Similar sequences but dissimilar biological functions of GDF11 and myostatin

Growth differentiation factor 11 (GDF11) and myostatin (MSTN) are closely related TGFβ family members that are often believed to serve similar functions due to their high homology. However, genetic studies in animals provide clear evidence that they perform distinct roles. While the loss of Mstn leads to hypermuscularity, the deletion of Gdf11 results in abnormal skeletal patterning and organ development. The perinatal lethality of Gdf11-null mice, which contrasts with the long-term viability of Mstn-null mice, has led most research to focus on utilizing recombinant GDF11 proteins to investigate the postnatal functions of GDF11. However, the reported outcomes of the exogenous application of recombinant GDF11 proteins are controversial partly because of the different sources and qualities of recombinant GDF11 used and because recombinant GDF11 and MSTN proteins are nearly indistinguishable due to their similar structural and biochemical properties. Here, we analyze the similarities and differences between GDF11 and MSTN from an evolutionary point of view and summarize the current understanding of the biological processing, signaling, and physiological functions of GDF11 and MSTN. Finally, we discuss the potential use of recombinant GDF11 as a therapeutic option for a wide range of medical conditions and the possible adverse effects of GDF11 inhibition mediated by MSTN inhibitors.


Introduction
Cytokines of the transforming growth factor β (TGFβ) family, including activins, growth differentiation factors (GDFs), bone morphogenetic proteins (BMPs), and TGFβs, have been extensively implicated in the regulation of developmental patterning, cellular proliferation and differentiation, and the maintenance of tissue homeostasis 1 . Among the TGFβ family members, there are two highly homologous proteins, GDF11 and myostatin (MSTN), which share 89% sequence identity in their mature form but exhibit distinct endogenous functions. While Gdf11 is expressed broadly in numerous tissues, Mstn is expressed primarily in skeletal muscle [2][3][4] . The functional divergence of GDF11 and MSTN is indicated by the fact that their mutation in animals leads to the development of largely dissimilar features. For instance, while the genetic deficiency of MSTN leads to a hypermuscular phenotype in various species [4][5][6][7][8] , homozygous deletion of Gdf11 generates defects in axial skeletal patterning and organ development in mice 9 . However, unlike the relatively consistent reports of the function of MSTN in suppressing skeletal muscle growth, the reports of GDF11 function, particularly those examining the postnatal role of GDF11, remain highly controversial. One of the main reasons for this controversy lies in the fact that Gdf11-null mice, unlike Mstn-null mice, show perinatal lethality 9 , leading most studies to utilize recombinant proteins that cannot fully recapitulate the complex endogenous functions of GDF11. Importantly, in contrast to studies that utilized recombinant GDF11 or MSTN proteins, those that applied genetic knockdown, knockout, or conditional knockout techniques revealed relatively unvarying results despite their being fewer in number, and most have reported the positive roles of GDF11 and the negative roles of MSTN in the regulation of the development of various tissues. In this review, we first present the similarities and differences between GDF11 and MSTN from an evolutionary point of view and summarize the insights obtained to date regarding the biological processing, signaling mechanisms, and physiological functions of GDF11 and MSTN during development, adulthood, and aging. We also discuss the potential of recombinant GDF11 protein as a therapeutic option for various clinical conditions and the possible adverse effects of GDF11 inhibition mediated by MSTN inhibitors.

Evolutionary analysis of GDF11 and MSTN
The remarkable sequence similarity between GDF11 and MSTN has led to the assumption that they were derived from the same ancestral gene through gene duplication. Indeed, analysis of multiple invertebrate species revealed that they harbor a single homologous gene corresponding to GDF11 and MSTN 10 . For instance, in Caenorhabditis elegans, daf-7 was shown to encode a homolog of GDF11 and MSTN, while in fruit flies (Drosophila melanogaster), myoglianin (Myo) was found to exhibit the highest sequence homology to GDF11 and MSTN [10][11][12][13] . An important question that arose from these identifications was whether the divergence of GDF11 and MSTN occurred at the time of the emergence of vertebrates. To provide an explanation, a phylogenetic study was conducted in various invertebrate and vertebrate species, and importantly, the amphioxus (Branchiostoma belcheri) 14 , which is an invertebrate known to be the closest relative of the vertebrates, was included in the analysis ( Fig. 1a and Table 1). Additionally, the amino acid sequences of the full-length protein, the propeptide with the signal peptide, and C-terminal peptide were separately compared ( Fig. 1b−d). All phylogenetic trees demonstrated a clear separation between the GDF11 and MSTN clusters that appeared after the divergence of vertebrates from the amphioxus, confirming that the gene duplication event occurred at the time when vertebrates and invertebrates split (Fig. 1b−d). Notably, unlike the single isoform of the MSTN gene observed in mammals, two isoforms of the mstn gene have been detected in fish 10 . The reason for and functional significance of the divergence of the two mstn genes in fish remains to be clarified. Interestingly, many of the reported functions of the invertebrate MSTN/GDF11 protein are very different from the well-established suppressive role of vertebrate MSTN in the development of multiple tissues, and the broad expression pattern of the ancestral protein more closely resembles the expression pattern of vertebrate GDF11 11,13,[15][16][17][18][19] . These observations imply that MSTN most likely emerged from the ancestral gene to allow more specific control of skeletal muscle growth in vertebrates, although the relatively small amount of information available on the function of invertebrate MSTN/ GDF11 limits further interpretation. The reported physiological roles of the ancestral protein in invertebrates will be discussed in more detail later.

Proteolytic processing of GDF11 and MSTN
Both GDF11 and MSTN, like the other members of the TGF-β family, are initially synthesized as precursor proteins and are subsequently cleaved by proteases to produce biologically active mature ligands. More specifically, following the removal of the signal peptides by signal peptidases, furin-like proteases recognize and cleave the conserved RSRR residues of GDF11 and MSTN, generating N-terminal propeptides and C-terminal mature peptides 20 . The different types of furin-like proprotein convertases and their substrates are listed in Table 2. The proprotein convertase PC5/6 was demonstrated to specifically cleave GDF11 by recognizing the RSRR↓N cleavage motif, which is not present in MSTN 21 . Accordingly, mice deficient in PC5/6 were shown to phenocopy Gdf11null mice by exhibiting anterior homeotic transformations of the vertebrae, the lack of a tail, kidney agenesis, and retarded ossification 21 . After the cleavage of the RSRR site by a furin-like protease, the propeptide and mature peptide remain noncovalently associated with each other, forming a latent complex that is unable to bind receptors. However, a recent study showed that the latent MSTN complex can also become capable of binding receptors after being exposed to acidic conditions. Exposure to acidic conditions led to a conformational change of the latent MSTN complex and stimulated it to become in a triggered state, in which the pro-and mature domains still remain associated but were capable of signaling 22 . The fact that MSTN can exist in both fully latent and triggered states further demonstrates the complexity of its activation mechanism. Nonetheless, to achieve full signaling activity, both the latent GDF11 and MSTN complexes require additional cleavage of the N-terminal propeptides by BMP1/tolloid (TLD)-like metalloproteinases, which dissociate the propeptides from the mature C-terminal dimers, thus freeing the ligands for receptor binding (Table 3) 20 . Mature dimers can also be inhibited by the addition of propeptides both in vitro and in vivo 23 .
To examine the rates of the evolutionary changes of the residues of GDF11 and MSTN, we utilized a recently developed webtool, Aminode 24 , and analyzed the evolutionarily constrained regions (ECRs) of the proteins ( Fig. 2a and Supplementary Table S1). As expected, the mature domains of GDF11, MSTN, activins, and TGF-βs were remarkably well-conserved among vertebrate species, displaying extremely low rates of amino acid substitution in most positions (Fig. 2a). Surprisingly, only GDF11 presented a striking degree of sequence conservation in the prodomain, emphasizing the functional significance of this region (Fig. 2a). In fact, while GDF11 and MSTN share 89% amino acid sequence identity in their mature domains, which differ by only 11 residues (Fig. 2b, c), their prodomains share only 48% amino acid sequence identity. This suggests the strong possibility that GDF11 prodomains may be associated with distinct and crucial extracellular regulatory mechanisms and biological functions that are not observed for the prodomains of MSTN, which warrants further investigation that may uncover significant differences that were previously unnoticed for the mature ligands. Phylogenetic tree generated by full-length protein sequence comparison. c Phylogenetic tree generated by N-terminal (propeptide with signal peptide) sequence comparison. d Phylogenetic tree generated by C-terminal peptide sequence comparison. Multiple sequence alignments were performed using MEGA X software 127 and the MUSCLE (multiple sequence comparison by logexpectation) algorithm 128 . All phylogenetic trees were constructed using MEGA X software by applying the neighbor-joining method, bootstrap method (1000 replicates), and Jones−Taylor−Thornton model. Gaps and missing data were treated as complete deletions. The numbers at the tree nodes indicate the percentage bootstrap values. Scale bars represent the number of substitutions per site.

Molecular mechanisms of GDF11 and MSTN signaling
The mature GDF11 and MSTN ligands first bind to activin type 2 receptors (ACVR2A or ACVR2B) and subsequently recruit type 1 receptors, activin receptor-like kinase 4 (ALK4) or ALK5 to form a heteromeric receptor complex to elicit downstream signaling via phosphorylation of SMAD2 and/or SMAD3 (Fig. 3) 20 . Both GDF11 and MSTN were recently revealed to be capable of also recruiting ALK7, while GDF11 signaled more potently than MSTN through this receptor 25 . Structural analysis also demonstrated that mature GDF11 and MSTN share identical type 2 receptor-binding residues but exhibit differences in the prehelix loop and wrist helix of the type 1 receptor-binding site 20 . Indeed, GDF11 was shown to signal more effectively and induce a greater SMAD3dependent signal through all type 1 receptors than MSTN, and substitution of the residues in the wrist helix of the MSTN type 1 interface with those of GDF11 significantly enhanced the potency of MSTN by improving the stability of the interaction between the prehelix loop and wrist helix 25 . In addition to stimulating SMAD2/3 phosphorylation, recent studies described that GDF11 can strongly activate SMAD1/5/9 phosphorylation in endothelial cells and osteoblasts to promote their proliferation and differentiation, respectively 26,27 . GDF11 was shown to utilize ALK1 receptors to elicit signal transduction through SMAD1/5/9 phosphorylation, which was effectively suppressed by siRNA-mediated knockdown of GDF11 27 . Via another layer of complexity, GDF11 and MSTN may signal through noncanonical pathways by activating other non-SMAD proteins, such as p38 MAPK, ERK, and JNK 20 .
The activities of mature GDF11 and MSTN are tightly modulated by different extracellular binding proteins, including follistatin (FST), follistatin-like 3 (FSTL3/ FLRG), growth and differentiation factor-associated serum protein 1 (GASP1), GASP2, latent TGF-β binding protein 3 (LTBP3), and decorin 20 . In contrast to the FSTlike proteins that antagonize a variety of molecules of the TGF-β family, GASP1 and GASP2 more selectively inhibit mature GDF11 and MSTN 28 . While GASP1 was shown to more potently bind MSTN/GDF11 than GASP2 in vitro 29 , GASP2 was shown to more specifically regulate GDF11 based on the similarity of the embryonic expression patterns of Gasp2 and Gdf11 and the phenotype of Gasp2 knockout mice, which exhibited posterior homeotic transformations indicative of GDF11 overactivity 30 . Recently, Parente et al. 31 demonstrated that transgenic mice ubiquitously overexpressing GASP1 and GASP2 present distinct phenotypes with contrasting expression patterns of Gdf11 and Mstn. The study illustrated that Mstn expression was significantly upregulated in GASP1overexpressing mice, which showed an increase in low oxidative muscle fibers and impaired metabolic homeostasis, but only Gdf11 but not Mstn expression was significantly elevated in GASP2-overexpressing mice, which exhibited an increase in fast glycolytic muscle fibers without metabolic defects 31 . These results provide evidence that distinct extracellular regulatory mechanisms  and endogenous functions are associated with GDF11 and MSTN. Both GDF11 and MSTN circulate in the blood, and maternal deficiency of MSTN was shown to stimulate additional muscle growth in Mstn-knockout pups 32 , implying that GDF11 and MSTN may function as endocrine signaling molecules. However, our previous findings in mosaic mice in which Mstn was deleted exclusively in posteriorly located muscles highlighted the important paracrine function of MSTN in addition to its endocrine action in regulating muscle mass 33 . Furthermore, whether circulating GDF11 levels have physiological relevance was formerly questioned based on a result showing that the molar concentration of circulating GDF11 was approximately 500 times less than that of MSTN 34 . Because GDF11 and MSTN circulate mostly in inactive, latent forms 35 , their local activation patterns and localization of antagonists may largely contribute to distinct physiological effects of GDF11 and MSTN. Therefore, due to the complex modes of action of GDF11 and MSTN, considerable caution is required for the interpretation of the results of tissue-specific deletion of GDF11 or MSTN in conditional knockout mice.

Functions of MSTN/GDF11 in invertebrates
The physiological roles of ancestral MSTN/GDF11 in invertebrates, despite the availability of sequence information, are much less well known than those of GDF11 and MSTN in vertebrates. It should be noted that most of the invertebrate studies that utilized genetic mutations and RNA interference (RNAi) methods have provided evidence that ancestral Mstn/Gdf11 positively regulates the development of diverse tissues and functions similarly to vertebrate GDF11 rather than MSTN (Table 4). Ancestral Mstn/Gdf11 was also shown to exhibit a broad expression pattern, which is similar to that of vertebrate GDF11 but different from the muscle-specific expression pattern of vertebrate MSTN. For instance, C. elegans daf-7, a homolog of GDF11 and MSTN expressed in ASI neurons, has been shown to promote the reproductive growth and development of worms 11,19,36 . Accordingly, daf-7-mutant worms exhibited a slower growth rate and increased dauer entry 11 . Genetic mutations or RNAimediated knockdown of daf-7 also resulted in an increase in fat accumulation 18 , despite a reduction in the feeding rate, and declines in germ cell production and sperm motility 37,38 . However, there have been conflicting reports regarding the role of DAF-7 in the regulation of lifespan. While Shaw et al. 39 reported that daf-7 mutants and wildtype worms treated with daf-7 RNAi exhibited an increased lifespan, Fletcher and Kim 12 more recently demonstrated that DAF-7 promotes lifespan extension in response to dietary restriction and that age-dependent reduction in daf-7 expression impairs the sensitivity of aged worms to the effects of dietary restriction on lifespan.
The insect gene myo, which is a homolog of GDF11 and MSTN, is strongly expressed in muscle and glial cells and has been shown to promote neuronal development and remodeling 13,40,41 , prevent age-related muscular dysfunction 16 , and extend the lifespan in Drosophila 16,42 . Specifically, RNAi-mediated knockdown of myo in glia 13 or muscle 16 in Drosophila resulted in neuronal remodeling defects or exacerbated age-related climbing defects accompanied by premature death, respectively. Furthermore, a recent study suggested that Myo extends lifespan in flies by exerting protective functions in muscle homeostasis through regulating 26S proteasome function 42 . However, whether Myo regulates muscle mass in flies requires further investigation due to the existence of conflicting reports. As an illustration, while Demontis et al. 16 showed no changes in muscle mass, body weight, and feeding behavior upon either muscle-specific Myo suppression or overexpression, Augustin et al. 43 demonstrated that muscle-specific silencing of Myo increased larval weight and body-wall muscle size. More recently, Upadhyay et al. 44 re-examined the same mutant flies and suggested that Myo deficiency did not result in larger muscles and that Myo is functionally distinct from vertebrate MSTN in terms of regulating muscle size. They also reported that Myo promotes imaginal disc growth in Drosophila 44 . Meanwhile, depletion of Myo through RNAi injection in third-instar cricket nymphs prevented the normal molting cycle and metamorphosis and led to reductions in nymph body size and weight, although the extended developmental period of the RNAi-treated nymphs eventually led them to exhibit a larger final insect size 45 . Injection of RNAi targeting myo into either fifthor sixth-instar nymphs resulted in developmental  MSTN  INHBA  INHBB  TGFB1   TGFB3   TGFB2   GDF11  MSTN  INHBA  INHBB  TGFB1   TGFB3 TGFB2 arrest and death, highlighting the crucial role of Myo in promoting proper insect development 45 . Similar functions of Myo were also reported for cockroaches 46 .

H. sapiens TGFB2
In shrimp, the ancestral Mstn/Gdf11 gene is expressed in diverse tissues, including muscle, hepatopancreas, eyestalk, heart, gill, and stomach, with the highest expression detected in the heart 17 . Endogenous expression of Mstn/Gdf11 in shrimp muscle has been shown to peak immediately after molting, a period when significant growth occurs without restriction by a hard exoskeleton 17 . Interestingly, downregulation of the shrimp Mstn/Gdf11 gene by tail-muscle injection of sequence-specific dsRNA led to a significantly impaired growth rate (68% reduction in final shrimp mass) 17 , an effect opposite to that observed after suppression of MSTN in vertebrates. Likewise, a separate study on shrimp revealed that silencing of the Mstn/Gdf11 gene by tail-muscle injection of dsRNA significantly diminished growth and the survival rate 15 , indicating that ancestral MSTN/GDF11 in invertebrates is a positive regulator of growth and development, unlike vertebrate MSTN. Moreover, Zhuo et al. 47 identified long (428 amino acids) and short (420 amino acids) forms of banana shrimp MSTN and FmMSTN, and the long form was positively correlated with a larger size in shrimp. Injection of dsRNA targeting FmMstn into these shrimp impaired their normal molting cycle but also resulted in the enlargement of the pleopod muscles 47 , which contradicts earlier findings. Further analysis and quantitation of the muscle fiber size in different muscle types are required to fully elucidate the effects of shrimp MSTN/ GDF11 on controlling muscle development.

Functions of GDF11 and MSTN during vertebrate development
During the embryonic development of vertebrates, GDF11 and MSTN exhibit distinct expression patterns Fig. 3 Processing, extracellular regulation, and signaling mechanisms of GDF11 and MSTN. GDF11 and MSTN are initially synthesized as precursor proteins that undergo proteolytic processing to generate biologically active mature dimers. After removal of the signal peptide (S) by a signal peptidase, pro-GDF11/MSTN is cleaved by a furin-like protease to produce a latent complex, in which the mature domain (MD) is noncovalently associated with the prodomain (PD). The latent complex is additionally cleaved by a BMP1/TLD-like metalloproteinase to generate the mature, disulfide-linked dimer (marked in dark gray) that elicits signal transduction. The latent complex has been shown to be capable of existing in a triggered state, allowing it to initiate signaling events 22 . The mature GDF11/MSTN ligand binds to activin type 2 receptors (ACVR2A/2B) that subsequently recruit activin type 1 receptors (ALK4/5/7) to signal through the canonical SMAD2/3 pathway. Activation of noncanonical pathways, including ERK, JNK, and p38 MAPK, has also been reported 20 . In addition, GDF11 has been demonstrated to activate the SMAD1/5/9 pathway in endothelial cells and osteoblasts 26,27,129,130 . Dotted lines with an arrowhead indicate noncanonical pathways, and solid lines with an arrowhead indicate canonical pathways. • Exerts neuroprotective effects [100][101][102][103][104][105]165,166 • Improves vascularization 101,167 • Improves skin health and repair [168][169][170] • Improves kidney regeneration 171 • Ameliorates colitis 172 • Promotes bone/cartilage development 115,173 • Protects against inflammatory arthritis 174 • Improves metabolic homeostatsis [175][176][177] • Improves endothelial function 178,179 • Promotes blood antioxidant enzyme activities 180 • Inhibits skeletal muscle growth/ regeneration [70][71][72][73]90 • Induces cachexia that leads to premature death/severe lethargy 90 • Induces pathological hypertrophy in ventricular myocytes 89 • Inhibits neurogenesis 56,106,107 • Impairs liver regeneration 181,182 • Induces kidney fibrosis/failure 183 • Inhibits bone development and titanium implant healing [116][117][118] • Inhibits chondrogenesis and callus formation 119 • Induces myocardial cell death 90 • Inhibits erythroid maturation 184 and functions. In mice, Mstn is initially expressed in the myotome compartment of somites at E9.5 and continues to be expressed in skeletal muscles to repress hyperplasia or increase the number of muscle fibers during development 4 . On the other hand, Gdf11 is predominantly expressed in the mouse tail bud at E9.5 and specifies the positional identity of the skeleton along the anteriorposterior axis 9 . Correspondingly, Gdf11-null mice exhibit anterior homeotic transformations of the vertebrae by displaying an increase in the number of thoracic and lumbar vertebrae and vertebrosternal ribs 9,48 . It should be noted that Gdf11 and Mstn double-mutants (Mstn −/− ; Gdf11 −/− ) exhibited more extensive homeotic transformations of the axial skeleton than Gdf11-null mice, indicating that GDF11 and MSTN have some redundant functions related to the control of skeletal patterning 49 . GDF11 has also been shown to mediate proper craniofacial development, as Gdf11-null mice display high (60%) penetrance of cleft palate 48,50 . In further support of this role of GDF11, a recent study identified a family with orofacial clefting and vertebral/rib hypersegmentation harboring a dominant-negative missense mutation in GDF11, in which an arginine is substituted for a glutamine at the furin protease cleavage site (R298Q) 51 . Additional analysis confirmed that mutant GDF11 (R298Q) is not processed into the active form, indicating that this mutation is the underlying cause of the phenotypes observed in this family 51 . An earlier study reported that GDF11 also promotes tooth development and that electroporation-mediated transfer of the Gdf11 gene to the amputated pulp of canine teeth enhances reparative dentin formation 52 . Furthermore, our group recently demonstrated that GDF11, in contrast to MSTN, facilitates osteogenesis during embryonic development and showed that compared to that in newborn wild-type mice, bone mass is diminished in newborn Gdf11-null mice and enhanced in newborn Mstn-null mice 26 .
Multiple studies that utilized Gdf11-null embryos demonstrated that GDF11 plays a crucial role in the development of various organs. Specifically, analysis of Gdf11-null mice revealed that the majority of these mice experience complete renal agenesis and failure of ureteric bud outgrowth from the Wolffian duct 53 . These mice were further shown to exhibit malformations of the stomach characterized by a two-fold reduction in the thickness of the gastric wall and a decreased number of gastric rugae (epithelial folds), a smaller spleen, and an abnormally shaped pancreas 54 , implying that GDF11 is essential for proper morphogenesis of the foregut-derived organs. Gdf11 deficiency also resulted in the greater expansion of islet progenitor cells as well as the impairment of β-cell differentiation in the pancreas 54 . In the olfactory epithelium and retina, GDF11 was shown to inhibit neurogenesis by either repressing progenitor cell • Inhibits neurogenesis 56 • Inhibits bone development 112,194,195 • Inhibits chondrogenesis 190 • Impairs metabolic homeostasis 196 Human GDF11 • Promotes proper orofacial development 51 • Rejuvenates endothelial progenitor cells 197

NR
• Enhances skin cell function 168 • Promotes expansion of liver progenitor cells 161 • Rejuvenates endothelial progenitor cells 197 • Inhibits myoblast differentiation 70 • Inhibits erythroid maturation 198 • proliferation or altering progenitor cell fate 55,56 . This conclusion was supported by the significantly increased number of olfactory epithelium neurons and retinal ganglion cells in mice lacking GDF11 and the contrasting patterns observed in mice deficient in FST, an antagonist of GDF11. However, a delay in neuronal differentiation and gliogenesis was later reported in the spinal cord of Gdf11-null mice, suggesting that GDF11 promotes the temporal progression of neurogenesis in the developing spinal cord 57 . In addition, using chicken embryos, Gamer et al. 58 demonstrated that implantation of beads soaked in human recombinant GDF11 protein into early wing buds led to a dramatic truncation of the limbs due to suppression of both myogenesis and chondrogenesis. In contrast, a recent analysis of Gdf11-null embryos at E15.5 and Gdf11-null sternal chondrocytes revealed that chondrocyte maturation was impaired under Gdf11-deficient conditions 26 . Moreover, skeletal muscle-specific deletion of Gdf11 using conditional knockout techniques resulted in no differences in muscle mass and fiber type, indicating that the functions of GDF11 and MSTN in the skeletal muscles are most likely divergent 49 . Additional investigation of the skeletal muscles of Gdf11-null embryos will further clarify the role of GDF11 in myogenesis during development.

MSTN in skeletal muscle
The primary function of MSTN became evident when mice homozygous for Mstn deletion were shown to have a substantial increase in skeletal muscle mass, with individual muscle groups growing to approximately twice the normal size 4 . A significant increase in muscle mass was also observed in humans, cattle, sheep, and dogs with naturally occurring mutations in the MSTN gene [5][6][7][8] . Further analysis of Mstn-null mice revealed that MSTN inhibits both skeletal muscle fiber hyperplasia during early development and hypertrophy in adults. The direct role of MSTN in postnatal suppression of muscle fiber hypertrophy was demonstrated by the severe loss of muscle mass induced by systemic overexpression of MSTN in adult mice 35 and the increase in muscle mass in adult mice treated with a monoclonal anti-MSTN antibody 59 . Paradoxically, circulating MSTN levels were shown to decrease with age in humans, implying that this decline is likely a secondary effect of age-related muscle loss 60 .
Previous studies that applied recombinant MSTN proteins have presented mixed results regarding the role of MSTN in satellite cells. For instance, while several early studies showed that treatment with recombinant MSTN proteins inhibited C2C12 myoblast proliferation 61,62 , Rodgers et al. 63 more recently argued that recombinant MSTN proteins stimulate C2C12 proliferation, emphasizing that the source of the recombinant MSTN protein can impact the outcome of an experiment. Furthermore, primary myoblasts isolated from both Mstnnull mouse embryos and adult mice were shown to exhibit a significantly increased proliferation rate 64 , and skeletal muscle regeneration after toxin-induced injury was significantly enhanced in Mstn-null mice 65,66 , indicating that endogenous MSTN suppresses satellite cell proliferation, differentiation, and muscle regeneration.

GDF11 in skeletal muscle
In contrast to Mstn-null mice, which survive to adulthood, Gdf11-null mice die shortly after birth, causing difficulties in identifying the role of GDF11 in adult tissue homeostasis. To overcome this limitation, Sinha et al. 67 injected recombinant GDF11 proteins into aged mice and demonstrated that GDF11, in contrast to MSTN, acts as a rejuvenating factor in skeletal muscle. The aged mice treated with recombinant GDF11 proteins displayed striking improvements in muscle regeneration, exercise endurance, grip strength, myofibrillar and mitochondrial morphology, neuromuscular junctions, and the genomic integrity of muscle stem cells 67 . Furthermore, a more recent investigation in annual fish revealed that the application of GDF11 recombinant proteins boosted antioxidant enzyme activity in muscle, thus prolonging the lifespan 68 . Gdf11 expression levels were also shown to increase in slow-twitch muscles of aged mice after 6 weeks of treadmill running 69 . However, multiple studies have failed to reproduce these results, showing that GDF11 is deleterious towards muscle repair. For example, Egerman et al. 70 argued that treatment with recombinant GDF11 proteins significantly impaired muscle regeneration and satellite cell expansion in mice through a downstream signaling pathway identical to that utilized by MSTN. Likewise, Hinken et al. 71 showed that the recombinant GDF11 protein repressed satellite cell expansion, while Hammers et al. 72 demonstrated that recombinant GDF11 and MSTN proteins decreased the myotube diameter through the canonical SMAD2/3 pathway. Zhou et al. 73 also observed that injection of recombinant GDF11 protein into older rats significantly hindered muscle regeneration and function and induced tissue fibrosis. In addition, several other recent studies have shown that exogenous GDF11 treatment inhibits muscle growth 74,75 and reduces strength 76 , while overexpression of the GDF11 propeptide, which antagonizes both mature GDF11 and MSTN, exerts beneficial effects 23,77 .
Although most reports strongly suggest that exogenous GDF11 supplementation exerts an inhibitory effect on skeletal muscle growth and regeneration, further studies that employ genetic loss-of-function approaches are needed to fully elucidate the endogenous function of GDF11 in skeletal muscles. Interestingly, Gdf11 expression levels were shown to peak in skeletal muscles in mice between 4 and 8 weeks of age, which is when the most dramatic postnatal muscle development occurs 20 . This expression pattern is similar to that observed in shrimp, in which the expression of Mstn/Gdf11, which was shown to promote growth unlike vertebrate MSTN, peaks immediately after molting 17 . Moreover, Gdf11 expression levels were revealed to increase even further in Mstn-null mice during periods of rapid muscle growth 20 . To date, only one study has applied conditional knockout techniques in mice to investigate the postnatal functions of GDF11 in regulating skeletal muscle mass 49 . The study demonstrated that skeletal muscle-specific targeting of Gdf11 had no significant effect on muscle mass, fiber number, or fiber type, demonstrating that GDF11 and MSTN exhibit distinct functions in controlling muscle size 49 . Additional examinations focusing on genetic approaches will further advance the understanding of the role of GDF11 in muscle biology.

MSTN in heart
Despite its establishment as a potent inhibitor of skeletal muscle growth, MSTN has also been implicated in the regulation of cardiac tissue growth and function. A recent study that analyzed hearts of adult Mstn-null mice revealed that the absence of MSTN had no effect on heart weight but significantly decreased the end systolic diameter and increased fractional shortening 78 . Lim et al. 79 consecutively demonstrated that Mstn-null mice subjected to ligation of the left anterior descending artery to induce myocardial infarction (MI) exhibited accelerated recovery of the ejection fraction, reduced cardiac fibrosis, and lower mortality, indicating that MSTN negatively affects cardiac function. Likewise, senescent MSTN-deficient mice were shown to display improved fractional shortening, smaller left ventricular diastolic and systolic diameters, and decreased cardiac fibrosis 80 , although an earlier study reported that MSTN has no significant effect on cardiac hypertrophy or fibrosis 81 . Additionally, transgenic mice overexpressing MSTN in cardiomyocytes exhibited interstitial fibrosis and impaired cardiac function 82 . Surprisingly, tamoxifen-induced, cardiomyocyte-specific deletion of Mstn in adult mice was reported to provoke severe cardiac hypertrophy and heart failure 83 . The authors of the study assumed that the much greater severity of the cardiac phenotype observed in the conditional knockout mice than in the straight knockout mice was due to the distinct modes of compensation 83 . To clarify the function of MSTN in hearts and whether the cardiac phenotypes of MSTN deficiency are influenced by the enhancement of skeletal muscle mass, additional examinations are needed.

GDF11 in heart
In 2013, Loffredo et al. 84 performed heterochronic parabiosis experiments in mice and identified GDF11 as a rejuvenating factor that circulates in plasma. The group utilized both proteomics (SOMAmer) and western blot analysis to determine that the circulating levels of GDF11 decline with age, reporting that restoration of youthful levels through daily intraperitoneal injections of recombinant GDF11 proteins reverses age-related cardiac hypertrophy 84 . Specifically, GDF11 rather than MSTN stimulated the dose-dependent inhibition of cardiac myocyte hypertrophy in vitro 84 . In further support of these results, Poggioli et al. 85 showed that circulating GDF11/MSTN levels diminish with age in multiple mammalian species, and administration of recombinant GDF11 proteins dose-dependently decreased cardiac mass in both young and old mice after only 9 days. Likewise, Du et al. 86 demonstrated that Gdf11 expression levels decline in aged hearts and that either targeted myocardial delivery of the Gdf11 gene or recombinant GDF11 protein enhanced cardiac function and effectively reduced infarct size after ischemic injury in aged mice, providing support for the anti-aging function of GDF11. The association of plasma GDF11/MSTN levels with cardiovascular outcomes and overall deaths in humans was also reported using SOMAmer technology, which revealed that in patients with stable ischemic heart disease, increased GDF11/MSTN levels were associated with decreased rates of cardiovascular events, left ventricular hypertrophy, and overall death 87 . Mechanistically, GDF11 was shown to increase intracellular calcium levels and activate SMAD2/ 3 to prevent cardiomyocyte hypertrophy 88 .
However, other groups have failed to observe the rejuvenating effects of GDF11 in cardiac tissues. After following the protocol used in a previous report 84 , Smith et al. 89 showed that treatment of old mice with recombinant GDF11 proteins had no effect on cardiac mass, structure, or function. Moreover, recombinant GDF11 protein caused pathological hypertrophic signaling in neonatal rat ventricular myocytes, contradicting the classification of GDF11 as an anti-aging factor 89 . The same group subsequently published a dose-range study (0.5, 1.0, or 5.0 mg/kg) performed in young mice that underwent transverse aortic constriction (TAC) surgery and reported that although treatment with recombinant GDF11 proteins reduced pathological cardiac hypertrophy and fibrosis and improved cardiac function, the highest dose (5.0 mg/kg) led to severe cachexia and premature death, and they also issued a warning against the use of recombinant GDF11 proteins as a therapy 90 . Recombinant GDF11 protein treatment was also recently shown to increase the levels of reactive oxygen species in isoproterenol-treated H9C2 cells (rat heart-derived cardiomyoblast cell line) 91 and impair cardiac function in old mice 75 . In addition, Egerman et al. 70 pointed out that the SOMAmer and antibody used in the initial study by Loffredo et al. 84 were nonspecific, claiming that circulating GDF11 levels actually increase with age and are a proaging factor. However, Poggioli et al. 85 later proposed that the levels of GDF11 detected by Egerman et al. 70 were in fact the levels of immunoglobulin light chain, generating further controversy regarding the circulating levels of GDF11. Applying a novel immunoplexed liquid chromatography with tandem mass spectrometry (LC-MS/MS) assay, Schafer et al. 60 more accurately measured the circulating levels of GDF11 and reported that GDF11 levels remain constant in healthy adults throughout the lifespan. The study revealed that in older adults with severe aortic stenosis, higher GDF11 levels were associated with comorbidity and frailty 60 . Adding further controversy, a novel detection method using a parallel reaction monitoring (PRM) LC-MS/MS assay combined with immunoprecipitation recently showed that circulating levels of both GDF11 and MSTN significantly decline with age in female mice 92 .
In contrast to the large number of studies that investigated the effects of recombinant GDF11 proteins, only a single recent study has addressed the function of GDF11 based on cardiac-specific genetic deletion in mice. Using a Myh6-Cre transgene, Garbern et al. 93 generated a conditional knockout mouse model in which Gdf11 was targeted exclusively to cardiomyocytes and demonstrated that the mice exhibited progressive left ventricular dilation and a decrease in left ventricular systolic function at the age of 6 months. However, the authors also noted the adverse effects of the Cre recombinase itself and the possible compensatory expression of Gdf11 in noncardiomyocytes, which prevented the clear interpretation of the mechanism underlying the above results 93 . Apparently, further genetic analysis with avoidance of Cre toxicity is required to delineate the endogenous role of GDF11 in cardiac tissues.

MSTN in the brain
Despite the scarcity of information on the function of MSTN in the brain, a recent study showed that Mstn is broadly expressed throughout the adult rat central nervous system, including most neurons, axons, oligodendrocytes, astrocytes, and ependymal cells, suggesting that MSTN may play a crucial role in the brain 94 . Regarding the role of MSTN in the nervous system, examination of adult Mstn-null mice revealed that these mice exhibit increases in the number and size of axons and a delay in their age-related reduction 95 . Furthermore, MSTNdeficient mice were shown to display enhanced myelin thickness in motor axons and an increase in the number of sensory axons 96 . These mice were also shown to present a smaller brain size than wild-type mice at the age of 4 months, but the mechanism of brain size regulation by MSTN remains unclear 97 . Meanwhile, conflicting reports exist regarding the effects of recombinant MSTN proteins on neuronal cells. For instance, while Kerrison et al. 98 showed that recombinant MSTN proteins dosedependently enhanced the survival of retinal ganglion cells and neurite outgrowth, others demonstrated that recombinant MSTN proteins decreased the formation of neuronal colonies 56 or suggested that MSTN inhibits neurogenesis in the olfactory system 99 .

GDF11 in the brain
The perinatal lethality observed in Gdf11-deficient mice has led to multiple studies that investigated the effects of recombinant GDF11 proteins on adult neurogenesis, demonstrating that GDF11 is a pro-neurogenic and proangiogenic factor. Shortly after Loffredo et al. 84 reported GDF11 as a rejuvenating agent that protects the aged heart, the same group proposed that GDF11 exerts antiaging effects on the brain, which was supported by the improvement of the cerebral vasculature and the enhancement of neurogenesis after the treatment of old mice with recombinant GDF11 proteins 100 . A separate experiment also showed that systemic delivery of recombinant GDF11 proteins enhanced hippocampal neurogenesis and vasculature in old mice by acting on brain endothelial cells, and only GDF11 but not MSTN promoted VEGF secretion in brain endothelial cells 101 . Likewise, a single injection of recombinant GDF11 protein was shown to improve short-term visual memory in middle-aged mice through upregulation of SOX2 expression 102 . Furthermore, treatment with recombinant GDF11 proteins was shown to promote neurogenesis and angiogenesis in mouse and rat models of stroke 103,104 and in a mouse model of Alzheimer's disease, revealing GDF11 as a potential therapeutic option for neurodegenerative disorders 105 . In contrast, in vitro data on the effects of recombinant GDF11 protein exposure to neural stem cell lines demonstrated that GDF11 suppresses cell proliferation and migration, suggesting that GDF11 should be a target for pharmacological blockade 106,107 . While the majority of studies presented the beneficial effects of recombinant GDF11 treatment on the mature nervous system, additional analysis utilizing genetic knockdown or conditional knockout of Gdf11 will further advance the understanding of the role and action mechanism of endogenous GDF11 in the adult brain.

MSTN in bone
The deficiency of MSTN has been described to result in not only an enlargement in skeletal muscle mass but also an increase in bone mass. In this regard, Mstn-null mice were shown to exhibit enhanced bone mineral density in various parts of the skeleton 26,80,108 . In humans, genetic polymorphisms in MSTN were demonstrated to be associated with peak bone mineral density 109 . The effects of MSTN on bone may be both direct and indirect through the influence of skeletal muscle. While the indirect positive effect of enhanced skeletal muscle mass on bone strength was evidenced in Mstn-null mice 110 , the direct inhibition of osteoblast differentiation and stimulation of osteoclast formation by MSTN were also reported 26,[111][112][113][114] . Surprisingly, despite the relatively low expression of Mstn in primary mouse osteoblasts and osteoclast precursors under physiological conditions, siRNA-mediated knockdown or genetic knockout of Mstn noticeably altered the differentiation rate of these cells 26,113 , highlighting the significant role of MSTN in the direct regulation of bone cells.

GDF11 in bone
As opposed to the consistent reports on the inhibitory function of MSTN on osteogenesis, the reports of the effects of GDF11 on adult bone homeostasis are controversial. In 2015, Zhang et al. 115 demonstrated that circulating GDF11 levels were significantly diminished in both aged humans and patients with osteoporosis, and Gdf11 expression levels were substantially downregulated in the bone marrow of aged mice and mice with osteoporosis. Additionally, the group showed that treatment with recombinant GDF11 proteins significantly promoted osteoblast differentiation and inhibited adipogenesis of bone marrow mesenchymal stem cells 115 , emphasizing the pro-osteogenic role of GDF11, which is in contrast with the function of MSTN. However, Lu et al. 116 subsequently published results showing the opposite results, indicating that recombinant GDF11 proteins inhibited osteoblast differentiation of bone marrow mesenchymal stem cells through a downstream signaling pathway identical to that of MSTN and that injection of recombinant GDF11 proteins suppressed bone formation in mice. In the same year, Liu et al. 117 also reported similar findings and demonstrated that recombinant GDF11 treatment led to bone loss in both young and aged mice through impairment of osteoblast differentiation and increased osteoclast formation. Later, the same group further showed that recombinant GDF11 negatively affects bone healing in mice 118,119 . Moreover, in postmenopausal women, increased levels of circulating GDF11 were associated with decreased bone mineral density, demonstrating the inhibitory effect of GDF11 on bone formation 120 .
In an attempt to avoid the controversy surrounding the effects of recombinant GDF11 proteins, our group has recently applied conditional knockout strategies in mice to examine the endogenous function of GDF11 in osteogenesis 26 . Our findings revealed that both time-specific ubiquitous deletion and limb mesenchyme-specific deletion of Gdf11 resulted in diminished bone mass in young adult mice, suggesting that GDF11 endogenously promotes bone development, in contrast to MSTN 26 . Furthermore, both Gdf11-null osteoblasts and wild-type osteoblasts subjected to siRNA-mediated Gdf11 knockdown exhibited impaired differentiation and mineralization, while the opposite effects were observed in wild-type osteoblasts transfected with full-length GDF11 cDNA 26 . Whether there is a difference in the cellular and physiological outcomes of treatment with the full-length cDNA and mature forms of the GDF11 protein requires further investigation. A previous study demonstrated that skeleton-specific transgenic overexpression of the GDF11 propeptide, which is capable of inhibiting both mature forms of GDF11 and MSTN, enhanced bone formation in mice during embryogenesis and postnatal development 121 . However, these mice were shown to exhibit a posterior homeotic transformation in the cervical vertebra, resulting in transformation of the seventh cervical vertebra into a thoracic vertebra, which is the exact opposite of the anterior homeotic transformations observed in Gdf11-null mice 9,48 . Furthermore, the authors did not provide information on the relative expression of endogenous Gdf11 and the transgene during developmental stages, limiting the potential for the clear interpretation of the results 122 . Additional examination focused on genetic studies in mice will lead to additional insights into the endogenous mechanism of action of GDF11 in postnatal bone remodeling.

Recombinant GDF11 protein as a therapeutic option
After the initial reports of GDF11 as a rejuvenating agent for the heart, skeletal muscle, and brain, numerous groups have evaluated the effects of recombinant GDF11 protein administration on various tissues (Table 5). However, despite using similar treatment and dosage regimens, multiple groups have produced widely varying results, especially in skeletal muscle and heart. Regarding this issue, Poggioli et al. 85 pointed out the existence of batch-to-batch variations in the concentrations of recombinant GDF11 proteins, which was also confirmed by the manufacturer, and suggested that differences in protein sources, protein refolding efficiencies, and protein concentrations all possibly contributed to the disparity in the outcomes. Likewise, Rodgers and Eldridge 34 highlighted the possible significant influence of the source and quality of recombinant proteins on experimental results and indicated that both Sinha et al. 67 and Egerman et al. 70 , who reported opposing results, utilized bacterially generated recombinant proteins that may be less effective or even produce different effects depending on the folding status. In support of this claim, Rodgers et al. 63 demonstrated that recombinant MSTN proteins produced in Table 5 The effects of in vivo recombinant GDF11 protein treatment on skeletal muscle, heart, brain, and bone.  bacteria and eukaryotes behave differently in the regulation of C2C12 myoblast proliferation. Most importantly, even though differences in the type 1 receptor-binding residues and signaling potency between mature GDF11 and MSTN have been reported 25 , it is difficult to rule out the fact that recombinant GDF11 and MSTN proteins cannot be effectively distinguished due to their high sequence similarity, revealing the possibility that the responses mediated by recombinant GDF11 protein treatment actually reflect the endogenous functions of MSTN. Apparently, due to the large discrepancy in the reported effects of recombinant GDF11 protein treatment, further establishment of experimental settings that generate more reliable outcomes as well as different strategies for GDF11 supplementation will be needed for further consideration of GDF11 as a therapeutic option.

Potential adverse effects of targeting GDF11
Ample studies that noted the detrimental effects of recombinant GDF11 protein injection have also indicated GDF11 as a potential target for pharmacological blockade. However, the relatively little information is available on the endogenous functions of GDF11 in regulating adult physiology, which indicates the need for a cautious approach in the development of GDF11 inhibitors. In fact, sotatercept (ACE-011), an ACVR2A fusion protein originally designed by Acceleron Pharma to increase bone mineral density 123 , unexpectedly promoted rapid increases in hematocrit, hemoglobin, red blood cells, and late-stage erythropoiesis, which were suggested to be caused by suppression of endogenous GDF11 124 , although recent studies using conditional knockout techniques refuted this mechanism 125,126 . In addition, our group has recently demonstrated that transgenic overexpression of FST, an endogenous inhibitor of MSTN, GDF11, and activins, substantially enhanced muscle mass but induced spontaneous tibial fractures due to a reduction in bone mineral density, implying that inhibition of GDF11 may have adverse effects on bone 26 . Therefore, possible side effects triggered by both exogenous administration and endogenous inhibition of GDF11 and the means to resolve them should be evaluated with caution in order to enhance the potential for GDF11 to be applied in clinical settings.

Conclusion and future perspectives
The remarkable sequence similarity between GDF11 and MSTN led to the assumption that the two molecules are functionally redundant. However, multiple genetic studies in mice provide clear evidence that they play distinct roles under a range of physiological conditions. Notably, the perinatal lethality observed in Gdf11-null mice, in contrast to the long-term viability of MSTNdeficient mice, led to complications in characterizing the role of GDF11 in adult tissues and caused many groups to utilize recombinant GDF11 proteins to identify its postnatal function. However, difficulties in biochemically distinguishing between the recombinant GDF11 and MSTN proteins as well as variations in the quality of recombinant proteins aroused much controversy regarding the effects of GDF11 treatment. Indeed, while numerous studies have presented the beneficial physiological outcomes after supplementation with recombinant GDF11 proteins, providing a rationale for its therapeutic application, an equally large number of studies have also underscored its harmfulness, demonstrating GDF11 as a potential therapeutic target for inhibition. Therefore, future studies should focus on implementing genetic knockdown or conditional knockout techniques, which may be more promising approaches to differentiate the endogenous functions of GDF11 and MSTN and their regulatory mechanisms. Furthermore, reliable research strategies to improve the consistency of test results are needed to support the progression of GDF11 therapy or GDF11 inhibitors to clinical trials.