The increased osteocyte death by oxidative stress (OS) during aging is a major cause contributing to the impairment of bone quality and bone loss. However, the underlying molecular mechanism is largely unknown. Here, we show that H2O2 induced cell death of primary osteocytes and osteocytic MLO-Y4 cells, and also caused dose-dependent decreased expression of gap junction and hemichannel-forming connexin 43 (Cx43). The decrease of Cx43 expression was also demonstrated with the treatment of other oxidants, rotenone and menadione. Antioxidant reversed the effects of oxidants on Cx43 expression and osteocyte cell death. Cx43 protein was also much lower in the osteocytes from 20-month-old as opposed to the 5-week-old or 20-week old mice. Dye transfer assay showed that H2O2 reduced the gap junction intercellular communication (GJIC). In contrast to the effect on GJIC, there was a dose-dependent increase of hemichannel function by H2O2, which was correlated with the increased cell surface expression of Cx43. Cx43(E2) antibody, an antibody that specifically blocks Cx43 hemichannel activity but not gap junctions, completely blocked dye uptake induced by H2O2 and further exacerbated H2O2-induced osteocytic cell death. In addition, knockdown of Cx43 expression by small interfering RNA (siRNA) increased the susceptibility of the cells to OS-induced death. Together, our study provides a novel cell protective mechanism mediated by osteocytic Cx43 channels against OS.
Osteocytes, representing approximately 95% of all bone cells, are embedded within the bone matrix and are characterized by the presence of long dendritic processes through which they communicate with neighboring osteocytes and with cells on the bone surface. Ability of bone to remodel efficiently, maintain quality, and repair damage depends on the presence of viable osteocytes.[1-3] The number of viable osteocytes in human femur decreases from 88% at age 10 to 29 to 58% at age 70 to 89. Furthermore, osteocyte lacunar density decreased with age in both men and women. Various adverse conditions including oxidant stress (OS), glucocorticoid excess,[7, 8] estrogen loss, and microdamage contribute to osteocyte apoptosis and subsequent bone loss. Furthermore, osteocyte cell death increases progressively with age in mice, and OS during aging seems to be the seminal mechanism contributing to the death of the bone cells (reviewed in Manolagas and Parfitt).
Gap junctions formed by two juxtaposed connexons or hemichannels regulate intercellular communication between osteocytes, osteoblasts, and osteoclasts, and thus play a critical role in bone formation and remodeling. Although three different gap junction–forming proteins, connexins Cx43, Cx45, and Cx46, are expressed in the bone cells, Cx43 is the major connexin isoform expressed in bone cells. Cx43 plays a critical role in many aspects of bone cell function, including proliferation, survival, and differentiation of osteoblasts, skeletal development, and postnatal bone mass acquisition.[13, 15] Furthermore, Cx43 is shown to be required for the antiapoptotic effect of bisphosphonates on osteoblasts and osteocytes in vivo. Hemichannels formed by Cx43 have been shown to regulate the release of NAD+, prostaglandin E2 (PGE2), and ATP in response to mechanical stimulation in osteocytes and mesenchymal stem cells.[17-20]
Recent reports have implicated the role of hemichannels and gap junctions in regulating susceptibility of cells to OS-induced cell death. Cigarette smoke extract and H2O2 are shown to induce hemichannel opening, which leads to cell death in Marshall and L2 cells. Cx43 hemichannels cause cadmium-induced cell death of renal epithelial cells. In contrast to the effect of hemichannels, Cx43 gap junction channels conferred protection to human retinal pigment epithelial cell line against tert-butyl hydroperoxide (t-BOOH)-induced cell death.
In this study, we investigated the effects of OS on Cx43 expression, gap junction, and hemichannel function, as well as the role of Cx43 channels in protection of osteocytes against OS. We observed differential effects of OS on gap junction and hemichannels. Furthermore, our findings point to a novel protective function of Cx43 and Cx43 hemichannels against OS-induced cell death of osteocytes.
Materials and Methods
MLO-Y4 cells were kindly provided by Dr. Lynda Bonewald, University of Missouri. α-MEM media and Alexa 488 were obtained from Invitrogen (Carlsbad, CA, USA). Rotenone was kindly provided by Dr. Brian Herman, University of Texas Health Science Center at San Antonio (UTHSCSA, San Antonio, TX, USA). The Cx43(E2) antibody was developed in our laboratory, targeting the second extracellular loop domain of Cx43. This antibody has been used for immunoblotting and immunocytochemistry, and also for blocking hemichannel functions[20, 22, 24-26] and was affinity-purified as described. The chemiluminescent substrates were from GE Healthcare (Fairfield, CT, USA). The Cx43 small interfering RNA (siRNA) was from Ambion (Life Technologies, Carlsbad, CA, USA). The annexin V/propidium iodide (PI) detection kit was from Clontech (Mountain View, CA, USA). The immunohistochemical ABC staining reagent was from Vector Laboratories (Burlingame, CA, USA). Alexa 488 was from Invitrogen (Eugene, OR, USA). All other reagents were obtained either from Sigma or Fisher Scientific with the highest grade available.
MLO-Y4 cells were cultured on collagen-coated (rat tail collagen type I, 0.15 mg/mL) surfaces and were grown in α modified essential medium (α-MEM) supplemented with 2.5% fetal bovine serum (FBS) and 2.5% bovine calf serum (BCS), and incubated in a 5% CO2 incubator at 37°C, as described.
Isolation of primary osteocytes from chicken calvaria
Calvarial osteocytes from 16-day-old embryonic chicks were isolated based on published protocol. Briefly, calvarial bone was minced and bone pieces were treated with collagenase to remove soft tissues and osteoid followed by decalcification using EDTA. Finally, osteocytes were released from the bone chips by treating with collagenase and vigorous agitation.
Annexin V fluorescein isothiocyanate and PI staining
Primary osteocytes and MLO-Y4 cells were seeded in 35 mm culture plates and incubated overnight at 37°C. After exposure to different doses of H2O2 (for 5 hrs, the cells were trypsinized and stained with annexin V and PI using the Apoalert annexin V apoptosis Kit (Clontech, CA) and MLO-Y4 cells were subjected to FACS analysis.
Cell surface biotinylation
MLO-Y4 cells were seeded in collagen-coated 100-mm plastic dishes and were treated with and without H2O2 for different time periods. After treatment, cells were labeled with or without 1 mg/mL EZ-link Sulfo-NHS-LC-Biotin (Pierce) in Dulbecco's phosphate-buffered saline (DPBS) (Invitrogen) with 1.2 mM Ca2+, 1 mM Mg2+ at 4°C for 30 minutes. The cells were washed with DPBS, and then were incubated with 15 mM glycine in DPBS for 30 minutes, after which cell lysate was collected in 0.5 mL of radioimmunoprecipitation assay (RIPA) buffer (100 mM NaCl, 10 mM EDTA, 25 mM Tris-HCl, 0.25% Triton X-100, and 1% SDS, pH 7.76) containing protease and phosphatase inhibitors. Cell lysates were then homogenized by passing through a 26-gauge needle syringe 20 times. Then 20 µL of each sample was subjected to Western blotting analysis using affinity-purified Cx43(E2) antibody. The band intensity of Cx43 was quantified by densitometry (ImageJ software; NIH). Based on similar levels of Cx43 protein (preloaded), different amounts of total lysates were mixed with 50 mM Tris, pH 7.8, to bring up the volume to 1 mL, and then incubated with 100 µL of monomeric avidin beads for 1 hour at 4°C. The beads were then washed with RIPA buffer without SDS, and biotinylated proteins were eluted by boiling the beads for 5 minutes in sample loading buffer containing 1% SDS and 2% 2-mercaptoethanol, and the eluted proteins were analyzed by Western blotting using affinity-purified Cx43(E2) antibody (biotinylated). The intensity of Cx43 bands was quantified by densitometry (ImageJ).
Cell lysates were resolved by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) on 10% or 15% Bis-Tris gels with Tris-Glycine running buffer, then were electroblotted onto nitrocellulose membranes. Western blotting using appropriate primary antibodies and peroxidase-conjugated suitable secondary antibodies was performed on the nitrocellulose membranes. Chemiluminescent substrates (Pierce) were used to detect antigen–antibody complexes on the membrane. Densitometry was performed on immunoblots using Scion image.
Eight-micrometer (8-µm) sections of long bone from 5-week-old, 20-week-old, and 20-month-old mice were collected using a Leica 2265 Microtome (Bannockburn, IL, USA). After deparaffinization and rehydration, sections were blocked in background terminator (Biocare Medical, Concord, CA, USA) at room temperature for 1 hour and then labeled with 1:100 dilution of affinity purified Cx43(E2) antibody overnight and followed by incubation with the secondary antibody for 1 hour at room temperature. Then the slides were incubated with ABC reagent (Vector Laboratories) at room temperature. Alkaline phosphatase substrate solution was used to visualize immunoreaction sites.
Dye coupling assay
MLO-Y4 cells attached to the plate in recording medium (HCO3−-free α-MEM medium buffered with 10 mM HEPES) were microinjected using a micromanipulator (InjectMan NI 2 and Femtojet, Eppendorf) at 37°C with Alexa 488 (10 mM in PBS). Alexa 488 fluorescence dye was injected into single cells and after 2 minutes of the injection, the dye coupling with neighboring cells was observed under an inverted microscope equipped with Lambda DG4 device (Sutter Instrument, Novato, CA, USA) with mercury arc lamp illumination and a Nikon eclipse (Nikon, Japan) filter (fluorescein isothiocyanate [FITC]) (excitation wavelengths 450–490 nm; emission wavelengths above 520 nm). The extent of dye coupling was quantified by counting the number of cells to which the dye was transferred.
Dye uptake assay
Dye uptake was evaluated using time-lapse measurements or snapshot images. MLO-Y4 cells were plated on collagen-coated plates and were bathed in recording medium (HCO3−-free α-MEM medium buffered with 10 mM HEPES) containing 25 µM ethidium bromide (Etd+) for time-lapse or 50 µM Etd+ for snapshot. In the time-lapse recording, fluorescence was recorded at regions of interest in different cells with Nikon eclipse filter (rhodamine B filter) (excitation wavelengths 540–580 nm; emission wavelengths 600–660 nm). Images were captured with a CoolSNAP HQ2 fast cooled monochromatic digital camera (16-bit) (Photometrics, Tucson, AZ, USA) every 2 minutes (exposure time = 100 ms, gain = 1) and image processing was performed offline with ImageJ software. The collected data was illustrated as ΔF, fold difference of initial fluorescence and fluorescence at the time of interest versus the basal fluorescence.
For snapshot images, MLO-Y4 cells were treated with H2O2 and then were exposed to 50 µM of Etd+ for 5 minutes, rinsed three times with PBS, and fixed with 2% formaldehyde. At least three microphotographs of fluorescence fields were taken with a 10× dry objective in an inverted microscope (Carl Zeiss) with a rhodamine filter. Image analysis was done with ImageJ software. The average of pixel density of 30 random cells was measured.
MLO-Y4 cells were transfected with either scrambled or Cx43 siRNA using Neon Transfection System (Invitrogen, Grand Island, NY, USA). This transfection method can achieve efficiency of 90% to 95%. Forty-eight hours after transfection, cells were treated with 0.5 mM H2O2 for 5 hours and were then subjected to fluorescence-activated cell sorting (FACS) analysis with annexin V-FITC and PI.
All the data were analyzed using GraphPad Prism 5.04 software (GraphPad Software, La Jolla, CA, USA). One-way ANOVA and Student-Newman Keul's test were used for more than two compared groups and paired Student's t test was used for comparison between two groups. Unless otherwise specified in the figure legends, the data are presented as the mean ± SEM of at least three determinations. Asterisks indicate the degree of significant differences: *p < 0.05, **p < 0.01, and ***p < 0.001.
OS induced cell death and decreased Cx43 expression in osteocytic cells
To elucidate the effect of OS on osteocytes, we treated MLO-Y4 cells with different doses (0–0.5 mM) of H2O2, and the cell death and apoptosis indicated by PI and annexin V staining, respectively, were quantified by FACS analyses. Treatment with H2O2 induced cell death, indicated by increased percentage of PI-positive cells (Fig. 1A). H2O2 mostly induced necrotic cell death as opposed to apoptotic cell death because there was no apparent increase in annexin V–positive cells with various doses of H2O2. Furthermore, treatment of these cells with different doses of the oxidant also failed to induce apoptotic cleavage of nuclear lamin B (Supplemental Fig. S2). The lack of apoptotic phenotype is unlikely due to the use of higher doses of the oxidant because the lower doses used in the study failed to induce apoptotic cell death in the osteocyte cell line. Similar to the MLO-Y4 cells, treatment of the primary calvarial osteocytes with H2O2 mostly induced necrosis because these cells were both annexin-positive and PI-positive (Supplemental Fig. S1A). H2O2 also caused a significant decrease in expression of Cx43 in a dose-dependent manner starting from the concentration at 0.3 mM (Fig. 1B, quantification shown on lower panel). The decrease in Cx43 expression is not through increased degradation of the protein by the proteasomal or lysosomal pathway because the inhibitors of these pathways, MG132 (Fig. 1C, upper and lower panel) and bafilomycin A1 (Baf A) (Fig. 1D, left and right panel), failed to rescue the decrease of Cx43 expression caused by the oxidant.
To determine if OS induced by other oxidants affects Cx43 in a similar manner in osteocytes as H2O2, we treated MLO-Y4 cells with rotenone (Fig. 2A), a mitochondrial complex I inhibitor, and menadione (vitamin K3) (Fig. 2B), a redox cycling quinine. Both rotenone and menadione resulted in significant reduction in the expression of Cx43, similar to H2O2 treatment. These results suggest that elevated OS in general, regardless of the type of oxidants, leads to a decrease in Cx43 expression. Menadione, the other oxidant, induced both necrotic and apoptotic cell death in these cells (Fig. 2C).
To establish the specific role of OS in regulating Cx43 expression and cell death, we treated MLO-Y4 cells with an antioxidant, N-(2-mercaptopropionyl)-glycine (NMPG), prior to the addition of H2O2. NMPG not only inhibited the effect of H2O2 on Cx43 expression (Fig. 3A), but also protected cells from death determined by percentage of PI-positive and annexin V-positive cells using FACS analysis (Fig. 3B). This data further suggests that OS is the underlying cause for decreased Cx43 expression and cell death.
Cx43 expression was reduced in older animals as compared to younger animals
Because OS is known to be a fundamental mechanism of aging of the bone (reviewed in Manolagas and Parfitt), we examined the expression of osteocytic Cx43 in bone sections from young and old mice by immunohistochemistry. In order to detect any difference of Cx43 expression, we performed immunohistochemistry with moderate exposure time. Positive signals for Cx43 were mostly observed in osteocytes. Under our experimental conditions, we did not observe much signal in other cell types including osteoblasts and bone marrow cells because these cells express lesser Cx43 protein than osteocytes. The results showed that positive staining of Cx43 protein was less visible in the osteocytes of 20-month-old mice as opposed to 5-week-old or 20-week-old mice (Fig. 4, arrowheads). The quantification data showed a significant decrease of osteocytic Cx43 expression in 20-month-old mice as compared to that of 5-week-old and 20-week-old mice (Fig. 4, lower panel).
H2O2 treatment decreased gap junction function while it increased hemichannel activity
Gap junction function assay was performed to determine if decreased Cx43 expression resulting from H2O2 treatment also affected gap junction coupling. Gap junction intercellular communication (GJIC) was evaluated using the microinjection dye coupling assay. Intercellular transfer of Alexa 488 dye from the microinjected to the neighboring cells was captured by snapshot over time. Alexa 488 coupling was reduced progressively after 30 minutes of 0.3 and 0.5 mM of H2O2 addition and complete uncoupling after 60 minutes. Quantification analyses showed that treatment with H2O2 significantly decreased the function of GJIC in a time-dependent manner (Fig. 5).
To determine the effect of OS on hemichannel activity, we performed Etd+ dye uptake assay. We measured uptake of 25 µM Etd+ by time-lapse recording under resting conditions (basal) and after the addition of different concentrations of H2O2. Under the basal condition the uptake rate was 3.6 × 10−4 ± 1.9 × 10−4 during all 90 minutes of recording. In MLO-Y4 cells treated with 0.3 or 0.5 mM of H2O2, the Etd+ slope (ΔF/min) between 10 to 40 minutes after addition of H2O2 was (6.7 ± 0.14) × 10−4 and (7.3 ± 0.76) × 10−4, respectively. Thus, H2O2 increased the Etd+ uptake rate (Fig. 6A). The Etd+ uptake induced by H2O2 was completely prevented with Cx43(E2) antibody (E2 Ab), a specific Cx43 hemichannel blocker, suggesting that the increase of membrane permeability is mediated by Cx43 hemichannels (Fig. 6A). To determine if increased hemichannel activity induced by H2O2 treatment was a result of increased surface expression of Cx43, we analyzed the level of Cx43 on the plasma membrane by cell-surface biotinylation. H2O2 increased the levels of cell-surface Cx43 in a time-dependent manner, reaching maximal level after 60 minutes of the treatment (Fig. 6B). These results suggest that H2O2 induces the accumulation of Cx43 on the cell surface. This increased uptake of Etd+ was not a result of overall membrane leakage from necrosis because uptake of PI was undetectable under conditions that open hemichannels, but uptake of Etd+ at same concentration was easily detectable (Supplemental Fig. S3).
Cx43 protects osteocytes from OS-induced cell death
As shown above, OS decreased Cx43 expression and gap junction function. To determine the functional role of Cx43 in response to OS, we used Cx43 siRNA. Knockdown of Cx43 by the siRNA (Fig. 7A) increased the susceptibility of osteocytes to H2O2-induced cell death (Fig. 7B). This result suggests that Cx43 plays a cell protective role against OS-induced osteocyte cell death.
Increased Cx43 hemichannel function by OS plays a critical role in osteocyte protection
We showed that OS induced the opening of hemichannels. To elucidate the functional importance of hemichannel opening in response to OS, we blocked Cx43 hemichannel activity using the Cx43(E2) antibody, a potent antibody that specifically blocks Cx43 hemichannel activity but not gap junction or any other channel activity.[20, 22, 25] The evoked Etd+ uptake was prevented to basal levels by Cx43(E2) antibody throughout the recording period (Fig. 6A), suggesting that Cx43-composed hemichannels play a critical role in the increase of membrane permeability induced by H2O2. Furthermore, incubation with Cx43(E2) antibody, but not control immunoglobulin G (IgG), further exacerbated H2O2-induced death of the osteocyte cells, as indicated by the increase of the number of PI-positive cells (Fig. 8A, B), suggesting the functional involvement of Cx43 hemichannels in protecting osteocytes against OS. Primary osteocytes were also more susceptible to H2O2-induced death when they were treated with the oxidant in the presence of hemichannel blocking Cx43(E2) antibody (Supplemental Fig. S1B).
The percentage of osteocyte death has been shown to be increased upon skeletal aging, which is associated with accumulation of reactive oxygen species. Osteocytes that disappear from their lacunae are commonly observed in the elderly. We carried out this study in order to dissect the molecular mechanism underlying OS-induced death in osteocytes and the role of connexin-forming channels in this process. In this study, we report several novel findings. First, we show decreased expression of Cx43 protein expression by OS in osteocytes and that this decrease could be rescued by antioxidant. Similarly, gap junction activity was also reduced by OS. The decreased Cx43 expression was also observed in osteocytes from old mice as compared to young mice. Second, in contrast to gap junctions and overall expression of Cx43, OS increased hemichannel activity with the enhancement of cell surface expression of Cx43. Finally, Cx43 and functional hemichannels play important roles in protecting osteocytes from OS-induced cell death.
We found that not only H2O2, but also other oxidants, such as rotenone, a mitochondrial oxidant, and menadione, also resulted in similar reduction in the expression of Cx43. These adverse effects by oxidants were reversed by an antioxidant, NMPG, further confirming the regulation of Cx43 by OS. These results suggest a common underlying mechanism for regulation of Cx43 expression by OS. In contrast to our findings, Jilka and colleagues show no difference between the level of Cx43 mRNA between young and old vertebral bones of mice. It is possible that the decrease of Cx43 expression at old age may be due to an overall decrease in the level of Cx43 protein, but not the level of Cx43 mRNA. The lack of difference in Cx43 mRNA and protein between young and old animals in the study by Genetos and colleagues could be a result of their observations in rat osteoblastic cells as opposed to the osteocytes. In line with this in vitro study, the low expression of Cx43 in osteocytes from older mice as compared to younger mice could be due to the increased OS in older animals as compared to younger ones. The elevated level of OS in bone cells has been shown to be directly associated with the aging process.[11, 12] The decreased expression of Cx43 and corresponding reduction of GJIC could partially contribute to the decreased bone strength in old age. Indeed, previous in vivo studies reveal the importance of Cx43 in bone function and development. In mouse models with the genetic deletion of Cx43, the quality of the bone is compromised associated with retardation in embryonic osteoblast differentiation, low bone mineral density (BMD), thin cortical bone, decreased bone strength, and attenuated response to parathyroid hormone.[31, 32]
We showed that OS induced by H2O2 not only caused decrease expression of Cx43 but also resulted in decreased gap junctional coupling and death of the osteocytes. Intriguingly, even the low dosage of oxidants primarily caused necrosis but less apoptosis in both primary osteocytes and MLO-Y4 cells. Lack of a strong apoptotic phenotype could possibly be a result of the lower activation of the apoptotic machinery by the oxidant. It is interesting to note that 0.3 mM of H2O2 caused a significant decrease in Cx43 expression but failed to induce any death, suggesting that decreased Cx43 expression as such does not result in cell death. It is possible that absence of Cx43 in the presence of threshold levels of OS increases the susceptibility of cells to death. Decreased GJIC could be attributed to an overall decrease in the amount of Cx43 protein available for assembly of functional gap junction channels. This result is consistent with a previous study showing the reduced presence of Cx43 by OS in junctional plaques associated with decreased gap junctional coupling in myocardium. A similar protective effect of Cx43 GJIC was reported in a human retinal pigment epithelial cell line when exposed to OS. Our observation of the increased susceptibility of Cx43 knockdown cells to OS-induced cell death demonstrates the important role of Cx43 in protecting osteocytes against OS.
In contrast to the effect on gap junction function, OS increased hemichannel activity in osteocytes. Several mechanisms could contribute to increased hemichannel activity, including increased opening of hemichannels already present on the cell surface, increased permeability of already open channels, increased cell surface expression, or decreased internalization. Our cell-surface biotinylation results suggested an increase in cell-surface expression of Cx43, which could account for increased hemichannel activity in response to OS. In accordance with this data, previous reports showed that OS caused by metabolic inhibition or reactive oxygen species and cigarette smoke increases permeation of Cx43 hemichannels[21, 34] and surface expression. It is possible that either decreased internalization of cell surface Cx43 or increased trafficking may contribute to increased surface expression of Cx43 by OS.
The treatment with hemichannel blocking Cx43(E2) antibody further augments the effects of H2O2 on osteocytes, demonstrating the functional importance of the hemichannel opening when exposed to OS. The osteocyte protective mechanism of hemichannels could have been a result of the release of small molecules and/or the activation of other signaling pathways. Opening of hemichannels in chondrocytes by cyclic loading has been shown to cause the release of ATP, and mechanical stimulation of corneal endothelial cells also results in hemichannel opening and ATP release. Furthermore, opening of hemichannels in MLO-Y4 osteocyte cells induced by oscillating fluid flow has been shown to cause release of ATP. Bisphosphonate-induced opening of hemichannels promoted cell survival by activation of src and extracellular signal-regulated kinase (ERK). Opening of osteocytic hemichannels by fluid flow shear stress has been shown to cause the release of PGE2, which has been shown to protect osteocytes from glucocorticoid-induced apoptosis. It is possible that factor(s) released from the osteocytes through hemichannel activation by OS further activates downstream cellular survival pathways. Our finding of a cell-protective function of hemichannel opening appears to be contradictory to previously published studies. Opening of the hemichannels by OS has been indicated to be the cause of the death of Marshall cells. In addition, a recent study shows that Cx43 sensitizes cells to Cd2+-initiated cytotoxicity through hemichannel-mediated effects on intracellular OS. These differences could be caused by cell specific regulation of hemichannels and the molecules passing through these channels. We showed that even though OS caused increased hemichannel activity, though being cell protective, osteocytes still succumbed to death. This is possibly due to decreased gap junctional coupling, which may prevent the passage of survival signals to neighboring cells. Alternatively, sensitivity of hemichannels to OS could be heterogeneous; ie, cells with highly responsive hemichannels may have stronger self-protective capability or vice versa. Together, the present study reveals regulatory mechanism and roles of Cx43 and Cx43 hemichannels in preventing osteocyte cell death from OS. Elucidation of key molecules passing through these channels and associated signaling mechanisms warrants further investigation.
All authors state that they have no conflicts of interest.
This work was supported by Welch Foundation grant AQ-1507 and National Institutes of Health grants AR46798 and EY012085. We thank Dr. Lynda Bonewald at the University of Missouri for kindly providing MLO-Y4 cell line and Dr. Brian Herman at UTHSCSA for kindly providing rotenone. We thank members of Dr. Jiang's laboratory for critical reading of the manuscript.
Authors' roles: Study design, RK, MAR, and JXJ. Data acquisition: RK, MAR, and SW; Drafting of manuscript: RK and JXJ.