Structural characteristics

The X-ray crystal structure of the myostatin growth factor was recently determined (Cash, Rejon et al. 2009). High resolution three-dimensional structures have been determined for at least nine other TGF-β superfamily members (Keah and Hearn 2005).

Myostatin displays the traditional TGF-β family architecture which is hand-shaped (Fig. 1.6a) (Cash, Rejon et al. 2009) with each monomer consisting of 4 curved β- strands or ‘fingers’, a cysteine knot motif in the ‘palm’ region and a major helix or ‘wrist’. Myostatin has the intricate cysteine knot motif where two disulphide bridges connect two neighbouring chain segments to form a ring structure and a third disulphide bridge penetrates the ring segment to cross-link two additional chain segments (Figure 1.6b) (Muller and Heiring 2002). Each myostatin monomer contains four intramolecular disulphides; three are involved in the cysteine knot (Fig. 1.6c, cysteines 15-74, 43-106 and 47-108). Two myostatin monomers come together palm to palm in an antiparallel direction to generate concave and convex surfaces (Fig.

1.6a) and dimerise via a ninth intermolecular disulphide bond, which provides additional stabilization of hydrophobic interactions that are the major driving force for dimerisation (Venkataraman, Sasisekharan et al. 1995). The cysteine knot and dimerisation are considered to be the major determinants of protein stability in the TGF-β superfamily (Muller and Heiring 2002), although the dimerisation disulphide is not essential in all family members, such as the vascular endothelial growth factor (VEGF) (Muller and Heiring 2002) and GDF-9 (Chang, Brown et al. 2002). Members of the TGF-β superfamily exhibit high thermal and structural stability (Brownh, Wakefiel et al. 1990), properties attributed to the cysteine knot motif and common to a range of cysteine-knot containing proteins such as collagen type III (Barth, Kyrieleis

et al. 2003) and peptide toxins (Craik, Daly et al. 2000).

Figure 1.6 Structural characteristics of the myostatin growth factor.

a. The X-ray crystal structure of the myostatin growth factor (Cash, Rejon et al. 2009).

b. The cysteine knot motif of vascular endothelial growth factor (VEGF) (Muller and Heiring 2002). c. The myostatin amino acid sequence showing secondary structure, taken from the RSCB. Cysteines involved in intramolecular disulphide bonds (green dotted lines) are shown in yellow. Secondary structure is as follows: yellow arrow, β-sheet; pink wave, α-helix; purple loop, β-turn.

The published structures of the TGF-β growth factors point to three modes of dimerisation (Thompson, Woodruff et al. 2003). TGF-β1, TGF-β2, BMP-2 and BMP- 7 show the common elongated dimer structure where the monomers dimerise in a head-to-tail fashion with the α-helix of one monomer interacting with the concave β- strand surface of the other monomer. TGF-β3 and GDNF exhibit a more open conformation and the activin/inhibin dimers have a more compact conformation than TGF-βs 1 and 2. In addition, there are at least two modes of receptor binding (Keah and Hearn 2005). Crystallographic analysis has revealed that BMPs and activins bind the ectodomains of type II receptors on their convex surface and type I receptors on their concave surface (Thompson, Woodruff et al. 2003). TGF-βs also bind type I receptors on their concave surface but bind type II receptors more distally towards the fingertip region; in this case, the two types of receptors interact with each other, which is necessary for high-affinity ternary complex formation and signaling (Keah and Hearn 2005).

Traditionally, myostatin was thought to belong to the activin class of the TGF-β superfamily due to 40% amino acid sequence identity with activin and binding of the activin type II receptors and the activin inhibitor follistatin (Cash, Rejon et al. 2009). The X-ray crystal structure however bears similarities to the TGF-β class as well. Although the myostatin structure is similar to other TGF-β family members, there are significant differences in the N-terminus and region preceding the wrist helix (the prehelix loop) which is situated in the type I receptor binding site. The prehelix loop is most similar to TGF-β; the activin prehelix loop contains a number of glycine and serine residues and is flexible whereas myostatin has several large hydrophobic residues in this area.

The myostatin precursor, propetide and latent complex

No structure of the precursor protein, the propeptide region or latent complex for any TGF-β family member has been published. However, circular dichroism (CD), denaturation studies and mutagenesis provide information on secondary structure, stability and latent complex interactions respectively.

Recombinant TGF-β1 propeptide has a CD spectrum dominated by a minimum at 206 nm and a shoulder at 223 nm (Fig. 1.7a) (McMahon, Dignam et al. 1996). Although the authors state that these minima are characteristic of β-sheet structure, a number of

published CD reviews state these represent α-helical structures (Kelly, Jess et al.

2005). The latent complex has a spectrum composed primarily of α-helix with some β-sheet. An addition spectrum of the β-sheet predominant TGF-β1 growth factor and the TGF-β1 propeptide differs from that of the latent complex with a reduction in intensities of both structural minima suggesting that structural rearrangement accompanies the formation of the latent complex (McMahon, Dignam et al. 1996). Circular dichroism of the BMP-2 precursor protein and isolated propeptide showed similar results with minima at 212 and 218 nm for the precursor and a predominance of α-helix for the propeptide domain (Hillger, Herr et al. 2005).

Figure 1.7 A conformational change accompanies TGF-β family latent complex formation.

a. CD analysis of TGF-β1 latent complex formation (McMahon, Dignam et al. 1996). CD spectra of recombinant TGF-β1 growth factor (long-dash line), propeptide (solid line), latent complex (short-dash line), summed spectra of TGF-β1 and propeptide (dash-dot-dash line) and difference spectra between latent complex and summed spectra (dotted line).

b. A proposed model for precursor dimerisation, latent complex formation and post-secretion structure (Walton, Makanji et al. 2009).

Thermodynamic stability of BMP-2 species, meausured by urea and guanidine HCl- induced denaturation, indicates reduced structural stability of the propeptide in comparison to the growth factor, and that the growth factor domain stabilizes the structure of the propeptide in the precursor protein (Hillger, Herr et al. 2005).

Site-specific mutagenesis of the TGF-β1 propeptide characterized residues 50-85 as important for interaction with the TGF-β1 growth factor (Sha, Yang et al. 1991). Similarly, mutagenesis of the myostatin propeptide domain characterized the inhibitory region to be residues 42-115, implying that this region contains residues important for interaction with the growth factor in the latent complex (Jiang, Liang et al. 2004). The C-terminus of the myostatin propeptide was suggested to confer propeptide stability.

Investigation into the interactions involved in the inhibin/activin latent complex showed that hydrophobic interactions involving the N-terminus of the propeptide were important for growth factor binding in the precursor protein (Walton, Makanji et al. 2009); the outer convex surface of the growth factor is postulated to be involved in this interaction. A conserved hydrophobic motif (Hyd-Hyd-X-X-Hyd-X-Hyd) containing the identified residues in the propeptide domains of other TGF-β family members implies a common mechanism for interaction, for example, the TGF-β1 propeptide is able to form a complex with the growth factor regions of other family members (McMahon, Dignam et al. 1996).

A proposed model for precursor dimerisation, latent complex formation and secretion is shown in Figure 1.7b. Hydrophobic interactions between the N-terminal portion of the propeptide and the ‘finger regions’ of the growth factor maintain the molecule in a conformation competent for dimerisation. Cleavage by furin generates the latent complex which consists of a growth factor dimer non-covalently associated to two propeptide domains, and is secreted from the cell (Walton, Makanji et al. 2009).