Adaptation of Connexin 43-Hemichannel Prostaglandin Release to Mechanical Loading*

Bone tissues respond to mechanical loading/unloading regimens to accommodate (re)modeling requirements; however, the underlying molecular mechanism responsible for these responses is largely unknown. Previously, we reported that connexin (Cx) 43 hemichannels in mechanosensing osteocytes mediate the release of prostaglandin, PGE2, a crucial factor for bone formation in response to anabolic loading. We show here that the opening of hemichannels and release of PGE2 by shear stress were significantly inhibited by a potent antibody we developed that specifically blocks Cx43-hemichannels, but not gap junctions or other channels. The opening of hemichannels and release of PGE2 are magnitude-dependent on the level of shear stress. Insertion of a rest period between stress enhances this response. Hemichannels gradually close after 24 h of continuous shear stress corresponding with reduced Cx43 expression on the cell surface, thereby reducing any potential negative effects of channels staying open for extended periods. These data suggest that Cx43-hemichannel activity associated with PGE2 release is adaptively regulated by mechanical loading to provide an effective means of regulating levels of extracellular signaling molecules responsible for initiation of bone (re)modeling.

Bone tissues respond to mechanical loading/unloading regimens to accommodate (re)modeling requirements; however, the underlying molecular mechanism responsible for these responses is largely unknown. Previously, we reported that connexin (Cx) 43 hemichannels in mechanosensing osteocytes mediate the release of prostaglandin, PGE 2 , a crucial factor for bone formation in response to anabolic loading. We show here that the opening of hemichannels and release of PGE 2 by shear stress were significantly inhibited by a potent antibody we developed that specifically blocks Cx43-hemichannels, but not gap junctions or other channels. The opening of hemichannels and release of PGE 2 are magnitude-dependent on the level of shear stress. Insertion of a rest period between stress enhances this response. Hemichannels gradually close after 24 h of continuous shear stress corresponding with reduced Cx43 expression on the cell surface, thereby reducing any potential negative effects of channels staying open for extended periods. These data suggest that Cx43-hemichannel activity associated with PGE 2 release is adaptively regulated by mechanical loading to provide an effective means of regulating levels of extracellular signaling molecules responsible for initiation of bone (re)modeling.
The skeleton regulates its architecture and mass to meet structural and metabolic needs. To fulfill its structural functions, this complex tissue must adapt to loading and unloading while simultaneously regulating the metabolic demands of the skeleton. Numerous in vivo animal studies show the essential role of mechanical loading for bone formation and remodeling; however, the underlying molecular mechanisms, in particular how bone cells adapt to mechanical stimulation, remain largely uncharacterized.
When mechanical forces are applied to bone, several potential stimuli occur including changes in hydrostatic pressure, direct cell strain, fluid flow, and electric potentials. These changes lead to fluid movement through the bone (1)(2)(3). Shear stress induced by mechanical loading facilitates the exchange of nutrients and bone modulators, and elicits biochemical responses. Osteocytes are well positioned in the bone to sense the magnitude of mechanical strain and are essential for the skeleton adaptive response to load. Experimental studies have shown that osteocytes are sensitive to stress applied to both intact bone tissue and in cell culture (4 -6). Encased within mineralized tissue, their dendritic morphology allows them to connect through small tunnels called canaliculi to form a threedimensional network not only with adjacent osteocytes but also to connect to cells on the bone surface and bone marrow.
Connexins (Cx), 2 gap junction-forming proteins, belong to a multigene family expressing four transmembrane domains. The regions corresponding to transmembrane and extracellular domains are highly conserved. Cx43 has been identified in most types of bone cells (7)(8)(9)(10)(11)(12)(13) and is the major connexin expressed in osteocyte-like MLO-Y4 cells and primary osteocytes (14 -16). A role for connexins, independent of the formation of gap junction channels, has been suggested by the existence of hemichannels recently shown to exist in the nonjunctional plasma membrane (17). Functional Cx43hemichannels have been observed in neural progenitor cells and neurons (18,19), astrocytes (20,21), heart (22), and especially, osteoblasts and osteocytes (23)(24)(25). Cx43-hemichannels appear to function as transducers of bisphosphonates, drugs used in the treatment of bone diseases by opening Cx43hemichannels (24). We recently demonstrated that fluid flow shear stress (FFSS) induced the rapid opening of hemichannels, which in turn mediate the release of PGE 2 in osteocytic MLO-Y4 cells (25). A recent study showed the release of ATP and PGE 2 in MLO-Y4 cells induced by oscillating fluid flow (26). Prostaglandins have been shown to be involved in the response of bone tissue and cells to mechanical loading (27). More importantly, PGE 2 can increase bone mass in mammals when used at the appropriate doses, and bone formation and remodeling caused by mechanical loading are attenuated by inhibition of prostaglandin biosynthesis (28 -30). However, how the release of prostaglandins is regulated by changes of mechanical loading is unclear. In the present work, the dynamics of hemichannel opening associated with PGE 2 release in response to FFSS was examined, and an antibody we developed specific for hemichannels, and not gap junctions, was used to validate this biological response to FFSS according to this mechanism.

EXPERIMENTAL PROCEDURES
Cell Culture and PGE 2 Measurement-MLO-Y4 cells were cultured on collagen-coated (rat tail collagen type I, 0.15 mg/ml) surfaces and were grown in ␣-modified essential medium (␣-MEM) supplemented with 2.5% fetal bovine serum and 2.5% BCS and incubated in a 5% CO 2 incubator at 37°C as described previously (15). After fluid flow treatment at specific time and shear stress levels, the conditioned media were collected, and the extracellular PGE 2 released into the media was measured using a PGE 2 EIA kit according to the manufacturer's instructions (Cayman Biochemicals, Ann Arbor, MI). Intracellular PGE 2 was measured after the cells were thoroughly washed three times with PBS and lysed. PGE 2 release after poststress incubation was determined by subjecting the MLO-Y4 cells to FFSS and then incubating the cells in media at 37°C, 5% CO 2 . The medium was collected at the respective time points as indicated and measured using the kit described above. The PGE 2 measurement at FFSS (0 h) was then subtracted from the accumulated amount at each time point yielding a net PGE 2 measurement.
Fluid Flow Shear Stress-FFSS experiments were conducted as previously described (31), and stress levels of 4, 8, and 16 dynes/cm 2 were generated. Controls consist of MLO-Y4 cells not subjected to FFSS. A cell surface area of 5 cm 2 was used for dye uptake, and 90 cm 2 was used for PGE 2 measurement and for the biotinylation assay. During post-stress chase experiments, the cells were maintained in ␣-MEM at 37°C.
Dye Uptake Assay-MLO-Y4 cells were grown at low initial density to ensure that the majority of the cells were not physically in contact. After treatment, dye uptake experiments were conducted in the presence of 0.2% Lucifer yellow (LY) or 0.2% Alexa 594 along with 0.2% rhodamine dextran (RD) for 5 min, and cells were rinsed and fixed with 1% paraformaldehyde. LY and Alexa 594 were used as a tracer for hemichannel activity, and RD was used as a negative control. Similar fields were observed under the fluorescence microscope. Approximately 7 images per slide per treatment group were collected. From the phase images, the total number of cells was counted up to 300 cells. The number of fluorescent cells was then counted from the corresponding fluorescent images. Dye uptake is presented as the number of fluorescent cells divided by the total number of cells and normalized to 1 for control.
Immunofluorescence Labeling and Fluorescence Microscopy-The cells cultured on the microscopic slides were fixed in 2% paraformaldehyde in PBS for 30 min at room temperature. The cells were incubated in blocking solution (2% normal goat serum, 2% fish skin gelatin, and 1% bovine serum albumin in PBS) in the absence or presence of 1% Triton X-100 for 30 min and then overnight at 4°C with affinity-purified antibodies against Cx43(E2), Cx43(CT), preimmune (1:300 dilutions), or anti-Cx43 antibody pre-absorbed with antigen Cx43 (E2) fusion protein (0.6 mg/ml) in blocking solution. Cells were then incubated for 1 h with fluorescein-conjugated goat anti-rabbit IgG (1:300 dilution) in blocking solution. Fluorescence microscopy was performed using an Olympus B-MAX microscope (Olympus, Tokyo, Japan) and recorded on a "Spot II" digital camera (Diagnostic Instruments, Tokyo, Japan). For dye uptake, LY was detected using the filter set for fluorescein at excitation wavelengths of 505-525 nm, and Alexa 594 and RD by using the filter set for rhodamine with a corresponding excitation wavelength of 560 nm. Mean fluorescence intensity was acquired by taking representative fluorescence images from various regions of each microscope slide. Using Image J program (Image J Software, NIH) the threshold was adjusted to clearly distinguish fluorescent cells from non-fluorescent cells. Threshold conditions were kept constant for each experimental condition. The program measured the intensity of the signal coming from all fluorescent cells in the field; the output is given as mean fluorescent intensity.
Cell Surface Biotinylation-Biotinylation of MLO-Y4 cells was performed based on a modification of previously published procedures (25,32). Briefly, MLO-Y4 cells were treated with and without FFSS for 30 min and then post-FFSS incubated for 30 min, 2, 4, 8, and 24 h or continuous FFSS for 0, 2, 4, and 24 h. These cells were labeled with or without 1 mg/ml EZ-link Sulfo-NHS-LC-Biotin (Pierce Biotechnology) in PBS at 4°C for 20 min. The cells were washed with PBS containing Ca 2ϩ , Mg 2ϩ , and glycine and lysed in lysis buffer (133 mM NaCl, 5 mM KCl, 1% dextrose, and 20 mM HEPES) plus 1% sodium deoxycholate, 1% Triton X-100, 0.1% SDS, and protease inhibitors. Cell lysates were mixed with monomeric avidin beads and incubated for 1 h at room temperature. The beads were then washed with PBS, and biotinylated proteins were eluted by boiling for 5 min in sample loading buffer containing 1% SDS and 2% 2-mercaptoethanol, and proportional amounts of sample prior to and after avidin bead elution were loaded on SDS-PAGE and analyzed by Western blotting by using affinity-purified anti-Cx43 antibody. The intensity of Cx43 bands was quantified by densitometry (NIH Image J).
Generation of Cx43(E2) Antibody and Use in Blocking Experiments-An antibody (Cx43(E2)) against the second extracellular loop region, E2, of Cx43 between amino acid residues 185 and 206 was generated. Initially, the corresponding cDNA fragment was amplified by PCR, subcloned into pGEX-2T vector (Amersham Biosciences), sequenced, and the protein was expressed in Escherichia coli. The fusion protein was isolated using glutathione-Sepharose 4B beads (Amersham Biosciences) and was used to generate antiserum in rabbits (Pocono Rabbit Farm & Laboratory, Canadensis, PA). Both preimmune sera and anti-Cx43 were collected from the same rabbit. Cx43(E2) was affinity-purified using a GST column, followed by a Cx43(E2) fusion protein-conjugated column with sequential washing with PBS followed by elution buffer containing 0.1 M glycine at pH 2.5. MLO-Y4 cells were incubated with affinity-purified anti-Cx43(E2) (28 g/ml), anti-Cx43 C-terminal domain(CT) (30 g/ml), anti-Cx43(E2) antibody pre-absorbed with antigen Cx43 (E2) fusion protein (0.6 mg/ml), or preimmune antibodies for 30 min prior to FFSS at 16 dynes/cm 2 for 30 min. Dye uptake experiments were conducted as described above.
Scrape-loading Dye Transfer Assay for Gap Junction-mediated Intercellular Communication-The scrape-loading dye transfer was based on a published protocol (33). Briefly, cells were scratched in the presence of two types of fluorescence dyes: RD (M r ϳ 10,000) and LY (M r ϳ 457). RD is too big to pass through gap junction channels and therefore serves as a tracer molecule for the cells originally receiving dyes. MLO-Y4 cells were treated in the absence or presence of 28, 56, and 100 g/ml of anti-Cx43(E2) antibody for 4 h, and then 1% RD and 1% LY were applied to the cells after scratching lightly with a surgical blade. After incubation for 5 min, cells were washed with Hank's balanced salt solution three times and finally fixed in freshly made 2% paraformaldehyde (from 16% stock) for 30 min. The dye transfer results were examined with an Olympus fluorescence microscope (Tokyo, Japan), and the degree of dye transfer was calculated based on the percentage of cells transferring LY dye.
Isolation of Chick Primary Osteocytes-Preparation of primary osteocytes from chick calvaria was based on previously published procedures (34) with some modifications. Briefly, calvarial bone was dissected from 16-day embryonic chicks and minced into small pieces. The soft tissues and osteoid were removed by collagenase treatment followed by decalcification using EDTA. The final particles were treated with collagenase and vigorously agitated to release osteocytes. The larger sized osteoblast/fibroblast cells were separated from osteocytes by incubating the cells with collagen-coated culture plates for 10 min, of which osteoblast/fibroblast were attached to the plates.
Statistical Analysis-Data were analyzed with one-way analysis of variance and Newman-Keuls multiple comparison test along with a biostatistics program (Prism). Data are presented as the mean Ϯ S.E. of multiple measurements. Asterisks represent the degree of significance compared with control (*, p Ͻ 0.05; **, p Ͻ 0.005; ***, p Ͻ 0.001). Some of the data presented were normalized to controls for easy comparison, and the variables were analyzed as described above.

Hemichannels Are Dynamic and Finely Regulated by Mechanical
Loading-We quantitated hemichannel activity with increasing magnitudes of FFSS. The opening of hemichannels, as detected by LY (M r ϳ 457) dye uptake, was directly proportional to the magnitude of FFSS applied (Fig. 1, A and B). Similar dye uptake results were obtained when these experiments were repeated with a larger probe, Alexa 594 (M r ϳ 760) (Fig. 1C), showing that hemichannel function was not limited by either the properties or the smaller size of LY. The lack of RD uptake (M r ϳ10,000), a dye too big to pass through hemichannels, excludes the possibility of membrane leakage or the cells being unhealthy. In addition, a close correlation between hemichannel opening with increasing FFSS and increased release of PGE 2 was also observed (Fig. 1D). To determine the effect of different types of FFSS on hemichannel function, steady (16 dynes/cm 2 ) and pulsatile (8 Ϯ 8 dynes/cm 2 ) flow were applied (Fig. 1E). As compared with non-treated controls, both steady and pulsatile FFSS treatment showed a similar increase in hemichannel opening. There was no significant difference in hemichannel opening between these two different flow types. Together, these results suggest that hemichannel opening, associated with the release of PGE 2 , is directly correlated to the magnitude of FFSS applied.
To determine the function of hemichannels after FFSS, we assessed hemichannel opening at various incubation periods after exposure to FFSS (Fig. 2). For all post-shear stress incubation periods (0 -8 h), except 24 h, a similar number of cells took up the LY dye ( Fig. 2A). Although the relative number of cells taking up the dye was significantly increased over control, there was no significant difference in dye uptake between 0 and 8 h after FFSS. At 24 h, however, the number of cells taking up dye returned to control levels. To assess the relative level of functional hemichannels in individual cells after FFSS, the intensity of fluorescence signals was measured (Fig. 2B). The levels of dye uptake reached a maximum immediately following FFSS (0 h) and gradually decreased over time (0.5-8 h), but was still elevated at significant levels compared with control and eventually returned to control levels after 24 h. These results suggest that up to 8 h after FFSS, even though the total number of cells that have functional hemichannels remains the same, the number of open hemichannels in individual cells gradually declines. The release of PGE 2 appeared to be closely associated with hemichannel function (Fig. 2C). This decrease in hemichannel function could be caused by either the gradual closure of hemichannels or/and decreased expression of Cx43 protein and therefore decreased numbers of hemichannels on the cell surface. To distinguish between these possibilities, Cx43 protein exposed on the cell surface was determined using a biotinylation assay (Fig. 2D). A significant increase in Cx43 protein on the cells surface was not observed until 2-h post-shear stress, which returned to control levels by 24-h post-shear stress. The results shown in Fig. 2 suggest that there is little direct correlation between Cx43 expression with either number of LYϩ cells uptaking dye ( Fig. 2A) or with dye intensity (Fig. 2B), not with PGE 2 release (Fig. 2C). These results suggest that decreased function of hemichannels up to 8 h after post-stress is primarily caused by a decrease in numbers of open hemichannels. However at 24 h, hemichannel closure does appear to correspond to the decreased cell surface expression of Cx43. Reduced expression of Cx43 may be partially responsible for the 24-h response.
To determine if response to reloading would affect hemichannel function, FFSS (16 dynes/cm 2 for 30 min) was applied to MLO-Y4 cells after 0.5, 4, and 24 h incubation after FFSS. The reapplication of 30 min of FFSS significantly re-stimulated the function of hemichannels as detected by dye uptake in osteocytes that showed modest (0.5 and 4 h) or significant (24 h) reduction in dye uptake with post-FFSS incubation (Fig. 3A). Although the effects of re-applying FFSS are similar and significant in cells from each of the post-stress periods, they are more dramatic after the longer post-FFSS incubation periods, i.e. after the hemichannels have had longer time to recover (Fig. 3, 24 h (Ϫ) versus 24 h (ϩ)). However, reapplication of FFSS results in a similar level of response regardless of length of the post-FFSS period.
To investigate how hemichannels respond to sustained mechanical stimulation, MLO-Y4 cells were subjected to continuous FFSS for various time periods. As shown in Fig. 1, low shear stress resulted in lower dye uptake than that from high stress (Fig. 1B), but still above control levels. We subjected MLO-Y4 cells to FFSS at 4 dynes/ cm 2 to ensure cell attachment during the 24 h of treatment. Compared with control, FFSS increased dye uptake significantly at 0.5 and 2 h of FFSS (Fig. 3B). However, the induction of hemichannel opening by FFSS decreased with longer periods of shear stress (4, 8, and 24 h). The surface expression of Cx43 in response to continuous FFSS was examined by cell surface biotinylation (Fig. 3C). Biotinylated Cx43 protein increased at a significantly in response to continuous FFSS at 2 and 4 h of FFSS, but decreased to the control level at 24 h. (No biotinylation was detected for intracellular ␤-actin, therefore. the possibility of potential biotinylation of intracellular Cx43 is unlikely.) Similar to the biotinylation results shown in Fig. 2D, the pattern of cell surface-expressed Cx43 in response to continuous FFSS does not exactly correlate with hemichannel activity, further suggesting that the opening of hemichannels is not solely regulated by the level of Cx43 protein expressed on the cell surface. Together, the data suggest re-application of FFSS to previously stressed osteocytes leads to the re-activation of hemichannels as early as 30 min after the All data are presented as mean Ϯ S.E. and n ϭ 3. B, the intensity of the fluorescence signal from cells at the same time points as A was analyzed using ImageJ software (NIH). Post-shear stress periods 0, 0.5, 2, 4, 8, and 24 h versus untreated control, C: ***, p Ͻ 0.001 and *, p Ͻ 0.05. All data are presented as mean Ϯ S.E. and n ϭ 3. C, conditioned media were collected after various post-FFSS periods, and the amount of PGE 2 was measured using a PGE 2 EIA kit. Post-shear stress periods 0, 0.5, 2, 4, 8, and 24 h versus untreated control, C: ***, p Ͻ 0.001 and **, p Ͻ 0.01. All data are presented as mean Ϯ S.E. and n ϭ 3. D, biotinylation was conducted for MLO-Y4 cells at various time points after FFSS at 16 dynes/cm 2 for 30 min. The cells were lysed, and equal amounts of total protein of control, C and FFSS samples were applied to avidin-conjugated beads. The biotin-labeled samples were purified after binding to avidin-conjugated beads. Cell lysates prior to loading on avidin-conjugated beads (preloaded, upper panel) and bound to avidin beads (biotinylated, lower panel) were detected by Western blot by using anti-Cx43 antibody. The relative ratio of biotinylated to total Cx43 was quantified using densitometric measurements of the band intensity (lower panel). The data are presented as mean Ϯ S.E. and n ϭ 3.
first round of FFSS stimulation. However, hemichannels appear to become less responsive to continuous FFSS and close over time in response to extended application of continuous shear stress. Similar to the post-stress experiments shown in Fig. 2, an increase in both cell surface protein expression and dye uptake occurs up to 4 h of continuous FFSS, but after 24 h, the decrease of dye uptake appears to correlate with a return to control levels of cell surface expression of Cx43, These results support the concept that a hemichannel gating mechanism is likely to be involved during the shorter periods of continuous FFSS, but reduced levels of cell surface-expressed Cx43 may be partially responsible for the decreased activity of hemichannels after 24 h of continuous FFSS.
Cx43-forming Hemichannels Play a Major Role in Regulating Osteocyte Response to FFSS-Clarification of the functional role(s) played by Cx43 hemichannels compared with gap junctions and other channels would be greatly advanced with specific blocking reagents for Cx43 hemichannels. Therefore, we generated a specific Cx43-hemichannel blocking antibody that targets the second extracellular loop (E2) of Cx43 (Fig. 4A). The generation of this antibody was not trivial as extracellular domains of connexins are highly conserved among different types of connexins from various species. This antibody (Cx43(E2)) recognizes Cx43, whereas no reactivity was observed using the preimmune antibody negative control (Fig.   4B). At the same dilution, the binding signal appeared to be even stronger than antibody against the C-terminal tail of Cx43 (Cx43(CT)), generated previously using a similar approach (35). Preincubating this antibody with its antigen, a GST fusion protein with the Cx43 E2 domain, but not GST protein alone, totally abolished this reactivity with Cx43 (Fig. 4C). This antibody failed to detect other connexins including Cx32, Cx26, Cx50, and Cx46 (data not shown). Immunostaining shows that in the presence of Triton X-100 (TX100) (Fig. 4D, panels D and F), comparable fluorescence signals are observed with Cx43(E2) antibody compared with that of Cx43(CT) antibody. While in the absence of detergent (panels C and E), only the labeling of Cx43(E2) antibody was observed, but no labeling for Cx43(CT) antibody, suggesting that the Cx43(E2) antibody recognizes the extracellular domain of Cx43 (Fig. 4D). The specificity of the Cx43(E2) antibody was further confirmed by preimmune and Cx43(E2) antibody pre-absorbed with antigen (Fig. 4D, panels A and B), where no specific signals were detected. Our previous preliminary study shows that gap junction activity appears not to be affected by this Cx43(E2) antibody (25). We showed here that gap junction function detected by scrape-loading dye transfer was not affected by increased concentrations of Cx43(E2) antibody (Fig.  4E). Cx43(E2) antibody significantly inhibited dye uptake induced by FFSS, but not by antibody against the cytosolic C-terminal domain of Cx43(CT) and preimmune antibody (Fig.  5A). In contrast, preincubating with antigen removed the capacity of Cx43(E2) antibody to block dye uptake induced by FFSS (Fig. 5B), further confirming the specificity of this antibody for hemichannels. Similar to MLO-Y4 cells, FFSS-induced dye uptake in primary osteocytes was significantly blocked by Cx43(E2) antibody, but not by preimmune antibody (Fig. 5C), showing that the MLO-Y4 cells function similarly to primary bone osteocytes. We have tested the specificity of this antibody in blocking hemichannnels, but not other membrane pores of similar sizes such as ATP-gated P2X 7 -associated pannexin channels (36). P2X 7 channels were reported to be involved in bone cell response to mechanical stimulation (37). Treatment with 1 mM ATP induced dye uptake, possibly mediated through pannexin channels because of the activation of P2X 7 receptor (36) (Fig. 5D). The opening of this channel was blocked by a P2X 7 channel inhibitor, oxidized ATP (1 mM), but not by the Cx43(E2) blocking antibody, demonstrating that this antibody is specific for blocking Cx43-hemichannels but not pannexin or The cells were lysed, and equal amounts of total protein from various continuous FFSS samples were applied to avidin-conjugated beads. Cell lysates prior to loading on avidin-conjugated beads (preloaded) and bound to avidin beads (biotinylated) were detected by Western blot by using anti-Cx43 or anti-␤-actin antibody (left panel). The relative ratio of biotinylated to total Cx43 was quantified using densitometric measurements of the band intensity (right panel). 2 and 4 h continuous FFSS versus non-flow control (0 h): ***, p Ͻ 0.001 and **, p Ͻ 0.01. All data are presented as mean Ϯ S.E. and n ϭ 3. P2X 7 channels. These results suggest that we have developed a potent antibody specific for Cx43-forming hemichannels and that FFSS-induced dye uptake is indeed mediated by the opening of Cx43-forming hemichannels.
Similar to the effects shown for dye uptake, FFSS-induced release of PGE 2 from hemichannels was significantly inhibited by Cx43(E2) but unaffected by preimmune antibody (Fig. 6A). Intracellular PGE 2 was also increased in response to FFSS (FF) (Fig. 6B); however, there was no difference in intracellular PGE 2 levels in Cx43(E2) or preimmune antibody-treated cells, which implies a potential feedback mechanism in maintaining stable levels of PGE 2 inside the cell. To test this hypothesis, the intracellular PGE 2 levels in cells exposed to FFSS for various time periods was measured (Fig. 6C). Interestingly, intracellular PGE 2 levels were increased even after 10 min of stress and continued to increase after 30 min and 1 h in the absence and presence of Cx43(E2) antibody. After 1 h of FFSS, intracellular PGE 2 in the Cx43(E2) antibody-treated cells was higher than preimmune antibody-treated cells, whereas no significant difference was observed during other time periods. These data suggest that intracellular PGE 2 levels are tightly regulated, maintaining a constant, possibly saturated level through a feedback inhibition of biosynthesis by the substrate or by degradation mechanisms. Together, these results show we have developed a potent antibody specific for blocking hemichannels formed by Cx43, which will prove invaluable to identify other functional aspects of hemichannels. More importantly, Cx43-forming hemichannels play an essential role in regulating the release of prostaglandins in response to mechanical stimulation.

DISCUSSION
In this study, we found that Cx43-hemichannels are dynamic channels capable of responding to different forms, magnitudes, and duration of mechanical loading. The opening of hemichannels is closely correlated with the release of PGE 2 , a factor directly involved in and essential for bone formation and remodeling in response to mechanical loading. Modulated release of cytoplasmic molecules, such as PGE 2 , by functional hemichannels in response to mechanical stimulation is likely to play a novel, important role in regulating anabolic/catabolic activities of bone cells. Moreover, an understanding of the dynamic aspects of hemichannels is important for finding associations between osteocyte response and bone adaptation, as they are shown here to be important for the temporal response to mechanical stimuli.
We show here that both uptake and release are magnitude-dependent with increasing response to increased applied shear stress most likely to accommodate an increased demand for exchange of signaling molecules in response to shear stress. Several theoretical models have merged over the years suggesting that bone cells can distinguish between different types of flow patterns, steady or pulsatile fluid flow (38 -40). Similar to our previous observations for gap junction channels (15), we observed no significant differences in hemichannel response with either steady or pulsatile fluid flow. These results do not support any one flow pattern as an exclusive player in regulating hemichannel response to mechanical loading.
During post-FFSS periods, although the number of cells that have functional hemichannels remained relatively constant up to 8 h after flow, the number of open hemichannels in individual cells, however, appeared to decrease. Using a biotinylation assay, we found a delayed increase of Cx43 protein on the cell surface compared with the opening of hemichannels during the same incubation time after FFSS. This delayed response may be due to the time for Cx43 to migrate to the cell surface to accommodate the demand for hemichannel function. In addition, our previous studies showed increased Cx43 protein expression during post-stress periods (15), reducing the possibility of reduced hemichannel closure because of a degradation of Cx43 protein, although the possibility of the relocation of Cx43 cannot be ignored. These results indicate that closure of hemichannels may occur via two phases; acute and chronic, through the decreased cell surface expression of Cx43. The acute early phase of channel opening involves Cx43 already expressed at FIGURE 4. Generation of an antibody specific for the extracellular domain (E2) of Cx43 which showed no effect on gap junction channels. A, using a GST-fusion protein approach, a potent Cx43-hemichannel blocking antibody that targets the second extracellular loop domain (E2) of Cx43 was generated. B, MLO-Y4 cell lysates were immunoblotted with anti-Cx43 C-terminal domain (Cx43(CT)), anti-Cx43 E2 domain (Cx43(E2)), or preimmune (Pre) antibody. C, MLO-Y4 cell lysates were immunoblotted with anti-Cx43(E2) antibody (Cx43(E2)) (1.06 l/ml), Cx43(E2) antibody (1.06 g/ml) premixed with GST-Cx43 E2 domain fusion protein (Cx43(E2)ϩE2) (11.2 g/ml) Cx43(E2) antibody premixed with GST protein (Cx43(E2)ϩGST) (11.2 g/ml) or anti-␤-actin antibody. D, MLO-Y4 cells were treated in the absence (panels C and E) or presence (panels A, B, D, and F) of Triton X-100 (TX100), then immunostained with anti-Cx43(E2) (panels C and D), anti-Cx43(CT) (panels E and F), preimmune (Pre) (panel A) antibody, or Cx43(E2) antibody pre-absorbed with antigen (Cx43(E2)AbϩE2) (panel B), and followed by labeling with FITC-conjugated anti-rabbit IgG secondary antibody (n ϭ 4). E, scrape-loading dye transfer analysis was conducted in the absence (0) or presence of anti-Cx43(E2) antibody (Ab) at 28, 56, and 100 g/ml for 4 h. the plasma membrane, whereas the chronic, delayed phase is related to trafficking of additional Cx43 molecules to the membrane in response to FFSS. This is most likely the reason for a lack of close correlation between activities of hemichannels and levels of cell surface-expressed Cx43. Another possibility may involve an association of Cx43 with other protein(s) capable of regulating hemichannel functions. This area of research is currently under investigation in our laboratory.
The tightly regulated opening/closing of Cx43-hemichannels observed in this study may have important implications for bone function in vivo. A continuously opened hemichannel may lead to continuous release of prostaglandin, which could be detrimental to bone. Although prostaglandins are known to be skeletal anabolic agents, they also have catabolic effects on bone and have been shown to stimulate osteoclast formation, activation, and resorption at higher concentrations and for longer durations (41)(42)(43)(44). Inappropriate opening of hemichannels may increase metabolic stress and subsequently lead to cell death, which would compromise the mechanosensory function of the osteocyte network.
Functionally adaptive remodeling is a process by which bone is continuously optimized to serve its mechanical needs throughout life (45). Previous in vivo models demonstrate a positive linear relationship exists between applied strain magnitude and a change in bone mass (46,47). Additional studies showed that bone can recover its sensitivity to mechanical stimuli to the same degree as previous exposure to load with the inclusion of rest periods (48). We demonstrate that this homeostatic control mechanism occurs at the level of the individual osteocyte and involves hemichannel function, when repeated mechanical loading enhanced the hemichannel response after the cells were permitted to recover. The highly responsive nature of hemichannels to mechanotransduction is likely to modulate the ability of osteocytes to adapt to repeated mechanical stimulation.
Hemichannel opening was decreased with increased duration of continuous flow. Similar to post-FFSS periods, the pattern of cell surface-expressed Cx43 does not exactly correlate with the activity of hemichannels. However, the reduction of hemichannel opening 24 h of flow could be attributed to a return to control levels of Cx43 protein on cell surface. The hemichannel response to continuous stimulation may be a mechanism to prevent persistent opening of these channels that could be detrimental to osteocytes. As the strain sensor in bone, exposure of osteocytes to short changes in FFSS may be all that is necessary to trigger a substantial osteogenic response. In addition, several in vivo studies have shown static strains fail to cause an osteogenic response in bone tissue and cultured cells (48 -50). Collectively, these studies support the hypothesis that hemichannels could function as mechanosensors of osteocytes, adjusting their activities in response to insertion of rest periods or to continuous shear, ultimately accommodating the needs of bone remodeling.
One of the major challenges for hemichannel research is related to the development of hemichannel, but not gap junction, specific reagents. A diverse range of chemical compounds and connexin mimetic peptides inhibit gap junction intercellular communication, but also act on hemichannels (25,(51)(52)(53). Connexin extracellular loop domain mimetic peptides, GAP26 and GAP27, were developed and shown to block hemichannels when treated for a short period of time (10 -30 min), while longer treatment (more than 1 h) inhibits gap junctions as well (54,55). In contrast to smaller molecules, such as chemical compounds and peptides, an immunoglobulin is a much larger protein (150K); therefore, access to gap junctional space is almost impossible. We found that 4 h of treatment with higher A, MLO-Y4 cells were treated with or without preimmune (Pre), Cx43(E2), or Cx43(CT) antibody in the absence (C), or presence of FFSS at 16 dynes/cm 2 (FF) for 30 min. Immediately after FFSS, LY dye uptake was performed, and the number of cells with dye uptake were counted and quantified. FFϩCx43(E2) versus FF, FFϩCx43(CT), and FFϩPre: ***, p Ͻ 0.001. B, cells were treated with or without premix of Cx43(E2) antibody with GST-Cx43 E2 domain fusion protein (Cx43(E2)ϩE2) in the absence and presence of FFSS at 16 dynes/cm 2 (FF) for 30 min. FF and FFϩCx43(E2)ϩE2 versus C: ***, p Ͻ 0.001. C, primary chick osteocytes were treated with or without Cx43(E2) antibody in the absence (C), or presence of FFSS at 16 dynes/cm 2 (FF) for 30 min. FFϩCx43(E2) versus FF, *, p Ͻ 0.05. D, MLO-Y4 cells were treated with or without (control, C) 1 mM ATP, 1 mM ATP plus Cx43(E2) antibody (ATPϩCx43(E2)), or 1 mM ATP plus 10 mM oxidized ATP (ATPϩoATP). LY dye uptake was conducted, and the numbers of cells uptaking dye were counted and quantified. ATPϩCx43(E2) and ATP versus C, ATPϩoATP: **, p Ͻ 0.01. All data are presented as mean Ϯ S.E. and n ϭ 3.
concentrations of Cx43(E2) antibody was still specific for blocking hemichannel, but not gap junction functions. Use of this antibody Cx43(E2) provides definitive evidence that the exit of PGE 2 from osteocytes is mediated through hemichannels based on the following: 1) specificity of the antibody only for hemichannels, 2) no reactivity or interference with gap junctions nor with other channels of comparable pore size, such as pannexin hemichannels; 3) the antibody is effective in blocking dye uptake, and 4) highly effective in blocking PGE 2 release. Consistent with our previous studies using general gap junction/hemichannel inhibitors (25), dye uptake and PGE 2 release were not completely blocked by this antibody under both basal and shear stress conditions. This could be interpreted as the incomplete effectiveness of blocking antibody or inhibitors used. Alternatively, this could be due to expression and function of other connexins even though Cx43 is the major connexin expressed in these cells. We previously found that the uptake and release of PGE 2 were not attenuated by inhibitors of other channels such as oxidized ATP for the purinergic receptor P2X 7 or the prostaglandin transporter PGT (25). The possible mechanism involved in mediating such low levels of dye uptake in MLO-Y4 osteocytes that is not blocked by this hemichannel-specific antibody will be pursued in future investigations.