The TGFβ superfamily consists of 33 unique signaling ligands which control delicate and indispensable cellular functions ranging from reproduction and development to homeostasis and wound repair. These ligands can be further subdivided into three classes: the TGFβs, for which the family is named, the bone morphogenetic proteins (BMPs), with roles in bone development, growth, and homeostasis, and the activins, which play deterministic roles in reproduction and muscle development. Family proteins are synthesized (A) as longer precursors containing a large N-terminal prodomain (~250 residues) and a smaller C-terminal growth factor domain (~110 residues) which are proteolytically cleaved (B) during processing but can remain noncovalently associated during trafficking and after secretion (C). A single ligand dimer can assemble a signaling complex (D) by association with two type II receptors and two type I receptors. when assembled as such, the constitutively active type II receptors can phosphorylate and activate the type I receptors (E), and in turn the type I receptors can phosphorylate and activate intracellular Smads (F) which then translocate to the nucleus and induce or repress the transcription of target genes. The 8 Smad proteins are divided into three subclasses: receptor-regulated Smads (R-Smads) 1, 2, 3, 5, 9, common partner Smad (Co-Smad) 4, and inhibitory Smads (I-Smads) 6 & 7. When TGFβ signaling is active, heterotrimeric complexes of two R-Smads and one Co-Smad accumulate in the nucleus and proceed modulate transcriptional activity directly, by binding DNA, or indirectly by influencing the action of other transcription factors.
Questions of activation, regulation, and specificity are inherent when studying TGFβ family ligands. Distinct extra- and intracellular mechanisms have evolved to regulate each class at several levels, providing exquisite control over structurally related signaling molecules. Within the family, there are only five type II receptors (TβR2, ActRIIA, ActRIIB, BMPR2, AMHR2) and seven type I receptors (ALK1-7) available. Furthermore, the receptor composition of the signaling assembly ligand dependent: this creates a receptor bottleneck for the more numerous ligands. Furthermore, many classes of ligands are exclusive to their type II receptor; TGFβs, for example, are exclusive for TGFBR2 and AMH is mutually exclusive with AMHR2. Since its inception, the long-term objective of our laboratory – funded by multiple R01 awards – has been to provide a molecular understanding of the mechanisms of TGFβ family signaling and its regulation. In the investigation of these ligands, the application of structural techniques coupled with biophysical studies has provided foundational knowledge of the molecular mechanisms governing their ubiquitous, high impact pathways. In response, there has been a surge of interest from the drug development industry to modulate points along these pathways for therapeutic gain.
Our lab currently concentrates on the structural determination and biophysical examination of proteins within the TGFβ superfamily. We are most interested in questions concerning extracellular modulation of BMP and activin-class signaling and how these mechanisms, once understood, might be manipulated to enhance human life. Using a combination of X-ray crystallography and binding analysis coupled with in vitro cellular assays, the objective of our laboratory is to define the molecular mechanisms of ligand-receptor interactions incorporated to differentiate signaling. Furthermore, our laboratory is characterizing the interactions of extracellular antagonists, which neutralize ligands by blocking ligand-receptor interactions. Similarly, we aim to understand how the N-terminal prodomain of certain ligands renders the growth factor latent or otherwise regulates it, with a focus on deciphering the molecular mechanisms of activation. Most recently, we have begun to incorporate outlier members of the TGFβ family with therapeutically relevant reproductive roles, such as AMH, into our investigative wheelhouse. In addition, the literature has shown that heterodimeric ligands can form spontaneously and are biologically relevant in certain cases, even more so than the cognate homodimeric ligands. Thus, our laboratory is investigating the structure, function, and synthesis of ligand heterodimers. In a disparate project, we have characterized the structure and function of apolipoproteins with the intent to understand how they transition from a lipid free state to a lipid bound state in the biogenesis of lipoprotein particles.
Structure-Function Studies of Ligand-Receptor Interactions
To better understand the general mechanisms of TGFβ signaling it is important to understand, at the molecular level, how ligands bind and interact with their signaling receptors. This information illuminates how certain ligands are able to signal through specific combinations of type I and type II receptors. Within the three ligand classes, differences in receptor affinity exist such that BMP ligands bind type I receptors with high-affinity while both TGFβ and activin ligands display low affinity for type I receptors. Differences also occur where ligand classes utilize different receptor combinations and can be more or less promiscuous. For example, while all TGFβ class ligands signal using ALK5, activin class ligands are capable of signaling through ALK4, ALK5 and ALK7. Structural studies have revealed differences in the mechanisms for how TGFβ and BMP ligands assemble type I and type II receptors. For TGFβ class ligands, a cooperative assembly paradigm is utilized where the type I and type II receptors directly interact in the presence of the ligand. This is in contrast to BMP class ligands, where the receptors bind independently. While structures of TGFβ and BMP ligands bound to their receptors are available, we have limited information for how activin class members — Activin A, Activin B, GDF8, & GDF11 — and AMH assemble a ternary signaling complex.
While the biological functions of Activin A and activin B have been well characterized, little is known about the biological functions of Activin C or Activin AC. One thought is that the Inhibin βC chain functions to interfere with Activin A production by forming less active Activin AC heterodimers. We have shown that Activin C and Activin AC activate SMAD2/3-dependent signaling via the type I receptor ALK7. Relative to Activin A and Activin B, Activin C exhibits lower affinity for the cognate activin type II receptors and is resistant to neutralization by the extracellular antagonist, follistatin. Collectively, these results establish that Activin C and Activin AC are active ligands that exhibit a distinct signaling receptor and antagonist profile compared to other activins.
Mullerian Inhibiting Substance (MIS) or Anti-Mullerian Hormone (AMH), originally identified for its role in male sex differentiation during development, has emerged as a significant molecule in female reproduction. For example, gonadal AMH plays a major role in regulating follicle development while blood serum levels are clinically valuable as a measure of ovarian reserve or as a diagnostic tool. Mutations in AMH are associated with both male and female reproductive disorders, including Persistent Mullerian Duct Syndrome (PMDS) in males and Polycystic Ovary Syndrome (PCOS) in females. As a member of the TGFβ family, AMH signals through a type I and type II receptor. Uniquely, AMH signals through its own type II receptor AMHRII, but will utilize the type I receptor ALK2, which is shared by multiple ligands. While previous studies have detailed TGFβ family ligand interactions, how AMH interacts at the molecular level with AMHRII and ALK2 is unknown. Our lab seeks to understand how AMH complexes with its cognate receptors and how specific AMH signaling is generated using these receptors. A more complete understanding of this interaction can provide a platform for developing reagents that modify AMH activity with the potential for future application in reproductive therapies.
Relevant Publications:
Goebel, E. J. et al. The orphan ligand, activin C, signals through activin receptor-like kinase 7. eLife 11, e78197 (2022).
Howard, J. A., Hart, K. N. & Thompson, T. B. Molecular Mechanisms of AMH Signaling. Front Endocrinol (Lausanne) 13, 927824 (2022).
Goebel, E. J. et al. Structures of activin ligand traps using natural sets of type I and type II TGFβ receptors. iScience 25, 103590 (2022).
Hart, K. N. et al. Structure of AMH bound to AMHR2 provides insight into a unique signaling pair in the TGF-β family. PNAS 118, (2021).
Hart, K. N., Pépin, D., Czepnik, M., Donahoe, P. K. & Thompson, T. B. Mutational Analysis of the Putative Anti-Müllerian Hormone (AMH) Binding Interface on its Type II Receptor, AMHR2. Endocrinology 161, (2020).
Aykul, S. et al. Activin A forms a non-signaling complex with ACVR1 and type II Activin/BMP receptors via its finger 2 tip loop. Elife 9, (2020).
Goebel, E. J. et al. Structural characterization of an activin class ternary receptor complex reveals a third paradigm for receptor specificity. Proc. Natl. Acad. Sci. U.S.A. 116, 15505–15513 (2019).
Walker, R. G. et al. Structural basis for potency differences between GDF8 and GDF11. BMC Biol. 15, 19 (2017).
Walker, R. G. et al. Biochemistry and Biology of GDF11 and Myostatin: Similarities, Differences, and Questions for Future Investigation. Circ. Res. 118, 1125–1141; discussion 1142 (2016).
Extracellular Ligand Modulators
Due to the potency of TGFβ ligands and the wide variety of biological targets they regulate, the modulation of TGFβ signaling is of key importance. Ligands are regulated by over 20 structurally diverse extracellular protein antagonists. Over the past 15 years, our lab has made significant contributions towards understanding the diversity of the extracellular protein antagonist and how they neutralize ligands. Our initial efforts were focused on the follistatin family where we solved a battery of follistatin-ligand structures, including follistatin:activin A, FSTL3:activin A, follistatin:GDF8, FSTL3:GDF8, and follistatin:GDF11. These studies emphasized that ligand-antagonist interactions are complex, using conformational plasticity to accommodate differences at the ligand-antagonist interface. Over the years, we have extended our studies to include the DAN-family and WFIKKN-family antagonists. Similar to follistatin, WFIKKN is a multi-domain antagonist, however it has much more limited specificity – it specifically antagonizes GDF8 and GDF11. In contrast, the DAN-family consists of a small, single domain antagonists that neutralize ligands of the BMP subclass.
Activin-class Antagonism by Follistatin and Structurally-related Proteins
TGFβ family ligands are regulated by a wide array of structurally diverse extracellular protein antagonists. Protein antagonists consist of varying domain architectures and inhibit ligand-receptor interaction through different binding mechanisms. The majority function by blocking both the two type I and two type II receptor binding sites to prevent receptor assembly and signaling. In 2005, Dr. Thompson solved the second ever published antagonist-ligand complex between follistatin and activin A. Follistatin is a multidomain protein containing an N-terminal Domain (ND) and three tandem follistatin domains (FSDs), two molecules of which bind the activin A dimer in a ring-like inhibitory structure. The ND of each follistatin occupies the type I receptor binding site with FSD1 and FSD2 covering the type II receptor site. The two follistatin molecules interact in a head-to-tail mechanism where the ND of one Follistatin interacts with the FSD3 of the neighboring Follistatin.
Subsequent structures generated by our lab of follistatin in complex with other ligands have highlighted a conformational selection model where both the antagonist and ligand adopt a preferred conformation for binding. In particular, the ND of follistatin can adopt different conformations to accommodate the wrist region of different ligands. This mechanism provides the basis for the promiscuity of follistatins and their targeting of multiple ligands, including activin (ActA/B, GDF8/11) and BMP (BMP2/4/6/7) class ligands. Similarly, Follistatin-Like 3 (Fstl3), which lacks the third FSD, was solved in complex with both activin A and GDF8. Here, the ND ligand interactions are less flexible, partially explaining why Fstl3 antagonizes fewer family members. Collectively, these structures have been important for understanding how an antagonist binds different ligands, as well as how a single ligand is neutralized by different antagonists.
We have a particular interest in the ligand GDF8, also known as myostatin. Myostatin is a strong negative regulator of muscle growth – disruption of myostatin signaling results in animals with massive gains in muscle, an outcome observed across many animal species, including humans. Therefore, the development of potent and selective inhibitors to myostatin with the goal of increasing muscle mass or preventing degradation might be one way to fight muscle wasting/atrophy in muscular dystrophies, cancer, cachexia, and sarcopenia. Therefore, the body’s natural antagonists, such as follistatin and FTSL3, are of significant therapeutic relevance.
Relevant Publications:
Walker, R. G. et al. Heparin-mediated dimerization of follistatin. Exp Biol Med (Maywood) 246, 467–482 (2021).
Goebel, E. J., Hart, K. N., McCoy, J. C. & Thompson, T. B. Structural biology of the TGFβ family. Exp. Biol. Med. (Maywood) 244, 1530–1546 (2019).
McCoy, J. C., Walker, R. G., Murray, N. H. & Thompson, T. B. Crystal structure of the WFIKKN2 follistatin domain reveals insight into how it inhibits growth differentiation factor 8 (GDF8) and GDF11. J. Biol. Chem. 294, 6333–6343 (2019).
Walker, R. G. et al. Structural basis for potency differences between GDF8 and GDF11. BMC Biol. 15, 19 (2017).
Walker, R. G., Angerman, E. B., Kattamuri, C., Lee, Y.-S., Lee, S.-J. & Thompson, T. B. Alternative binding modes identified for growth and differentiation factor-associated serum protein (GASP) family antagonism of myostatin. J. Biol. Chem. 290, 7506–7516 (2015).
Cash, J. N. et al. Development of a small-molecule screening method for inhibitors of cellular response to myostatin and activin A. J Biomol Screen 18, 837–844 (2013).
Cash, J. N., Angerman, E. B., Keutmann, H. T. & Thompson, T. B. Characterization of follistatin-type domains and their contribution to myostatin and activin A antagonism. Mol. Endocrinol. 26, 1167–1178 (2012).
Cash, J. N. et al. Structure of myostatin·follistatin-like 3: N-terminal domains of follistatin-type molecules exhibit alternate modes of binding. J. Biol. Chem. 287, 1043–1053 (2012).
Cash, J. N., Rejon, C. A., McPherron, A. C., Bernard, D. J. & Thompson, T. B. The structure of myostatin:follistatin 288: insights into receptor utilization and heparin binding. EMBO J. 28, 2662–2676 (2009).
Stamler, R., Keutmann, H. T., Sidis, Y., Kattamuri, C., Schneyer, A. & Thompson, T. B. The structure of FSTL3.activin A complex. Differential binding of N-terminal domains influences follistatin-type antagonist specificity. J. Biol. Chem. 283, 32831–32838 (2008).
Thompson, T. B., Lerch, T. F., Cook, R. W., Woodruff, T. K. & Jardetzky, T. S. The structure of the follistatin:activin complex reveals antagonism of both type I and type II receptor binding. Dev. Cell 9, 535–543 (2005).
BMP-class Antagonism by Differential Screening-selected Gene in Neuroblastoma (DAN) family Proteins
The DAN-family represents the largest single family of extracellular antagonists with 7 total members. Each protein consists of a core DAN domain that is composed of four conserved disulfide bonds, adopting a growth-factor like shape. Differences in family members occur in both the length and composition of N- and C-terminal segments which flank the DAN domain. Our initial work showed that DAN-family members, specifically Gremlins, formed very stable non-disulfide bonded dimers. Furthermore, we identified a set of core hydrophobic residues that are important for interactions with BMP ligands. A subsequent structure of NBL1 confirmed the unique dimeric structure and the presence of a central BMP binding epitope.
A major breakthrough came when we solved structure of gremlin-2 in complex with the ligand GDF5. The structure revealed that DAN antagonists use a completely novel binding mode compared to other known antagonists, using an H-like mechanism, where gremlin-2 binds perpendicular to GDF5. Having these structures shows the molecular transitions that occur upon ligand binding where, in the unbound state the N-terminus partially shields a set of hydrophobic residues that are liberated during complex formation when the N-terminus dissociates from the core and integrates into the type I receptor binding pocket. These represent a foundational set of bound and unbound antagonist structures, giving insight into the molecular transitions that occur upon ligand binding. Currently, we are studying the DAN family members, SOST and SOSTDC1.
Relevant Publications:
Gipson, G. R. et al. Structural perspective of BMP ligands and signaling. Bone 140, 115549 (2020).
Gipson, G. R., Kattamuri, C., Czepnik, M. & Thompson, T. B. Characterization of the different oligomeric states of the DAN family antagonists SOSTDC1 and SOST. Biochem J 477, 3167–3182 (2020).
Kattamuri, C., Nolan, K. & Thompson, T. B. Analysis and identification of the Grem2 heparin/heparan sulfate-binding motif. Biochem. J. 474, 1093–1107 (2017).
Nolan, K., Kattamuri, C., Rankin, S. A., Read, R. J., Zorn, A. M. & Thompson, T. B. Structure of Gremlin-2 in Complex with GDF5 Gives Insight into DAN-Family-Mediated BMP Antagonism. Cell Rep 16, 2077–2086 (2016).
Nolan, K. et al. Structure of neuroblastoma suppressor of tumorigenicity 1 (NBL1): insights for the functional variability across bone morphogenetic protein (BMP) antagonists. J. Biol. Chem. 290, 4759–4771 (2015).
Nolan, K. & Thompson, T. B. The DAN family: modulators of TGF-β signaling and beyond. Protein Sci. 23, 999–1012 (2014).
Nolan, K. et al. Structure of protein related to Dan and Cerberus: insights into the mechanism of bone morphogenetic protein antagonism. Structure 21, 1417–1429 (2013)
Kattamuri, C. et al. Members of the DAN family are BMP antagonists that form highly stable noncovalent dimers. J. Mol. Biol. 424, 313–327 (2012).
Kattamuri, C., Luedeke, D. M. & Thompson, T. B. Expression and purification of recombinant protein related to DAN and cerberus (PRDC). Protein Expr. Purif. 82, 389–395 (2012).
Structure-function studies of Heterodimers
Most all TGFβ family ligands form disulfide-bonded dimers. Over the past three decades, the field has primarily focused on homodimers formed from 2 identical subunits. However, emerging evidence indicates that heterodimers play a vital role in signaling. Indeed, a recent study concluded that BMP9 and BMP10 signaling in human plasma is almost exclusively generated by BMP9/10 heterodimers. In collaboration with Martin Matzuk (Baylor College of Medicine), we showed that GDF9:BMP15 heterodimers were essential for folliculogenesis. Further studies have suggested that only heterodimers signal during development. Given that there are 33 distinct ligand monomers, the number of potential heterodimers is over 500, providing a vast new dimension to TGFβ signaling. Therefore, we have initiated projects in our laboratory that will address the function, formation, and signaling of TGFβ heterodimers.
More recently, we have characterized other TGFβ family heterodimers, including a synthetic BMP class heterodimer and the naturally-occurring inhibitory heterodimer, inhibin A. In the BMP study, we have determined that mature GDF5 can be combined with mature BMP2 or BMP4 to form BMP2/GDF5 and BMP4/GDF5 heterodimer. Intriguingly, this combination of a BMP2 or BMP4 monomer, which exhibit high affinity to heparan sulfate characteristic to the BMP class, with a GDF5 monomer with low heparan sulfate affinity produces a heterodimer with an intermediate affinity. Furthermore, the X-ray crystal structure of BMP2/GDF5 heterodimer was determined, highlighting the formation of two asymmetric type 1 receptor binding sites that are both unique relative to the homodimers. Ultimately, this method of heterodimer production yielded a signaling molecule with unique properties relative to the homodimeric ligands, including high affinity to multiple type 1 and moderate heparan binding affinity.
In our contribution to the inhibin story, we isolated the inhibin A:follistatin 288 complex, showing that it forms in a 1:1 stoichiometric ratio, different from previously reported homodimeric ligand:follistatin complexes, which bind in a 1:2 ratio. Small angle X-ray scattering coupled with modeling provided a low-resolution structure of inhibin A in complex with follistatin 288. Characterizing the inhibin A:follistatin 288 complex in an activin-responsive luciferase assay and by surface plasmon resonance indicated that the inhibitor complex readily dissociated upon binding type II receptor activin receptor type IIb, allowing both antagonists to inhibit activin signaling. Additionally, injection of the complex in ovariectomized female mice did not alter inhibin A suppression of FSH. Taken together, this study shows that while follistatin binds to inhibin A with a substochiometric ratio relative to the activin homodimer, the complex can dissociate readily, allowing both proteins to effectively antagonize activin signaling.
Relevant Publications:
Gipson, G. R. et al. Formation and characterization of BMP2/GDF5 and BMP4/GDF5 heterodimers. BMC Biol 21, 16 (2023). PDB: 8E3G
Kappes, E. C. et al. Follistatin Forms a Stable Complex With Inhibin A That Does Not Interfere With Activin A Antagonism. Endocrinology 164, bqad017 (2023).
Peng, J., Wigglesworth, K., Rangarajan, A., Eppig, J. J., Thompson, T. B. & Matzuk, M. M. Amino acid 72 of mouse and human GDF9 mature domain is responsible for altered homodimer bioactivities but has subtle effects on GDF9:BMP15 heterodimer activities. Biol. Reprod. 91, 142 (2014).
Peng, J. et al. Growth differentiation factor 9:bone morphogenetic protein 15 heterodimers are potent regulators of ovarian functions. Proc. Natl. Acad. Sci. U.S.A. 110, E776-785 (2013).
Prodomain-Ligand Interactions
Ligands within the TGFβ family are synthesized with a prodomain that facilitates proper assembly of the TGFβ dimer before cleavage by the protease furin. Evidence suggests that most ligands remain noncovalently associated with their prodomains. In a number of cases, this interaction does not inhibit signaling – in others it can even potentiate ligand signaling. Conversely, the prodomain can also render the ligand latent, keeping it trapped in a non-productive prodomain-ligand complex or procomplex. This occurs for both TGFβ class ligands and the activin class members GDF8 and GDF11. For TGFβ, activation occurs through a complex set of protein interactions where the prodomain is pulled in opposing directions by association with the extracellular matrix and integrin receptors. This is in stark contrast to the activation mechanism of the GDF8 and GDF11 procomplex where the tolloid-protease family liberates GDF8/11 by cleaving the prodomain at a specific scissile bond.
Two of our recent manuscripts characterized the interactions of the prodomain with GDF8 and GDF11. We tested the hypothesis that similar structural features of the prodomain, between TGFβ and GDF8/11, are important for latency. Interestingly, this hypothesis was only partially true. While both had similar prodomain components used to bind the ligand, the GDF8/11 complex showed a much more open conformation as compared to TGFβ. We also discovered that the procomplex can exist in three states; 1) a fully latent state which requires tolloid cleavage for activation, 2) a semi-active state where the prodomain is bound but only dampens signaling, and 3) a fully active state. In our study, we show that these states are not reversible and have referred to them as “spring loaded” and “triggered” procomplex states. We subsequently identified mutations in the prodomain which alleviated latency and bypassed the need for tolloid activation.
Relevant Publications:
McCoy, J. C., Goebel, E. J. & Thompson, T. B. Characterization of tolloid-mediated cleavage of the GDF8 procomplex. Biochem J (2021).
Cotton, T. R. et al. Structure of the human myostatin precursor and determinants of growth factor latency. EMBO J. 37, 367–383 (2018).
Walker, R. G. et al. Molecular characterization of latent GDF8 reveals mechanisms of activation. Proc. Natl. Acad. Sci. U.S.A. 115, E866–E875 (2018).
Experimental Techniques
“structure without function is a corpse; function without structure is a ghost” -Vogel and Wainwright, 1969
While our research program includes various and diverse research practices, the backbone and basis of our research is structural biology. While historically focused on X-ray crystallography (XRC), the lab is currently incorporating single particle Cryogenic Electron Microscopy (Cryo-EM) into many of our projects. We consider the structural information we generate only a starting point, and we approach functional validation with equal effort. our typical studies include cell-based endpoint signaling assays and protein-based binding measurement using Surface Plasmon Resonance (SPR). We are currently expanding our structural toolset to include more advanced computational techniques and modeling, meanwhile we are incorporating kinetic signaling assays with the through a collaboration with the Elowitz lab at Caltech.
In addition to our core capabilities, we also utilize these other biophysical techniques (some in collaboration):
Circular Dichroism (CD)
Small-angle X-ray Scattering (SAXS)
Fluorescence Polarization (FP)
Isothermal Titration Calorimetry (ITC)
Analytical Ultracentrifugation (AUC)
Funding
R01HD105818 - Structure-function analysis of Mullerian Inhibiting Substance (MIS)
R01AG072087 - Regulation of GDF11 by extracellular mechanisms
R35GM134923 - Structural/functional characterization of TGFβ superfamily signaling and regulation
S10OD030388 - Glacios 200 kV cryogenic transmission electron microscope (cryo-TEM)
