InteractiveFly: GeneBrief

Fs: Biological Overview | References

Gene name - Follistatin

Synonyms - follistatin

Cytological map position - 51E7-51E10

Function - activin inhibitor activity

Keywords - secreted activin inhibitor activity

Symbol - Fs

FlyBase ID: FBgn0259878

Genetic map position - 2R: 11,111,061..11,129,084 [-]

Classification - Kazal type serine protease inhibitors and follistatin-like domains

Cellular location - secreted

NCBI links: Precomputed BLAST | EntrezGene

Follistatin (FS) is one of several secreted proteins that modulate the activity of TGF-β family members during development. The structural and functional analysis of Drosophila Follistatin (dFS) reveals important differences between dFS and its vertebrate orthologues: it is larger, more positively charged, and proteolytically processed. dFS primarily inhibits signaling of Drosophila Activin (dACT) but can also inhibit other ligands like Decapentaplegic (DPP). In contrast, the presence of dFS enhances signaling of the Activin-like protein Dawdle (DAW), indicating that dFS exhibits a dual function in promoting and inhibiting signaling of TGF-β ligands. In addition, FS proteins may also function in facilitating ligand diffusion. Mutants of daw are rescued in significant numbers by expression of vertebrate FS proteins. Since two PiggyBac insertions in dfs are not lethal, it appears that the function of dFS is non-essential or functionally redundant (Bickel, 2008).

Polypeptide cytokines of the transforming growth factor β (TGF-β) family control a wide range of developmental and physiological functions in higher eukaryotes. This diverse group of signaling molecules provides positional information required for axis formation and tissue specification, controls various processes such as tissue growth, cell death, and pathfinding of axons in the nervous system, and prevents differentiation of embryonic stem cells. Many components of this pathway have been linked to tumor formation in humans. The highest degree of sequence conservation between various family members is found within the C-terminal domains, which are released as dimers by proteolytic processing. Similarities in sequence and biological activities allow these factors to be divided into at least two distinct subgroups: Bone Morphogenetic Proteins (BMPs) and Activins/Inhibins/TGF-βs (Newfeld, 1999). The latter group exhibits an additional intramolecular disulfide bond at the N-terminus after processing. In Drosophila, there are four Activins/TGF-βs, Drosophila Activin (dACT), Dawdle (DAW, also known as Activin-like protein at 23, ALP23, and Anti-Activin, AACT), Myoglianin (MYO), and Maverick (MAV), and three BMP-type ligands, Decapentaplegic (DPP), Screw (SCW), and Glass Bottom Boat (GBB). Each ligand dimer forms a complex with two type II and two type I receptor serine/threonine kinases that phosphorylate SMAD transcription factors. BMP-type ligands signal primarily through the type I receptors thick veins (TKV) and Saxophone (SAX) and activate Mothers against DPP (MAD). Activins/TGF-β-type ligands are believed to signal through the type I receptor Baboon (BABO), which in turn activates primarily dSMAD2 but to a minor extent also MAD (Gesualdi, 2007; Bickel, 2008)

TGF-β signaling is regulated by various extracellular proteins. Antagonists like Follistatin (FS), Noggin, Chordin/Short Gastrulation, and DAN/Cerberus bind ligands and prevent interactions with receptors and signaling. In some species, they exhibit overlapping and redundant functions. Recently, it was shown that the simultaneous depletion of FS, Noggin, and Chordin in Xenopus tropicalis (Khokha, 2005) results in transformation of ventral into dorsal tissue during embryogenesis (Bickel, 2008).

Follistatin was first identified as an inhibitor of Activin in vertebrates. Subsequent studies showed that it also binds other ligands with lower affinities including BMP 2, 4, 6, 7, and Myostatin (Abe, 2004; Canalis, 2003). Knockout mice of fs die shortly after birth. They are smaller and exhibit defects in skeletal and muscle development (Matzuk, 1995). Recently, the crystal structure of the human FS-Activin complex was resolved (Thompson, 2005). It provides valuable insight into the function of the different FS domains and a basis to explain the mechanism of ligand inhibition (Bickel, 2008)

This study has analyzed the function of Drosophila Follistatin (dFS). Like vertebrate FS proteins, dFS is subdivided into a N-terminal domain (N) and three FS domains (FS1-3). However, dFS is substantially larger than its vertebrate homologues due to a large basic insertion into FS1. Interestingly, dFS is proteolytically processed, and small processed forms of dFS are able to bind to ligands like dACT. This result suggests a possible different inhibitory mechanism: ligands bound to processed dFS can bind to type II receptors but cannot recruit type I receptors. Consequently, processed dFS might not only sequester ligands but also prevent unbound ligands from interacting with receptor complexes. Among the seven Drosophila TGF-β ligands, dFS primarily inhibits dACT but can also inhibit signaling of other ligands like DPP. In contrast, dFS can augment signaling of the TGF-β member DAW. These results suggest that dFS might exhibit dual functions in facilitating and inhibiting TGF-β signaling. Analysis of two PiggyBac insertions in dFS reveals that they affect dfs expression. Since homozygous animals of these lines are viable and phenotypically wild type, it is assumed that the function of dFS is non-essential or functionally redundant. Taken together, this study reveals interesting differences between the mechanisms of modulating TGF-β signaling by dFS and its vertebrate orthologues (Bickel, 2008)

The primary structure of dFS shows both similarities and differences compared to its vertebrate orthologues (Haerry, 2002). dFS is divided into a N-terminal domain (N) and three FS domains (FS1-3) that can be further subdivided into EGF-like and Kazal protease inhibitor-like domains. Compared to its vertebrate orthologues, dFS is substantially larger due to an insertion of 260 amino acids. The insertion contains 55 positively charged amino acids (pI > 10) and is located after the heparin binding site in FS1 (Inouye, 1992). In contrast to Drosophila, vertebrates produce a second, long isoform of FS by alternative splicing (FS-315). This form contains a C-terminal extension with many negatively charged amino acids that reduce adhesion to sulfates of proteoglycans at the cell surface. FS-315 (Schneyer, 2004) is the major mammalian circulating form (Bickel, 2008)

To analyze the function of dFS, dfs cDNA was cloned into expression vectors for tissue culture cells and transgenic flies. From comparison to vertebrate orthologues, the start site was assumed to be at nucleotide 577. At this position, the SignalP server website predicts a signal sequence (MALLIGLLLLNFRLTAA-GTCW) that is cleaved equivalently to the vertebrate FS proteins. However, expression of a construct that contains this predicted signal sequence but lacks the first 279 nucleotides does not result in a translated protein in culture cells and growth inhibition in transgenic flies. In contrast, the complete cDNA is translated in tissue culture cells and inhibits growth in vivo. There are three potential start codons upstream of the predicted signal peptide sequence. Based on the consensus for Drosophila start sites, the most likely start codon is at nucleotide 181-183 (CAACA-ATG). It is noted that there is a full-length dfs cDNA that is 671 nucleotides longer, GH04473 (NM_144119). This cDNA extends the coding sequence by 62 additional amino acids. Although conserved between different Drosophila species, the absence of this additional sequence in the shorter cDNA does not prevent translation and secretion of a growth inhibitory protein. Interestingly, the full-length cDNA does not encode an N-terminal signal peptide sequence either. Instead, several protein analysis programs predict three potential transmembrane domains within the 211 amino acid leader. Thus, in contrast to vertebrate FS proteins, the secretion of dFS might involve an initial membrane-anchoring step (Bickel, 2008)

There are several differences in the structure of dFS and its vertebrate orthologues. One unique characteristic of dFS is the unusual signal trailer that contains three potential transmembrane domains. The results suggest that cleavage likely occurs after the third transmembrane domain at the equivalent position of vertebrate FS proteins. No evidence was found that this feature affects secretion or alters the function of dFS (Bickel, 2008)

A major difference between dFS and its vertebrate orthologues is the large basic insertion. This modification is likely consequential and may alter the activity of dFS. Based on protein folding prediction and the structure of the human FS-Activin complex (Thompson, 2005), the insertion is not well structured and is located after the heparin binding site within the EGF sub-domain of FS1 (Innis, 2003; Inouye, 1992). Since this area projects away from Activin, the positively charged amino acids likely increase the affinity of dFS to heparin sulfate proteoglycans on the cell surface and do not contribute to Activin binding. It is conceivable that this feature leads to a reduction of diffusion and an increase in local concentration of dFS. In addition, it may also reduce diffusion of ligands and enhance the stability of ligand-receptor complexes. Vertebrates have adopted an opposite strategy. They generate a long form of FS by alternative splicing, which contains a negatively charged C-terminal tail that reduces interactions with heparin sulfate. The long form is the major endocrine form in humans. Since it does not efficiently interact with the cell surface, it is not clear whether this form primarily inhibits Activin signaling or actually enhances Activin circulation by preventing unwanted interactions with neighboring cells (Bickel, 2008)

The FS-Activin complex shows that the N domain binds to the wrist region of Activin blocking interaction with the type I receptor. The FS2 domain binds to the type II receptor binding site of Activin. In humans, two FS proteins can encircle an Activin dimer preventing interactions with both types of receptors. It was found experimentally that Activin bound to FS still can interact with the type II receptor. This result is explained if only one molecule of FS is bound to an Activin dimer leaving the second Activin monomer free to bind to receptors. Unlike vertebrate FS proteins, this study found that dFS is proteolytically processed at the C-terminus into two major forms. Based on the migration of epitope tagged forms of dFS on Western blots, the small form contains at least the N and FS1 domains and lacks FS3 and probably part of FS2. Since the small form is immunoprecipitated by dACT, it appears that it can bind to dACT. What is the possible role of this small processed form? The following model suggests that the small form could potentially be a stronger inhibitor: if two small processed dFS proteins without FS2 bind to dACT monomers, they prevent interactions with type I but not type II receptors. A dFS-dACT-PUT complex is inactive, since dACT cannot recruit the type I receptor BABO. In this scenario, the small dFS forms would not only inactivate bound dACT but also reduce interactions of free dACT with type II receptors. If type II receptors were limiting, the small dFS forms would be able to reduce dACT signaling more efficiently than the full-length protein. This hypothesis can be tested by expressing transgenes that encode the small form of dFS. To perform these experiments, the structures of the small and large forms of dFS are currently being determined (Bickel, 2008)

In most experiments, dFS was seen to function as an inhibitor of TGF-β signaling. As a major exception, dFS enhances rather than inhibits DAW signaling in tissue culture assays. Interestingly, unlike other ligands, DAW does not increase but reduces the viability of animals that over-express dFS. This observation suggests that over-expression of dFS may not only be lethal due to reducing the activities of various TGF-β family members but also due to increased signaling of ligands like DAW. If DAW and dFS physically interact, the question arises how dFS can increase ligand signaling. In the early embryo, it has been shown that Short Gastrulation (SOG), the Drosophila orthologues of the vertebrate Chordin protein, not only inhibits signaling of DPP but also facilitates diffusion, which is necessary for the formation of the peak levels of the DPP gradient. The formation of the DPP gradient depends on the equilibrium between bound and unbound DPP by SOG. Similarly, it is conceivable that ligand binding by the positively charged dFS can reduce ligand diffusion. If the concentration or the affinity of dFS for a ligand is high, the ligand is not released or bound again, and signaling is inhibited. Alternatively, if the concentration of dFS or its affinity for a ligand is low, dFS binding can locally increase ligand concentration, while subsequent ligand release will enhance signal transduction. Consistent with such a model, in tissue culture assays dFS rather increases than decreases the level of MAD activation by DPP, while presumably higher levels of dFS inhibit DPP signaling in the wing. Such a dual function could probably be at work during dorsal-ventral axis formation. While dact is not expressed in the early embryo, dfs is present in a dorsal stripe. This pattern corresponds with the highest levels of DPP signaling. Expression of dfs is rather late in the formation of the DPP gradient and is potentially regulated by DPP. It is conceivable that initial low levels of dFS could, like SOG, function to increase the local concentration of DPP and contribute to the formation and maintenance of the dorsal peak activity of DPP signaling. In contrast, higher levels of dFS protein that accumulate by the end of the dorsal-ventral axis formation process may inhibit DPP and contribute to terminating the dorsal DPP signal. Recently, a computational model that described ligand distribution and signaling in the presence of a cell surface BMP-binding protein was developed that supports this idea. A dual function of FS proteins in facilitating and inhibiting TGF-β signaling is also supported by the finding that fs mutants in mice exhibit overlapping phenotypes with activin knock out animals (Matzuk, 1995). Consequently, FS was originally described as enhancer of Activin signaling. Finally, facilitating diffusion and redistribution of dACT and potentially other ligands is probably the best mechanism to explain why dFS and vertebrate FS proteins from frogs and humans can partially compensate for the lack of DAW. It is speculated that the vertebrate FS proteins exhibit lower affinities for Drosophila ligands like dACT and diffuse better since they lack the basic insertion of dFS. Most likely, these differences could explain why expression of the vertebrate but not dFS can rescue small but significant numbers of daw mutant animals. Taken together, it is possible that distinct affinities for dFS account for the different interactions seen with various ligands. Further studies are necessary to investigate such a possible dual function of dfs in early embryos and during development (Bickel, 2008).

Two PiggyBac insertions affect dfs transcription. Interestingly, both lines are homozygous viable and do not show any obvious pattern defects. These results suggest that the function of dFS not essential for viability. In mice, a mutation in fs results in various defects with lethal consequences. In contrast, FS, Chordin, and Noggin exhibit overlapping functions in X. tropicalis. It is necessary to reduce all three inhibitors to transform ventral into dorsal tissue during embryogenesis of this species (Khokha, 2005). Although there is no Noggin-like protein in Drosophila, it is possible that dFS also shares overlapping functions with SOG. Expression of sog overlaps with dfs in many tissues, and it is possible that SOG can substitute for dFS in its absence (Bickel, 2008)

The analysis of dfs RNA levels in homozygous f00897 flies shows that they are substantially reduced. Thus, this insertion can be regarded as a hypomorphic allele. It is conceivable that the amount of protein synthesized in this mutant is sufficient for normal development. However, further reduction of dFS by combining f00897 with the deficiency Df(2R)Exel7135 that entirely removes dfs does not affect viability either. In contrast to f00897, the PiggyBac insertion in e03941 clearly disrupts transcription of a full-length mRNA. Based on the location of the insertion, the truncated mRNA likely encodes an altered protein that still contains the N, FS1, and FS2 domains. It lacks the entire FS3 domain and likely contains additional C-terminal amino acids due to altered or absence of splicing. Since the small form is still able to bind dACT, a partially functional dFS protein could still be present in e03941 flies, if proper processing would occur. However, it is unlikely that such an altered protein is processed correctly into the small form of dFS. Since no homozygous lethal lines were obtained in the FRT-mediated deletion screen, it appears that dfs is not lethal. Taken together, the lack of any obvious phenotypes in homozygous f00897 and e03941 lines suggests that dFS is not essential for normal development or is redundant like in some vertebrate species (Bickel, 2008)


Search PubMed for articles about Drosophila Follistatin

Abe, Y., Minegishi, T. and Leung, P. C. (2004). Activin receptor signaling. Growth Factors 22(2): 105-10. PubMed ID: 15253386

Bickel, D., Shah, R., Gesualdi, S. C. and Haerry, T. E. (2008). Drosophila Follistatin exhibits unique structural modifications and interacts with several TGF-beta family members. Mech. Dev. 125(1-2): 117-29. PubMed ID: 18077144

Canalis, E., et al. (2003). Bone morphogenetic proteins, their antagonists, and the skeleton. Endocr. Rev. 24: 218-235. PubMed ID: 12700180

Gesualdi, S. C. and Haerry, T. E. (2007). Distinct signaling of Drosophila activin/TGF-β family members. Fly 1: 212-221

Haerry, T. E. and O’Connor, M. B. (2002). Isolation of Drosophila Activin and Follistatin cDNAs using novel MACH amplification protocols. Gene 291: 85-93. PubMed ID: 12095682

Innis, C. A. and Hyvonen, M. (2003). Crystal structures of the heparan sulfate-binding domain of follistatin. Insights into ligand binding. J. Biol. Chem. 278: 39969-39977. PubMed ID: 12867435

Inouye, S., Ling, N. and Shimasaki, S. (1992). Localization of the heparin binding site of follistatin. Mol. Cell Endocrinol. 90: 1-6. PubMed ID: 1301390

Khokha, M. K., Yeh, J., Grammer, T. C. and Harland, R. M. (2005). Depletion of three BMP antagonists from Spemann’s organizer leads to a catastrophic loss of dorsal structures. Dev. Cell 8: 401-411. PubMed ID: 15737935

Matzuk, M. M., et al. (1995). Multiple defects and perinatal death in mice deficient in follistatin, Nature 374: 360-363. PubMed ID: 7885475

Newfeld, S. J. Wisotzkey, R. G. and Kumar, S. (1999). Molecular evolution of a developmental pathway: phylogenetic analyses of transforming growth factor-ß family ligands, receptors and Smad signal transducers. Genetics 152: 783-795. PubMed ID: 10353918

Schneyer, A. L. Wang, Q., Sidis, Y. and Sluss, P. M. (2004). Differential distribution of follistatin isoforms: application of a new FS315-specific immunoassay. J. Clin. Endocrinol. Metab. 89: 5067-5075. PubMed ID: 15472207

Thompson, T.B., et al. (2005). The structure of the follistatin:activin complex reveals antagonism of both type I and type II receptor binding. Dev. Cell 9: 535-543. PubMed ID: 16198295

Biological Overview

date revised: 8 June 2008

Home page: The Interactive Fly © 2008 Thomas Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.