Ubiquitin-protein ligase Ubr5 cooperates with hedgehog signalling to promote skeletal tissue homeostasis

Mammalian Hedgehog (HH) signalling pathway plays an essential role in tissue homeostasis and its deregulation is linked to rheumatological disorders. UBR5 is the mammalian homologue of the E3 ubiquitin-protein ligase Hyd, a negative regulator of the Hh-pathway in Drosophila. To investigate a possible role of UBR5 in regulation of the musculoskeletal system through modulation of mammalian HH signaling, we created a mouse model for specific loss of Ubr5 function in limb bud mesenchyme. Our findings revealed a role for UBR5 in maintaining cartilage homeostasis and suppressing metaplasia. Ubr5 loss of function resulted in progressive and dramatic articular cartilage degradation, enlarged, abnormally shaped sesamoid bones and extensive heterotopic tissue metaplasia linked to calcification of tendons and ossification of synovium. Genetic suppression of smoothened (Smo), a key mediator of HH signalling, dramatically enhanced the Ubr5 mutant phenotype. Analysis of HH signalling in both mouse and cell model systems revealed that loss of Ubr5 stimulated canonical HH-signalling while also increasing PKA activity. In addition, human osteoarthritic samples revealed similar correlations between UBR5 expression, canonical HH signalling and PKA activity markers. Our studies identified a crucial function for the Ubr5 gene in the maintenance of skeletal tissue homeostasis and an unexpected mode of regulation of the HH signalling pathway.


Introduction
Ubiquitin ligases target proteins for ubiquitination which can modulate protein function by regulating protein degradaton, protein-protein interactions, and protein localization [1][2][3][4], and thus, provide important post-translational mechanisms essential for a variety of cellular processes. The Drosophila homologue of the mammalian Ubiquitin Protein Ligase E3 Component N-Recognin 5 (UBR5), designated as hyperplastic discs (Hyd), was originally identified as a Drosophila tumor suppressor protein [5][6][7] and regulator of Hedgehog (HH) signalling [6]. Physical and genetic interactions with established components of the HH signalling pathway [7,8] strengthened Hyd's role as a regulator of HH signalling. We previously addressed a possible conserved role for UBR5 in HH-mediated processes in mice [9]. Although no overt effects were seen in patterning of the developing limb bud in mouse embryogenesis; here, we show that the coordinated action of Ubr5 with HH signalling is crucial to maintain skeletal tissue homeostasis associated with the appendicular skeleton postnatally and in adult mice.
HH signalling regulates cell processes that are critical for skeletal tissue development, growth and homeostasis [10]. Two HH ligands, Sonic-and Indian-Hedgehog (SHH and IHH, respectively) are widely expressed and function as extracellular signalling molecules that bind to cells expressing HH receptors such as patched-1 (PTCH1). Binding to PTCH1 results in derepression of the G protein-coupled receptor, smoothened (SMO), and activation of SMOassociated canonical and non-canonical signalling pathways [11][12][13]. Activation of the SMOassociated canonical pathway results in stimulation of GLI-mediated transcription and expression of crucial target genes [7]. Activation of the recently identified SMO-associated noncanonical pathway relies on SMO's GPCR activity [14,15] and results in inhibitory heterotrimeric G protein-mediated inhibition of adenylate cyclase and a concomitant reduction in cyclic AMP (cAMP) levels [14,16,17].
Although not yet experimentally addressed, non-canonical signalling may also contribute to many of the well-described roles for canonical HH signalling in normal skeletal formation, maturation and maintenance [10,18].
At birth, IHH is the ligand that drives HH signalling within the growing limbs. Expression of Ihh is localized to a zone of postmitotic, prehypertrophic chondrocytes immediately adjacent to the zone of proliferating chondrocytes [18][19][20] and is essential for endochondral ossification but also induces osteoblast differentiation in the perichondrium [21]. Dysregulation of this signalling pathway is detrimental to musculoskeletal tissue homeostasis [22,23]. Notably, studies have shown that increased HH signalling can drive pathological ectopic cartilage and bone formation in soft tissues [10] through the process of heterotopic chondrogenesis and heterotopic ossification (HO) [24]. Upregulation of HH signalling is believed to contribute to the rare disorder, progressive osseous heteroplasia (POH), which includes in its phenotypic spectrum soft tissue ossification. POH is caused by loss-of-function of GNAS, a G protein alpha subunit and activator of adenylate cyclase. A murine model of POH demonstrated that increased HH signalling as a consequence of GNAS loss-of-function in mesenchymal limb progenitor cells drove heterotopic ossification [25]. Similarly, synovial chondromatosis, a disease resulting in ossification of synovial tissue is associated with increased canonical HH signalling [26]. However, in contrast with cartilage and bone gain, elevated HH signalling is also associated with the cartilage degradation and loss [27,28]. Hence, appropriate HH signalling is normally involved in the suppression of ectopic, and genesis and maintenance of normtopic, cartilage and bone.
Here, we show that the loss of Ubr5 function in Ubr5 mt mice resulted in diverse musculoskeletal defects including spontaneous, progressive and tissue-specific patterns of ectopic chondrogenesis and ossification as well as articular cartilage degeneration and shedding. Surprisingly, reducing SMO function in UBR5-deficient mice led to a dramatic reduction in the age of onset and increased severity of the Ubr5 mt phenotype. These observations challenge the existing dogma by highlighting an important role for Smo, in the absence of UBR5, in suppressing, rather than promoting, ectopic chondrogenesis, tissue calcification/ossification and articular cartilage damage. We, therefore, reveal a previously unknown physiological role for Ubr5 and highlight its genetic interaction with Smo in regulating cellular and tissue-homeostasis. These findings may influence current therapeutic approaches modulating HH signalling for the treatment of degenerative musculoskeletal conditions such as osteoarthritis and heterotopic ossification.

Loss of Ubr5 function causes skeletal heterotopias at 6 months
To overcome the embryonic lethality associated with germline mutant animals [29], we combined a Ubr5 conditional loss-of-function gene trap (Ubr5 gt ) [9] with Prx1-Cre [30] (Prx1-Cre; Ubr5 gt/gt animals henceforth, referred to as Ubr5 mt ) to ensure that adult tissues derived from early limb bud mesenchyme, predominantly bone and connective tissue, were Ubr5 deficient. Since the HH pathway affects embryonic limb patterning and bone growth, the Ubr5 deficient fetuses (at E15.5) were initially examined and bones and joints appeared to develop normally [9]. However, the HH pathway continues to function in postnatal bone growth and homeostasis [10] and thus, at approximately 6 months of age, we noticed that mice began to display defects in locomotion. Control animals normally remained supported by their hindlimbs ('sprung'), whereas, Ubr5 mt animals rested their posteriors directly upon the floor ('squat') (S1A-S1C Fig). Considering the tissue targeted by the conditional mutation, the observed phenotype indicated a potential musculoskeletal system defect which prompted the examination of hindleg bone and joint structures.
At 6 months of age, X-ray imaging revealed that Ubr5 mt animals exhibited abnormally shaped and/or ectopic signals around knee and ankle joints (S1D-S1G Fig). 3D micro-computed tomography (μCT) revealed that, whereas Prx1-Cre control joints appeared normal with no evidence of ectopic structures (Fig 1A), the knees of all Ubr5 mt mice (n = 10) exhibited isolated ectopic signals clearly separated from the adjacent femoral condyles and tibia ( Fig 1B). Surface rendering of the μCT scans demonstrated that the array of knee-associated sesamoid bones (patella and fabella) and calcified menisci (Fig 1C and 1D) were abnormal. Ubr5 mt knees presented with large ectopic structures on all four faces of the knee joint, as well as enlarged and irregularly shaped fabella and patella sesamoid bones (Fig 1E and 1F). In addition, the Ubr5 mt animals exhibited multiple ectopic signals around the ankle joint (Fig 1G and  1H), with the most striking one appearing consistently on the dorsal side running parallel to the long axis of the tibia (Fig 1H, open arrows) associated with the Achilles tendon (AT). This ectopic signal remained isolated from the calcaneus and tibia. Other ectopic structures included two ectopic U-shaped signals on the ventral and lateral sides of the tibia (Fig 1H).
Following Prx1-Cre mediated recombination of the Ubr5 gt gene-trap construct, the reporter gene lacZ (encodes β-galactosidase [β-gal] activity) is activated and expressed under the influence of Ubr5 gene regulators [9] acting as proxy for Ubr5 expression. β-gal activity is initially detected in limb mesenchyme at early embryonic stages [9] and at the postnatal stages examined here, activity is restricted to limb tissue derived from this embryonic mesenchyme. Analysis of lacZ expression in 20 week-old control and Ubr5 mt mice indicated that Ubr5 is expressed around the knee (Fig 1I and 1K) and ankle (Fig 1J and 1L) joints. Further analysis revealed strong β-gal activity around the periphery of the menisci and synovium (Fig 1 M and 1N). The ankle also revealed β-gal activity within the AT and superficial digital flexor tendon and in a large ectopic structure within the AT midbody (Fig 1O and 1P). Thus, the tissues that exhibit Cre-mediated expression of the lacZ gene are affected in the mutant phenotype.

Ubr5 mt -associated ectopic structures exhibit chondrogenesis and calcification
The morphology of these ectopic structures around the knee (see Fig 1D and 1F) and ankle (see Fig 1H) joints were further investigated to determine the cellular composition and possible derivation of these ectopias. As shown by μCT, both knee (Fig 2A and 2B) and ankle ( Fig  2D and 2E) ectopic structures harbored X-ray dense internal structures indicative of bone. Accordingly, in these joints (Fig 2C and 2F) von Kossa staining, which is widely used to detect abnormal calcium deposits, revealed staining in these ectopic structures. Von Kossa staining in the Ubr5 mt knee joints revealed positive stained structures within the synovium deep in the patellar tendon ( Fig 2C) and near the ankle joint staining within the AT (Fig 2F). In addition, in the control AT, the expected ordered stacking of tenocytes along the anterior-posterior axis ( Fig 2G) and an absence of toluidine blue staining associated with proteoglycans ( Fig 2H) was observed. In contrast, these regions of the Ubr5 mt AT were devoid of tenocytes, which were replaced by long columns of proteoglycan-expressing hypertrophic chondrocytes (Fig 2I and  2J). Histological analysis of control knee joints revealed a synoviocyte-rich intimal layer of the synovium (Fig 2K), whereas Ubr5 mt knee joints exhibited bone-( Fig 2L) and cartilage-like ( Fig  2M) ectopic structures. The observed mutant morphology was consistent with the formation of ectopic bone around the knee and in the AT, which was further associated with the presence of ectopic chondrocytes and the production of extracellular matrix.Thus, an abnormal phenotype consisting of ectopic chondrogenesis, calcification and ossification (hereafter, referred as ECCO) of the synovium and tendons in Ubr5 mt tissues was observed. We concluded that Ubr5 normally functions to prevent spontaneous ectopic formation of chondrocytes in tissues and calcification and/or ossification in cartilage.

Loss of Ubr5 function causes articular cartilage degradation
μCT analysis of 6-month old control ( Fig 3A) and Ubr5 mt (Fig 3D) knee revealed the previously described overt morphological differences (see above), but further analysis at the surface of the joints, showed a significantly increased volume of high subchondral bone density (designated by red in Fig 3B, 3C, 3E and 3F) in the mutant. The increase in volume as a percentage of total subchondral bone (depicted in S1H-S1J Fig) is quantified in Fig 3G. In addition, histological assessment showed a dramatic loss of articular cartilage (AC) from the lateral tibial and femoral surfaces of all Ubr5 mt knee joints assessed (Fig 3H and 3I); a condition not detected in any control mice at this stage. Further examination of the exposed subchondral bone in these Ubr5 mt mice revealed abnormal intermixed bone and cartilage within this region (Fig 3J and (Fig 1Q and 1R) which corresponds with a role for Ubr5 in this tissue. Thus an overview of the hindlegs at 24 weeks reveals a diverse range of cartilaginous defects which include firstly, metaplastic conversion of connective tissue associated with the knee and ankle and secondly, severe degradation of AC causing exposure of the subchondral bone at the joint surface.

Ubr5 deficiency results in a postnatal, progressive phenotype
To establish the approximate age at which this striking ECCO phenotype was initially detectable, a timed series of in vivo μCT scans on ageing, live animals was followed.  and Ubr5 mt which exhibited extended regions that lacked AC and exposed subchondral bone (dashed lines). Peripheral regions retained some AC (white arrowheads in I). (J) 26-week-old control tibial AC was normal and (K) Ubr5 mt AC revealed exposed subchondral bone (black arrowheads) and intercalated cartilage (white arrowheads)).
https://doi.org/10.1371/journal.pgen.1009275.g003 at 3-weeks of age revealed no marked abnormal phenotype in knee or ankle joints (S2A-S2D Fig), suggesting that the ectopic structures did not form during fetal development but rather formed postnatally. Between 6 and 12 weeks of age, the ectopic structures began to emerge ( Fig  4A and 4B), initially on the ventral side of the tibia. Dorsally located ectopic signals associated with the AT emerged by 16 weeks of age ( Fig 4C) and all ectopic structures were enlarged by 24 weeks of age ( Fig 4D). Ubr5 deficiency, therefore, led to enhanced, progressive chondrogenesis and osteogenesis in the connective tissue of the knee and ankle of the aging Ubr5 mt mouse.
These metaplastic conversions within the connective tissue supporting the knee and ankle, however, contrasted with the changes demonstrated in the AC which manifested as a degenerative phenotype. To investigate the timing of AC degradation, we examined mice at 3 and 6 weeks. No gross structural disruption of the AC in the Ubr5 mt animals at 3-weeks of age was detected (Fig 4E and 4F). By 6-weeks of age, Ubr5 mt articular cartilage exhibited an irregular osteochondral interface (Fig 4G and 4I), clusters of large, hypertrophic-like chondrocytes ( Fig  4H and 4J) and a reduction in the number of superficial chondrocytes ( Fig 4K). Ubr5 mt articular cartilage also exhibited multiple tidemarks and regions of strongly eosin positive nuclei indicative of necrosis ( Fig 4L and 4M) that were absent in controls. Thus, progressive damage was found in the AC due to the loss of Ubr5 but in this case, early cellular and extracellular abnormalities occurred prior to AC degradation, which included increased subchondral bone density and exposure of subchondral bone Despite loss of UBR5 in early limb mesenchyme, these data indicated that the ectopic structures arose postnatally and subsequently progressed with age. To directly address if postnatal UBR5 function was required to suppress ECCO and the degradation of the AC, we utilised a mouse line carrying a tamoxifen-inducible, conditional Cre, pCAGG-CreERT2 [30]. Control pCAGG-CreERT2 (pCAGG-Con) or pCAGG-CreERT2;Ubr5 gt/gt (pCAGG-Ubr5 mt ) animals were treated with tamoxifen (administered on two consecutive days) at six weeks of age. Staining for β-gal activity, although more broadly distributed, confirmed tamoxifen-mediated recombination of the Ubr5 mt gene trap and its associated β-gal expression in tissues that included muscles and tendons (S2E and S2F Fig), and within the midbody ectopia at the AT (S2G Fig). μCT analysis at 8 weeks revealed that tamoxifen-treated control animals exhibited no ectopic signals (Fig 4N), whereas pCAGG-Ubr5 mt animals exhibited AT -associated ectopic signals ( Fig 4O). Scoring ( Fig 4P) and heterotopic ossification (HO) volumetric analysis ( Fig  4Q) confirmed that only tamoxifen-treated pCAGG-Ubr5 mt animals exhibited ectopic signals. Comparison of 12 week control to treated pCAGG-Ubr5 mt (Fig 4R and 4S) knees revealed Ubr5 mt -associated apical acellular layer ( Fig 4S and 4T), damage to the apical surface, multiple tidemarks, reduced superficial zone chondrocytes ( Fig 4V) and increased numbers of empty lacunae (Fig 4U and 4W). We concluded that postnatal Ubr5 function was both necessary and sufficient to maintain AC homeostasis and prevent ECCO.

Inhibition of Smo promotes Ubr5 mt -associated ECCO and enhances Ubr5 mt -mediated AC degradation
As UBR5/HYD regulates HH signalling in Drosophila [7,8], we next used a genetic approach to address whether aberrant HH signaling contributed to the Ubr5 mt ECCO and AC phenotypes. The Smo gene encodes a core membrane component, regulated by the HH receptor PTCH1, that initiates the downstream signalling cascade leading to GLI-dependent transcription (canonical signalling) or G i protein-dependent events that are tissue specific (non-canonical signalling). We reasoned that reduction in Smo expression levels would sensitize the HH pathway; thus, heterozygosity for a Smo loss of function allele (Smo LoF ) [31] was used in a cross to Ubr5 mt to create Prx1-Cre;Ubr5 gt/gt ;Smo LoF/+ animals (Ubr5 mt +Smo LoF ). In contrast to our expectations, μCT analysis of 12-week Ubr5 mt +Smo LoF mice exhibited significantly more severe defects than those of age-matched Ubr5 mt (Fig 5A-5C) and Smo LoF/+ mice (which were indistinguishable from wildtype), with multiple, large ectopic signals apparent around the knee (Fig 5A-5F) and ankle joints (Fig 5H-5N). Volumetric analysis revealed a significant increase in the volume of Ubr5 mt +Smo LoF femoral-associated ectopic bodies compared to Ubr5 mt alone ( Fig 5G) and the ankles harboured a 20-fold increase in the volume of ectopic signals (Fig 5O). In agreement, histological analysis of the Ubr5 mt +Smo LoF joints revealed an enhanced phenotype to that described in Ubr5 mt (Figs 1 and 2). Ubr5 mt +Smo LoF synovium harboured large ectopic tissue masses (Fig 6A) with extensive vascularisation ( Fig  6B) and chondrocytes lining the surface (Fig 6C) with deeper calcified cartilage and vascularization (Fig 6D). Sagittal sectioning through the ankle revealed large ectopic structures within the superficial digital flexor tendon (Fig 6E), consisting of bone and cartilaginous tissue ( Fig  6F and 6H), and at the tendon interface (Fig 6G). Large swathes of chondrocytes were present within the superficial digital flexor and AT that coincided with an absence of tenocytes (Fig 6I  and 6J), as previously reported in the Ubr5 mt (Fig 2). In addition, the AC in Ubr5 mt +Smo LoF knee joints exhibited extensive loss over both tibial and femoral surfaces at this young age ( Fig  6M and 6N), while Ubr5 mt knee joints exhibited only tears within the AC (Fig 6K, 6L and quantification in 6O). Importantly, the loss of a single copy of Smo alone (Prx1-Cre;Smo LoF/+ ) resulted in no structural or AC damage (Fig 6P).

Ubr5 suppresses canonical HH signalling and PKA activity
A functional link between UBR5 activity and HH signalling was further examined in 6-week old Ubr5 mt mice. At this age ectopic structures were not detectable (Fig 4), thereby increasing the likelihood of detecting potential causative changes in expression patterns. Immunohistochemistry on Ubr5 mt knee intimal (Fig 7A-7F) and subintimal synovium (Fig 7G-7L) revealed increased Gli1 expression in comparison to Prx1-Cre control animals (Fig 7B, 7E, 7H and 7K; respectively), indicative of increased canonical HH signalling. qRT-PCR analysis also confirmed increased expression of markers of canonical HH signalling in RNA from isolated synovium (Gli1 and Ptc1) (Fig 7M). Additionally, intimal and sub-intimal Ubr5 mt synovium exhibited increased phosphorylated PKA substrate (PPS) staining suggesting decreased G i proteins activation, characteristic of non-canonical HH signalling (Fig 7C, 7F, 7I and 7L). Consistent with the observations in the synovium, Ubr5 mt AC exhibited markers of increased canonical (Fig 8A-8D) and decreased non-canonical HH signalling (Fig 8E and 8F). Although little change for PTCH1 was detected ( Fig 8G) there was significant differences for Gli1 expression and PKA substrate staining (Fig 8H and 8I).

UBR5 mt AC and damaged human AC exhibits both aberrant expression of markers of chondrogenesis and HH signalling
As seen in murine Ubr5 mt AC, osteoarthritic AC from patients also exhibits markers of increased canonical HH signalling [32]. We next addressed (i) UBR5 expression and (ii) markers of decreased non-canonical HH signalling (PPS) in human AC. Graded samples from   To further delineate whether mammalian Ubr5 could influence markers of canonical and non-canonical HH signalling, murine NIH3T3 cells were engineered to either exhibit increased (cDNA overexpression) or decreased (shRNA knock-down) Ubr5 expression. Cells were then transfected with constructs encoding (i) Shh, (ii) constitutively active Smo mutant (Smo-M2) or (iii) Gli1. Canonical pathway activity was measured using a Gli-responsive luciferase reporter assay. While perturbation of Ubr5 expression had no effect on Shh-or Smo-M2-mediated signalling (Fig 9A and 9B), Ubr5 overexpression caused a significant reduction (Fig 9A, P<0.001), and Ubr5 shRNA-mediated knockdown caused a significant increase ( Fig  9B, P<0.05), in Gli1-mediated luciferase activity. However, Ubr5-overexpression did not perturb the expression level of endogenous or exogenous GLI1 protein (Fig 9C), excluding a role for UBR5-mediated degradation. Therefore, UBR5 appeared to only suppress canonical HH signalling associated with overexpression of GLI1.
We then addressed whether loss of Ubr5 function would also affect cAMP production as a readout of G i protein activity, an indirect marker of non-canonical HH signalling. Ubr5 shRNA cells showed an~2-fold increase in maximal cAMP production in response to forskolin, an adenylate cyclase agonist (Fig 9D) [33]. Moreover, simultaneous addition of forskolin and purmorphamine, a SMO agonist, lowered maximal cAMP generation, but its effect was suppressed by Ubr5 shRNA (Fig 9D). Together, the in vitro findings suggest that Ubr5 loss results in reduced stimulation of G i proteins by Smo, leading to increased cAMP/PKA activity levels. Overall, these data supported our in vivo observations that Ubr5 normally acts to suppress GLI1 activity while promoting PKA activity.

Ubr5 mutation causes musculoskeletal tissue defects
We report a role for mammalian Ubr5 in adult skeletal homeostasis that impacts upon and genetically interacts with, components of the HH signalling pathway. These findings add to the emerging importance of the N-end rule ligases in regulating important signaling and cellular processes in human, and animal health and disease [34,35]. Loss of the Ubr5 gene in early limb mesenchyme resulted in postnatal defects in and around joints within the fore and hindlimb. The mutant phenotype displayed a degree of variability between individual mice; however, the abnormalities were pervasive in the mutants and were never observed in wildtype or control mice. The defects that were consistently observed included ectopic bone and cartilage formation, and articular cartilage degradation. The spectrum of defects observed are summarized in S4 Fig. Our data indicates metaplastic production of chondrocytes and/or ectopic endochondral ossification as a major component of Ubr5 mt -associated ECCO. Comparison of the Ubr5 mtassociated ECCO phenotype with that of human inherited HO diseases reveals some similarities and differences. Within the ECCO-prone tissues there were distinct tissue-specific responses; for example, the knee-associated synovium underwent ectopic chondrogenesis, calcification and ossification to produce bone, whereas the Achilles tendon only underwent ectopic chondrogenesis and calcification. The abnormalities of the knee-associated synovium which display heterotopic chondrogenesis are reminiscent of human benign bone tumours called osteochondromas [36], whereas the heterotopic tissue calcification without ossification seen in the AT resembles a form of calcific tendinopathy [37]. The mouse Ubr5 mutation, thus, provides a genetic model for the generation of these bone abnormalities and suggests that the processes of chondrogenesis, tissue calcification and ossification represent discrete, albeit interrelated, steps that when deregulated can individually, or collectively, contribute to distinct tissue pathologies.
Our findings also demonstrated an important role for Ubr5 in regulating AC homeostasis, where its loss led to dramatic cellular, extracellular and structural defects. The observed defects and 8xGLI-Firefly and pTK-Renilla luciferase reporters in growth medium (DMEM with 10% FBS). After 24 h, serum was reduced to 0.5% and the Firefly/Renilla luciferase activity was measured 48 h later. Bars represent mean +/-s.e.m. of n = 3 independent experiments. (B) A similar GLI-luciferase assay was carried out in NIH3T3 cells stably expressing Ubr5 shRNA (black bars) or scrambled shRNA (grey bars). Bars represent mean +/-s.e.m. of n = 3 independent experiments. (C) NIH3T3 cells were co-transfected with pN21-Ubr5 (Ubr5) or empty vector (Control) and Gli1-myc or empty pcDNA3.1, followed by Western blot analysis of Gli1 expression (arrowhead) using β-actin as loading control. (D) Stable knockdown of Ubr5 impaired readouts of non-canonical HH signalling. Production of cAMP by control scrambled shRNA (black bars) or Ubr5 shRNA stable cells (grey bars) following acute treatment with the adenylate cyclase activator forskolin (For) or forskolin plus the SMO agonist purmorphamine (For/Pur), compared to DMSO vehicle as control (-). Forskolin-stimulated cAMP production in Ubr5 shRNA cells was significantly elevated compared to control cells (p = 0.0368; t-test). Purmorphamine suppressed forskolin-mediated cAMP production in both scramble control (p = 0.0318; t-test) and Ubr5 (p = 0.0160; t-test) shRNA cell lines. in HH signalling could have been causative in nature as HH signalling is intimately linked to both stem cell [22] and chondrocyte biology [10]. One of the more distinctive Ubr5 mt AC defects was the tearing along the tidemark between non-calcified and calcified cartilage. This focal failure suggested the interface was prone to transverse shear forces and 'slipping' of one layer (i.e., non-calcified cartilage) relative to the other (i.e., calcified cartilage). Interestingly, this mode of AC shedding and the associated regions of necrosis mirrored defects observed in mammalian osteochondrosis [38,39].
Recently, increased Ubr5 expression was correlated with muscle growth, hypertrophy [40] and recovery from injury [41] suggesting a role in muscle maintenance. These reports suggest that Ubr5 may have a broad role in the musculoskeletal system of the limbs affecting homeostasis not only in the cartilaginous connective tissue but also in the associated musculature.

UBR5 influences markers of canonical and non-canonical HH signalling
Based on the current dogma, we hypothesized that the Ubr5 mt -associated ECCO was caused by increased HH signalling. In contrast, the introduction of Smo LoF heterozygosity into a Ubr5 mt background both (i) exacerbated Ubr5 mt -associated defects as well as elicited novel defects not observed by loss of Ubr5 function alone (e.g., ECCO of the calcaneal periosteum and the superficial digital flexor tendons and increased volume and altered shape of normotopic sesamoid bones). This combined ability to influence both normotopic and heterotopic bones (S4 Fig), highlights the importance of UBR5 in normal and pathological skeletal tissue homeostasis. Furthermore, our genetic analysis exposed a pro-homeostatic function for SMOand by extension HH signaling-in suppressing Ubr5 mt ECCO.
In vivo and in vitro observations identified a loss of Ubr5 associated with predictors of increased (GLI1 activity) and decreased (PKA activity) canonical HH signalling. Based on the current dogma, it is difficult to reconcile increased GLI activity in the context of increased PKA activity, given that PKA phosphorylates other GLI family members, GLI2 and GLI3, targeting them for processing into transcriptional repressors [14,42]. However, the evolving breadth of the HH pathway ( Fig 9E) provides potential mechanistic explanations for this apparently paradoxical observation.
Recent evidence expanded the role of PKA to promote canonical HH signalling by promoting BRD4-mediated stimulation of GLIs transcriptional activity (Fig 9E) [43][44][45]. Interestingly, HO-associated with increased HH signalling was suppressed by the BRD4 inhibitor JQ1 [46], which clearly demonstrated a role for a cAMP-PKA-BRD4-GLI1 axis in skeletal tissue homeostasis. A non-canonical role of SMO as a G protein-coupled receptor (14,15) provides a mechanism to control PKA activity. Upon stimulation, SMO activates heterotrimeric G i proteins, which, upon dissociation, inhibit adenylate cyclase through the Gα subunit to reduce cAMP production and PKA activation [15,47,48]. Therefore, SMO inhibition can lead to increased cAMP-mediated PKA activity accounting for SMO modification of the Ubr5 mt phenotype, as impairment of either UBR5 or SMO leads to increased cAMP-mediated PKA activity-with their combined impairment leading to either additive or synergistic effects. Interestingly, our preliminary research (personal communication NDGR) supports a role for UBR5 in regulating readouts of non-canonical HH signalling other than PKA (i.e.; RhoA) [16]. Although our data reveal a genetic interaction between UBR5 and an essential component of the HH signalling pathway, we cannot fully establish the underlying mechanism(s) driving Ubr5 mt -associated ECCO. Future work will require developing the tools to differentiate between causative individual, or combined, contributions of aberrant canonical or non-canonical HH signalling. The addition of Smo LoF into a Ubr5 mt background would have exacerbated a pre-existing imbalance between the pathway outputs to drive ECCO.
The importance of balanced canonical and non-canonical HH signalling was recently demonstrated in osteogenesis [49]. Loss of the cilia regulatory protein IFT80 resulted in impaired osteoblast differentiation and coincided with (i) decreased expression of canonical target genes and (ii) increased non-canonical activity. The authors proposed that the non-canonical HH pathway prevented, and the canonical pathway promoted, formation of osteoblasts. Due to the emerging importance of non-canonical HH signalling [12], we also propose that the combined effects on canonical and non-canonical HH signalling contributed to the observed loss of tissue homeostasis in Ubr5 mt animals. Overall, our detection of Ubr5 mt -associated increased canonical (GLI1 activity) and indications of decreased non-canonical HH signalling (cAMP-PKA) are in general agreement with a reported pro-osteogenic environment conducive to HO [49]. UBR5 may therefore join IFT80 [49] and DYRK1B [50] as differential regulators of canonical and non-canonical HH signalling. Our future work will involve establishing which of the various non-canonical, SMO's GPCR-associated downstream effectors (e.g., PKA, RHOA, RAC1, PI3K etc.) [51,52] drive ECCO.
In summary, we reveal a previously unknown role for Ubr5 in influencing HH signalling, tissue homeostasis and preventing spontaneous ECCO. A role for UBR5 in regulating HH signalling and tissue homeostasis supports the classification of human UBR5 as a Tier 1 human cancer susceptibility gene (Sanger Cancer Gene Consensus). We believe the Ubr5 mt mouse model could assist in uncovering mechanisms that lead to disorders including characterisation of early pathological events and elucidation of pro-homeostatic mechanisms capable of promoting general bone health. In the future, manipulation of human UBR5 and SMO function could potentially provide a means of preventing pathological, and promoting beneficial, chondrogenesis and ossification in both the clinic and in biomedical engineering applications.

Ethics statement
Human material. Human AC was obtained from knee joint arthroplasty specimens with ethical approval from the Lothian Research Ethics Committee. Written formal consent was obtained.
Murine studies. All animal experiments were reviewed and approved by the University of Edinburgh Animal Welfare and Ethics Committee and were conducted with appropriate licensing under Animals (Scientific Procedures) Act 1986 (license number PPL 60/4424).
Prx1-Cre;Ubr5 gt/gt experimental animals (referred to as Ubr5 mt ) and their respective littermate controls were generated and all experiments were conducted in accordance with the ARRIVE guidelines. Animals were routinely weighed and there were no significant differences between experimental and control animals. Males animals were used for analysis unless otherwise stated. Tamoxifen (0.1mg/kg body weight) in corn oil, or vehicle only, were administered i.p to six-week-old animals on two consecutive days. For X-gal staining, embryos and postnatal hind limbs were dissected, fixed in 4% formaldehyde (from paraformaldehyde [PFA]) at 4˚C, washed and stained in X-Gal stain solution (XRB supplemented with 1mg/ml X-Gal) overnight [20].

Histology
Hindlimbs were fixed in 4% formaldehyde (fromPFA)) for 72hrs at 4˚C before being decalcified 0.5M ethylenediaminetetraacetic acid (EDTA) pH7.4 at 4˚C. Samples were embedded in paraffin wax blocks and 5μm sagittal sections cut. For cryotome sectioning, samples were equilibrated in a 30% sucrose/phosphate buffered saline (PBS) solution at 4˚C and then embedded in OCT compound (Fisher Scientific, Loughborough, UK) before 10μm sagittal sections were cut. For human material, 8x3mm blocks of AC were cut from femoral tibial condyles and fixed in neutral buffered formalin and then paraffin wax embedded. Histological staining with Von Kossa (Abcam, Cambridge, UK), toluidine blue (Sigma) and haematoxylin and eosin (Sigma) were carried out according to standard procedures. All histological scoring was carried out on the lateral tibial condyle with AC damage determined by a binary scoring system, of 'normal' or 'damaged'. At least three slides separated by 25μm were analysed for each limb. For cell and immunohistochemical scoring, cell-types or positive staining cells were expressed as a percentage of the total chondrocyte count. The number of empty lacunae were expressed per mm of AC analysed. Paraffin sections were de-waxed, blocked for endogenous peroxidase and underwent antigen retrieval in 10mM sodium citrate pH6 at 80˚C for 30-60 minutes. Slides were blocked with serum-free pan-species block (DAKO, Glostrup, Denmark), incubated with primary antibodies overnight at 4˚C, and incubated with biotinylated secondary antibodies for 45mins at room temperature. Sections underwent streptavidin-mediated signal amplification (ELITE ABC, Vectorlabs, Burlingame, US) prior to incubation with peroxidase substrate kit DAB (Vectorlabs).

μCT monitoring
For longitudinal studies, eight-week-old mice were anesthetized with isoflurane prior to invivo μCT scanning using a Skyscan 1076 (Bruker, USA, MA) at a resolution of 18μm isotropic voxel size. Procedures were repeated every four weeks and carried out under local ethical approval (ERF WGH-14-74). Fixed limbs were imaged at 18μm resolution using a Skyscan 1076 (Brucker, USA, MA) at 65kV, 110uA, 0.5mm Al filter, 700ms exposure, 0.6-degree rotation step, 180-degree rotation, 2 frame averaging.

μCT image processing
Raw μCT image stacks was reconstructed using GPU-based NRecon (Bruker, USA, MA) using identical parameters within each study type (10% beam hardening and image conversion range of 0.003-0.1125. Reconstructed image stacks were imported into CTAn (Bruker, USA, MA) for selecting regions of interest and acquiring 2D density maps, volumetric quantification of ectopic structures and generation of surface rendered 3D models. Surface rendered 3D were visualized in CTVol (Bruker, USA, MA). 3D volume rendered models were generated in CTVox using a colour-coded transfer functions to identify low, medium and high densities. Longitudinal studies used IMARIS (Bitplane, UK) to create, view and perform volumetric quantification of 3D models.

RNA extraction and q-RT-PCR analysis
Individual joint components were micro-dissected and stored in liquid nitrogen. RNA was extracted using Trizol reagent (Life Technologies), according to manufacturer's instructions. RNA was reverse-transcribed using QuantiTect Reverse Transcription Kit (Qiagen). The qRT-PCR was performed using LightCycler 480 SYBR Green I Master (Roche, Germany) and target gene expression normalized to Rpl5 and analysed using the ΔΔCT method [53].
The complete Ubr5 cDNA was synthesised from murine embryonic stem cells total RNA [9] and cloned into a modified pcDNA5/FRT vector (Life Technologies) containing an aminoterminal 2×HA/2×Strep. NIH3T3 cells (American Type Culture Collection, USA) were seeded at a density of 100,000/ml and transfected after 24hr with pcDNA3.1 alone, UBR5 and pcDNA3.1, Gli1 and pcDNA3.1, or Gli1 and Ubr5 using FuGENE6 (Roche). After 48hrs, the medium was replaced by DMEM/0.5% FCS, and cells were lysed 24 hrs later in Laemmli buffer. Whole cell lysate was separated on a 6% SDS-PAGE and transferred onto nitrocellulose membranes. Membranes were blocked in 5% non-fat milk, incubated with primary antibodies overnight at 4˚C at 1:1,000 dilution for GLI1 (Cell Signalling) or 1:10,000 dilution for β-actin (Sigma). Secondary HRP-conjugated-anti-mouse antibody was applied at a 1:2,000 dilution for 1hr at room temperature. The membranes were developed using the Clarity western ECL substrate (BioRad, USA, CA).

Retrovirus production and stable Ubr5 silencing
Previously validated shRNA-encoding oligos targeting murine Ubr5 and or a scrambled sequence were cloned into pLKO.1-puro (Sigma). shUBR5 and shScrambled-pLKO.1-puro were co-transfected with pCMV-VSV-G and pCMV-dR8.2 dvpr plasmids into HEK 293T cells using TransIT293 reagent (Mirus Bio LLC, USA). To generate stable silenced shUBR5 cells, NIH3T3 cells were seeded at 120,000 cells/ml and infected with 0.5 ml shScramble or shUBR5 retroviral supernatant in the presence of 8 mg/ml polybrene (Sigma). The media was changed after 24 hrs and cells were selected with 2 mg/ml puromycin 48 hrs post-infection.

cAMP assay
Control (scrambled) or knock down (Ubr5 shRNA) NIH3T3 cells were seeded at 130,000 cells/ ml, serum starved overnight, and stimulated with 10μM forskolin (FORSK) for 5min. Cells were pre-incubated with 5μM purmorphamine for 10min before addition of FORSK. Cells were processed according to Parameter cAMP Enzyme Immune Assay (R&D Systems) instructions.

Statistical analysis
Data analysis and statistics was performed using PRISM software (GraphPad, La Jolla, US). Count data was analysed using a contingency table and either two-sided Chi square or Fisher's exact tests according to count size. Continuous data was analysed using unpaired, two-tailed Students t-tests. The level of significance for all tests was set at p = <0.05. Overview of the metaplastic events of various Ubr5 mt tissues. Red = non-cartilaginous tissues (Synovium, AT and Superficial Digital Flexor tendon); Orange = cartilaginous tissues (retinaculum); Yellow = calcified cartilage; Green = normotopic bone; Blue = heterotopic or enlarged normotopic bone. Arrows indicate the direction of metaplasia, with the arrowhead indicating the tissue type in 24-week-old Ubr5 mt and/or Ubr5 mt +Smo LoF animals. Metaplastic tissue events unique to Ubr5 mt +Smo LoF are indicated by an asterisk. (TIF)