SUMOylation in Skeletal Development, Homeostasis, and Disease

The modification of proteins by small ubiquitin-related modifier (SUMO) molecules, SUMOylation, is a key post-translational modification involved in a variety of biological processes, such as chromosome organization, DNA replication and repair, transcription, nuclear transport, and cell signaling transduction. In recent years, emerging evidence has shown that SUMOylation regulates the development and homeostasis of the skeletal system, with its dysregulation causing skeletal diseases, suggesting that SUMOylation pathways may serve as a promising therapeutic target. In this review, we summarize the current understanding of the molecular mechanisms by which SUMOylation pathways regulate skeletal cells in physiological and disease contexts.


Introduction
The emergence of a skeletal system was a leap forward in evolution, for it created a strong framework for the body, protecting vital organs, facilitating movement, establishing niches for hematopoiesis, and serving as a mineral reservoir.
The skeletal system develops from mesenchymal cells originating from the ectoderm and mesoderm through one of two types of ossifications processes: intramembranous or endochondral ossification. In intramembranous ossification, mesenchymal cells directly differentiate into osteoblasts to generate the flat bones of the skull and lateral clavicles [1]. Endochondral ossification, which gives rise to the bones at the base of the skull and the long bones, starts from mesenchymal cell condensation followed by primary and secondary ossification [2]. Condensed mesenchymal cells first undergo chondrogenic differentiation to form cartilage templates [2][3][4]; next, chondrocytes in the center of the cartilage templates mature and differentiate into hypertrophic chondrocytes that secrete factors to promote vascular invasion [2][3][4][5]. This brings in hematopoietic cells from the blood and osteogenic progenitors from the perichondrium [2][3][4][5]. Next, osteoblasts, derived from either osteogenic progenitors or hypertrophic chondrocytes, produce bone matrix to replace the cartilage templates generated by the apoptotic hypertrophic chondrocytes [2][3][4][5][6][7]. At the same time, bone-absorbing osteoclasts derived from the hematopoietic lineage remodel the bone and form the bone marrow cavity [8]. Secondary ossification areas form at the center of the cartilage at both ends of long bones in a process similar to primary ossification [8,9], dividing cartilage into two parts: the growth plate, which contains growth plate chondrocytes (GPCs); and articular cartilage, which consists of articular cartilage chondrocytes (ACCs). The finely controlled, directional chondrocyte proliferation and differentiation in the growth plate propels bone elongation. The coupling between osteoblast-mediated bone formation and osteoclast-mediated bone resorption continues throughout life to maintain bone tissue homeostasis [10,11].
The development and homeostasis of the skeletal system require diverse and responsive signaling and cell-cell communication, which heavily rely on dynamic posttransla- (1) The n SUMO precursor protein is proteolytically cleaved to expose its C-terminal Gly-Gly motif by S (2) Mature SUMO is then activated by a heterodimer consistent of SAE1 and SAE2, the E1 co in an ATP dependent reaction, resulting in the formation of a thioester bond between SUM SAE2. (3) The activated SUMO is transferred to the E2 enzyme, UBC9. (4) With or without th of an E3 ligase, UBC9 conjugates the SUMO group to the substrate protein by forming an isop bond on a Lys residue. (5) SUMO modifications are removed from the target protein by SEN In addition to proteolyzing the SUMO precursor, SENPs can also remove SUMO teins from their targets, a process known as deSUMOylation [31]. Seven SENP pr have been identified in humans (SENP1-3, SENP5-7, and SENP8 [19]). SENP1, 2, 3, catalyze both SUMO maturation and deconjugation, whereas SENP6 and 7 do not ca SUMO maturation but instead have a poly-SUMO chain-editing function [28,37,38 sides the SENP family, three additional SUMO proteases have been identified in hu desumoylating isopeptidase 1 and 2 (DeSI1 and DeSI2) [39] and ubiquitin-specific ase-like 1 (USPL1) [40]. These desumoylases share little sequence homology wi SENP proteases, and their functions are less well characterized [41].
The effects of SUMO modifications on their target proteins are diverse and are m classified into three categories [13]: first, the attachment of the SUMO group can binding sites of the target protein, thus impairing its interaction with other mol [13,42]; second, SUMOylation can introduce novel binding sites within the target pr thus conferring novel molecular interactions [13,42]; finally, SUMO can change the ture of the target protein, thereby affecting its function [13,42]. The SUMOylatio SUMOylation equilibrium regulates many cellular processes, including DNA dama sponse, mitochondrial dynamics, cell growth, proliferation, senescence, and apop Disruption of this balance is associated with many diseases, including cancer, n degenerative diseases, heart disease, and skeletal diseases, such as osteoarthritis (OA rheumatoid arthritis (RA) [29,[43][44][45].

SUMOylation in Skeletal Cell Differentiation, Homeostasis, and Disease
Osteoblasts, chondrocytes, and osteoclasts are the major cell types of the skeleta tem and cooperate seamlessly to regulate bone development and homeostasis [46,4 (1) The nascent SUMO precursor protein is proteolytically cleaved to expose its C-terminal Gly-Gly motif by SENPs.
(2) Mature SUMO is then activated by a heterodimer consistent of SAE1 and SAE2, the E1 complex, in an ATP dependent reaction, resulting in the formation of a thioester bond between SUMO and SAE2. (3) The activated SUMO is transferred to the E2 enzyme, UBC9. (4) With or without the help of an E3 ligase, UBC9 conjugates the SUMO group to the substrate protein by forming an isopeptide bond on a Lys residue. (5) SUMO modifications are removed from the target protein by SENPs.

SUMOylation in Skeletal Cell Differentiation, Homeostasis, and Disease
Osteoblasts, chondrocytes, and osteoclasts are the major cell types of the skeletal system and cooperate seamlessly to regulate bone development and homeostasis [46,47].

SUMOylation in Osteogenesis, Osteoblast Homeostasis, and Bone Mass Regulation
SUMOylation regulates key signaling pathways and transcription factors of osteogenesis and osteoblasts; the requirement for this PTM is demonstrated by the dysregulation of bone development and homeostasis when SUMOylation is disrupted (selected examples are illustrated in Figure 2).
We reported that postnatal and ubiquitous loss of SENP6 leads to kyphosis, a sign of premature skeletal aging [48]. Furthermore, mice with OCP-specific Senp6 knockout have small skeletons and decreased trabecular bone mass and cortical thickness, as well as delayed secondary ossification center formation [48]. OCP-derived cell lineages lacking Senp6 undergo severe apoptosis and cellular senescence. Mechanistically, Senp6 loss results in excessive SUMOylation of the multifaceted protein TRIM28, which is involved in chromatin silencing, transcriptional repression, and p53 inhibition. SUMOylation destabilizes TRIM28 and weakens TRIM28-mediated p53 repression, leading to OCP/chondrocyte apoptosis and senescence [48] (Figure 2A). SUMOylation regulates key signaling pathways and transcription factors of osteogenesis and osteoblasts; the requirement for this PTM is demonstrated by the dysregulation of bone development and homeostasis when SUMOylation is disrupted (selected examples are illustrated in Figure 2). We reported that postnatal and ubiquitous loss of SENP6 leads to kyphosis, a sign of premature skeletal aging [48]. Furthermore, mice with OCP-specific Senp6 knockout have small skeletons and decreased trabecular bone mass and cortical thickness, as well as delayed secondary ossification center formation [48]. OCP-derived cell lineages lacking Senp6 undergo severe apoptosis and cellular senescence. Mechanistically, Senp6 loss results in excessive SUMOylation of the multifaceted protein TRIM28, which is involved in chromatin silencing, transcriptional repression, and p53 inhibition. SUMOylation destabilizes TRIM28 and weakens TRIM28-mediated p53 repression, leading to OCP/chondrocyte apoptosis and senescence [48] (Figure 2A).
Importantly, SUMOylation regulates TGF-β/BMP signaling, a fundamental and diverse signaling network that controls embryonic skeletal development and postnatal bone homeostasis [49][50][51][52]. TGF-β/BMP superfamily ligands interact with their heteromeric receptor complexes and transmit extracellular signals to the nucleus via SMAD proteins [49][50][51][52]. In the human Saos-2 osteosarcoma cell line, SMAD4 interacts with and is SUMOylated by UBC9. Knockdown of Ubc9 decreases the levels of SMAD4 protein and phosphorylated SMAD1, prevents the nuclear accumulation of SMAD1 and 4, and decreases the expression of osteogenic transcription factors downstream of BMP (Runx2, Dlx5, Msx2, and Osx) [53]. In contrast, Ubc9 knockdown can elevate BMP signaling and enhance osteogenic differentiation in C2C12 mouse myoblasts and ST2 mouse bone-marrow derived stromal cells (BMSCs) [54]. Mutation of the SMAD4 SUMOylation site (K158R) increases SMAD4 transcriptional activity [54]. It is unclear why UBC9 loss has the opposite effect on BMP/SMAD signaling in these studies. One explanation would be that these cells express different SUMO E3 ligases, which SUMOylate SMAD at various sites, leading to different signaling outcomes. The loss of SUMOylation, then, would have grossly different cellular effects. This possibility needs further investigation. Importantly, SUMOylation regulates TGF-β/BMP signaling, a fundamental and diverse signaling network that controls embryonic skeletal development and postnatal bone homeostasis [49][50][51][52]. TGF-β/BMP superfamily ligands interact with their heteromeric receptor complexes and transmit extracellular signals to the nucleus via SMAD proteins [49][50][51][52]. In the human Saos-2 osteosarcoma cell line, SMAD4 interacts with and is SUMOylated by UBC9. Knockdown of Ubc9 decreases the levels of SMAD4 protein and phosphorylated SMAD1, prevents the nuclear accumulation of SMAD1 and 4, and decreases the expression of osteogenic transcription factors downstream of BMP (Runx2, Dlx5, Msx2, and Osx) [53]. In contrast, Ubc9 knockdown can elevate BMP signaling and enhance osteogenic differentiation in C2C12 mouse myoblasts and ST2 mouse bone-marrow derived stromal cells (BMSCs) [54]. Mutation of the SMAD4 SUMOylation site (K158R) increases SMAD4 transcriptional activity [54]. It is unclear why UBC9 loss has the opposite effect on BMP/SMAD signaling in these studies. One explanation would be that these cells express different SUMO E3 ligases, which SUMOylate SMAD at various sites, leading to different signaling outcomes. The loss of SUMOylation, then, would have grossly different cellular effects. This possibility needs further investigation.
Hormones and their receptors, especially the androgen receptor (AR), are important regulators of skeletal development. AR knockout dramatically reduces trabecular and cortical bone mass [55]. SUMOylation of ARs is necessary for bone mass maintenance, as mutations (K381R and K500R) within the AR SUMOylation site result in significantly decreased trabecular bone and cortical bone mass [56] ( Figure 2B). Of note, while loss of AR SUMOylation decreases osteoblast numbers, the number of osteoclasts is unaffected [56].
Our own studies have shown that inhibition of SUMOylation can yield profound effects on BMSC fate determination between osteogenesis and adipogenesis. We reported that ginkgolic acid, a SUMOylation inhibitor that binds to E1 ligase to prevent the formation  [67]. Consistent with our findings, PPAR-γ SUMOylation inhibits PPAR-γ transcriptional activity in BMSCs [68]. When stimulated with GDF11 (a TGFβ family member), PPAR-γ SUMOylation attenuates adipogenesis in favor of osteogenesis [68].
SUMOylation is also implicated in the epigenetic regulation of osteogenesis. In human dental follicle stem cells, SENP3 binds to and deSUMOylates RBBP5 [69], an important component of several histone methyltransferase complexes [70][71][72]. This facilitates the formation of active MLL1/MLL2 histone methyltransferase complexes that methylate H3K4 residues on the promoters of DLX3 (an osteogenic transcription factor) and a subset of other HOX genes, thus enhancing osteogenic differentiation [69].
The above studies demonstrate that SUMOylation has crucial roles in regulating osteogenesis, osteoblast homeostasis, and bone mass via broad mechanisms, including regulation of growth factor signaling, hormone receptors, transcription factors, and epigenetic mechanisms.

SUMOylation in Chondrogenesis, Chondrocyte Homeostasis, and Osteoarthritis
Chondrocytes of healthy cartilage are formed by the differentiation of skeletal progenitor/stem cells (SSCs) into GPCs through an intermediate and bipotent osteochondroprogenitor, or into ACCs via a multipotent joint progenitor [4,73,74]. GPCs proliferate and produce the extracellular matrix template for subsequent ossification, thus allowing for fast elongation of bone elements [4,73,74]. In contrast, ACCs are mostly quiescent but secrete and maintain extracellular matrix to sustain the cartilage integrity in response to outside stimuli and tissue damage and to provide a smooth and lubricated surface for articulation [75,76].
SUMOylation also likely regulates the maintenance of heterochromatin structure in articular cartilage. For instance, DGCR8 -which maintains heterochromatin through interactions with TRIM28 and HP1γ -is stabilized to prevent its degradation via the ubiquitinproteasome pathway by SUMO1 modification at the K707 residue [89,90]. DGCR8 stabilizes heterochromatin and reduces senescence of human MSCs. Overexpression of DGCR8 alleviates OA symptoms in mice [89,90]. CLOCK, the core component of the mammalian circadian rhythm machinery, is another protein involved in maintaining heterochromatin. CLOCK transcriptional activity is increased by SUMOylation at residues K67 and K851 [91]. Overexpression of CLOCK1 prevents human MSC aging, and lentivirus-delivered CLOCK expression promotes cartilage regeneration in aged mice [92]. Future studies will need to determine whether SUMOylation of DGCR8 and CLOCK is required for regulating chondrocyte differentiation and homeostasis. Use of mouse genetic or chondrocyte models with DGCR8 or CLOCK1 SUMO sites nullified would be one way to evaluate this.
Osteoarthritis (OA) is a disease that is characterized by progressive loss of cartilage, the formation of bone spurs, and chronic synovial inflammation [93]. OA severely impairs joint function and often causes joint pain [93]. The onset and progression of OA are associated with various risk factors, including gender, genetic predisposition, obesity, joint malalignment, sports injury, and aging [93]. Several lines of evidence suggest that enhanced SUMOylation promotes OA pathogenesis. A large genome-wide association analysis in Europe identified the rs9350591 C/T single nucleotide polymorphism (SNP), located upstream of the SENP6 locus, as one of the most strongly OA-associated SNPs [94]. SENP6 expression is significantly decreased in OA cartilage even in the absence of rs9350591, suggesting that a deficiency in SENP6 desumoylase activity may be a widespread phenomenon in OA [95]. Further support for SUMOylation promoting OA comes from studies where IL-1β treatment of human ACs was shown to induce SUMO1 modification of S100A4 (a member of the Ca 2+ -binding S100 proteins that modulates p53 transcriptional activity), resulting in S100A4 nuclear translocation. Nuclear S100A4 binds to the MMP13 promoter region and increases expression of a major OA-associated protease that degrades cartilage, MMP13 [96] ( Figure 3A).

SUMOylation in Osteoclastogenesis and Osteoclast Function
Osteoclasts differentiate from the hematopoietic cell lineage upon induction by cytokines, such as m-CSF and RANKL, present in the bone and bone marrow microenvironment [101,102]. Osteoclast progenitors differentiate, fuse, and form multinucleated mature osteoclasts, which produce acid and matrix-degrading proteases and serve as dedicated bone-resorbing cells of the skeletal system [101,102].
Recent studies revealed the regulatory role of SUMOylation in osteoclast formation and function. For instance, SENP3 suppresses osteoclastogenesis. Mice with the Lyz2-Cre-mediated Senp3 deletion in bone marrow-derived monocytes exhibit decreased bone mass [103]. These knockout mice also have aggravated bone loss after ovariectomy due to overactivation of osteoclasts. Mechanistically, Senp3 deletion increases SUMO3 modification of IRF8 and weakens the ability of IRF8 to suppress NFATc1 (a master regulator of osteoclastogensis) expression [103] (Figure 2B). A similar effect on osteoclastogenesis is observed in transgenic mice overexpressing the SUMO E3 ligase PIAS3 [104]. The mice exhibit an osteopetrotic phenotype caused by impaired osteoclast differentiation [104]. PIAS3 overexpression inhibited c-Fos and Nfatc1 expression in RAW264.7 cells, thereby blunting RANKL-induced osteoclastogenesis [104]. Likewise, in a bone marrow monocyteosteoblast co-culture system, PIAS3 overexpression in osteoblasts downregulated IL6induced RANKL expression and inhibited osteoclast formation. Downregulation of PIAS3 in osteoblasts in the same system increased RANKL expression [104]. Thus, PIAS3 inhibits osteoclastogenesis either by intrinsically inhibiting osteoclast differentiation or by indirectly suppressing the expression of osteoclastogenic cytokines, such as RANKL, from osteoblasts. However, as PIAS3 can also affect transcriptional regulators (such as NFκB and STAT3 signaling [105]) independent of SUMO ligase activity, it is unclear whether the activity of PIAS3 in osteoclastogenesis depends upon its E3 ligase function or not.

Split Hand/Split Foot Malformation (SHFM)
SHFM is a rare limb malformation characterized by clefts in the middle of the hands and feet, as well as syndactyly and aplasia/hypoplasia of phalanges, metacarpals, and metatarsals [106]. P63α mutations are associated with SHFM [107,108]. Notably, SUMO1 is conjugated to K549 and K637 of P63α, following the binding of UBC9 to the C-terminal domain of P63α [109,110]. The SHFM-associated P63α mutation, Q634X, disrupts the interaction between P63α and UBC9. K549E and K637E mutations of P63α, both of which block P63α SUMOylation, markedly increase the transcriptional activity of TAP63α (an isoform of P63α containing the N-terminal transactivation domain) [110]. At the same time, these mutations inhibit the dominant-negative effect of the naturally occurring N-terminus truncated isoform of P63α, ∆NP63α. Both SUMOylation and ubiquitylation are required for the efficient degradation of ∆NP63α [111]. One downstream molecular consequence of loss of P63α SUMOylation was determined using cells expressing mutant P63α lacking the two SUMOylation sites. In these cells, the expression of genes related to bone and tooth development, such as Runx2 and Mint [110], were decreased. These data emphasize the functional importance of SUMOylation of P63α in limb development.

Craniofacial Disorders
Craniofacial disorders are one of the most common human birth defects. Cleft lip and palate are the most frequent types of craniofacial disorders [112], and several studies have linked SUMO1 deficiency to these disorders. First, a balanced chromosomal translocation 46,XX,t(2;8)(q33.1;q24.3) resulting in SUMO1 haploinsufficiency was identified in a patient with isolated cleft lip and palate [113] ( Figure 3B). Second, a 4-SNP SUMO1 haplotype was found significantly associated with non-syndromic cleft lip with or without cleft palate (NSCLP) from a study of 181 patients and 162 healthy controls of Han Chinese origin [114]. Other studies have also related SUMO1 to cleft lip with or without cleft palate, cleft palate only, or NSCLP in Poland [115], Ireland [116], and western China [117] ( Figure 3B). In addition, transcription factors such as TBX22, MSX1, SATB2, P63, PAX9, TRPS1, and EYA1, which contribute to the development of the lip and palate, have all been identified as substrates of SUMO modification [118]. For example, SUMOylation regulates the subnuclear localization, stability, and transcriptional activity of SATB2, affects subnuclear localization of MSX1, modulates the transcriptional activity and stability of P63 (see above section on SHFM), facilitates the transcriptional repressor activity of TBX22, and regulates the transcriptional suppression function of TRPS1 [119][120][121][122][123][124]. In summary, the formation of the lip and palate appears to be particularly sensitive to changes in SUMOylation [118].

SUMOylation in rheumatoid arthritis
Rheumatoid arthritis (RA) is a chronic systemic, inflammatory disease characterized by joint stiffness and destruction [125,126]. Synovial inflammation is a hallmark of RA and the main driver of cartilage degradation. The main cellular features of RA include synovial hyperplasia, increased vascularity, and inflammatory cell infiltration [125,126].
A direct relationship between the SUMOylation pathway and RA was first reported in 2000 [127]. SUMO1 mRNA was found to be highly expressed in synovial specimens from RA patients, predominantly in the synovial fibroblasts of the lining layer and at the sites where cartilage is invaded by synovium [127]. The expression of SUMO1 in RA synovial fibroblasts (RASFs) is over 30 times higher than that found in OA synovial fibroblasts or normal fibroblasts [127]. Furthermore, the expression of the E3 ligase PIAS3 is also increased in RAFLSs and RA synovial tissues [128]. A recent study found that SUMO1 knockdown inhibits the migration and invasion of RA fibroblast-like synoviocytes (RAFLSs) and RAFLS expression of MMP1 and MMP3. Mechanistically, SUMO1 deficiency suppresses the activity of the Rac1/PAK1 pathway, which normally promotes cell motility [129]. PIAS3 promotes the SUMOylation of Rac1 and activates the expression of Rac1 downstream targets, such as PAK1 and JNK [128]. Decreased PIAS3 expression in RAFLS also inhibits the invasion and migration of RAFLSs and the expression of MMP3, MMP9, and MMP13 [128] (Figure 3C). SUMO E1 conjugating enzymes SAE1 and SAE2 are also increased in FLSs and synovial tissues of RA patients [130]. Knockdown of SAE1 or SAE2 by siRNA results in a less aggressive phenotype and reduced inflammation of RAFLSs [130]. SAE1-and SAE2mediated SUMOylation of pyruvate kinase M2 (PKM2) promotes its phosphorylation and nuclear translocation, resulting in the suppression of pyruvate kinase activity, contributing to synovial glycolysis and joint inflammation [130] (Figure 3C).
Further support for increased SUMOylation in RA comes from the finding that expression of the SENP1 desumoylase is decreased in RASFs [131,132]. Mechanistic studies revealed that overexpression of SENP1 can deSUMOylate nuclear promyelocytic leukemia (PML) nuclear bodies and inhibit the recruitment of DAXX, a FADD (Fas-associated death domain)-interacting protein, to PML nuclear bodies, thus promoting the Fas-mediated apoptosis of RASFs [131] ( Figure 3C). In addition, SENP1 suppresses MMP1 expression by promoting HDAC4 binding to the MMP1 promoter, further weakening the invasiveness of RASFs [132] ( Figure 3C).
These studies show that increased SUMOylation is positively related to RA, suggesting that downregulation of SUMOylation may have therapeutic benefits. In support of this, in a mouse collagen-induced arthritis model, downregulation of UBC9 using siRNA can reduce arthritis intensity scores and joint destruction [133]. RA-related markers, including serum levels of anti-collagen (CII) antibodies, VEGF-A, MMP3, and MMP9, were also decreased. Moreover, downregulating UBC9 expression in ex vivo human RAFLS cultures inhibits TNF-α-stimulated secretion of VEGF-A, MMP-3, and MMP-9, and blocks RAFLS proliferation and migration [133] ( Figure 3C).
The only study supporting the anti-RA function of SUMOylation reported that the expression of SUMO2 in RA tissue or RASFs is significantly higher than that of OA tissues and is increased in the synovium and synovial fibroblasts of human TNF-transgenic (hTN-Ftg) mice, a common RA model [134]. TNF-α treatment promotes the expression of SUMO2 in vitro, while SUMO2 knockdown significantly increases the expression of MMP3 and MMP13 induced by the TNF-α-and IL-1β-stimulated NF-κB pathway [134] (Figure 3C).
Most of these studies show that gross alteration of SUMOylation in the joint contributes to the development of OA and RA. Although the detailed mechanisms are still not well understood, some insight may be derived from studies in other disease conditions or cell types, which suggest that SUMOylation regulates inflammation by modulating the NFκB pathway, the PPARγ pathway, among others [68,[135][136][137][138].

SUMOylation in Osteosarcoma
Osteosarcoma is the most common cancer type in the human skeletal system. It occurs in humans in a biphasic pattern, i.e., with a peak in adolescence and another in patients over 60 years of age [139,140]. SUMOylation of proteins has a crucial role in regulating the cell cycle, genome stability, and the expression of oncoproteins and tumor suppressors [141,142], and has been linked to the development of osteosarcoma [82,[143][144][145][146][147][148][149][150][151] (Figure 3D). However, there is no consensus on whether SUMOylation is pro-or anti-tumorigenic in osteosarcoma, as this is likely dependent on the specific proteins and/or the specific SUMOylation site within them that are SUMOylated.

Studies Supporting a Pro-Tumorigenic Effect of SUMOylation
Several studies have linked increased SUMOylation to osteosarcoma. For example, UBC9 is overexpressed in osteosarcoma tissues and cell lines [143]. UBC9 knockdown inhibits the proliferation and migration of osteosarcoma cells and markedly increases the sensitivity of these cells to the combination treatment of herpes simplex virus thymi-dine kinase/ganciclovir (HSV-TK/GCV) [143]. The integrity of gap-junction-mediated intercellular communication (GJIC) is required for the HSV-TK/GCV-induced tumor repression. Ubc9 knockout decreases SUMO1 modification and increases the free protein level of connexin 43 (CX43), a component of gap junctions [143]. Thus, UBC9 deficiency sensitizes osteosarcoma cells to chemotherapy by reconstructing and promoting GJIC [143] ( Figure 3D).
In addition, SENP1 expression is decreased in osteosarcoma tissues, cell lines, and osteosarcoma stem cells compared to non-cancer cells and stem cells [144]. Low SENP1 is essential for maintaining the stemness of osteosarcoma stem cells, and overexpression of SENP1 markedly decreases the stemness of osteosarcoma cells while sensitizing them to apoptosis induced by HSV-TK/GCV combination treatment [144] ( Figure 3D). This highlights the potential for using SENP1 activation for the treatment of osteosarcoma. SENP2 expression is also significantly decreased in osteosarcoma compared with adjacent normal tissue [82]. SENP2 overexpression inhibits osteosarcoma cell proliferation, migration, and invasion, while SENP2 knockout by CRISPR-Cas9 has the opposite effect [82]. Mechanistically, SENP2-mediated deSUMOylation promotes SOX9 ubiquitylation and degradation [82]. SOX9 knockdown greatly reduces the proliferation and invasiveness of the SENP2 knockout osteosarcoma cells [82]. This study suggests that SENP2 acts as an osteosarcoma suppressor by destabilizing SOX9 ( Figure 3D).
Talin is a key component of focal adhesions [152] and can be modified by SUMOylation in U2OS osteosarcoma and MDA-MB-231 breast cancer cells. Using ginkgolic acid (GA) to inhibit SUMOylation increases the number and size of talin-containing focal adhesions [145]. Inhibition of SUMOylation can significantly reduce the migration of MDA-MB-231 breast cancer cells, but this effect was not studied in U2OS cells [145]. Cumulatively, these studies indicate that SUMOylation can promote osteosarcoma proliferation, invasion, and migration, and that targeting it may be a relevant point of therapeutic intervention.

Studies Supporting an Anti-Tumorigenic Effect of SUMOylation
In contrast to what was presented above, several studies suggest that SUMOylation can have anti-osteosarcoma effect. For example, the desumoylase SENP5 is highly expressed in osteosarcoma cells and tissues [146]. Silencing SENP5 expression in two osteosarcoma cell lines, U2OS and Saos-2, significantly inhibits growth and colony formation and promotes apoptosis [146]. The inhibition of osteosarcoma growth following SENP5-knockdown is likely via an increase in caspase-3/-7 activity (apoptosis activators) and a decrease in the expression of the cell cycle gene, cyclin B1 [146] ( Figure 3D).
The expression of the E3 ligase, PIASxα, is lower in osteosarcoma compared to adjacent tissue [147]. Notably, PIASxα overexpression can significantly inhibit osteosarcoma cell proliferation and increase apoptosis [147], whereas PIASxα silencing in U2OS cells increases the expression of cyclin D kinase genes. Moreover, PIASxα overexpression weakens the tumorigenic potential of U2OS cells in nude mice [147]. Because of the aforementioned pleiotropic functions of PIASxα, further studies are needed to determine whether the anti-tumor effects observed here depend on SUMO-E3 ligase activity of PIASxα.
As another example, all-trans-retinoic acid (ATRA), is an anti-cancer drug that induces osteosarcoma cell differentiation, which is used as a prognostic indicator of weakened osteosarcoma malignancy and tumor progression [148]. SUMO1 deletion blocks the antiosteosarcoma efficiency of ATRA, demonstrating that SUMO1 is required for the prodifferentiation effect of ATRA [149]. In addition, the target of ATRA, retinoic acid receptor α (RARα), is stabilized by SUMOylation at K399 [149]; and mutation of K399 of RARα impairs ATRA-induced osteosarcoma cell differentiation [149]. These suggest that SUMO1 acts as an anti-osteosarcoma molecule by targeting RARα (Figure 3D).
In a recent study, higher expression of SENP1 was found in human osteosarcoma tissue than in adjacent normal tissue (53/60 vs. 28/60) [151]; and levels of SENP1 derived from patient plasma exosomes directly correlate with osteosarcoma tumor size, location, necrosis rate, pulmonary metastasis, and surgical stage [151]. Furthermore, patients with higher plasma levels of exosome-derived SENP1 had worse tumor malignancy and overall survival rate. Notably, the prognostic value of plasma exosome-derived SENP1 levels in osteosarcoma was found to be better than plasma SENP1 [151]. In the human osteosarcoma cell line MG-63, a hypoxic environment (similar to that typically found in tumors) induces the expression of high amounts of SENP1 [150], and inhibiting SENP1, in turn, reduces the expression of two major hypoxia-induced genes, HIF1α and VEGF (vascular endothelial growth factor). This reduction in HIF1α normalizes hypoxia-induced SENP1 expression [150]. SENP1 knockdown accelerates apoptosis by decreasing Bcl-2 expression while increasing Bax expression. The ultimate effect of this is the reduction of MG-63 cell invasiveness by suppressing epithelial-mesenchymal transition (EMT) genes [150]. These findings suggest a positive feedback loop between SENP1 and HIF1α in regulating proliferation, invasion, and EMT of osteosarcoma cells in hypoxic conditions ( Figure 3D).
In summary, changes in both SUMOylation and deSUMOylation enzymes are related to osteosarcoma. UBC9, the only SUMO E2 enzyme, promotes osteosarcoma, suggesting that a global increase in SUMOylation favors osteosarcoma development. However, several SENP deSUMOylases show diverse roles in osteosarcoma, reflecting that the SUMOylation status of their specific targets are the determinants of osteosarcoma development. While we could not make a simple generalization whether SUMOylation is osteosarcoma-promoting or -inhibiting, these studies demonstrate SUMOylation as a crucial PTM in osteosarcoma tumor initiation and progression.

SUMOylation in Chondrosarcoma
SUMOylation is also associated with malignant tumors that form from bone cartilage, known as conventional chondrosarcoma [139,153]. SUMO1 and SUMO2/3 expression are positively correlated with increased aggressiveness of chondrosarcomas, and patients with high SUMO2/3 expression have poorer survival outcomes [153]. The authors of that study suggest that SUMO expression could be a useful prognostic marker in chondrosarcoma. Future studies are warranted to explore this and to identify the SUMOylation targets that promote chondrosarcoma development.

Summary and Future Perspectives
PTM by SUMOylation regulates signaling pathways and transcription factors that are crucial for skeletal cell differentiation, development, and homeostasis ( Figure 2). Dysregulation of SUMOylation is associated with skeletal diseases, such as OA and RA, craniofacial defects, and bone tumors ( Figure 3). Thus, targeting SUMOylation/deSUMOylation pathways is a promising strategy for the development of new treatments for these disorders. However, this requires a better characterization of the SUMOylation/deSUMOylation machinery and identifying regulators and effectors (substrates) of SUMOylation/deSUMOylation. The establishment of tissue-and disease-specific mouse genetic models that manipulate the expression of SUMO pathway components in skeletal cells will be a valuable resource to achieve this goal.
Targeting SUMOylation for disease treatments is in a more preliminary state than that of targeting ubiquitination, but some basic studies suggest that this is a promising pursuit. For example, the SUMO E1 inhibitor ML792 reduces proliferation, viability, and colony formation of chondrosarcoma cell lines [153]. The SUMOylation inhibitor of TRPS1 (a transcriptional repressor of genes important for bone and cartilage development and maintenance), GSK145A(23), has been developed to treat bone and cartilage diseases [154]. A novel SENP1 inhibitor, senpPNA-R8, is also being investigated in a clinical trial for treating osteosarcoma [155].
Finally, SUMOylation is involved in regulating stress response, epigenetics, and senescence, all of which are closely associated with aging. Future studies dissecting the relationship between SUMOylation and aging and tissue rejuvenation will likely also bring forth new approaches to promote skeletal health.