Mechanically stimulated ATP release from murine bone cells is regulated by a balance of injury and repair

Bone cells sense and actively adapt to physical perturbations to prevent critical damage. ATP release is among the earliest cellular responses to mechanical stimulation. Mechanical stimulation of a single murine osteoblast led to the release of 70 ± 24 amole ATP, which stimulated calcium responses in neighboring cells. Osteoblasts contained ATP-rich vesicles that were released upon mechanical stimulation. Surprisingly, interventions that promoted vesicular release reduced ATP release, while inhibitors of vesicular release potentiated ATP release. Searching for an alternative ATP release route, we found that mechanical stresses induced reversible cell membrane injury in vitro and in vivo. Ca2+/PLC/PKC-dependent vesicular exocytosis facilitated membrane repair, thereby minimizing cell injury and reducing ATP release. Priming cellular repair machinery prior to mechanical stimulation reduced subsequent membrane injury and ATP release, linking cellular mechanosensitivity to prior mechanical exposure. Thus, our findings position ATP release as an integrated readout of membrane injury and repair.


Introduction
The mechanical environment is an important determinant of bone health, as emphasized by a consistent bone loss in astronauts exposed to microgravity or in paralyzed or bedridden patients (Nagaraja and Jo, 2014). Gravitational and muscle forces act on the skeleton during physical activity resulting in a complex combination of shear forces, strains and pressures. Bone-embedded osteocytes and bone-forming osteoblasts are widely regarded as the mechanosensitive cells in the skeletal system (Weinbaum et al., 1994).
Following mechanical stimulation of rodent and human osteoblasts, transient intracellular free calcium ([Ca 2+ ] i ) elevations and ATP release are among the earliest detectable events, which result in autocrine and paracrine purinergic (P2) receptor signaling (Robling and Turner, 2009;Romanello et al., 2001;Genetos et al., 2005). Vesicular release of lysosomes or secretory vesicles, or conductive release via channels such as Maxi anion channels, Volume-regulated anion channels (VRAC), Connexins or Pannexins are the main mechanisms of regulated ATP release in mammalian cells (Mikolajewicz et al., 2018). Pathological ATP spillage also occurs from traumatically damaged cells (Burnstock and Verkhratskii, 2012). Of interest, non-lethal membrane injury has been demonstrated in vivo in several tissues (McNeil and Steinhardt, 2003), including bone (Yu et al., 2017), under physiological conditions. The mechanism of facilitated cell membrane repair has been described and involves Ca 2+ /PKC-dependent vesicular exocytosis (Togo et al., 1999). However, the contribution of non-lethal cell injury to ATP release and related mechanotransductive purinergic signaling remains unclear.
The goal of this study was to examine the mechanism of ATP release from mechanically stimulated cells of the osteoblastic lineage. Since we have previously demonstrated that transient membrane disruption is required to induce global [Ca 2+ ] i elevations in osteoblasts (Lopez-Ayon et al., 2014), we were particularly interested in understanding the contribution of membrane injury to mechanically induced ATP release. Mechanical forces were applied by local membrane deformation or turbulent fluid shear stress in vitro to BMP-2 transfected C2C12 osteoblastic cells (C2-OB), primary bone marrow (BM-OB) and compact bone (CB-OB)-derived osteoblasts and changes in [Ca 2+ ] i , vesicular exocytosis, membrane permeability and ATP release were assessed. The prevalence of membrane injury in osteocytes at physiological and supraphysiological mechanical strain levels was investigated following in vivo cyclic compressive tibial loading of 10-week-old female C57Bl/6J mice.

Mechanically stimulated osteoblasts release ATP that induces calcium responses in non-stimulated neighboring cells
Osteoblasts from three different sources, C2-OB, CB-OB, and BM-OB, were loaded with [Ca 2+ ] i dye Fura2 and mechanically stimulated with a glass micropipette, which induced qualitatively similar transient global [Ca 2+ ] i elevations, consistent with prior work (Robling and Turner, 2009;Romanello et al., 2001;Genetos et al., 2005) (Figure 1A-C, Figure 1-video 1). L-type voltagesensitive calcium channel (VSCC) inhibitor Nifedipine and P2 antagonist PPADS significantly reduced the amplitude of mechanically-stimulated [Ca 2+ ] i transients ( Figure 1D). L-type VSCC activation occurred gradually ( Figure 1E) while the P2 receptor-driven component of the response peaked within seconds of stimulation ( Figure 1F). Together, L-type VSCC and P2 receptor-driven component accounted for~50% of the mechanical stimulated [Ca 2+ ] i transient. Consistent with previous reports (Robling and Turner, 2009;Romanello et al., 2001;Genetos et al., 2005), shortly after a single osteoblast was mechanically stimulated, neighboring cells exhibited delayed secondary [Ca 2+ ] i responses ( Figure 1G). Pharmacological interventions revealed that P2 receptors mediated the secondary response in all three osteoblast models, while a tendency for GAP junction involvement was eLife digest Athletes' skeletons get stronger with training, while bones weaken in people who cannot move or in astronauts experiencing weightlessness. This is because bone cells thrive when exposed to forces. When a bone cell is exposed to a physical force, the first thing that happens is the release of the energy-rich molecule called ATP into the space outside the cell. This molecule then binds to the neighboring cell to unleash a cascade of responses. ATP can exit the cell either through special canals in the cell membrane or released in tiny pod-like structures called vesicles. It is known that strong forces can injure the cell membrane and cause ATP to spill out. However, it is less clear how ATP is released when cells are subjected to regular forces. Mikolajewicz et al. investigated whether ATP exits through injured membranes of cells experiencing regular forces. Bone cells grown in the laboratory were gently poked with a glass needle or placed in a turbulent fluid to simulate forces experienced in the body. Dyes and fluorescent imaging techniques were used to observe the movement of vesicles and calculate the concentration of ATP in these cells.
The experiments showed that regular forces in the body do indeed injure the cell membranes and cause ATP to spill out. But importantly, the cells repaired the injuries quickly by releasing vesicles that patch the wound. As soon as the membrane is sealed, ATP stops coming out. From the first injury, cells adapted and quickly strengthened their membrane and repair system to be more resilient against future forces. This process was also seen in the shin bones of mice.
These results are important because knowing how bone cells sense, respond and convert physical forces can help us develop treatments for astronauts, the injured and aged.   Figure 1, *significance between treatment and control by ANOVA. Source data for Figure 1 is provided in Figure 1- observed in BM-OB responses ( Figure 1H). Puff application of 10 mM ATP mimicked the appearance of secondary responders in C2-OB ( Figure 1I).
We used a luciferin fluorescence-based imaging technique (Sørensen and Novak, 2001) to directly measure ATP release ([ATP] e ) ( Figure 2-figure supplement 1). Simultaneous recording of [ATP] e and [Ca 2+ ] i demonstrated that ATP release occurred within seconds after mechanical stimulation (time to peak release: 1.64 ± 0.36 s) and preceded the onset of [Ca 2+ ] i responses in neighboring cells ( Figure 2). We measured pericellular [ATP] e concentrations in the range of 0.05-80.5 mM corresponding to a release of 70 ± 24 amole ATP from a single mechanically stimulated osteoblast ( Figure 2B). Pericellular [ATP] e significantly correlated with the percentage of secondary responders ( Figure 2C). Vesicular ATP released upon mechanical stimulation is not the primary source of extracellular ATP Vesicular ATP release was proposed to be a major source of extracellular ATP (5). Using confocal microscopy, we found that an acidophilic marker quinacrine and a fluorescent ATP analog MANT-ATP co-localized in intracellular granular compartments ( Figure 3A; Figure 3-video 1-2). Quinacrine-positive vesicular pool was entirely released after treatment with Ca 2+ ionophore ionomycin, and only partially by mechanical stimulation ( Figure 3B; Figure 3-video 3-5), while negligible vesicular release was observed in unstimulated cells. Interestingly, vesicular release peaked within 20 s of mechanical stimulation ( Figure 3C, D), while ATP release within~2 s ( Figure 2A). Basal vesicular density was 70 ± 4Â10 À3 vesicles/mm 2 ( Figure 3E) and mechanical stimulation resulted in the  ]-depleted physiological solution, and micropipette-stimulated when indicated (+). Vesicular density (E) and cumulative release (F), secondary responsiveness (G) and [Ca 2+ ] response activation rates (H) were determined, n = 7-37 primary cells. (I) tFSS was applied by replacing 50% media volume n times. (J, K) ATP released per cell (left axis) or as percent of cellular ATP content (right axis) was measured following CB-OB stimulation by tFSS (J, black dashed line: rational function fit; red dashed line: corresponding asymptote) or after indicated pre-treatments followed by tFSS (K, 10x media displacements, +), n = 6-8 independent cultures. For Figure 3, means ± SEM, *significance compared to vehicle (E-H), basal ATP release (J) or to tFSS-stimulated vehicle (K) by ANOVA. Source data for Figure 3 is provided in Figure 3-source data 1.  cumulative release of 5 ± 1Â10 À3 vesicles/mm 2 over 100 s post-stimulation (7.2 ± 1.2% of the basal vesicular pool) ( Figure 3F). In [Ca 2+ ] e -free conditions vesicular density was unaffected, whereas mechanically evoked vesicular release was markedly suppressed ( Figure 3E, F). Treatment with N-ethylmaleimide (NEM) caused a 33 ± 2% increase in the vesicular density ( Figure 3E), consistent with a NEM-induced increase in the readily releasable vesicular pool (Lonart and Südhof, 2000), and a 55 ± 8% increase in mechanically-evoked vesicular exocytosis ( Figure 3F). Surprisingly, the percentage of secondary responders and activation rates for Ca 2+ responses (response amplitude divided by time for signal to increase from 10 to 90% of peak) were unaffected in [Ca 2+ ] e -free conditions, whereas NEM induced a significant decline in the percentage of secondary responders and activation rates ( Figure 3G, H). We measured bulk ATP release following mechanical stimulation by media displacement, a model for turbulent fluid shear stress (tFSS, Rumney et al., 2012, Figure 3I). Proportional to the extent of tFSS, 21 ± 11 to 422 ± 97 amol ATP/cell were released, approaching 31.5% of total intracellular ATP at high tFSS ( Figure 3J). tFSS (10x media displacements) induced release of 324 ± 50 amol ATP/cell, while ionomycin only led to release of 94 ± 5 amol ATP/cell. Moreover, tFSS-induced ATP release was significantly potentiated under [Ca 2+ ] e -free conditions (961 ± 88 amol ATP/cells) and inhibited in NEM-treated cells (41 ± 24 amol ATP/cells) ( Figure 3K). We also investigated conductive channels, however, pharmacological inhibition of Maxi-anion channels, P2X7, Connexins, Pannexins, T-type VSCC, TRPV4 or Piezo1 channels did not affect tFSS-stimulated ATP release (Figure 3-figure supplement 1). Although inhibition of L-Type VSCC had a small inhibitory effect (Figure 3-figure supplement 1), our data suggest that conductive channels cannot account for ATP release observed from mechanically stimulated osteoblasts. Importantly, our data show that increases in vesicular release were consistently associated with a reduction in mechanically stimulated ATP release, while inhibition of vesicular exocytosis increased ATP release from mechanically-stimulated osteoblasts.

Mechanical stimuli regularly and reversibly compromise the integrity of the osteoblast membrane
We hypothesized that mechanically stimulated ATP release is related to membrane injury. Adopting a Fura2-based dye-leakage assay (Togo et al., 1999), we examined changes in 340 ex/510 em Fura2 fluorescence following mechanical stimulation of osteoblasts ( Figure 4A). In~40% of micropipettestimulated osteoblasts, intracellular Fura2 fluorescence returned to pre-stimulation baseline (minor or no injury; mIn). However, in 30-40% of osteoblasts post-stimulation fluorescence was partially reduced (intermediate injury; iIn) and in 15-20% of osteoblasts completely lost (severe injury; sIn) ( Figure 4B). The extent of injury sustained by a micropipette-stimulated cell correlated with the magnitude and duration of [Ca 2+ ] i elevations ( Figure 4C-E), as well as the percentage of secondary responders ( Figure 4F), suggesting that ATP release is related to the extent of membrane disruption. In a dye uptake assay, we added membrane impermeable dyes, trypan blue (TB) or rhodamineconjugate dextran (R-dextran), to osteoblasts before or after mechanical stimulation ( Figure 4G). When the dyes were applied prior to tFSS, the proportion of dye-permeable osteoblasts positively correlated with the degree of tFSS. The extent of dye uptake was higher for TB (~900 Da) compared to R-dextran (~10 kDa) ( Figure 4H,I). However, when membrane permeability was examined 300 s after stimulation, a significantly lower percentage of cells were permeable to the dyes ( Figure 4H,I).
Hemichannel involvement was excluded, since hemichannel blockers did not affect mechanicallystimulated [Ca 2+ ] i elevations, membrane permeability or ATP release (Figure 4-figure supplement 1). Consistent with transient membrane disruption, lactate dehydrogenase (LDH,~140 kDa) leaked into the extracellular media proportionally to tFSS magnitude ( Figure 4J), even though cell viability was minimally affected 1 hr after tFSS stimulation ( Figure 4K). Examining TB uptake following micropipette stimulation, we have found that all C2-OB were TB-permeable immediately upon stimulation, decreasing to 75 ± 22% within 10 s, to 40-50% by 180 s and to 18 ± 12% by 300 s ( Figure 4L). Comparing these data to dye leakage experiments suggested that minor injuries resealed within 10 s, intermediate injuries within 20-180 s and severe injuries did not reseal by 300 s. ( Figure 4L). Based on predicted molecular radii of the dyes (Erickson, 2009), we estimated the membrane lesion radius to be on the nanometer scale, sufficiently large to permit the efflux of smaller ATP molecules (~507 Da) ( Figure 4M). Osteocyte membranes are reversibly disrupted during mouse in vivo tibia loading To determine whether mechanically induced repairable membrane disruptions occur in vivo, 10week-old female C57Bl/6J mice were injected with lysine-fixable Texas Red-conjugated dextran (LFTR-Dex,~10 kDa) either 30 min prior to or 20 min after cyclic compressive tibial loading to strain magnitudes at the upper level of physiological activities (600 me) or at supraphysiological osteogenic levels (1200 me) (Willie et al., 2013) ( . Arrows indicate direction of load. (C) The triangle waveform included 0.15 s symmetric active loading/unloading, with a 0.1 s rest phase (À1 N) between load cycles and a 5 s rest inserted between every four cycles. A maximum force of À5.5 N or À11 N was applied, which engenders 600 me or 1200 me, respectively at the periosteal surface of the tibia mid-diaphysis in these mice. (D) Experimental design schematic: animals were injected with LFTR-Dex 30 min before (red) or 20 min after (blue) in vivo cyclic tibial loading (5 min). (E) Proportion of cells exhibiting LFTR-Dex uptake in calvariae (n = 5 animals) and control tibiae (n = 20 from 10 animals), normalized to average total cell number, means ± SEM, **p<0.01. (F) Dye uptake in tibia injected before (red) or after (blue) loading at 600 me or 1200 me, means ± SEM, n = 5/ strain/dye protocol, normalized to non-loaded contralateral tibia, *significance compared to non-loaded tibia, † † significance between dye protocols, by ANOVA. (G) Contrast enhanced images of dye uptake in control (left), and 1200 me loaded tibia injected with dye before (middle) or after (right) loading. Source data for Figure 5 is provided in Figure 5-source data 1.  Figure 5E). Cyclic compressive tibial loading in the presence of LFTR-Dex resulted in significantly increased osteocyte dye uptake compared to unloaded bone, demonstrating that in vivo mechanical loading results in cellular membrane disruption ( Figure 5F,G, red). Importantly, when LFTR-Dex was administered 20-50 min after 600 me loading, the levels of cellular dye uptakes were similar to those in unloaded tibia, suggesting that damage evident immediately after loading was repaired within this time period ( Figure 5F, light blue). Of interest, when LFTR-Dex was administered 20-50 min after 1200 me loading, the proportion of osteocyte exhibiting dye uptake was significantly lower compared to non-loaded controls, suggesting that adaptive improvements in membrane integrity occurred in response to the supraphysiological strain ( Figure 5F,G, dark blue).

PKC regulates vesicular exocytosis and membrane resealing
Vesicular exocytosis has been demonstrated to facilitate membrane repair and improve membrane integrity in a PKC-dependent manner (Togo et al., 1999). We examined the role of Ca 2+ /PLC/PKC pathway during mechanically induced membrane injury and resealing. Pharmacological activation of PKC ( Figure  Together these data suggest that mechanically stimulated membrane injury and repair, vesicular release and ATP release are regulated by Ca 2+ /PLC/PKC-dependent signaling. We next examined which PKC isoform is involved in the mechanoresponse. Immunoblot analysis demonstrated that mechanical stimulation (using tFSS) of CB-OB activated conventional PKCa/b and PKD/PKCm, but not atypical PKCz/l or novel PKCd/q isoforms ( Figure 6E, Figure 6-figure supplement 4). PMA stimulated phosphorylation of conventional and PKD/PKCm isoforms, but not atypical or novel PKC isoforms, while only basal PKD/PKCm phosphorylation had a tendency (p=0.09) to be inhibited by Bis. Total PKC levels were unaffected by pharmacological interventions or mechanical stimulation ( Figure 6-figure supplement 4G). These findings suggest that the PKD/PKCm isoform regulates membrane resealing in osteoblasts.
To establish the relationship between vesicular release, membrane disruption and ATP release, we pooled data for the mechanically induced responses from all treatments studied in CB-OB ( Figure 6F-H). Mechanically induced vesicular release negatively correlated with the extent of membrane disruption ( Figure 6F) and the amount of ATP released ( Figure 6G), while membrane disruption positively correlated with amount of ATP released in response to mechanical stimulation ( Figure 6H). These data support a model in which mechanical disruption of the cell membrane leads to ATP spillage, while the extent of membrane resealing, regulated by Ca 2+ /PLC/PKC-dependent vesicular exocytosis, limits the amount of total ATP released (Figure 7).

Discussion
This study establishes a novel mechanism of mechanotransduction in bone cells, which positions ATP release as the integrated outcome of membrane injury (reflective of the destructive potential of mechanical forces) and membrane repair (countering and adapting to the damage). We report that mechanical stimuli in vitro and in vivo induced reversible cell membrane disruption. This disruption resulted in ATP release predominantly through the compromised membrane, rather than through vesicular or conductive mechanisms. Ca 2+ /PLC/PKC-dependent vesicular release was necessary for successful membrane repair and prevention of further ATP release. Importantly, prior activation of PKD/PKCm or vesicular priming improved membrane integrity and repair, thereby minimizing membrane injury and ATP release upon subsequent mechanical stimulation, and serving as a mechanism of cellular mechano-adaptive memory.

Membrane injury
We have demonstrated that non-lethal cell membrane injury is routine in mechanically stimulated cells. In vitro, single cell membrane deformation or fluid shear stress transiently compromised cell membrane integrity in 2-30% of cells, which was repaired within 10-100 s. Consistent with findings observed using in vitro models of mechanical stimulation, in vivo cyclic compressive tibial loading engendering physiological and supraphysiological strains resulted in transient osteocyte membrane disruption in 5-10% of cells. Interestingly, supraphysiological mechanical strains improved membrane integrity post-stimulation compared to baseline levels. Previously, non-lethal reversible cell wounding was shown to occur in mechanically active tissues, with membrane-impermeable marker uptake reported in 3-20% of cells in skeletal muscle, 3-6% in skin, 25% in cardiac muscle, 2-30% in lung and aorta (McNeil and Steinhardt, 2003). Consistent with our findings, membrane disruption was reported in 20-60% of long bone osteocytes after in vivo treadmill loading (Yu et al., 2017). The higher prevalence of disruption reported by Yu et al. in long bones was likely a consequence of cumulative tracer uptake over 18 hr (Yu et al., 2017), while our study examined membrane disruption over a 20-50 min period. Our data demonstrated that reparable membrane injuries are common and significantly contribute to ATP release from mechanically stimulated osteoblasts and osteocytes.

Membrane resealing
Vesicular exocytosis is believed to promote membrane repair via membrane insertion, thereby reducing membrane tension and allowing exposed hydrophobic residues to reseal (McNeil and Steinhardt, 2003). We have shown that vesicular release was critical for the repair of mechanically induced membrane injury and that the Ca 2+ /PLC/PKC pathway regulates vesicular exocytosis in osteoblasts. We interrogated the role of vesicular release in the mechanoresponse using NEM, a modulator of vesicular docking machinery (Lonart and Südhof, 2000). While NEM is commonly used as an inhibitor of vesicular release (Genetos et al., 2005;Kowal et al., 2015), it has also been shown to increase vesicular release probability (Kirmse and Kirischuk, 2006) through accumulation of readily-releasable vesicles (Lonart and Südhof, 2000). Using real-time imaging, we demonstrated that NEM increased the basal vesicular density and potentiated mechanically stimulated vesicular exocytosis in osteoblasts. In the presence of NEM, mechanically induced membrane injury and ATP release were significantly reduced. Vesicular exocytosis has been implicated as a mechanism of ATP release from osteoblasts and osteocytes (Genetos et al., 2005;Kringelbach et al., 2015;Romanello et al., 2005) based on the assumed inhibition of vesicular exocytosis by NEM or vesicular trafficking by brefeldin A, bafilomycin or monensin. However, our data demonstrated that the role of vesicles cannot be inferred without a direct validation of the effect of NEM on vesicular release. In our study, [Ca 2+ ] e depletion abolished mechanically stimulated vesicular release, interfered with membrane repair and increased ATP release. Calcium is one of the first mechanically induced signals that initiates downstream responses (Robling and Turner, 2009), including exocytosis-mediated membrane repair (Cooper and McNeil, 2015). It was previously shown that during in vivo loading of mouse bone, the number of osteocytes exhibiting [Ca 2+ ] i elevations increased proportionally to the magnitude of applied strain (Lewis et al., 2017). We found that PLC and PKCm, known downstream targets of Ca 2+ signaling (Mochly-Rosen et al., 2012), regulated basal vesicle abundance and membrane resealing. Our study suggests that the osteoporotic phenotype reported in PKCm-deficient bones (Li et al., 2017) may be associated with defective mechanoresponsiveness. Consistent with mechanisms of facilitated membrane resealing (Togo et al., 1999;Cooper and McNeil, 2015), we propose that mechanical stimulation results in membrane disruption, which leads to ATP efflux and Ca 2+ influx, triggering vesicular exocytosis and membrane repair that then limits ATP release (Figure 7).

Mechano-adaptive memory
Our findings suggest that osteoblast mechano-adaptive status may be influenced by PKCm/vesicular signaling. Activation of PKC prior to mechanical stimulation or priming the vesicular pool for release with NEM resulted in drastic decreases in membrane injury and ATP release in response to subsequent mechanical stimulation. Prior work has demonstrated that mechanical loading induces the production and release of LAMP-1-positive vesicles from osteocytes (LAMP-1 colocalizes with the quinacrine tracer-dye used in this study (Cao et al., 2014), which was critical for mechano-adaptive bone formation (Morrell et al., 2018). This mechano-adaptive effect may underlie cellular accommodation in the bone response to mechanical loading previously reported (Schriefer et al., 2005), and explain the reduced responsiveness of osteocytes exposed to higher frequency in vivo mechanical loading, compared to lower frequency loading of the same magnitude (Lewis et al., 2017). Our findings suggest that during repeated mechanical stimulation, the priming of PKCm/vesicular pathway by earlier stimuli will determine osteoblast resilience and responsiveness to subsequent cycles of mechanical stimulation, representing a mechanism for cellular accommodation.

Concluding remarks
In this study, we established the regulatory mechanisms involved in controlling the amount of ATP released following reversible cellular injury. We have demonstrated that membrane injury and repair are common under physiological stresses in vitro and in vivo, and determined the critical role for PKC-regulated vesicular exocytosis in bone cell membrane repair. We suggest that rather than delivering ATP to extracellular space, exocytosis of ATP-containing vesicles limits the much larger efflux of intracellular ATP through damaged membranes. We propose a model of biological adaptation to mechanical forces, which directly links mechanosensation through reversible membrane injury and ATP release to the development of adaptive resilience against the destructive potential of mechanical forces. PKC-mediated vesicular release provides a target for potential therapeutic interventions to modulate mechano-responsiveness and mechano-resilience in humans habitually experiencing altered mechanical environment, such as astronauts or paralysis patients.

Cell cultures
Three osteoblast models were used in this study.
All key experiments required to support our proposed model were performed using primary CB-OB cells; however, additional data from another primary source (bone-marrow-derived; BM-OB) and cell-line (C2-OB) were reported to emphasize the validity of our findings and demonstrate how different osteoblast models perform in this study.

Bone-marrow-derived osteoblasts (BM-OB)
To obtain bone-marrow-derived osteoblasts, 4-6 week C57Bl/6J mice were sacrificed and femur and tibia were collected. Bone pieces were sterilized in 70% EtOH and washed (3x, PBS). Bones were cut in half and placed into longitudinally-cut 1 mL pipette tips, with the exposed marrow end of the bone facing down. Pipette tips were centrifuged in Eppendorf tubes at 12 000 rpm (3x, 30 s) to isolate bone marrow cells. Bone marrow cells were plated at a density of 5Â10 4 cells per cm 2 in osteoblastic differentiation medium (aMEM supplemented with 10% FBS, 1% sodium pyruvate, 1% penicillin streptomycin and 50 mg/mL ascorbic acid) and media was refreshed every 2-3 days. Osteoblast phenotype was observed within 7-11 days.

Osteoblast phenotype
Cells were fixed with formalin (pH 7.4, 8 min) and rinsed with PBS (3x). Alkaline phosphatase (ALP) staining solution was prepared by combining 4.5 mg fast red violet salt dissolved in 3.75 mL H 2 O and 3.75 mL 0.2 M Tris-HCl, pH 8.3 with 0.75 mg Naphthol AS-MX phosphate disodium salt dissolved in 30 mL N,N-dimethylformamide. Sufficient staining solution was added to each well to ensure entire surface was covered (approx. 500 mL in 35 mm dish) and incubated at room temperature for 15 min, or until red precipitate formed. Cells were washed with H 2 O and observed under bright-field microscopy. Alkaline phosphate-positive cells were stained pinkish-red, while yellow hue was observed in alkaline phosphatase-negative cells.

Mechanical stimulation Local membrane deformation of a single cell
Single osteoblastic cells were stimulated by local membrane deformation with a glass micropipette using a FemtoJet microinjector NI2 (Eppendorf Inc.). The micropipette was positioned approximately 10 mm from the cell membrane at a 45˚angle from the horizontal plane and mechanically stimulated at a speed of 250 mm/s with a contact duration of 60 ms. Glass micropipettes were custom-made from glass capillary tubes (1.5 mm outer diameter, King Precision Glass Inc.) using a Sutter Pipette Puller (Flaming/Brown, model P-87).

Bulk stimulation by turbulent fluid shear (tFSS)
50% of the culture medium volume was manually withdrawn and replaced at an approximate speed of 2 resuspensions per second for an indicated number of n times (e.g., 10x tFSS corresponded to 50% vol displacement performed 10 times).

Intracellular calcium recording and analysis
Cells plated on glass-bottom plates (MatTek Corporation) were loaded with a ratiometric fluorescent calcium dye Fura2-AM (r.t., 30 min), washed with PS and allowed to acclimatize for 10 min to reduce the effects of mechanical agitation. [Ca 2+ ] i was recorded at a sampling rate of 2 images per second (2 Hz) using a Nikon T2000 fluorescent inverted microscope with 40x objective (Tiedemann et al., 2009). Recordings consisted of~10 s baseline [Ca 2+ ] i and additional 110-170 s following application of mechanical stimulation or treatment. The excitation wavelength was alternated between 340 and 380 nm using an ultra-high-speed wavelength switching illumination system (Lambda DG-4, Quorum Technologies). Regions of interest (ROI) were manually defined and the ratio of the fluorescent emission at 510 nm, following 340 and 380 nm excitation (f340/f380), was calculated and exported using the imaging software (Volocity, Improvision). All data were imported into an excel spreadsheet for characterization using a MATLAB algorithm previously described (Mackay et al., 2016). The following parameters were obtained for statistical analysis: amplitude (amp; magnitude of response), activation time (t 10%-90% ; time between 10% and 90% of maximum response), and decay constant (t decay ) of exponential decay region of deactivation phase of transient response. For micropipette stimulation experiments, secondary responsiveness was calculated as the percentage of neighboring cells exhibiting [Ca 2+ ] i elevations.

Vesicle labeling, imaging, and analysis Localization
To visualize the intracellular distribution and colocalization of quinacrine and MANT-ATP, cells were loaded with MANT-ATP (5 mM, overnight, 37˚C) followed by quinacrine (10 mM, 15 min, room temperature). Cells were washed with PS and imaged using a Zeiss LSM 780 laser scanning confocal microscope. An argon and diode laser were used as light sources (ex. 405 nm and 458 nm) and the detection wavelength ranges were set to 410-470 nm and 490-570 nm to ensure separation of overlapping emission spectra, for MANT-ATP and quinacrine, respectively. Images were acquired using a Zeiss Plan-Apochromat 63x/1.40 Oil immersion objective lens. Subsequent analysis was carried out using ImageJ (National Institutes of Health, USA). Vesicular density was determined using ImageJ by counting number of intracellular quinacrine-positive puncta and normalizing count to cell surface area.

Kinetics
To study vesicular release kinetics, time-lapse recordings of quinacrine-loaded cells were acquired with a Nikon T2000 fluorescent inverted microscope after a 10-min period of post-stain acclimatization to minimize the effects of mechanical agitation. Quinacrine-rich vesicles were imaged using a 40x objective at a sampling frequency of 2 Hz for 2 min. To minimize the phototoxic effects of quinacrine, the excitation intensity was set to 33% of peak intensity and the shutter was shut in between 100 ms exposures. Image sequences were exported using the imaging software Volocity (Improvision) for subsequent analysis in ImageJ. To quantify the kinetics of vesicular release, sudden losses of fluorescence were monitored. Vesicles that gradually came in and out of focus did not qualify as release events. SparkSpotter (Babraham Bioinformatics, UK), an ImageJ plugin, was used to identify release events in temporally reversed and contrast-enhanced image stacks, so that the loss of fluorescence would instead be detected as a sudden increase by the spark identifying algorithm. Acquired data was visually evaluated and represented as either release events per second or cumulative vesicular release for indicated time frame, normalized to unit area.

ATP measurements
Fluorescence recordings of [ATP] e 10 nM D-luciferin-and 2 mg/mL fire-fly luciferase-containing physiological solution (PS) was added to C2-OB and extracellular luciferin fluorescence was recorded (380 ex/510 em). Using an ATP standard calibration curve (Figure 2-figure supplement 1), pericellular (<15 mm from stimulated cell) ATP concentrations [ATP] were obtained. The amount of ATP released by individual cells was estimated by fitting the diffusion profile obtained from the fluorescence measurements with the diffusion equation for release from an instantaneous plane source (Crank, 1975).
where [ATP](x,t) is ATP concentration at distance x from the source, at time t from the moment of release. M is the amount of ATP released at x = 0, t = 0. D is ATP diffusion coefficient, assumed to be 3.3Â10 -6 cm 2 /s at 22˚C (Newman, 2001).

Bioluminescent recordings of [ATP] e
Bioluminescence assay was used to measure ATP content in supernatant samples. Luciferin-luciferase reaction solution contained 0.45 mM D-luciferin and 40 mg/mL firefly-luciferase in bioluminsecence reaction buffer. 10 mL supernatants were sampled after tFSS stimulation, added into a glass cuvette containing 100 mL of luciferin-luciferase reaction solution, and measured using FB 12 single tube luminometer. Total cellular ATP content was measured in RIPA-lysed samples.

Membrane integrity assays Dye leakage
Fura2-loaded osteoblastic cells were mechanically stimulated with a glass micropipette and 510 nm emission intensity was monitored after 340 nm excitation (Togo et al., 1999). Unless stated otherwise, Fura2 leakage refers to the percentage of cells that were severely injured (sIn; complete loss of intracellular Fura2 fluorescence) following micropipette stimulation.
Dye uptake 0.08% TB or 10 mM R-dextran were added before or 5 min after mechanical stimulation, washed with PS, and the proportion of dye-positive cells were assessed by microscopy.

Cytosolic leakage
Extracellular lactate dehydrogenase (LDH) was measured using Cytotoxicity Detection Kit PLUS . Total cellular LDH content was measured in RIPA-lysed samples.

Cell viability
1 hr after tFSS, AlamarBlue assay was performed according to manufacturer's protocol. Cell viability was reported as percentage of total cells.

In vivo bone loading
10-week-old female C57Bl/6J mice were injected intraperitoneally with LFTR-Dex either 30 min before or 20 min after loading. In vivo cyclic compressive loading at 600 me or 1200 me was applied for 5 min (Willie et al., 2013). Mice were randomized into two groups, anesthetized, and in vivo cyclic compressive loading was applied to the left tibia (216 cycles at 4 Hz the mean mouse locomotor stride frequency [Clarke and Still, 1999]), delivering a maximum force of À5.5 N and À11 N, which engenders 600 me or 1200 me, respectively at the periosteal surface of the tibia mid-diaphysis in these mice determined by prior in vivo strain gauging studies (Birkhold et al., 2014). Previous reports have shown strain levels of 200-600 me are engendered on the medial tibia during walking in the mouse (De Souza et al., 2005;Sugiyama et al., 2012). Strain levels of 1200 me are considered supraphysiological resulting in a robust bone formation response after 5 days of mouse tibial loading (Willie et al., 2013;Brodt and Silva, 2010) and altered gene expression after only a single bout of loading (Zaman et al., 2010;Kelly et al., 2016). Mice ambulated freely after controlled loading, and were euthanized 50 min post-dye-injection, weighed ( Figure 5-figure supplement 1A) and tibiae and calvariae were dissected and fixed in formalin. Tibia lengths were measured ( Figure 5-figure supplement 1B) and mid-shaft tibiae were cut into fragments, which were immunofluorescently labeled for sclerostin, counterstained with DAPI, and visualized with Nikon T2000 fluorescence inverted microscope. Right non-loaded tibiae were used as an internal control. Female mice were used instead of males to minimize in-cage aggression and related variations in background mechanical-loading. Experimenter was blinded during image analysis.

GAP junction functional assay
Cells were scraped in the presence of GAP-junction permeable dye Lucifer yellow (LY, 10 mM), incubated for 2 min (37˚C), rinsed and fixed with formalin (pH 7.4, 8 min). Dye transfer was visualized using wide-field epifluorescence microscopy and cell-cell coupling was visually determined by bright-field microscopy. Cells that were not initially scraped but were physically coupled and LY-positive after 2 min of staining were reported as a percentage of total cells that were physically coupled to scraped neighbors.

Statistical analysis
Data are representative results or means ± standard errors (S.E.M). Curve fittings and [Ca 2+ ] i transient characterization was done in MATLAB (MathWorks). Sample sizes for in vitro experiments indicate the total number of biological replicates (specified in figure legends as number of stimulated cells or number of independent cultures, which were isolated from different mice for primary OB, or represent different plating dates for C2-OB) pooled across a minimum of three independent experiments. For in vivo experiments, sample sizes indicate number of animals. Statistical significance was assessed by ANOVA followed by Bonferroni post-hoc test, significance levels are reported as single symbol (*p<0.05), double symbol (**p<0.01) or triple symbol (***p<0.001).