Mechanical loading resets the circadian clocks in femoral head cartilage and IVD tissue
We hypothesized that if mechanical loading is a key entraining factor for circadian rhythm in skeletal tissues, physical activity during the mouse resting period should shift the phase of the clock. To test this, voluntary running (e.g. on a running wheel) is not suitable as mice will not normally run during the daytime. The treadmill is the methodology of choice because it simultaneously allows for precise control of the time of exercise, intensity, and volume. PER2::Luc reporter mice were gradually adapted to daily treadmill running (45 min at 15 m/min speed) to avoid a stress response. Running occurred at ZT 2 (2 h since the resting phase), the predicted trough of Per2 gene expression in skeletal tissues based on our RNAseq data (Fig. 1a). After 5 days of treadmill exercise, mice were sacrificed immediately after running and tissues (SCN, femoral head cartilage, and IVDs) were harvested for explant cultures. Recording of PER2::Luc bioluminescence from the tissues showed no effect of such exercise bouts on the circadian rhythm of the SCN. However, the circadian phase of the clocks in cartilage and IVD tissues was advanced by ~8 h, indicating a decoupling effect of physical exercise between the skeletal clocks and the central SCN clock (Fig. 1b, c). To exclude the potential metabolic or systemic effects of exercise on skeletal clocks, we tested directly the role of mechanical loading using an ex vivo PER2::Luc tissue explant culture model and the FlexCell compression system. A short loading regime (1 h of 1 Hz, 0.5 MPa compression) resulted in a robust amplitude increase of the circadian rhythm in cartilage which was maintained for at least three more days (Fig. 1d). To delineate the best response window within the 24-h cycle, we applied the same compression protocol at 4 time points 6 h apart. Only compression applied at the peak of PER2::Luc resulted in a significant increase of circadian amplitude, with a minimal phase shift. Compression at other phases resulted in significant phase delay or advance (p < 0.001, Fig. 1e, f). Compression applied at the trough disrupted circadian rhythm, highlighting the importance of the circadian phase in modulating the clock response to loading (Fig. 1e, f). The phase shift of the cartilage circadian rhythm was dependent on the magnitude of the force of compression (Fig. 1g).
The same loading protocol altered the PER2::Luc rhythm in IVD tissues in a similar manner (Fig. 1h, I and Fig. S2a). Next, we tested whether rhythmic loading patterns can entrain circadian rhythms of endogenous clock genes in IVD cells. To this end, rat IVD cells were subjected to oppositely phased cycles of mechanical loading at 0.5 MPa of 12-h loading/12-h unloading. qPCR showed that the expression of clock genes Bmal1 and Cry1 were driven ~180 degrees out of phase (Fig. 1j), further supporting rhythmic loading as an endogenous clock resetting cue for cells in the skeletal system.
Hyperosmolarity (but not hypo-osmolarity) phenocopies the effect of mechanical loading
The diurnal pattern of mechanical loading of the articular cartilage and IVDs is associated with daily changes in osmotic pressure. Under load, the pressurized interstitial fluid flows to regions of lower pressure, resulting in increased osmolarity. When unloaded, the process is reversed, causing a return to normal osmolarity. These fluctuations in osmolarity play an important role in normal skeletal physiology by promoting the synthesis of the extracellular matrix and stabilizing cellular phenotypes24,25,26. Therefore, we hypothesized that fluctuations in osmolarity associated with loading might mediate the exercise-induced clock changes. We first tested how circadian rhythms in skeletal tissues operate under different baseline osmolarity conditions. PER2::Luc IVD explants were cultured under static osmotic conditions of 230–730 mOsm for 3 days then resynchronized with dexamethasone and recorded in media at the adapted osmolarity as indicated. Here, despite the wide range of osmotic conditions, the intrinsic circadian pacemaking mechanism was still intact with equivalent circadian amplitude and phase, highlighting the ability of these skeletal tissues to adapt to static changes in their environmental osmolarity (Fig. 2a). The periodicity was maintained at close to 24 h despite drastic differences in baseline osmolarity, indicating that circadian period in these skeletal tissues is “osmolarity-compensated” (Fig. 2a).
Next, we analyzed how an acute change in osmolarity (as experienced by skeletal tissues on a daily basis) impacts skeletal clocks. Cartilage tissue explants were placed in iso-osmotic (330 mOsm) or hyperosmotic (530 mOsm) media and allowed to adapt for 3 days before synchronization and recording. When clocks gradually become desynchronized, the conditioned media were swapped between iso and hyper-adapted explants. The clock amplitude in explants that experienced an increase in osmolarity was increased to a level equivalent to the beginning of the recording, and a robust circadian rhythm continued for the next 4 days. In contrast, explants experiencing a decrease in osmolarity showed a gradual loss of rhythmicity, similar to the iso-control explants (Fig. 2b). After 4 days the media were swapped back between explants and again the tissues experiencing an increase in osmolarity showed enhanced circadian rhythm (Fig. 2b). Clocks in IVD explants showed a very similar response to that of cartilage, with improved oscillation amplitude by increased osmolarity (Fig. S2b and Movie S1). To assess whether hyper-osmolarity synchronizes clocks at an individual cell level and to evaluate the percentage of cells that respond, single-cell fluorescence imaging of PER2::Venus mouse primary chondrocytes were performed. We observed increased nuclear PER2 signal following hyperosmotic exposure (Fig. 2c, d) and clock re-synchronization in 66% of cells (Fig. 2e).
If altered osmolarity is a key mechanism through which mechanical loading acts on the skeletal clock, it should exert very similar phase- and dose-dependent effects to mechanical loading. Indeed, a hyperosmotic challenge elicited a very similar direction of phase shifts in a circadian phase-dependent manner comparable to loading (Fig. 2f, Fig. S3). The phase response curve (PRC) and the phase transition curve (PTC) demonstrated a type 1 resetting (typical response to a relatively mild physiological clock stimulus) (Fig. S2c, d). The hyperosmolarity-induced clock-synchronizing effect appears much stronger than loading with improved circadian amplitude at all timepoints except at the trough (Fig. 2g). Similar to loading, the circadian clock response to osmolarity is also “dose” dependent, with a phase delay of up to 9.5 h (at mid-descending phase) for the +400 mOsm increase (Fig. 2h). A significant amplitude effect was detectible at an increase as small as +20 mOsm, and maximal amplitude induction was observed with +300 mOsm condition, suggesting higher osmotic challenge may exceed the physiological range (Fig. 2h, Fig. S4). To further validate the responses of molecular circadian clocks to loading and osmolarity, we used cartilage and IVD tissue explants from a different clock reporter mouse model, the Cry1-Luc which is a promoter reporter27 as opposed to the PER2::Luc fusion protein reporter28. The dose-dependent clock resetting effect by both mechanical loading and osmolarity was also observed in cartilage and IVD explants from Cry1-Luc mouse (Fig. S5).
Daily hyperosmotic challenge synchronizes dampened circadian rhythms in both young and aging skeletal tissues
With the onset of the activity phase, mouse cartilage and IVD tissues experience roughly 12 h of loading within a 24-h day, leading to increased osmolarity22. Having determined that the acute increase in osmolarity is a clock synchronizing factor, we next used a protocol to approximate the diurnal osmotic cycles experienced by skeletal tissues by exposing cartilage and IVD explants to 12 h of hyperosmotic condition and returning them to 12 h of iso-osmotic medium. A single cycle with a hyperosmotic challenge as low as +100 mOsm synchronized the circadian rhythm in cartilage (Fig. 3a) and IVDs (Fig. 3b) from young mice (2 months old) that have been in culture for 5 days, although, this effect seems to be dependent on the extent of osmotic change, or the number of cycles applied, or both. When explants were exposed to two daily cycles of +200 mOsm challenges, they showed an even stronger clock amplitude (Fig. 3c, d).
We have previously shown that the circadian amplitude of IVD and cartilage rhythms dampen with aging18,20. Therefore, we tested whether clocks in aging skeletal tissues still respond to osmotic cycles. Indeed, exposure of aging explants to two osmotic cycles resulted in resynchronization in both articular cartilage (Fig. 3e) and IVDs (Fig. 3f).
Hyperosmotic challenge induces rhythmic global gene expression patterns in a cell-type-specific manner
To gain mechanistic insights into the clock resetting mechanisms, we initially investigated how chondrocyte circadian clocks responded to hyperosmotic challenge in vitro. Clock-unsynchronized primary chondrocytes from PER2::Luc mice exhibit only a low amplitude oscillation. Exposure to +200 mOsm increase in osmolarity augmented PER2::Luc amplitude (Fig. 4a). Similar observations were made in a human IVD annulus fibrosus cell line (Fig. S6a). Strikingly, the same hyper-osmotic stimuli disrupted circadian oscillation in U2OS cells, a human osteosarcoma cell line widely used as a cellular model of circadian clocks, as well as in keratinocytes (Fig. S6b and c). As such, the osmolarity-entrainment of circadian clock is likely cell-type dependent and could indicate cellular adaptation to their local niche.
We next determined to what extent hyperosmolarity can synchronize circadian rhythms of gene expression at the transcriptome level by circadian time-series RNA sequencing in primary mouse chondrocytes. The PER2::Luc reporter allowed us to track clock rhythms in parallel cultures (Fig. 4a). mRNA samples were collected every 4 h for 2 full circadian cycles, starting with samples just before the osmotic increase (ZT 0). Principal component analysis of the RNAseq results revealed the circadian time of sampling as the biggest factor separating the samples (Fig. 4b). Analysis revealed 1312 rhythmic genes using integrated p < 0.05 cut-off threshold (254 rhythmic genes with a BHQ < 0.05 cut-off), including most core clock genes (Fig. 4c, d; Supplementary Data 1). Following hyperosmolarity, 1557 genes showed significant differential expression (with >2-fold change) between T0 and T4 (Fig. 4d, Supplementary Data 2). Importantly, most core clock genes appeared as early response genes 4 h after osmotic stress, with significant upregulation of Bmal1 (Arntl), Tef, Per1, Clock, Cry1/2, Rora and Nfil3, and downregulation of Nr1d1, Per2/3, and Dbp (Fig. 4e).
Treadmill running elicits transcriptome-wide changes in mouse cartilage and IVD tissues
As previously, mice were gradually adapted to treadmill running for 5 days after which they run at ZT 2 for 45 min for 5 days. Tissues were collected on the last day immediately after treadmill running. Control sedentary littermates were collected at the same time. PCA analysis of the RNAseq results showed clear separation of sedentary and running samples both in cartilage and IVDs (Fig. S7a). Differential expression analysis showed 421 upregulated and 261 downregulated genes in cartilage and 253 upregulated and 470 downregulated in IVDs (Fig. 4f, h, Supplementary Data 7 and 8). Among clock genes Bmal1 (Arntl) and Npas2 were consistently downregulated and Per1/2, Nr1d2, Tef and Dbp consistently upregulated in both tissues (Fig. 4g, i). Next, we compared the differentially expressed genes in cartilage tissues from treadmill running mice with our published circadian cartilage transcriptome19. In cartilage close to 1/3 of genes differentially expressed in the treadmill experiment were rhythmic in the time-series dataset (Fig. S7b). Most importantly, the expression pattern of these genes in the treadmill running samples which were collected at ZT 2 resembles that of timeseries samples collected at ZT 14-18 (Fig. 4j and Fig. S7c, d), consistent with a global shift of gene expression by exercise timing. Ingenuity Pathway Analysis (IPA) analysis showed Osteoarthritis, HIF1α, circadian rhythm and PI3K/AKT pathways as significantly regulated by both osmotic stress and treadmill running (Fig. 4k and Supplementary Data 3, 5, 9, 11). Upstream regulators common to all three datasets (cartilage run, IVD run, and chondrocyte hyperosmolarity) included BMAL1-Clock complex, p38, ERK, PI3K, AKT, mTOR, TSC2, PP2A as well as CREB, Forskolin, cAMP, Ca2+, NFAT5, HSF1, TGFβ and FOXO1, 3 and 4 (Fig. 4l and Supplementary Data 4, 6, 10, 12).
PLD2-mTORC2-AKT-GSK3β as a convergent pathway for loading- and hyperosmolarity- induced clock resetting
Based on the IPA analysis we selected several candidate pathways for further analysis. Pharmacological inhibition of pathways such as cAMP/CREB, calcium channels, Rho/ROCK, p38 and ERK failed to block clock responses in tissues to osmolarity and mechanical loading (Fig. S8-S11). One pathway that featured prominently in Upstream Regulators analysis was mTOR (Fig. 4l). Therefore, we tested two mTOR inhibitors, Torin1 (blocking both mTORC1 and mTORC2 complex) and Rapamycin (primarily blocking mTORC1). While Torin1 completely blocked the increase in amplitude following hyperosmotic challenge in cartilage and IVD explants, Rapamycin had no effect (Fig. 5a, b), indicating involvement of mTORC2. AKT is a known phosphorylation target of mTORC229 and can be activated by hyperosmolarity in renal cells30. Indeed, pre-treatment of cartilage explants with an AKT inhibitor prevented the increase in circadian amplitude following hyperosmotic challenge (Fig. 5c). Western blotting showed an increase in phosphorylation of AKT at the mTORC2 site 8 h after an increase in osmolarity. Pre-treatment with Torin1 but not with rapamycin prevented this phosphorylation (Fig. 5d and Fig. S12). AKT is a known upstream regulator of GSK3β activity, a key clock-regulating kinase implicated in the regulation of the stability and/or nuclear translocation of PER2, CRY2, CLOCK, REV-ERBα and BMAL131,32. Indeed, inhibition of GSK3β activity by lithium had a synchronizing effect on IVD clocks similar to that of hyperosmolarity (Fig. 5e). Concordantly, hyperosmotic challenge increased GSK3β phosphorylation at Ser9 and Ser389 (corresponding to reduced activity) (Fig. 5f). Pre-treatment with Torin1 (but not Rapamycin) or an AKT inhibitor decreased GSK3β phosphorylation at Ser9 and Ser389 following hyperosmotic challenge (Fig. 5f and Fig. S10). We next verified whether the above mechanisms were also involved in clock resetting by mechanical loading. Pre-treatment with the mTORC2 or AKT inhibitor (but not Rapamycin) completely blocked the mechanical loading-induced increase in circadian clock amplitude (Fig. 5g, h, Fig. S10). Consistent with our findings, mechanical strain in mesenchymal stem cells initiates a signaling cascade that is uniquely dependent on mTORC2 activation and phosphorylation of AKT at Ser-473, an effect sufficient to cause inactivation of GSK3β33.
Finally, the lack of involvement of membrane calcium channels (Fig. S9a-d) and Rho/ROCK pathway (Fig. S9e) prompted us to explore what cell surface mechano-sensors might mediate the clock resetting effects of loading and osmolarity. PLD2 is a plasma membrane tension sensor that connects the mechanical forces involved in cell swelling and shrinking as well as external stimuli with the mTORC-AKT pathway34,35. PLD2 catalyses the conversion of PIP2 to Phosphatidic Acid which stabilizes the mTORC complexes36. Pre-treatment with the PLD2 inhibitor prevented amplitude increase for both the osmotic stress and mechanical loading of cartilage explants (Fig. 5i, j). Western blotting showed that the PLD2 inhibitor prevented phosphorylation of AKT at the mTORC2 site as well as phosphorylation of GSK3β after hyperosmotic stress (Fig. 5k).