Phospholipase Cγ1 (PLCγ1) Controls Osteoclast Numbers via Colony-stimulating Factor 1 (CSF-1)-dependent Diacylglycerol/β-Catenin/CyclinD1 Pathway*

Phospholipases Cγ (PLCγ) 1 and 2 are a class of highly homologous enzymes modulating a variety of cellular pathways through production of inositol 1,4,5-trisphosphate and diacylglycerol (DAG). Our previous studies demonstrated the importance of PLCγ2 in osteoclast (OC) differentiation by modulating inositol 1,4,5-trisphosphate-mediated calcium oscillations and the up-regulation of the transcription factor NFATc1. Surprisingly, despite being expressed throughout osteoclastogenesis, PLCγ1 did not compensate for PLCγ2 deficiency. Because both isoforms are activated during osteoclastogenesis, it is plausible that PLCγ1 modulates OC development independently of PLCγ2. Here, we utilized PLCγ1-specific shRNAs to delete PLCγ1 in OC precursors derived from wild type (WT) mice. Differently from PLCγ2, we found that PLCγ1 shRNA significantly suppresses OC differentiation by limiting colony-stimulating factor 1 (CSF-1)-dependent proliferation and β-catenin/cyclinD1 levels. Confirming the specificity toward CSF-1 signaling, PLCγ1 is recruited to the CSF-1 receptor following exposure to the cytokine. To understand how PLCγ1 controls cell proliferation, we turned to its downstream effector, DAG. By utilizing cells lacking the DAG kinase ζ, which have increased DAG levels, we demonstrate that DAG modulates CSF-1-dependent proliferation and β-catenin/cyclinD1 levels in OC precursors. Most importantly, the proliferation and osteoclastogenesis defects observed in the absence of PLCγ1 are normalized in PLCγ1/DAG kinase ζ double null cells. Taken together, our study shows that PLCγ1 controls OC numbers via a CSF-1-dependent DAG/β-catenin/cyclinD1 pathway.

Phospholipase C␥ (PLC␥) 4 family members PLC␥1 and PLC␥2 are critical regulators of signaling pathways down-stream of growth factor receptors, integrins, and immune complexes and modulate a variety of signaling pathways involved in cell differentiation, motility, and adhesion to name a few (1)(2)(3). The main function of the PLC␥ family is to cleave phosphatidylinositol 4,5-bisphosphate (PIP 2 ) into two secondary messengers, inositol 1,4,5-trisphosphate (IP 3 ) and diacylglycerol (DAG). IP 3 , in turn, increases intracellular calcium levels via binding to the IP 3 receptors on the endoplasmic reticulum, whereas DAG serves as an endogenous activator of protein kinase C (2,4).
Generation of knock-out animals has identified critical and non-redundant functions of the two PLC␥ enzymes. PLC␥1 is ubiquitously expressed, whereas PLC␥2 is mainly expressed in hematopoietic lineage cells (5). PLC␥1-deficient mice die soon after embryonic day 8.5, and embryos show impaired vasculogenesis and erythropoiesis (6). Interestingly, PLC␥2 is expressed in the PLC␥1 Ϫ/Ϫ embryos but is not sufficient to prevent the early lethality (7). By contrast, global PLC␥2 knockout mice are viable but have a variety of immunological defects, including impaired inflammatory responses in models of arthritis or infections (8 -10), developmental defects in the lymphatic system (11), and high bone mass due to defective osteoclastogenesis (12). Structure analysis of the two enzymes indicates over 90% homology in the catalytic domain, which mediates the conversion of PIP 2 into IP 3 and DAG, whereas 50 -60% homology is observed within the SH2 and SH3 adaptive motifs (2). Interestingly, however, the different phenotype of the two null mice suggests distinct, non-overlapping specific roles for each enzyme (5).
So far, the majority of the studies have indicated that each PLC␥ isoform has cell type-specific effects. For example, T cells express both PLC␥1 and PLC␥2, but only PLC␥1 modulates T cell receptor signaling and thus T cell functions (13). In B cells, deletion of PLC␥2 affects B cell differentiation (14), and there are no reports suggesting any role for PLC␥1 in this population. Similar findings were reported in dendritic cells (8), neutrophils (9), macrophages (15), platelets (16), and osteoclasts (12) where deletion of PLC␥2 impaired specific cellular functions regard-less of PLC␥1 expression. Interestingly, in platelets, enforced expression of PLC␥1 in PLC␥2 Ϫ/Ϫ cells restores glycoprotein VI-dependent aggregation and ␣IIb␤3-dependent spreading defects, suggesting that reduced PLC␥1 levels might be responsible for the lack of compensation in the in vivo setting (16). Whether the non-overlapping effects of PLC␥1 and PLC␥2 are due to different levels of expression or non-redundant functions remains to be established.
To delve into the specificity of PLC␥ signaling, we turned to the osteoclasts, which express both isoforms throughout osteoclast maturation. Osteoclasts are multinucleated giant cells derived from monocyte/macrophage lineage cells, attached to the bone surface, and responsible for bone degradation during normal bone remodeling and during pathological bone loss (17). The process of osteoclast differentiation requires activation of osteoclastogenic pathways through binding of RANKL to its receptor RANK and survival and proliferative cues activated by CSF-1 and its receptor CSF-1R (17,18). Interestingly, both PLC␥1 and PLC␥2 are expressed and phosphorylated during the osteoclast differentiation process (12), and it is believed that both isoforms contribute to IP 3 -mediated calcium fluxes and NFATc1 up-regulation in response to RANKL (12,19). However, studies from PLC␥2 Ϫ/Ϫ mice show a complete absence of NFATc1 expression and blockade of osteoclast differentiation despite normal expression of PLC␥1 (12,20). The current study was designed to answer two important questions: does PLC␥1 play any role during osteoclast differentiation, and why does PLC␥1 not compensate for the lack of PLC␥2 despite the high homology? Answering these questions will aid the design of better strategies to target the PLC␥ pathway in pathological bone loss and will improve our understanding of the specificity of PLC␥ signaling.

Results
PLC␥1 deficient mice exhibit early embryonic lethality (6), thus limiting our ability to study the effects of PLC␥1 deletion in osteoclasts. To overcome this issue, we screened various shRNA constructs targeting PLC␥1 but not PLC␥2. We identified five PLC␥1 shRNAs showing high knockdown efficiency and specificity for PLC␥1 that did not affect PLC␥2 (Fig. 1A). Next, we expressed the five shRNA constructs in primary OC precursors via lentivirus infection and examined the effects of PLC␥1 deletion on osteoclastogenesis. All five shRNA-infected OC precursors exhibited a severe defect in osteoclast differentiation (Fig. 1B), suggesting that PLC␥1, similarly to PLC␥2, is a critical regulator of osteoclast differentiation. Interestingly, however, shPLC␥1 cultures (also referred throughout the text as PLC␥1-knocked down or PLC␥1-deficient) showed no reduction in expression of NFATc1, c-Fos, and c-Src (Fig. 1C), all known osteoclastogenic pathways affected by PLC␥2 deficiency (12). This finding suggested that PLC␥1 governs osteoclast differentiation via a different mechanism than PLC␥2.
Interestingly, we observed that PLC␥1-knocked down OC precursor cultures had a lower number of cells compared with control shRNA, a defect that was not observed in the context of PLC␥2 deletion. To confirm these findings, we performed an MTT assay to compare the number of uninfected cells (ctrl) and PLC␥1-deficient cells cultured in CSF-1-containing medium for 6 days. We found a significant decrease in the number of viable cells in the absence of PLC␥1 (Fig. 1D), suggesting that the defect in OC differentiation might be due to reduced availability of OC precursors.
To clarify whether PLC␥1 deficiency inhibits osteoclast differentiation via limiting cell number, we plated increasing numbers of PLC␥1-deficient OC precursors in osteoclastogenic medium and analyzed their ability to undergo osteoclast differentiation. Strikingly, the osteoclastogenesis defect was rescued by plating higher numbers of shPLC␥1 cells (Fig. 1E).
Reduced cell number can be due to decreased cell proliferation or increased cell death (21). CSF-1 is well recognized to activate key signaling pathways promoting OC precursor FIGURE 1. PLC␥1 deficiency blocks osteoclastogenesis by limiting the number of OC precursors. A, knockdown efficiency of five different shRNAs targeting PLC␥1 in OC precursors was compared with ctrl and shctrl. Levels of PLC␥2 and actin are used as loading controls. B, OC precursors infected with the indicated shPLC␥1 constructs were cultured in osteoclastogenic medium for 5 days. Representative images (top) and quantification data (bottom) show that PLC␥1 deficiency leads to a severe osteoclastogenesis defect. C, infected OC precursors were subjected to Western blotting analysis for NFATc1, c-Src, c-Fos, and PLC␥1. D, infected OC precursors cultured in CSF-1-containing medium for 6 days were subjected to MTT assay to quantify the number of cells available. E, increasing numbers of shPLC␥1-3-and shPLC␥1-4-infected OC precursors were plated in osteoclastogenic medium, and OCs were enumerated after TRAP staining (OC quantification and representative images of TRAP-stained wells are shown). When indicated, error bars represent S.D., and asterisks represent p Ͻ 0.01 (**) and p Ͻ 0.001 (***). proliferation and survival (22). In previous studies, we documented that high doses of CSF-1 could rescue the osteoclastogenic defect of cells lacking the ␣v␤3 integrin (23) or the co-stimulatory molecule DAP12 (24). However, high concentrations of CSF-1 were not sufficient to rescue cell numbers in PLC␥1-deficient cultures ( Fig. 2A), thus suggesting that PLC␥1 is a major downstream effector of CSF-1 signaling in OC precursors.
To understand whether PLC␥1 deficiency induces apoptosis or affects cell proliferation, we examined cell death and BrdU incorporation by ELISA. As shown in Fig. 2B, no significant differences in cell death were observed between scrambled shRNA (shctrl) and shPLC␥1 OC precursors cultured in CSF-1-containing medium for 2, 4, or 6 days (Fig. 2B). By contrast, the BrdU signal was lower in shPLC␥1 cells (Fig. 2C) in the presence of both 30 and 100 ng/ml CSF-1. Confirming these findings, cell cycle analysis revealed that shPLC␥1 cells were arrested in G 1 phase with fewer cells entering the S phase following 24-h exposure to CSF-1 (Fig. 2D).
It is established that CSF-1 activates PI3K/AKT and Grb2/ ERK pathways in OC precursors (22,25). Therefore, we hypothesized that PLC␥1 modulates either AKT or ERK signaling cascades to promote cell proliferation. In contrast to our expectations, neither AKT nor ERK phosphorylation was altered by PLC␥1 deficiency in response to CSF-1 (Fig. 3A). CyclinD1, cyclinD2, and cyclinD3 are required for cell cycle G 1 /S transition, and their expression is increased upon CSF-1 stimulation (26). Interestingly, the levels of cyclinD1 were decreased in PLC␥1-deficient cells exposed to CSF-1 (Fig.  3B). We observed a similar result for cyclinD3 (not shown).
CyclinD1 is a downstream target of ␤-catenin (27). ␤-Catenin has also been reported to participate in CSF-1-induced OC precursor proliferation via a mechanism that is independent of ERK/AKT phosphorylation (28,29). Therefore, we wondered whether PLC␥1 controls ␤-catenin levels in OC precursors. Consistent with cyclinD1 expression, ␤-catenin levels were significantly reduced in cells lacking PLC␥1 (Fig. 3, C and D). The reduction in ␤-catenin was already observed in basal conditions and became more pronounced after short or longer exposure to CSF-1.
To clarify whether PLC␥1 mediates CSF-1-induced cyclinD1 expression via regulation of ␤-catenin, we used OC precursors from mice expressing a non-degradable form of ␤-catenin driven by the lysozyme M (Lyz2; herein indicated as LysM) promoter (LysM-Cre/Ctnnb1 WT/floxEx3 ; herein referred to as ␤-catenin-CA) (30). To avoid confounding variables due to activity of LysM-Cre in various myeloid populations, we isolated ␤-catenin-CA OC precursors and infected them ex vivo with PLC␥1 or ctrl shRNAs to measure cyclinD1 levels follow-FIGURE 2. PLC␥1 deficiency impairs CSF-1-induced proliferation. A, shctrl-, shPLC␥1-3-, and shPLC␥1-4-infected OC precursors were cultured in medium containing the indicated concentrations of CSF-1 for 6 days, and cell number was assessed by MTT assay. B, shctrl-and shPLC␥1-infected OC precursors were cultured in the presence of 100 ng/ml CSF-1 for the indicated days, and cell death was assessed by ELISA. C, shctrl-and shPLC␥1-infected OC precursors were cultured in the presence of CSF-1 at the indicated concentrations for 3 days, and cells were then subjected to BrdU analysis. D, shctrl, shPLC␥1-3, and shPLC␥1-4 cells were serum-and cytokine-starved for 24 h and then treated with 100 ng/ml CSF-1 for an additional 24 h before cell cycle analysis. When indicated, error bars represent S.D., and asterisks represent p Ͻ 0.05 (*), p Ͻ 0.01 (**), and p Ͻ 0.001 (***).
PLC␥1 enzymatic activity leads to increased calcium levels and DAG production (2). Because calcium is known to modulate NFATc1 and shPLC␥1 cells have normal NFATc1 levels, we hypothesized that reduced DAG production may be responsible for impaired ␤-catenin/cyclinD1 expression in PLC␥1deficient cells. To test this hypothesis, we turned to diacylglycerol kinase (DGK)-deficient OC precursors. DGK modulates DAG levels in OC precursors by converting DAG into phosphatidic acid, and thus DGK deficiency leads to DAG accumulation (31,32). We have recently documented that DGK deficiency increases osteoclast numbers (32), leading to the hypothesis that DAG accumulation could potentiate OC precursor proliferation. Indeed, MTT (Fig. 5A) and BrdU (Fig.  5B) incorporation assays revealed higher numbers of DGK Ϫ/Ϫ OC precursors compared with WT. This increase in cell number was observed in the presence of various concentrations of CSF-1 (Fig. 5, A and B). In line with this result, ␤-catenin and cyclinD1 expression was significantly higher in DGK Ϫ/Ϫ OC precursors compared with WT cells (Fig. 5C).
Finally, to determine whether increased DAG levels could rescue the OC differentiation defect observed in PLC␥1-deficient cultures, we cultured shPLC␥1/DGK Ϫ/Ϫ OC precursors in the presence of osteoclastogenic medium. Consistent with our previous report (32), DGK Ϫ/Ϫ OC precursors formed more and larger osteoclasts than WT, whereas fewer and smaller OCs were observed in the absence of PLC␥1 (Fig. 5F). Notably, OC differentiation was normalized in shPLC␥1/ DGK Ϫ/Ϫ OC cultures. Taken together, these results demonstrate that CSF-1 activates the PLC␥1/DAG/␤-catenin/cy-clinD1 pathway to support OC precursor proliferation, which in turn benefits osteoclastogenesis.
We previously reported that PLC␥2 modulates OC differentiation by regulating NFATc1 activation downstream of RANKL signaling (12). PLC␥2 has also been implicated in cytoskeletal reorganization in response to CSF-1 stimulation and  integrin engagement (33). Thus, we next asked whether PLC␥2 could also modulate CSF-1-dependent proliferative signals. To this end, we compared cyclinD1 expression in WT and PLC␥2 Ϫ/Ϫ OC precursors exposed to CSF-1 and found no differences (Fig. 6A). Consistent with this finding, no differences were noted in the proliferation rate between WT and PLC␥2 Ϫ/Ϫ cells (not shown). Furthermore, expression of shPLC␥1 in WT, PLC␥2 ϩ/Ϫ , and PLC␥2 Ϫ/Ϫ cells leads to a similar decrease in cell numbers, further suggesting that PLC␥2 is not sufficient to compensate for the loss of PLC␥1 (Fig. 6B).
The different response between PLC␥1 and PLC␥2 to CSF-1 stimulation made us hypothesize that PLC␥1 is the main modulator of CSF-1 signals required for OC precursor proliferation.
these results demonstrate that CSF-1R recruits PLC␥1 but not PLC␥2 in response to CSF-1 stimulation, and this could explain why PLC␥1 but not PLC␥2 modulates CSF-1-induced OC precursor proliferation.

Discussion
PLC␥1 and PLC␥2 are two highly homologous enzymes with cell-specific, non-redundant biological functions (5). Although both enzymes can be expressed in the same cell, often one appears to play a dominant role over the other. Various hypotheses have been proposed for the non-overlapping effects of these two enzymes, ranging from different levels of expression or inability to be recruited to the same molecular complexes (3,5). Interestingly, in the osteoclasts, both PLC␥1 and PLC␥2 are required for osteoclastogenesis but do not compensate for each other. Although our previous work demonstrated that PLC␥2 is activated downstream of RANK and regulates expression of OC markers, we now find that PLC␥1 is activated in response to CSF-1 and modulates OC precursor proliferation. To our knowledge, this is the first known cell type where deletion of either enzyme has major consequences on the cell differentiation by acting on two different signaling pathways.
CSF-1 is an important cytokine modulating critical pathways required for OC precursor proliferation, survival, and cytoskel-etal reorganization (17,34), and mice lacking CSF-1 develop osteopetrosis due to lack of osteoclasts (35). The two major signaling pathways activated by CSF-1 are ERK and AKT, which control cell proliferation and survival via the activation and stabilization of D-type cyclins. Intriguingly, however, although PLC␥1 deficiency suppresses OC precursor proliferation, it does not alter CSF-1-induced ERK or AKT phosphorylation. By contrast, PLC␥1 modulates cyclinD1 expression, a protein known to modulate cell proliferation downstream of ␤-catenin (36). Although the role of ␤-catenin in cell proliferation is well established (28,37), contradiction exists regarding the role of ␤-catenin in regulating osteoclast differentiation. In vitro studies with cells from ␤-catenin conditional knock-out or ␤-catenin constitutively active mice using peroxisome proliferator-activated receptor ␥ promoter driven Cre expression show defective OC differentiation (29). By contrast, deletion of ␤-catenin using LysM promoter-driven Cre expression led to osteopenia due to increased number of OCs (38,39). These data suggest that balanced ␤-catenin signals in osteoclast precursors may be required for proper osteoclastogenesis. In addition, treatment with Wnt3a, which induces ␤-catenin accumulation, did not affect RANKL-induced osteoclastogenesis (40 -42), further implicating that how ␤-catenin is activated might also FIGURE 6. PLC␥1 but not PLC␥2 is recruited to CSF-1R after CSF-1 stimulation. A, WT and PLC␥2 Ϫ/Ϫ OC precursors were stimulated with 100 ng/ml CSF-1 for 0, 8, and 16 h and subjected to Western blotting analysis for cyclinD1, PLC␥2, and ␤-actin. B, PLC␥2 ϩ/ϩ , PLC␥2 ϩ/Ϫ , and PLC␥2 Ϫ/Ϫ OC precursors were infected with shctrl and shPLC␥1; cultured in the presence of 100 ng/ml CSF-1 for 6 days; and subjected to MTT assay. C-E, WT OC precursors were stimulated with 100 ng/ml CSF-1 for 0, 0.5, 1, and 1.5 h; lysed; and subjected to immunoprecipitation (IP) using the following antibodies CSF-1R (C), PLC␥1 (D), and PLC␥2 (E) followed by Western blotting for the indicated proteins. When indicated, error bars represent S.D., and asterisks represent p Ͻ 0.001 (***). play a critical role in the OC differentiation process. We show that expression of ␤-catenin-CA in cells lacking PLC␥1 is sufficient to rescue cyclinD1 levels and cell proliferation, which in turn normalizes the number of mature OCs. Our finding places ␤-catenin downstream of CSF-1/PLC␥1 signaling and is consistent with the role of the DAP12/␤-catenin axis in promoting macrophage proliferation independently of AKT/ERK activation (28).
PLC␥1 is known for converting PIP 2 to IP 3 and DAG (2, 4). DAG is a second messenger that transduces signals modulating a variety of cellular functions, including proliferation, motility, and angiogenesis to name a few. We now demonstrate that DAG production is required for PLC␥1 effects on macrophage proliferation. Because DAG is unstable and deletion of a second messenger cannot be achieved, we decided to use DGK knockout mice, which lack the kinase converting DAG into phosphatidic acid. We previously reported that DGK knock-out mice exhibit lower bone mass and enhanced osteoclastogenesis compared with WT mice (32); however, little is known about the role of DAG in OC precursor proliferation. We now find that DGK-deficient OC precursors have an increased proliferation rate and higher cyclinD1 levels compared with WT mice. Most importantly, deletion of the DGK in PLC␥1-deficient cells rescues CSF-1-induced proliferation and osteoclastogenesis. Interestingly, we recently reported that DAG accumulation in DGK-null OC precursors leads to higher c-Fos expression (32), a mechanism conferring augmented osteoclast numbers. However, c-Fos levels appear to be normal in PLC␥1-deficient OC cultures. This result would suggest that DAG controls OC formation by modulating c-Fos expression and affects OC precursor proliferation via the ␤-catenin/cyclinD1 axis.
In conclusion, this study demonstrates that PLC␥1 works as a novel and critical signal for the promotion of OC precursor proliferation and thus osteoclastogenesis downstream of CSF-1. PLC␥1 effects on the OCs are different compared with PLC␥2, which is activated in response to RANKL. Clarifying the exact roles of PLC␥1 and PLC␥2 in the regulation of osteoclastogenesis has important implications that go beyond their effects on the OCs but also demonstrate how two highly homologous molecules can specifically activate distinct pathways in the same cell type.

Experimental Procedures
Primary Cell Cultures and Mice-OC precursors were isolated and cultured as described previously (43). Briefly, bone marrow was harvested from femora and tibiae of 6 -8-week-old C57/BL6 mice. Cells were then cultured in ␣-minimum essential medium (M0894, Sigma) containing 10% heat-inactivated fetal bovine serum (26140, Gibco), 100 IU/ml penicillin plus 100 g/ml streptomycin (15140, Gibco), 2 mM glutamine (25-005-Cl, Corning) (referred to as ␣-10 medium), and 10% CMG 14-12 cell-conditioned medium (44) containing the equivalent of 100 ng/ml CSF-1 for 3 days in Petri dishes. The attached cells represent OC precursors ready to use. In some experiments, purified CSF-1 (576404, Biolegend, San Diego, CA) was used to stimulate OC precursors. To generate mature osteoclasts, OC precursors were plated in a 96-well plate at a concentration of 5 ϫ 10 3 cells/well in the presence of 100 ng/ml RANKL and 1% CMG 14-12 (containing the equivalent of 10 ng/ml CSF-1) for 6 days changing medium every day. The differentiated cells were then fixed in 4% paraformaldehyde and subjected to TRAP staining using a commercial kit (387A, Sigma). All experiments were approved by the Washington University School of Medicine animal care and use committee. Male and female WT, DGK Ϫ/Ϫ , LysM-Cre/Ctnnb1 WT/WT and LysM-Cre/ Ctnnb1 WT/floxEx3 mice (C57BL/6 background), 6 -8 weeks of age, were used in the study. Mice were housed in cages and fed with food and water ad libitum with a 12-h light and 12-h dark cycle.
MTT Assay-shPLC␥1 or shctrl OC precursors were plated in a 96-well plate at a concentration of 2.5 ϫ 10 3 cells/well and cultured for few days in ␣-10 medium containing either 10% CMG 14-12 or purified CSF-1 at the indicated concentrations. Cells were then incubated with 0.5 mg/ml MTT (M2128, Sigma) for 2.5 h. The crystalline formation was dissolved in 150 l of DMSO (D5879, Sigma) followed by a spectrophotometric reading at A 540 (EL-800, Bio-Tek, Winooski, VT).
BrdU Assay-The BrdU assay was performed using a cell proliferation ELISA kit (11647229001, Roche Applied Science). shctrl-and shPLC␥1-infected cells or WT (DGK ϩ/ϩ ) and DGK Ϫ/Ϫ OC precursors were plated in 96-well plates at a concentration of 5 ϫ 10 3 /well and cultured in ␣-10 medium containing purified CSF-1 at the indicated concentrations for 3 days. Cells were then incubated with 0.1% BrdU for 4 h at 37°C, fixed for 30 min at room temperature, and incubated with blocking reagent for another 30 min. Peroxidase-labeled anti-BrdU antibody was then added for 90 min, and BrdU signal was measured at A 450 following reaction with the appropriate substrate as indicated in the manufacturer's instructions.
Cells were then resuspended in 200 l of lysis buffer and incubated for 30 min, and lysates were centrifuged at 200 ϫ g for 10 min. 20 l of supernatant was transferred into a streptavidincoated microplate strip with an additional 80 l of immunoreagent (72 l of incubation buffer, 4 l of anti-histone-biotin, and 4 l of anti-DNA-peroxidase). The mixture was then incubated on a microplate shaker under gentle shaking (300 rpm) for 2 h. After washing with incubation buffer, the immunocomplex was incubated with 100 l of 2,2Ј-azinobis(3-ethylbenzothiazoline-6-sulfonic acid solution for 20 min, and the reaction was stopped by adding 100 l of 2,2Ј-azinobis(3-ethylbenzothiazoline-6-sulfonic acid stop solution. A 405 and A 490 (reference) were recorded, and cell death was assessed by calculating (A 405 Ϫ A 490 ).
Statistical Analysis-Data are represented as mean Ϯ S.D. for absolute values as indicated in the vertical axis of the figures. Two-tailed one-type Student's t test was performed to analyze p values between experimental and control groups. All experiments with multiple parameters (groups, days, or concentrations) were analyzed by using a two-way analysis of variance followed by Bonferroni post-tests. p Ͻ 0.05 was considered to be statistically significant. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.
Author Contributions-Z. Y. and R. F. designed the study and analyzed the data. Z. Y., S. K., A. Z., and S. M. performed the experiments. Z. Y. and R. F. wrote the manuscript.