Gremlin-2 is a BMP antagonist that is regulated by the circadian clock

Tendons are prominent members of the family of fibrous connective tissues (FCTs), which collectively are the most abundant tissues in vertebrates and have crucial roles in transmitting mechanical force and linking organs. Tendon diseases are among the most common arthropathy disorders; thus knowledge of tendon gene regulation is essential for a complete understanding of FCT biology. Here we show autonomous circadian rhythms in mouse tendon and primary human tenocytes, controlled by an intrinsic molecular circadian clock. Time-series microarrays identified the first circadian transcriptome of murine tendon, revealing that 4.6% of the transcripts (745 genes) are expressed in a circadian manner. One of these genes was Grem2, which oscillated in antiphase to BMP signaling. Moreover, recombinant human Gremlin-2 blocked BMP2-induced phosphorylation of Smad1/5 and osteogenic differentiation of human tenocytes in vitro. We observed dampened Grem2 expression, deregulated BMP signaling, and spontaneously calcifying tendons in young CLOCKΔ19 arrhythmic mice and aged wild-type mice. Thus, disruption of circadian control, through mutations or aging, of Grem2/BMP signaling becomes a new focus for the study of calcific tendinopathy, which affects 1-in-5 people over the age of 50 years.

D iseases and injury to tendons are a major biomedical problem. Rotator cuff disease, repetitive motion injuries, complications following iatrogenic injury, and tendon rupture, collectively affect 1-in-3 people over the age of 40 years and incur a significant sociological and economic burden (see 1,2 and references therein). Calcium deposits in the mid-substance of the tendon are common in rotator cuff tendons, Supraspinatus, Achilles and patellar tendons resulting in calcific tendinopathy (CT), which has a prevalence of between 2.7% and 22% depending on age and gender (reviewed by [3][4][5][6][7]. Thus, tendinopathies are the most common of the arthropathy disorders. However, the pathophysiology is not always associated with overuse, indicative of an underlying defect in tissue homeostasis as a result of age. An under-researched aspect is the dramatic differences in physiological activity of tendons during a 24-hour cycle of being asleep and awake, and if these differences are influenced by aging and gene expression changes. The circadian clock coordinates the 24-hourly rhythms in biochemical and behavioral parameters that help prepare an organism for anticipated changes in its environment. In addition to the central pacemaker in the suprachiasmatic nuclei (SCN), many peripheral organs contain autonomous peripheral clocks 8,9 . Mammalian circadian clocks rely on the transcriptional/translational feedback loops consisting of the transcriptional activator complex CLOCK and BMAL1 (bone and muscle ARNT-like 1), and repressor complex PERIOD1/2 and CRY1/2. PERs/CRYs periodically inhibit the CLOCK/BMAL1-mediated transactivation of the Pers and Crys through Ebox elements in their promoters. The core oscillator then drives rhythmic clock-controlled genes through regulatory elements in their promoters (e.g. E-box, D-box, and ROREs).
Importantly, mice with a deletion of Bmal1 show age-related and non-inflammatory arthropathy, including ectopic tendon calcification 10 . It was unknown if the tendon phenotype was the result of Bmal1 gene-specific effects or to disruptions to the molecular clock, and if the latter, if tendon contains a functional clock that controls tissue homeostasis. Here, we show for the first time the presence of self-sustained circadian clocks in mouse tendon and human tenocytes, and, the circadian clock controls BMP signaling. The results indicate that tendon clocks control 4.6% of local transcripts including Grem2, which we show is an antagonist of BMP signaling and inhibits calcification of primary human tenocytes in culture. Disruption of the circadian rhythms in the arrhythmic CLOCKD19 mice 11 is associated with aberrant Grem2 expression and BMP signaling, calcific pathology and mechanically weak tendons. Finally, we show that aged wild-type mice exhibit a dampened and delayed tendon circadian rhythm associated with profound calcification.

Results
Autonomous circadian rhythm in tendon. To investigate if tendon has an intrinsic clock we examined Achilles and tail tendons from PER2::Luc reporter mice 12 and used real-time bioluminescence microscopy and a photomultiplier tube (PMT) to record light emission. This revealed robust circadian rhythms of PER2::Luc activity, with periods of 23.73 6 0.26 hr and 23.23 6 0.06 hr in Achilles and tail tendons, respectively, indicating that there is a circadian clock in Achilles and tail tendon ( Figure 1A and B, and Supplementary video). As expected, the rhythm dampened over time in culture but was effectively reinstated after a single treatment with dexamethasone, a known synchronizing agent for peripheral clocks.
We next examined the circadian rhythm of Achilles and tail tendons from the CLOCKD19 x PER2::Luc crosses. The CLOCKD19 mice harbor a deletion in exon 19 of the CLOCK gene producing a dominant negative mutant protein 11 . The results confirmed that CLOCKD19 tendons are arrhythmic ( Figure 1C). We also established that tendon cells isolated from wild type but not CLOCKD19 tail tendons have an autonomous circadian rhythm ( Figure 1D). To extend these studies to human tendon cells, primary human tenocytes were transduced with lentivirus carrying Bmal1-or Per2-promoter-driven luciferase reporters. Both reporters showed circadian oscillation in anti-phase to one another ( Figure 1E). The circadian transcriptome in tendon. Circadian transcriptome studies in muscle, liver and cartilage have demonstrated very low numbers of overlapping rhythmic genes, highlighting the tissuespecific functions of the peripheral circadian clocks (for discussion see 9 ). To gain further insights into the functions of the tendon clock, we performed a time-series microarray to identify the circadianlyregulated genes. Female C57BL/6 mice were kept in constant dark-ness for 39 hr to reveal genes driven by the endogenous circadian clock rather than by light-related cues. Tail tendon was taken at 4-hr intervals during 48 hr. In total, 16,000 genes (represented by 29,000 probes) were classified as expressed. To identify circadian transcripts, we used two well-recognized algorithms based on different statistical models as described previously 9 . CircWave Batch revealed ,1000 probes oscillating with a period of ,24 hr (Figure 2A and   Figure S1, q , 0.003). The JTKcycle algorithm identified 1,500 rhythmic probes with 20-28 hr period (q , 0.1). To be stringent, we considered only those genes identified by both methods to be circadianly regulated. From these analyses we identified 745 genes (4.6% of genes expressed in tendon) with rhythmic expression and a period of ,24 hr ( Figure 2A). The uniqueness of the tendon circadian transcriptome is shown schematically in Figure 2B in that only 14 genes are common between previously published circadianregulated genes in cartilage and skeletal muscle 8,9 . Among the 14 common genes were the core clock components Bmal1, Per2 and Nr1d1, whose circadian expression rhythms were validated by qPCR in mouse Achilles and tail tendon ( Figure 2C). Using a Gaussian process model with a Bayesian approach 9 , rhythmic tendon genes were clustered into 9 distinct clusters according to their inferred periodic functions ( Figure S2, Supplementary Table 2). Most genes peaked during late night and early day ( Figure S3). In addition, expression of endogenous core clock genes, Bmal1, Per2 and Nr1d1 were rhythmically expressed in primary human tenocytes ( Figure 2D). Thus, mouse and human tendon cells exhibit a cell autonomous circadian molecular oscillator.
Grem2 as a clock controlled gene in tendon. One of the rhythmic genes in tendon was Grem2, with expression peaks in the beginning of day and is down-regulated at night ( Figure 3A and B). The Xenopus Gremlin is a secreted protein that can block BMP signaling by binding BMPs, thereby preventing receptor activation and phosphorylation of Smad1/5 13 (reviewed by 14 and discussed in more detail below). Therefore, we predicted that the mammalian Grem2 might have a similar role in antagonizing BMP signaling. We used western blot analysis to assess levels of phosphorylated Smads in temporally-collected tail-tendon proteins ( Figure 3C). Phosphorylation of Smad1/5 was observed at all 6 time points across one circadian cycle but was noticeably less abundant at CT3, CT7 and CT11 and more abundant at CT15, CT19 and CT23. The total amount of Smad1, Smad5, Smad2 and phosphorylated Smad2 (a known target of TGFb) did not appear to vary. Target genes of BMP signaling include inhibitor of differentiation 1 (Id1), Id2 and Id3 and Bmp2k (BMP2-inducible kinase) 15,16 . The microarray expression profiles of Id1, Id3 and Bmp2k were lower in the day phase and higher in the night phase ( Figure 3D). Thus, the data support the hypothesis that BMP signaling in tendon is ''gated'' by the circadian clock, with an opposing pattern to the rhythmic expression of Grem2.
Gremlin-2 inhibits phosphorylation of Smad1/5 and calcium deposition. To determine if Gremlin-2 is capable of inhibiting BMP signaling we used an in vitro osteogenic differentiation model for human tenocytes. Addition of rhBMP2 to primary human tendon cells activated the phosphorylation of Smad1/5, which was reduced by pre-incubation with rhGremlin-2 ( Figure 4A). Next, human tendon cells cultured in an osteogenic medium for 14 days stained bright red by Alizarin red, indicative of calcium deposition ( Figure 4B). However, only low-level calcium deposition was observed in cells co-incubated with rhGremlin-2. Together, these data show that Gremlin-2 inhibits BMP signaling and calcium deposition in human tenocytes.
Aberrant BMP signaling and ectopic calcification in Achilles tendon of CLOCKD19 mice. The circadian control of Grem2 and BMP activity prompted us to hypothesize that the CLOCKD19 mouse will have aberrant BMP signaling, predisposing the tendons to calcification. To test this, we first compared the expression of Grem2 in CLOCKD19 with wild type tail tendons at two time points corresponding to the peak and trough expression of Grem2. Consistent with the loss of circadian rhythms in CLOCKD19 tendons ( Figure 1C), we observed a lack of time-dependent expression for Grem2 ( Figure 5A). Importantly, at the peak expression time of CT7 in wild type tendons, Grem2 expression in CLOCKD19 tendon was significantly low. The low level of Grem2 expression was associated with an aberrant pattern of phosphorylated Smad1/5 ( Figure 5B). Noteworthy, low level BMP signaling seen in wild-type tendon at CT7 was replaced by elevated BMP signaling in the CLOCKD19 tendons, suggesting an inhibitor function of the clock on BMP signaling. Alizarin Red staining of Achilles tendon showed the presence of calcification in tendons of CLOCKD19 mice but not in young age-matched wild type mice ( Figure 5C).
Aberrant clock in aged tendons is associated with ectopic calcification. To test if the tendon clock changes with ageing, we compared PER2::Luc rhythms in tendon from old (22-24 month old) and young adult (2-3 month old) mice (Fig 6A). There was a significant 40% reduction in the oscillation amplitude of PER2::Luc rhythms in aged tendon and the circadian phase was delayed by ,6 hr ( Figure 6B-D). Comparison of the mRNA levels of core clock genes in young and old tendon showed significant dampening in the time-of-day dependent expression of Bmal1 and Per2 and reduced levels of Nr1d1 ( Figure 6E). Importantly, in parallel to the clock gene changes, the rhythmic expression of  Grem2 seen in young tendon was severely dampened in the old tendon. Finally, the aged wild-type mice (22-24 month old) developed calcific deposits in their Achilles tendons ( Figure 6F).

Discussion
We have identified cell-autonomous oscillations of clock genes in murine and human tenocytes and shown that the circadian clock is a master regulator of BMP signaling. These results provide a basis for exploring the links between tendon clocks, ageing and tendinopathies.
Circadian clock disruptions have been linked to various pathologies and diseases, including defects in the musculoskeletal system. For instance, bone volume is significantly increased in the Per2 mutant and Cry2 knockout mice, caused by defects in osteoblasts and osteoclasts, respectively 17 . Furthermore, shorter bone length was reported in the Bmal1 knockout mouse because of defects in chondrocyte differentiation in the growth plate 18 . Moreover, maximal muscle force was reduced (by 30%) in the Bmal1-null and CLOCK mutant mice 19 . Consistently, by comparing the tendon circadian transcriptome to that of the cartilage 9 and muscle 8 , we found that these tissues had few rhythmic genes in common. In fact, the majority of the common rhythmic genes encoded components of circadian clocks.
Interestingly, the initial phenotypic characterization of the Mop3 (Bmal1) null mouse reported ectopic calcification in tendons and ligaments, among other signs of non-inflammatory arthropathy 10 . The underlying mechanisms for the calcified tendonopathies have never been addressed. Our results showing calcification in the CLOCKD19 mouse tendons, support a hypothesis that the CLOCK/ BMAL1 complex regulates key signaling pathways that inhibit ectopic calcification in normal rhythmic mice. Thus, when the circadian rhythm is disrupted, the tendons are predisposed to calcification.
BMPs are the largest class of signaling ligands in the TGFb superfamily and regulate a wide variety of processes during embryo patterning, skeletal and non-skeletal development, differentiation of cell fate, and have additional roles in cancer and tumor progression (reviewed by 20 ). The basic signaling mechanism is the same for all TGFb superfamily members and is initiated when ligand dimers engage with type I and type II serine/threonine kinase receptors. In the activated receptor complex, the type II receptor phosphorylates the type I receptor which provides a binding site for the receptorregulated SMADs (R-SMADs). The plethora of normal biological processes triggered by BMP signaling, and the involvement of the pathway in pathological conditions including cancer, is partly explained by the large number of BMP ligands (,20 in number), the choice of type I and type II receptors (12 in total) for ligand engagement, and the availability of SMAD dependent and SMADindependent signal transduction. With such multiple opportunities for pathway activation, the blockade of BMP ligand-receptor engagement by extracellular antagonists is a crucial on-off switch of the signaling pathway. Thus, the circadian clock could be a super-regulator of BMP signaling by regulating the expression of Grem2.
Gremlins are small, single domain proteins that contain a cysteine-rich fold that occurs in members of the DAN/Cerberus family of secreted BMP antagonists [21][22][23][24][25] . Gremlin-1 prevents binding of BMP receptor binding but does not affect the activity of other members of the TGFb superfamily 13 . Gremlin-2 (also known as Protein Related to Dan and Cerberus, PRDC) is the second member of the Gremlin family and can inhibit ligand signaling by BMP2 and BMP4 but has minimal effects on GDF-9, activin and TGFb signaling 26 . Previous studies have shown that Grem2 restricts BMP signaling to the ventral arches and thereby promotes dorsal and intermediate skeletal fates 27 , is necessary for cardiac laterality and atrial differentiation 28 , and can promote the differentiation of stem cells to cardiomyocytes 29 . However, a role for Grem2 in tendon homeostasis and calcification had not been previously described. BMP signaling has been shown to drive osteogenic differentiation of tendon-derived stem cells 30 . Furthermore, BMP2 and BMP7 can be detected in the subacromial bursa of patients with chronic rotator cuff degeneration at levels sufficient to induce osteogenic differentiation of mouse myoblast-derived C2C12 cells 31 .
Analysis of our microarray data showed that Grem2 (but not Bmp2, Bmp4 and Bmp7) was rhythmically expressed in tendon. The rhythmic expression of Grem2 was validated in mouse Achilles and tail tendon and primary human tendon cells. In further experiments we showed that at the peak expression time of CT7, Grem2 expression in CLOCKD19 mutant tail and Achilles tendons was significantly lower than in wild-type tendons, thus indicating a role of the molecular circadian clock in regulating Grem2 expression in tendon. We also showed that BMP-signaling target genes (Id1, Id3 and Bmp2k) are rhythmic, indicating circadian activity of the BMP signaling. A mechanistic link between Gremlin-2 and tendon calcification was evident when Gremlin-2 was capable of inhibiting BMP-induced calcification of human tenocytes in culture, revealing Gremlin-2 as a novel potent antagonist of BMP signaling in tendon.
Consistent with the inhibitory role of Gremlin-2, we observed a profound day/night difference in the Smad1/5 phosphorylation opposing the levels of Grem2 expression. Therefore, the circadian control of Grem2 is likely to be involved in ''gating'' the BMP signaling pathway in a rhythmic manner.
It is well established that systemic circadian rhythms dampen with aging as shown by body temperature cycles and hormones (see 32 and references therein). Likewise, we showed that tendon rhythms with altered phase and reduced oscillation amplitude are a feature of aging mice. The delayed phase implies that there is an internal desynchrony between tendon clocks and the central brain clock, which might disrupt the temporal coordination of the tendon rhythmic genes. Such desynchrony in gene expression is exacerbated by the reduced amplitude, leading to compromised tissue homeostasis and increased risk of pathology. As age is a risk factor for calcific tendinopathies in elderly humans, we envisage a scenario in which chronic circadian misalignment/disruption (e.g. during aging) predisposes tendon tissue to calcific changes. Future studies in humans and the use of conditional mutant mouse models may reveal more information about the tendon clock as an important regulator of tissue physiology and pathology.

Methods
Animals. All animal work was approved by The University of Manchester Breeding and Supply Unit and was in accordance with the 1986 Home Office Animal Procedures Act (UK), and following local ethical review.
Tendon Explant Cultures and Bioluminescence Recording. Tendon cultures from PER2::Luc mice were prepared by dissecting out the Achilles and tail tendons. Tendon explants were cultured on 0.4 mm cell culture inserts (Millipore) and bioluminescence was recorded in real-time using as previously described 33 . Dexamethasone (100 nM) was added to the recording media to re-initiate circadian rhythms. Baseline subtraction, period, phase and amplitude calculation was carried out as previously described 9 . Cultures were also visualized using a self-contained Luminoview LV200 microscope (Olympus) and recorded using a cooled Hamamatsu ImageEM C9100-13 EM-CCD camera. Images were taken every hour for 4 days, and combined in ImageJ.
Time Series Microarrays and Quantitative PCR. The circadian transcriptome studies in tendon were performed as described before 9 . Tail tendons were collected every 4 hr commencing at circadian time CT3 (3 hr into the light phase), for 48 hr. Tissues (6 samples per time point) were snap frozen in liquid nitrogen. Two samples per time point were used for microarray; the remainder was used for qPCR validation. RNA extractions were carried out using TRIzol reagent (Invitrogen) according to manufacturer's protocol. Mouse430_2 Affymetrix GeneChips were run according to manufacturers' instructions. Raw data were deposited in Array Express (accession number pending). CircWave v5 (by Dr. Roelof Hut, cut-off point of q , 0.003) and JTK-Cycle 34 (q-value of 0.1) were used to identify circadian transcripts as described previously 9 . Local false discovery rate was calculated using an R package 'fdrtool' 35 . To be stringent, genes identified using both methods are counted as positive. To model the data for clustering, we used a Gaussian process model 9 . Validation of time-series arrays and gene expression analyses was examined by qPCR (for primers see Supplementary Table 1) as described previously 36 .
Primary Mouse and Human Tenocytes Culture. Cells from tail tendons were released by treatment with 1000 U/ml bacterial collagenase type 4 (Lorne Laboratories) in 0.25% (wt/vol) trypsin (Invitrogen). Cells were grown at 37oC until 90-100% confluent prior to synchronization and PMT recording. Primary human tenocytes were obtained as described previously 37 . Cells from three preparations were used in the experiments described in this study. For time course experiments, confluent cells were synchronized with DMEM:F12 containing 50% (v/v) horse serum (Invitrogen), L-G and P/S for 2 hr at 37oC. RNA was isolated after 24 hr, at 4hr intervals for 20 hr. For treatment with recombinant human BMP2 (PHC7145, Invitrogen) and recombinant human Gremlin-2 (SRP4657, Sigma), 90-100% confluent cells were washed with PBS and incubated with serum-free media for 1 hr at 37oC. Then, cells were treated with BMP2 that had been pre-incubated for 1 hr at 4oC with Gremlin-2 in serum-free medium. Protein was isolated after 24 hr and analyzed by western blotting. Lentiviral Transduction. Lentiviral transduction of primary human tenocytes was performed using methods previously described 9 .
Calcium Deposition Assay and Alizarin Red Staining. Achilles tendons were fixed in 2% (w/v) PFA in PBS for 20 minutes, wax processed. 5-6 mm sections were dewaxed before staining. Calcium deposition in primary human tenocytes was induced by culturing cells in Osteogenic Medium (Lonza) according to the manufacturer's protocol. After 14 days, cells were fixed with 2% (w/v) PFA in PBS for 20 minutes. Fixed cells or tissue sections were washed with distilled water and stained with 2% (w/ v) Alizarin Red S, pH 4.2 for 10 min. Washed plates or sections were air-dried and imaged using a Pannoramic slide scanner and Pannoramic Viewer software (3DHISTECH).
Statistical Analyses. Mean and standard error of the mean were calculated for each data set from at least three biological replicates (indicated by the n in figures). Data was evaluated using two sample t-tests with GraphPad Prism (GraphPad Software). Differences were considered significant at the values of p , 0.05 (*), p , 0.01 (**) or p , 0.0001 (***).