OCs) 7 days after the addition of RANKL. The TPC2 deletion normally inhibited the osteoclast differentiation (B versus C) but promoted it under low-Mg 2ϩ conditions (D versus E) (normal/low Mg 2ϩ : Mg 2ϩ -free DMEM containing 2.5% FBS with/without 1 mM MgCl 2 , respectively). Asterisks show multinucleated osteoclasts. Scale bar, 100 m. F, TPC2 protein levels were diminished by targeting the TPC2 gene in BM by clustered regularly interspaced short palindromic repeats (CRISPR)/caspase-9 (CAS9). TPC2 WT, non-targeted by CRISPR/CAS9; TPC2 deletion, TPC2 gene targeted by CRISPR/CAS9. G-J, TRAP activity (G) and mRNA expression of the calcitonin receptor (H), integrin ␤3 (I), and cathepsin K (J) in TPC2-WT and TPC2-deletion BM-OCs in normal-and low-Mg 2ϩ culture media at 5 days after the addition of RANKL (n ϭ 9). *, p Ͻ 0.01. K-N, TRAP activity (K) and mRNA expression of the calcitonin receptor (L), integrin ␤3 (M), and cathepsin K (N) in BM-OCs cultured in the presence (ϩ) or absence (Ϫ) of the TPC2 inhibitor Ned19 in normal-or low-Mg 2ϩ culture medium, measured at 5 days after the addition of RANKL (n ϭ 9). *, p Ͻ 0.01. O-R, TRAP activity (O) and mRNA expression of the calcitonin receptor (P), integrin ␤3 (Q), and cathepsin K (R) in BM-OCs cultured in the presence (ϩ) or absence (Ϫ) of the PIKfyve inhibitor YM201636 in normal-or low-Mg 2ϩ culture medium, measured at 5 days after the addition of RANKL (n ϭ 9). *, p Ͻ 0.01; NS, not significant. S-V, TRAP activity (S) and mRNA expression of the calcitonin receptor (T), integrin ␤3 (U), and cathepsin K (V) in BM-OCs cultured in the presence (ϩ, addition of PI(3,5)P 2 ϩC3) or absence (Ϫ, addition of PBSϩC3) of PI(3,5)P 2 in normal-or low-Mg 2ϩ culture medium, measured at 5 days after the addition of RANKL (n ϭ 9). *, p Ͻ 0.01; NS, not significant. W, TRAP activity in TPC2-deficient BM-OCs in the presence (ϩ, addition of PI(3,5)P 2 ϩC3) or absence (Ϫ, addition of PBSϩC3) of PI(3,5)P 2 in low-Mg 2ϩ medium, measured at 5 days after the addition of RANKL (n ϭ 9).

Role of TPC2 in osteoclastogenesis under low-Mg 2؉ conditions myo-Inositol supplementation rescues bone loss induced by low-Mg 2؉ diet
Our results thus far suggested that PI(3,5)P 2 potentially reduced the low-Mg 2ϩ -induced bone loss through inhibition of osteoclast differentiation. Various foods contain myo-inositol, which is commonly used as a supplement (18). Moreover, myo-inositol is one of the stereoisomers of inositol, and it is phosphorylated by PI3K into phosphatidylinositol 3-phosphate, which is further phosphorylated by PIKfyve to PI(3,5)P 2 (19). Thus, we hypothesized that myo-inositol supplementation could increase PI(3,5)P 2 and affect bone remodeling under low-Mg 2ϩ conditions. To test this hypothesis, we conducted in vitro experiments as well as in vivo experiments, in which we combined a low-Mg 2ϩ diet and myo-inositol supplementation. First, changes in PI(3,5)P 2 levels were investigated in RAW cells with/without myo-inositol addition. The results of the dot-blot analysis showed that myo-inositol increased the PI(3,5)P 2 levels at 24 h after the addition (Fig. 4A). To confirm the effects of myo-inositol on osteoclast differentiation, myo-inositol was added into the culture medium under the normal-or low-Mg 2ϩ conditions. The TRAP activity decreased following the myoinositol addition under the low-Mg 2ϩ conditions but not under the normal-Mg 2ϩ conditions (Fig. 4B), which were similar to what was observed in the experiment used to examine the effects of PI(3,5)P 2 ( Fig. 1S and supplemental Fig. S1R). Moreover, the PI(3,5)P 2 levels increased in the total protein extracted from the entire tibia of the mice following the myo-inositol supplementation (supplemental Fig. S3A), as was observed in the experiments on RAW cells. The intake of water with/without myo-inositol did not differ among the four groups tested, normal-Mg 2ϩ diet (NMg), normal-Mg 2ϩ diet with myo-inositol (NMgϩIno), low-Mg 2ϩ diet (LMg), and low-Mg 2ϩ diet with myo-inositol (LMgϩIno) and was as follows (g/mouse/day): NMg, 5.5 Ϯ 0.6; NMgϩIno, 5.9 Ϯ 0.5; LMg, 5.8 Ϯ 0.6; and LMgϩIno, 5.4 Ϯ 0.3. The serum Mg 2ϩ levels were decreased by the intake of the low-Mg 2ϩ diet; however, supplementation of myo-inositol did not affect the serum Mg 2ϩ levels in the groups fed either the normal-or low-Mg 2ϩ diet (Fig. 4C). The low-Mg 2ϩ diet-induced trabecular bone loss (Fig. 4D), deterioration of the trabecular structure (Fig. 4, E-G), and increased bone resorption (Fig. 4, H and I). In the mice fed the normal-Mg 2ϩ diet (NMg versus NMgϩIno), myo-inositol supplementation did not affect the structural parameters such as the bone volume/tissue volume ratio (BV/TV; Fig. 4D), trabecular thickness (Tb.Th; Fig. 4E), trabecular number (Tb.N; Fig. 4F), and trabecular separation (Tb.Sp; Fig. 4G) or the bone resorption parameters such as the osteoclast surface area/bone surface area (Oc.S/BS) and osteoclast number/bone surface (Oc.N/BS) ratios (Fig. 4, H and I). By contrast, in the mice fed the low-Mg 2ϩ diet (LMg versus LMgϩIno), myo-inositol supplementation increased BV/TV, Tb.Th, and Tb.N and decreased Tb.Sp (Fig. 4, E-G). Moreover, the low-Mg 2ϩ diet-induced increase in the bone resorption parameters, Oc.S/BS and Oc.N/BS, was inhibited following the myo-inositol supplementation (Fig. 4, H and I). These results suggested that myo-inositol supplementation reversed the low-Mg 2ϩ -induced trabecular bone loss and inhibited the low-Mg 2ϩ -induced osteoclastogenesis.

Discussion
Bones store Ͼ50% of the total Mg 2ϩ of the body, and osteoclasts release Mg 2ϩ from the bone during bone resorption. Mg 2ϩ deficiency is considered to promote osteoclastogenesis and bone loss (3,5), but the underlying molecular pathways and associated molecules remain mostly unknown. This study demonstrated that TPC2 negatively regulated the low-Mg 2ϩinduced osteoclastogenesis through a Mg 2ϩ -dependent pathway (supplemental Fig. S3B). Furthermore, PI(3,5)P 2 inhibited the osteoclast differentiation through TPC2 and membrane potential changes only under low-Mg 2ϩ conditions. Accordingly, myo-inositol supplementation, which increased the PI (3,5)P 2 levels, rescued the low-Mg 2ϩ diet-induced trabecular bone loss and inhibited osteoclastogenesis.
TPC2 was initially identified as a lysosomal NAADP-activated Ca 2ϩ channel (11). Although numerous studies have suggested that TPC2 is related to NAADP and [Ca 2ϩ ] i in various cells (10,11,20,21), recent reports have confirmed that TPC2 is a PI(3,5)P 2 -activated Na ϩ channel (13)(14)(15). A previous study has provided insights into the mechanism of this discrepancy, suggesting that TPC2 is functionally converted in a Mg 2ϩ -and MAPK-dependent manner (16). In our previous report, we found that TPC2 promotes osteoclastogenesis through changes in [Ca 2ϩ ] i ; however, the effects of PI(3,5)P 2 were not examined (10). Here, we focused on functional attributes of TPC2 under low-Mg 2ϩ conditions and found that TPC2 inhibited osteoclast differentiation through PI(3,5)P 2 signaling. Although the direct effects of Mg 2ϩ on TPC2 were not elucidated in this study, our new findings provide additional insights, which can help resolve the controversy surrounding TPC2 and phospholipids. Further, low Mg 2ϩ does not affect osteoclastogenesis and/or bone mass directly via TPC2 because the effect of low Mg 2ϩ on osteoclastogenesis was much more profound than that observed in the presence of TPC2 under conditions of TPC2 deficiency (Fig. 1). Thus, low Mg 2ϩ appears to indirectly activate a functional effect of TPC2, which counteracts the bone loss induced by low Mg 2ϩ .
It is estimated that ϳ4 -5 mM of [Mg 2ϩ ] i in the cytosol exists as a complex with ATP in several types of mammalian cells (22). In a previous report (23), carbonyl cyanide p-(trifluorome-

Role of TPC2 in osteoclastogenesis under low-Mg 2؉ conditions
thoxy) phenylhydrazone, an inhibitor of ATP synthesis in the mitochondria, increased ϳ0.4 mM of [Mg 2ϩ ] i in PC12 cells. In another study (24), carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone increased ϳ20 -30% of the fluorescence intensity in KMG20-AM-loaded PC12 cells. Based on these reports and the linearity between KMG20 and concentration of Mg 2ϩ (25), we estimated that the changes in 0.1 mM of [Mg 2ϩ ] i are equivalent to the changes in 5.0 -7.5% of fluorescence intensity. In this study, the changes in fluorescence intensity by perfusion were found to be ϳ10 -20% (supplemental Fig. S1B). We speculate that the [Mg 2ϩ ] i might be decreased by ϳ0.2 mM during osteoclastogenesis by low-Mg 2ϩ condition. Jha et al. (16) showed that Mg 2ϩ induced the functional conversion of TPC2 at 0.3 mM and inhibited TPC2 currents with an apparent affinity of 0.13 mM. Therefore, changes in [Mg 2ϩ ] i in low-Mg 2ϩ culture medium could induce the functional conversion of TPC2, at least, in a part of the cells.
In animals and humans, the bone is the main storage site for both Mg 2ϩ and Ca 2ϩ (3,4). Although molecular pathways of Ca 2ϩ signaling in bone remodeling and bone cell physiology have been extensively investigated and are well understood, the molecules and signaling pathways related to Mg 2ϩ remain unknown. Here, we identified TPC2 as a Mg 2ϩ -dependent, functionally convertible molecule involved in osteoclastogenesis. Previous studies have shown that Mg 2ϩ deficiency results in various, considerably different disorders, including osteoporosis (5,26). Thus, our results could facilitate the discovery of molecular targets of the Mg 2ϩ pathway. In osteoclast biology in particular, the mechanism underlying low-Mg 2ϩ -induced osteoclastogenesis has remained unclear, but we now confirmed that Mg 2ϩ deficiency-related osteoclastogenesis includes an inhibitory mechanism involving TPC2.
In this study, myo-inositol supplementation positively affected bone remodeling under low-Mg 2ϩ conditions. A few previous studies have demonstrated beneficial effects of myo-inositol on the bone. myo-Inositol supplementation was reported to increase the Ca 2ϩ uptake in the bone (27), and a mixture of arginine, inositol, and silicone was shown to increase the bone mineral density and strength (28,29). Trabecular bone loss and reduced bone formation were observed in sodium/ myo-inositol cotransporter knock-out mice, in which myo-inositol was not detected in cells, although the osteoclast numbers did not change (30). Here, myo-inositol supplementation affected the bone mass and osteoclastogenesis in vivo and in vitro only under low-Mg 2ϩ conditions. Thus, myo-inositol likely exerts beneficial effects on bone remodeling under abnormal conditions such as Mg 2ϩ deficiency. We did not detect any effect of myo-inositol supplementation on the bone under normal conditions. To elucidate the beneficial effects of myo-ino-sitol, comparatively longer supplementation periods and further investigation are required. TPC2 was suggested as a possible target molecule for osteoporosis treatment in our previous study (10). However, this suggestion must be revised based on the new findings of the complex mode of TPC2 action in this study, in which TPC2 showed distinct behaviors under normal-and low-Mg 2ϩ conditions. Thus, TPC2 inhibition might not only suppress but also promote osteoclastogenesis, in particular, under conditions of Mg 2ϩ deficiency, such as alcohol abuse, diabetes, and severe diarrhea (3). The functional conversion of TPC2 is related to MAPKs such as p38/JNK (16). Although we did not investigate the relationship between TPC2 and p38/JNK in osteoclast differentiation, future studies focused on TPC2 inhibition/activation in vivo for targeting bone diseases should involve the MAPK pathway.
In summary, our results demonstrated a different role of TPC2 in osteoclasts under low-Mg 2ϩ conditions compared with that played under normal Mg 2ϩ conditions. TPC2 regulated the low-Mg 2ϩ -induced osteoclast differentiation in a manner distinct from that observed under normal conditions. Moreover, PI(3,5)P 2 inhibited osteoclastogenesis through TPC2 and membrane potential changes only under low-Mg 2ϩ conditions. Meanwhile, myo-inositol, which can increase the PI(3,5)P 2 levels, rescued the low-Mg 2ϩ diet-induced trabecular bone loss and concomitantly inhibited the increase in bone resorption. These new findings related to the Mg 2ϩ -dependent role of osteoclastic TPC2 and to that of PI(3,5)P 2 highlight a previously unrecognized aspect of lysosomal biology and skeletal disorders.

Experimental procedures
Cell culture RAW 264.7 mouse osteoclast precursor-like cells (American Type Culture Collection, Manassas, VA) were cultured in DMEM (GE Healthcare) supplemented with 10% FBS and 1% penicillin and streptomycin. To investigate the effects of Mg 2ϩ on osteoclastogenesis, we used Mg 2ϩ -free DMEM (GE Healthcare) with/without 1 mM MgCl 2 . Moreover, to avoid any effects of FBS-derived Mg 2ϩ , Mg 2ϩ -free DMEM containing a reduced amount of FBS (2.5%) was used for osteoclast development (normal/low-Mg 2ϩ media: Mg 2ϩ -free DMEM containing 2.5% FBS with/without 1 mM MgCl 2 , respectively). All cells were maintained at 37°C and 5% CO 2 in a humidified atmosphere. All chemicals were purchased from Wako Pure Chemical Industries (Osaka, Japan) unless noted otherwise.
For osteoclast differentiation, RAW 264.7 cells were seeded at 2.0 ϫ 10 5 cells/cm 2 with the addition of soluble RANKL

Role of TPC2 in osteoclastogenesis under low-Mg 2؉ conditions
(sRANKL, 50 ng/ml; Oriental Yeast, Tokyo, Japan). TRAP staining was performed as previously described (2,10). Ned19 (1 M; Tocris Bioscience, Bristol, UK), a TPC2 inhibitor, and YM201636 (100 nM; Tocris Bioscience), a PIKfyve inhibitor, were added to culture media in some experiments. For control, the same amount of DMSO was added. The shuttle PIP kit (catalog no. P-9035; Echelon Biosciences, Salt Lake City, UT) was used to add PI(3,5)P 2 (5 M) to investigate the relationship between PI(3,5)P 2 and TPC2 under low-Mg 2ϩ conditions. In the PI(3,5)P 2 experiments, the medium was replaced 2 h after the addition of PI(3,5)P 2 to avoid cell toxicity. Carrier3 (5 M; Echelon Biosciences) was used to deliver PI(3,5)P 2 into the cells. For the control, the same amount of PBS was added with Carrier3. TRAP-positive cells with more than three nuclei were defined as multinucleated osteoclasts. Cells were cultured in 3.5-cm dishes by subculturing from 10-cm dishes. Samples were extracted from each 3.5-cm dish. This procedure was carried out three times at different time points.

Bone marrow-derived osteoclasts
Mouse bone marrow cells were obtained from the tibia and femur (C57BL/6 mice, 4-week-old males; Charles River Japan, Kanagawa, Japan). Bone marrow stromal cells were depleted from the mouse bone marrow by using a Sephadex G10 (GE Healthcare) column as previously described (2,10). The obtained cells were cultured in Mg 2ϩ -free DMEM with/without 1 mM MgCl 2 in the presence of macrophage colony-stimulating factor (50 ng/ml; R&D Systems, Minneapolis, MN) and sRANKL (50 ng/ml).

Establishment of cell lines
Cell lines were established as previously described (2,10,17). Lentiviral constructs expressing two different artificial guide RNAs, targeting murine TPC2 (target sequence: 5Ј-GAACGC-CTGGCTCTCGCGGC-3Ј) and non-targeting gRNA, were prepared in a plasmid (lentiCRISPR v2 (one-vector system), a gift from Feng Zhang (Addgene plasmid catalog no. 52961)). This plasmid contains two expression cassettes, hSpCas9 and the chimeric guide RNA. Lentiviral constructs expressing WR or mCherry (control) were generated using the pLenti6.3/V5-DEST vector (Thermo Fisher Scientific). Viral transduction was performed as previously described (10). Briefly, lentiviral vectors and three packaging plasmids (pVSVG, pLG1, and pLG2) were transfected into 293FT cells using Lipofectamine 2000 (Thermo Fisher Scientific), and the culture medium (lentiviral solution) was collected after 48 h. Viral titers were estimated using Lenti-X quantitative reverse transcription-PCR (Chemicon, Temecula, CA). Viral transduction was performed by adding the lentiviral solution to the culture medium in the presence of 6 g/ml Polybrene (Sigma-Aldrich). For guide RNA-expressing cells, puromycin (5 g/ml) was added to the culture medium for 3 days. Then TPC2-deficient cell clones were selected by limited dilution. Stably transduced cells expressing WR or mCherry were selected using 5 g/ml blasticidin (Thermo Fisher Scientific).

Light stimulus experiment for osteoclast differentiation
The light stimulus experiment has been described previously (2,17). The light stimulus (480 Ϯ 10 nm, 2-s stimulus followed by a 1-s rest, repeated 10 times) was applied to cells under an IX71 inverted microscope (Olympus, Tokyo, Japan) equipped with a VMM-T1 fluorescence shutter controller (Uniblitz, Rochester, NY). RAW cells stably expressing WR or control cells (mCherry only) were used for the light stimulus experiments. Both WR and control cells were seeded at 2.0 ϫ 10 6 cells/cm 2 , and the experiment was repeated three times.

Measurement of Mg 2؉ concentration and TRAP activity
To measure the Mg 2ϩ concentration, the cells were quickly rinsed with cold PBS, and a cell lysate was obtained with a solution containing 50 mM Tris-HCl (pH 7.4) and 1% Nonidet P-40. Magnesium concentrations were determined using a magnesium test (Wako) according to the manufacturer's protocol.
TRAP activity was measured as previously described (2,10,17). Briefly, cells were lysed in extraction buffer (150 mM NaCl, 50 mM Tris, 1% Nonidet P-40, pH 8.0) supplemented with protease inhibitors (Sigma-Aldrich), and a 20-l aliquot of the lysate was added to 200 l of TRAP buffer (50 mM sodium acetate, 25 mM sodium tartrate, 0.4 mM MnCl 2 , 0.4% N,N-dimethylformamide, 0.2 mg/ml Fast Red Violet, 0.5 mg/ml naphthol AS-MX phosphate, pH 5.0). After incubation for 1 h (BM) or 2 h (RAW) at 37°C, absorbance was measured at 540 nm using a SpectraMax microplate reader (Molecular Devices, Sunnyvale, CA). The data were normalized to the total protein content, which was determined using a bicinchoninic acid kit (Thermo Fisher Scientific). Each experiment was performed in triplicate.

Gene expression assay
Total RNA was extracted using an RNeasy kit (Qiagen) according to the manufacturer's protocol. First-strand cDNA was synthesized from total RNA using a high-capacity cDNA reverse transcription kit (Thermo Fisher Scientific). The amount of total RNA used in each reaction of synthesis was 1 g. Quantitative real-time PCR was performed using a StepOne real-time PCR system (Thermo Fisher Scientific), SYBR Green (Fast SYBR Green master mix, Thermo Fisher Scientific), and specific forward and reverse primers. According to the manufacturer's protocol, the conditions for the reaction were 20 l/well, including 10 l of SYBR Green, 1.0 l of cDNA tem-

Fluorescence-based recording of [Ca 2؉ ] i , [Na ؉ ] i , [Mg 2؉ ] i , and membrane potential
Fluorescence-based recording of membrane potential has been described elsewhere (2,17). Cells were loaded with the FLIPR dye (FLIPR membrane assay kit; Molecular Devices) for 30 min in Hanks' balanced salt solution (HBSS) containing 1 mM CaCl 2 with/without 1 mM MgCl 2 at 37°C and 5% CO 2 in a humidified incubator. The membrane potential was recorded every 30 s for 120 min using a FLIPR Tetra system (Molecular Devices) at room temperature. To analyze the changes in membrane potential, fluorescence intensity in each well was quantified using the FLIPR software (Molecular Devices).
The protocol for recording [Na ϩ ] i , [Ca 2ϩ ] i , and [Mg 2ϩ ] i was described previously (2,17,24). Briefly, the cells were loaded with 5 M CoroNa (Thermo Fisher Scientific) for [Na ϩ ] i , Fluo8-AM (Nacalai Tesque, Kyoto, Japan) for [Ca 2ϩ ] i , and 10 M KMG-20-AM (Wako) for [Mg 2ϩ ] i for 30 min in HBSS containing 1 mM CaCl 2 with/without 1 mM MgCl 2 at 37°C and 5% CO 2 in a humidified incubator. After rinsing the cells, the images were captured every 30 s for 120 min for [Na ϩ ] i and [Ca 2ϩ ] i using a SpectraMax fluorescence microplate reader. To analyze the changes in [Na ϩ ] i and [Ca 2ϩ ] i in each well, the fluorescence intensity was quantified using the SoftMax Pro software (Molecular Devices). The replacement experiments for [Na ϩ ] i were initiated after the incubation with physiological solution containing 145 mM NaCl, 5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM D-glucose, 0.1% BSA, and 10 mM HEPES (pH 7.4). To confirm the effects of Na ϩ , NaCl was replaced with N-methyl-D-glucamine chloride (supplemental Fig. S2G). For [Mg 2ϩ ] i , images were captured every 6 s for 400 s using an LSM700 confocal fluorescence microscope (Carl Zeiss). HBSS (2.0 ml) was continuously perfused with/without MgCl 2 at a rate of 4 -5 ml/min. Exchange of the solutions could be accomplished within 30 s. To analyze [Mg 2ϩ ] i changes in individual cells, the fluorescence intensity was quantified using the LSM image software (Carl Zeiss). Approximately five to seven cells per field were analyzed in each culture, and the experiment was repeated several times. The experiments were initiated after incubation of HBSS at room temperature for 5-10 min. PI(3,5)P 2 with Carrier3 or PBS with Carrier3 were added to the solution. At the end of the experiment, cell viability was confirmed by adding a 150 mM KCl solution (for membrane potential) and 5 M ionomycin (for [Na ϩ ] i and [Ca 2ϩ ] i ) (supplemental Fig. S2, C-E).

Electrophysiology
Electrophysiological experiments were performed as previously described (2,17). For whole-cell recordings, the standard external solution contained 145 mM NaCl, 5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, 10 mM HEPES (pH 7.3), and 0.1% bovine serum albumin. The pipette solution contained 130 mM potassium gluconate, 20 mM KCl, 3 mM MgCl 2 , 1 mM EGTA, 1 mM ATP, 0.5 mM GTP, and 10 mM HEPES (pH 7.3). The pH levels of the bath and pipette solutions were adjusted with NaOH and KOH. The osmolality of the solution without CaCl 2 , MgCl 2 , glucose, and BSA or ATP and GTP was determined using a micro-osmometer type 13 (Roebling, Berlin, Germany). The osmolality of this solution was 280 -300 mOsm. The external and internal solutions with CaCl 2 , MgCl 2 , glucose, and BSA for the external solution or ATP and GTP for the internal solution had osmolalities of 320 -330 and 290 -300 mOsm, respectively.
The resistance of the used borosilicate glass pipettes was 5-8 M⍀. The reference electrode was an Ag-AgCl wire connected to the bath solution through a Ringer-agar bridge. The zero current potential before the formation of the gigaseal was taken as 0 mV. The current signals were recorded using an Axopatch 200A amplifier (Axon Instruments, Foster City, CA) digitized at 1-2 kHz with a Digidata 1200 analog-digital converter (Axon Instruments) and analyzed using the pCLAMP software (Axon Instruments). The membrane potential was recorded in the current-clamp mode.

Western blotting and quantification of PI(3,5)P 2 levels
The cells were rinsed with cold phosphate-buffered saline, and total protein was extracted with a solution containing 50 mM Tris-HCl (pH 7.4) and 1% Triton X-100. Protease inhibitors (Wako) were added to all solutions. Total protein was extracted from the entire tibia as described previously (31), and the protein concentration was determined using a bicinchoninic acid assay kit. SDS-PAGE was performed as previously described (32). Briefly, proteins were separated in 7.0% gels and transferred to polyvinylidene difluoride membranes. The membranes were blocked with EzBlock (Atto, Tokyo, Japan) and incubated with an antibody against TPC2 (1:2,000; Alomone Labs, Jerusalem, Israel) or with the antibody preincubated with an antigen (10 g/ml; control peptide antigen for anti-TPC2; Alomone Labs). The antibody was incubated with the antigen for 1-2 h at room temperature. The quantity of total protein loaded to each lane was adjusted to 10 g. For quantification of PI(3,5)P 2 levels, 10 l of the lysate was pipetted onto a nitrocellulose membrane. The quantity of total protein for each dot was adjusted to 10 g. To confirm the linearity of the signal intensity, PI(3,5)P 2 (range, 0.031-2.0 nmol/dot) was pipetted onto the membrane. After drying, the membrane was blocked with

Role of TPC2 in osteoclastogenesis under low-Mg 2؉ conditions
EzBlock and then incubated with an anti-PI(3,5)P 2 antibody (1:1,000; Echelon Biosciences). Images were captured using a CCD camera (ChemiDoc; Bio-Rad), and quantitative analysis was performed using the Image Lab software (Bio-Rad).

Animals
Sixteen 3-week-old male C57BL/6J mice (Charles River Japan) were allowed to acclimate for 1 week at 22 Ϯ 2°C and 60% humidity under a 12/12-h light/dark cycle (with light exposure from 7:00 a.m. to 7:00 p.m.). Drinking water with/without myo-inositol (Ino: 10 mg/ml) was available at all times, and the rats were fed a NMg or LMg (CLEA Japan, Tokyo, Japan), with the magnesium content of 80 mg/100 g (NMg) and 2.7 mg/100 g (LMg). The body weight of the mice was measured weekly, and all mice were found to remain healthy.
The mice were divided into four groups of four animals each at the age of 4 weeks as follows. Groups NMg and NMgϩIno received the NMg diet and NMg diet with drinking water containing Ino, respectively. Groups LMg and LMgϩIno received the LMg diet and LMg diet with drinking water containing Ino, respectively. The mice were sacrificed at the age of 8 weeks through exsanguination under ether anesthesia. All studies involving animals were performed in accordance with the procedures approved by the Tokyo Medical and Dental University and Osaka Dental University Animal Care and Use Committees.

Microcomputed tomography
Microcomputed tomography analysis was performed as previously described (31). The 3D microcomputed tomography images were captured using an MX-90 3D system (Medixtech, Tokyo, Japan) and analyzed using the TRI/3D-Bon software (Ratoc System Engineering, Tokyo, Japan). The X-ray source was set at 90 kV and 90 A for 3D imaging.

Bone histomorphometry
Bone histomorphometry was performed as previously described (33,34). Femoral bone samples were embedded in a mixture of methyl methacrylate, hydroxyglycol methacrylate, and 2-hydroxyethylacrylate, which was polymerized at 4°C. Sections were obtained and stained for TRAP. Histomorphometry was performed on a semiautomatic image analysis system connected to a light microscope (Cosmozone 1S; Nikon, Tokyo, Japan). For each section, the secondary spongiosa area was measured. We also measured the osteoclast surface area (m) and number and trabecular BS (m) as bone resorption parameters. TRAP-positive cells that formed resorption lacunae at the trabecular surface and contained three or more nuclei were identified as osteoclasts.

Statistical analysis
The values are expressed as the means Ϯ S.E. of the mean, and all values were analyzed using the Student's t test or analysis of variance. The Tukey's honest significant difference test was applied as a post hoc test. p Ͻ 0.05 was considered statistically significant. Analyses were performed using SPSS (SPSS Japan, Tokyo, Japan).