1 . A method for treating fatty liver disease in a patient in need thereof, the method comprising administering an effective amount of a compound selected from the group consisting of:
a. a polypeptide comprising an amino acid sequence that is at least 90% identical to the sequence of amino acids 29-109 of SEQ ID NO:2; and b. a polypeptide encoded by a nucleic acid that hybridizes under stringent hybridization conditions to the nucleic acid of SEQ ID NO: 3.
2 . The method of claim 1 , wherein the polypeptide is a fusion protein comprising a portion heterologous to ActRIIB.
3 . The method of claim 1 , wherein the polypeptide is a dimer.
4 . The method of claim 2 , wherein the polypeptide is fused to a constant domain of an immunoglobulin.
5 . The method of claim 2 , wherein the polypeptide is fused to an Fc portion of an immunoglobulin.
6 . The method of claim 5 , wherein the immunoglobulin is a human IgG1.
7 . The method of claim 1 , wherein the polypeptide comprises the sequence of SEQ ID NO:5 or 6.
8 . The method of claim 1 , wherein the patient has insulin resistance.
9 . The method of claim 1 , wherein the patient has insulin resistance and a metabolic disorder.
10 . The method of claim 1 , wherein the polypeptide comprises an amino acid sequence that is at least 95% identical to the sequence of amino acids 29-109 of SEQ ID NO:2.
11 . The method of claim 1 , wherein the polypeptide comprises an amino acid sequence that is at least 97% identical to the sequence of amino acids 29-109 of SEQ ID NO:2.
12 . The method of claim 1 , wherein the polypeptide comprises an amino acid sequence that is at least 99% identical to the sequence of amino acids 29-109 of SEQ ID NO:2.
13 . The method of claim 1 , wherein the polypeptide comprises an amino acid sequence that is at least 90% identical to the sequence of amino acids 25-131 of SEQ ID NO:2.
14 . The method of claim 1 , wherein the polypeptide comprises an amino acid sequence that is at least 95% identical to the sequence of amino acids 25-131 of SEQ ID NO:2.
15 . The method of claim 1 , wherein the polypeptide comprises an amino acid sequence that is at least 97% identical to the sequence of amino acids 25-131 of SEQ ID NO:2.
16 . The method of claim 1 , wherein the polypeptide comprises an amino acid sequence that is at least 99% identical to the sequence of amino acids 25-131 of SEQ ID NO:2.
17 . The method of claim 1 , wherein administration of the compound inhibits hepatic steatosis in the treated patient.
18 . The method of claim 17 , wherein the patient is treated for non-alcoholic fatty liver disease.
19 . A method for treating fatty liver disease in a patient in need thereof, the method comprising administering an effective amount of a compound selected from the group consisting of:
a. an antagonist of ActRIIB; b. an antagonist of myostatin; c. an antagonist of activin A and/or activin B; and d. an antagonist of GDF3.
20 . The method of claim 19 , wherein the compound is an antagonist of ActRIIB.
21 . The method of claim 20 , wherein the antagonist of ActRIIB is selected from the group consisting of: an antibody that binds to ActRIIB and a nucleic acid that hybridizes to a nucleic acid encoding ActRIIB and inhibits ActRIIB production.
22 . The method of claim 19 , wherein the compound is an antagonist of myostatin.
23 . The method of claim 22 , wherein the antagonist of myostatin is selected from the group consisting of: an antibody that binds to myostatin, a nucleic acid that hybridizes to a nucleic acid encoding myostatin and inhibits myostatin production, and a polypeptide comprising a myostatin propeptide or variant thereof.
24 . The method of claim 19 , wherein the compound is an antagonist of activin A and/or activin B.
25 . The method of claim 24 , wherein the antagonist of activin A and/or activin B is selected from the group consisting of: an antibody that binds to activin A and/or activin B and a nucleic acid that hybridizes to a nucleic acid encoding activin A and/or activin B and inhibits production of activin A and/or activin B.
26 . The method of claim 19 , wherein the compound is an antagonist of GDF3.
27 . The method of claim 26 , wherein the antagonist of GDF3 is selected from the group consisting of: an antibody that binds to GDF3, a nucleic acid that hybridizes to a nucleic acid encoding GDF3 and inhibits GDF3 production, and a polypeptide comprising a GDF3 propeptide or variant thereof.
 This application claims the benefit of Provisional Application Ser. No. 61/280,544, filed Nov. 3, 2009. All the teachings of the above-referenced application are incorporated herein by reference.
BACKGROUND OF THE INVENTION
 The transforming growth factor-beta (TGF-beta) superfamily contains a variety of growth factors that share common sequence elements and structural motifs. These proteins are known to exert biological effects on a large variety of cell types in both vertebrates and invertebrates. Members of the superfamily perform important functions during embryonic development in pattern formation and tissue specification and can influence a variety of differentiation processes, including adipogenesis, myogenesis, chondrogenesis, cardiogenesis, hematopoiesis, neurogenesis, and epithelial cell differentiation. The family is represented by proteins named, variously, the activins and inhibins, TGF-beta, Growth and Differentiation Factors (GDFs) and Bone Morphogenetic Factors (BMPs). Other members of the family are also known, such as Nodal and Lefty. By manipulating the activity of a member of the TGF-beta family, it is often possible to cause significant physiological changes in an organism. For example, the Piedmontese and Belgian Blue cattle breeds carry a loss-of-function mutation in the GDF8 (also called myostatin) gene that causes a marked increase in muscle mass. Grobet et al., Nat. Genet. 1997, 17(1):71-4. Furthermore, in humans, inactive alleles of GDF8 are associated with increased muscle mass and, reportedly, exceptional strength. Schuelke et al., N Engl J Med 2004, 350:2682-8.
 Changes in muscle, bone, fat, cartilage and other tissues may be achieved by agonizing or antagonizing signaling that is mediated by an appropriate TGF-beta family member. Fatty liver disease is a potentially serious condition often associated with insulin resistance, diabetes or alcoholism for which there are few therapeutic options. Thus, there is a need for agents that function as potent regulators of signaling by members of the TGF-beta superfamily for the treatment of fatty liver disease.
SUMMARY OF THE INVENTION
 In certain aspects, the present disclosure provides methods for treating fatty liver disease in patients by using antagonists of the ActRIIB signaling pathway. Such antagonists may be, for example, soluble ActRIIB proteins (e.g., ActRIIB-Fc fusion proteins), antagonists that bind to ActRIIB or inhibit ActRIIB expression, and antagonists that bind to or inhibit the expression of ligands that signal through ActRIIB and participate in fatty liver disease. Such ligands may include myostatin, GDF3, activins, BMP7, BMP2 and BMP4. As demonstrated herein, ActRIIB-Fc fusion proteins can be used to decrease liver adiposity with, in some instances, additional benefits such as a decrease in insulin resistance or an increase in adiponectin production. Accordingly, in certain embodiments, antagonism of the ActRIIB signaling pathway may be used to achieve beneficial effects on hepatic steatosis (lipid deposition), hypoadiponectinemia, and/or insulin resistance (including, for example, indicators of insulin resistance such as hyperinsulinemia).
 In certain aspects, the disclosure provides methods for treating fatty liver disease by administering to a patient in need thereof an effective amount of an ActRIIB-related polypeptide. An ActRIIB-related polypeptide may be an ActRIIB polypeptide (e.g., an ActRIIB extracellular domain or portion thereof) that binds to an ActRIIB ligand such as GDF3, BMP2, BMP4, BMP7, GDF8, GDF11, activin A, activin B or nodal. Optionally, the ActRIIB polypeptide binds to an ActRIIB ligand with a Kd less than 10 micromolar or less than 1 micromolar, 100, 10 or 1 nanomolar. A variety of suitable ActRIIB polypeptides have been described in the following published PCT patent applications, all of which are incorporated by reference herein: WO 00/43781, WO 04/039948, WO 06/012627, WO 07/053,775, WO 08/097,541, and WO 08/109,167. Optionally, the ActRIIB polypeptide inhibits ActRIIB signaling, such as intracellular signal transduction events triggered by an ActRIIB ligand. A soluble ActRIIB polypeptide for use in such a preparation may be any of those disclosed herein, such as a polypeptide having an amino acid sequence selected from SEQ ID NOs: 1, 2, 5, 6, 12 and 14, or having an amino acid sequence that is at least 80%, 85%, 90%, 95%, 97% or 99% identical to an amino acid sequence selected from SEQ ID NOs: 1, 2, 5, 6, 12 and 14. A soluble ActRIIB polypeptide may include a functional fragment of a natural ActRIIB polypeptide, such as one comprising at least 10, 20 or 30 amino acids of a sequence selected from SEQ ID NOs: 1, 2, 5, 6, 12 and 14 or a sequence of SEQ ID NO: 1, lacking the C-terminal 1, 2, 3, 4, 5 or 10 to 15 amino acids and lacking 1, 2, 3, 4 or 5 amino acids at the N-terminus. Optionally, polypeptides will comprise a truncation relative to SEQ ID NO:1 of between 2 and 5 amino acids at the N-terminus and no more than 3 amino acids at the C-terminus. Another polypeptide is that presented as SEQ ID NO:12. A soluble ActRIIB polypeptide may include one, two, three, four, five or more alterations in the amino acid sequence (e.g., in the ligand-binding domain) relative to a naturally occurring ActRIIB polypeptide. The alteration in the amino acid sequence may, for example, alter glycosylation of the polypeptide when produced in a mammalian, insect or other eukaryotic cell or alter proteolytic cleavage of the polypeptide relative to the naturally occurring ActRIIB polypeptide. A soluble ActRIIB polypeptide may be a fusion protein that has, as one domain, an ActRIIB polypeptide (e.g., a ligand-binding domain of an ActRIIB or a variant thereof) and one or more additional domains that provide a desirable property, such as improved pharmacokinetics, easier purification, targeting to particular tissues, etc. For example, a domain of a fusion protein may enhance one or more of in vivo stability, in vivo half life, uptake/administration, tissue localization or distribution, formation of protein complexes, multimerization of the fusion protein, and/or purification. A soluble ActRIIB fusion protein may include an immunoglobulin constant domain, such as an Fc domain (wild-type or mutant) or a serum albumin In certain embodiments, an ActRIIB-Fc fusion comprises a relatively unstructured linker positioned between the Fc domain and the extracellular ActRIIB domain. This unstructured linker may correspond to the roughly 15 amino acid unstructured region at the C-terminal end of the extracellular domain of ActRIIB (the “tail”), or it may be an artificial sequence of between 5 and 15, 20, 30, 50 or more amino acids that are relatively free of secondary structure. A linker may be rich in glycine and proline residues and may, for example, contain repeating or non-repeating sequences of threonine/serine and/or glycines (e.g., TG 4 , TG 3 , SG 4 , SG 3 , G 4 , G 3 , G 2 , G). A fusion protein may include a purification subsequence, such as an epitope tag, a FLAG tag, a polyhistidine sequence, and a GST fusion. Optionally, a soluble ActRIIB polypeptide includes one or more modified amino acid residues selected from: a glycosylated amino acid, a PEGylated amino acid, a farnesylated amino acid, an acetylated amino acid, a biotinylated amino acid, an amino acid conjugated to a lipid moiety, and an amino acid conjugated to an organic derivatizing agent. In general, it is preferable that an ActRIIB protein be expressed in a mammalian cell line that mediates suitably natural glycosylation of the ActRIIB protein so as to diminish the likelihood of an unfavorable immune response in a patient. Human and CHO cell lines have been used successfully, and it is expected that other common mammalian expression vectors will be useful.
 In certain aspects, a compound disclosed herein may be formulated as a pharmaceutical preparation for the treatment of fatty liver disease. A pharmaceutical preparation may also include one or more additional compounds such as a compound that is used to treat an ActRIIB-associated disorder. Preferably, a pharmaceutical preparation is substantially pyrogen free.
 In certain aspects, the disclosure provides nucleic acids encoding a soluble ActRIIB polypeptide, which do not encode a complete ActRIIB polypeptide. An isolated polynucleotide may comprise a coding sequence for a soluble ActRIIB polypeptide, such as described above. For example, an isolated nucleic acid may include a sequence coding for an extracellular domain (e.g., ligand-binding domain) of an ActRIIB and a sequence that would code for part or all of the transmembrane domain and/or the cytoplasmic domain of an ActRIIB, but for a stop codon positioned within the transmembrane domain or the cytoplasmic domain, or positioned between the extracellular domain and the transmembrane domain or cytoplasmic domain. For example, an isolated polynucleotide may comprise a full-length ActRIIB polynucleotide sequence such as SEQ ID NO: 4, or a partially truncated version, said isolated polynucleotide further comprising a transcription termination codon at least six hundred nucleotides before the 3′-terminus or otherwise positioned such that translation of the polynucleotide gives rise to an extracellular domain optionally fused to a truncated portion of a full-length ActRIIB. Other suitable nucleic acids that encode ActRIIB polypeptides are shown as SEQ ID NO: 3 and 15. Nucleic acids disclosed herein may be operably linked to a promoter for expression, and the disclosure provides cells transformed with such recombinant polynucleotides. Preferably the cell is a mammalian cell such as a CHO cell.
 In certain aspects, the disclosure provides methods for making a soluble ActRIIB polypeptide. Such a method may include expressing any of the nucleic acids (e.g., SEQ ID NO: 3 or 15) disclosed herein in a suitable cell, such as a Chinese hamster ovary (CHO) cell. Such a method may comprise: a) culturing a cell under conditions suitable for expression of the soluble ActRIIB polypeptide, wherein said cell is transformed with a soluble ActRIIB expression construct; and b) recovering the soluble ActRIIB polypeptide so expressed. Soluble ActRIIB polypeptides may be recovered as crude, partially purified or highly purified fractions using any of the well known techniques for obtaining protein from cell cultures.
 In certain aspects, a compound described herein may be used in the management of a variety of forms of fatty liver disease and complications thereof (e.g., nonalcoholic fatty liver disease, nonalcoholic steatohepatitis, alcoholic fatty liver disease, alcoholic steatohepatitis, hepatic fibrosis, cirrhosis) as well as related disorders such as hypoadiponectinemia, insulin resistance, or hyperinsulinemia Remarkably, as shown herein, ActRIIB polypeptides may be used to achieve positive effects on fatty liver disease while also having a positive effect on the related disorders of hypoadiponectinemia and insulin resistance.
 In certain aspects, the disclosure provides uses of a soluble ActRIIB polypeptide for making a medicament for the treatment of a disorder or condition as described herein.
 In certain aspects, the disclosure provides methods for treating fatty liver disease in a patient in need thereof, and such method may comprise administering an effective amount of a compound selected from the group consisting of: a polypeptide comprising an amino acid sequence that is at least 90% identical to the sequence of amino acids 29-109 of SEQ ID NO: 2 and a polypeptide encoded by a nucleic acid that hybridizes under stringent hybridization conditions to a nucleic acid of SEQ ID NO: 3. The polypeptide may be a fusion protein comprising a heterologous portion. The polypeptide may be a dimer. The polypeptide may be fused to a constant domain of an immunoglobulin. The polypeptide may be fused to an Fc portion of an immunoglobulin, such as an IgG1, IgG2, IgG3 or IgG4. The polypeptide may comprise an amino acid sequence that is at least 80%, 90%, 93%, 95%, 97%, 98%, 99% or 100% identical to the sequence of amino acids 29-109, 29-128, 29-131, 29-134, 25-109, 25-128, 25-131, 25-134 or 20-134 of SEQ ID NO:2. The polypeptide may comprise an amino acid sequence that is at least 80%, 90%, 93%, 95%, 97%, 98%, 99% or 100% identical to the sequence of amino acids of SEQ ID NO:5, 6, 12 or 14. A patient to be treated with such a compound may be one having a disorder described herein, including, for example, fatty liver disease or complication thereof (e.g., nonalcoholic fatty liver disease, nonalcoholic steatohepatitis, alcoholic fatty liver disease, alcoholic steatohepatitis, hepatic fibrosis, or cirrhosis), hypoadiponectinemia, insulin resistance, hyperinsulinemia,
 In certain aspects, the disclosure provides methods for treating fatty liver disease in a patient in need thereof, the method comprising administering an effective amount of a compound that inhibits the ActRIIB signaling pathway, either by targeting ActRIIB or a ligand that signals through ActRIIB. Examples of such compounds include antagonists of ActRIIB; antagonists of myostatin; antagonists of activin A; antagonists of activin B; antagonists of BMP2; antagonists of BMP4 and antagonists of GDF3. Antagonists of each of the foregoing may be an antibody or other protein that specifically binds to and inhibits such target (e.g., an antibody such as a monoclonal antibody, or a propeptide in the case of myostatin and GDF3). Antagonists of the foregoing may also be a compound, such as a nucleic acid based compound (e.g., an antisense or RNAi nucleic acid) that inhibits the expression of ActRIIB or the ligand. A patient to be treated with such a compound may one having a disorder described herein, including, for example, fatty liver disease (e.g., nonalcoholic fatty liver disease, nonalcoholic steatohepatitis, alcoholic fatty liver disease, alcoholic steatohepatitis, hepatic fibrosis, or cirrhosis), hypoadiponectinemia, insulin resistance, hyperinsulinemia
BRIEF DESCRIPTION OF THE DRAWINGS
 The patent or patent application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the nessisary fee.
 FIG. 1 shows the full amino acid sequence of ActRIIB(25-131)-hFc (SEQ ID NO:14). The TPA leader (residues 1-22) and truncated ActRIIB extracellular domain (native residues 25-131) are each underlined. Highlighted is the glutamate revealed by sequencing to be the N-terminal amino acid of the mature fusion protein.
 FIG. 2 shows a nucleotide sequence encoding ActRIIB(25-131)-hFc (SEQ ID NO:15) (the coding strand is shown at top and the complement shown at bottom 3′-5′). Sequences encoding the TPA leader (nucleotides 1-66) and ActRIIB extracellular domain (nucleotides 73-396) are underlined. The corresponding amino acid sequence for ActRIIB(25-131) is also shown.
 FIG. 3 shows the effect of ActRIIB(25-131)-hFc treatment for 60 days on hepatic steatosis in a mouse model of diet-induced obesity. Liver sections (all shown at the same magnification) stained with Oil Red 0 reveal pronounced lipid deposition under high-fat dietary conditions but not control conditions. Arrows indicate several of many densely packed lipid droplets, which are stained bright red but difficult to discern in black-and-white images. ActRIIB(25-131)-hFc inhibited formation of such lipid droplets and largely restored the appearance of liver tissue to that of mice fed the standard diet.
 FIG. 4 shows the effect of ActRIIB(25-131)-hFc treatment for 60 days on adiponectin mRNA levels in epididymal white fat in a mouse model of diet-induced obesity. RT-PCR data (in relative units, RU) are means±SEM; n=7 per group; *, p<0.05. In mice fed a high-fat diet, ActRIIB(25-131)-hFc increased adiponectin mRNA levels by more than 60%, thus contributing to elevated concentrations of circulating adiponectin in these mice.
 FIG. 5 shows the effect of ActRIIB(25-131)-hFc treatment for 60 days on serum adiponectin concentrations in a mouse model of diet-induced obesity. ELISA measurements detect all main oligomeric isoforms (total adiponectin), and data are means±SEM; n=7-8 per group; **, p<0.01; ***, p<0.001. ActRIIB(25-131)-hFc increased circulating adiponectin concentrations by more than 75% compared to high-fat diet controls and even raised such concentrations significantly above those observed in standard-diet controls.
 FIG. 6 shows the effect of ActRIIB(25-131)-hFc treatment for 60 days on serum insulin concentrations in a mouse model of diet-induced obesity. Data are means±SEM; n=7-8 per group; **, p<0.01. ActRIIB(25-131)-hFc reversed the effect of high-fat diet on insulin concentrations, indicative of increased insulin sensitivity in target tissues.
 FIG. 7 shows the effect of ActRIIB(25-131)-mFc treatment for 16 weeks on liver tissue density in a mouse model of atherogenesis, as determined by micro-computed tomography. Data (means±SEM) are expressed in standardized units based on a positive value for the bone mineral hydroxyapatite (HA) and a value of zero for water; therefore, lipid values are negative; **, p<0.01. Compared to vehicle treatment in this model, ActRIIB(25-131)-mFc increased liver density, which indicates a significant reduction in hepatic steatosis.
 Fatty liver disease encompasses a spectrum of liver conditions and is typically classified as either alcoholic or nonalcoholic. In either case, fatty liver disease ranges from simple hepatic steatosis (lipid accumulation and deposition) to steatohepatitis (ASH or NASH), which often progresses to hepatic fibrosis, cirrhosis, and probably hepatocellular carcinoma. Alcoholic (AFLD) and nonalcoholic fatty liver disease (NAFLD) are histologically indistinguishable; however, by definition NAFLD develops in patients who consume little or no alcohol. Instead, NAFLD is frequently found in individuals with obesity, metabolic syndrome, and type 2 diabetes and is closely linked to insulin resistance (Utzschneider et al., 2006, J Clin Endocrinol Metab 91:4753-4761). With the dramatic recent increase in the prevalence of obesity and insulin resistance, NAFLD has surpassed AFLD and viral hepatitis-induced liver disease as the most common chronic liver disease. It has been estimated that approximately 75% of those with obesity have NAFLD and as many as 20% may have NASH (Clark, 2006, J Clin Gastroenterol 40(Suppl 1):S5-S10; Lazo et al., 2008, Semin Liver Dis 28:339-350).
 Evidence has emerged of a close relationship between adiponectin and both types of fatty liver disease. Adiponectin, a fat-derived hormone whose concentration varies inversely with the mass of white adipose tissue, exerts important insulin-sensitizing actions in target tissues (Yamauchi et al., 2001, Nat Med 7:941-946; Maeda et al., 2002, Nat Med 8:731-737; Kadowaki et al., 2005, Endocr Rev 26:439-451). Across different human ethnic groups, the degree of hypoadiponectinemia (low circulating adiponectin concentrations) correlates even more closely with insulin resistance than with adiposity (Weyer et al., 2001, J Clin Endocrinol Metab 86:1930-1935). Moreover, in obese, ob/ob mice, adiponectin administration ameliorates hepatic steatosis and liver enlargement (hepatomegaly) (Xu et al., 2003, J Clin Invest 112:91-100). Finally, there is growing evidence of alcohol-mediated dysregulation of adiponectin signaling (You et al., 2009, Exp Biol Med 234:850-859), and adiponectin administration also ameliorates hepatic steatosis, hepatomegaly, and hepatic inflammation in a mouse model of AFLD (Xu et al., 2003, J Clin Invest 112:91-100).
 As described in the Examples, an ActRIIB-Fc fusion protein can be used to inhibit hepatic steatosis (lipid deposition), increase serum adiponectin concentrations, and normalize serum insulin concentrations in a mouse model of diet-induced obesity. Therefore, ActRIIB-derived agents and other compounds that inhibit ActRIIB signaling can be used to treat fatty liver disease while also achieving positive effects in part by increasing circulating adiponectin concentrations and/or decreasing insulin resistance in target tissues. Ligands that bind to ActRIIB which are implicated in the regulation of hepatic steatosis, circulating adiponectin concentrations, and insulin resistance include the activins (e.g., Activin A and Activin B), myostatin, GDF3, BMP7, BMP2 and BMP4. In certain aspects, the present invention relates to ActRIIB polypeptides. As used herein, the term “ActRIIB” refers to a family of activin receptor type IIB (ActRIIB) proteins and ActRIIB-related proteins, derived from any species. Members of the ActRIIB family are generally all transmembrane proteins, composed of a ligand-binding extracellular domain with cysteine-rich region, a transmembrane domain, and a cytoplasmic domain with predicted serine/threonine kinase specificity. The human ActRIIB precursor has the following amino acid sequence, with the signal peptide underlined, the extracellular domain indicated in bold, and the potential N-linked glycosylation sites boxed (SEQ ID NO: 2) (NM — 001106, 512 aa).
 The above wild type sequence, including the native leader, is used throughout this disclosure as the base sequence for numbering the amino acids of any of the various truncations, mature forms and variants of ActRIIB. The term “ActRIIB polypeptide” is used to refer to polypeptides comprising any naturally occurring polypeptide of an ActRIIB family member as well as any variants thereof (including mutants, fragments, fusions, and peptidomimetic forms) that retain a useful activity. For example, ActRIIB polypeptides include polypeptides derived from the sequence of any known ActRIIB having a sequence at least about 80% identical to the sequence of an ActRIIB polypeptide, and preferably at least 85%, 90%, 95%, 97%, 99% or greater identity.
 In a specific embodiment, the invention relates to soluble ActRIIB polypeptides. As described herein, the term “soluble ActRIIB polypeptide” generally refers to polypeptides comprising an extracellular domain of an ActRIIB protein. The term “soluble ActRIIB polypeptide,” as used herein, includes any naturally occurring extracellular domain of an ActRIIB protein as well as any variants thereof (including mutants, fragments and peptidomimetic forms) that retain a useful activity. For example, the extracellular domain of an ActRIIB protein binds to a ligand and is generally soluble. The following is an example of a soluble ActRIIB polypeptide (SEQ ID NO: 1) (116 aa).
 Other examples of soluble ActRIIB polypeptides comprise a signal sequence in addition to the extracellular domain of an ActRIIB protein (see Example 1). The signal sequence can be a native signal sequence of an ActRIIB, or a signal sequence from another protein, such as a tissue plasminogen activator (TPA) signal sequence or a honey bee mellitin (HBM) signal sequence.
 Two related type II receptors, ActRIIA and ActRIIB, have been identified as the type II receptors for activins (Mathews and Vale, 1991, Cell 65:973-982; Attisano et al., 1992, Cell 68: 97-108) as well as a variety of other BMPs and GDFs. Besides activins, ActRIIA and ActRIIB can biochemically interact with several other TGF-β family proteins, including BMP7, Nodal, GDF8, and GDF11 (Yamashita et al., 1995, J. Cell Biol. 130:217-226; Lee and McPherron, 2001, Proc. Natl. Acad. Sci. 98:9306-9311; Yeo and Whitman, 2001, Mol. Cell. 7: 949-957; Oh et al., 2002, Genes Dev. 16:2749-54). In certain embodiments, the present invention relates to antagonizing a ligand of ActRIIB receptors (also referred to as an ActRIIB ligand) with a subject ActRIIB polypeptide (e.g., a soluble ActRIIB polypeptide). Thus, compositions and methods of the present invention are useful for treating disorders associated with abnormal activity of one or more ligands of ActRIIB receptors. Exemplary ligands of ActRIIB receptors include some TGF-β family members, such as activin A, activin B, GDF3, Nodal, GDF8, and GDF11.
 Activins are dimeric polypeptide growth factors and belong to the TGF-beta superfamily. There are three activins (A, B, and AB) that are homo/heterodimers of two closely related 13 subunits (β A β A , β B β B , and β A β B ). In the TGF-beta superfamily, activins are unique and multifunctional factors that can stimulate hormone production in ovarian and placental cells, support neuronal cell survival, influence cell-cycle progress positively or negatively depending on cell type, and induce mesodermal differentiation at least in amphibian embryos (DePaolo et al., 1991, Proc SocEp Biol Med. 198:500-512; Dyson et al., 1997, Curr Biol. 7:81-84; Woodruff, 1998, Biochem Pharmacol. 55:953-963). Moreover, erythroid differentiation factor (EDF) isolated from the stimulated human monocytic leukemic cells was found to be identical to activin A (Murata et al., 1988, PNAS, 85:2434). It was suggested that activin A acts as a natural regulator of erythropoiesis in the bone marrow. In several tissues, activin signaling is antagonized by its related heterodimer, inhibin. For example, during the release of follicle-stimulating hormone (FSH) from the pituitary, activin promotes FSH secretion and synthesis, while inhibin prevents FSH secretion and synthesis. Other proteins that may regulate activin bioactivity and/or bind to activin include follistatin (FS), follistatin-related protein (FSRP), α 2 -macroglobulin, Cerberus, and endoglin, which are described below.
 Growth-and-Differentiation Factor-3 (GDF3), also known as Vg1-related 2, plays an important role in embryonic development and has also been implicated in adipogenesis during adulthood. In brief, expression of GDF3 in white adipose tissue is correlated with body mass or obesity (Weisberg et al., 2003, J Clin Invest 112:1796-1808), and adenovirus-mediated overexpression of GDF3 exaggerates the increase in adiposity observed under high-fat dietary conditions in wildtype mice (Wang et al., 2004, Biochem Biophys Res Commun 321:1024-1031). Importantly, mice with genetic ablation of GDF3 are healthy and essentially normal when maintained on a standard diet but are protected from obesity, and display an increased basal metabolic rate, when maintained on a high-fat diet (Shen et al., 2009, Mol Endocrinol 23:113-123). Taken together, these findings implicate GDF3 specifically in diet-induced obesity and more generally in the regulation of adiposity.
 Nodal proteins have functions in mesoderm and endoderm induction and formation, as well as subsequent organization of axial structures such as heart and stomach in early embryogenesis. It has been demonstrated that dorsal tissue in a developing vertebrate embryo contributes predominantly to the axial structures of the notochord and pre-chordal plate while it recruits surrounding cells to form non-axial embryonic structures. Nodal appears to signal through both type I and type II receptors and intracellular effectors known as Smad proteins. Recent studies support the idea that ActRIIA and ActRIIB serve as type II receptors for Nodal (Sakuma et al., Genes Cells. 2002, 7:401-12). It is suggested that Nodal ligands interact with their co-factors (e.g., cripto) to activate activin type I and type II receptors, which phosphorylate Smad2. Nodal proteins are implicated in many events critical to the early vertebrate embryo, including mesoderm formation, anterior patterning, and left-right axis specification. Experimental evidence has demonstrated that Nodal signaling activates pAR3-Lux, a luciferase reporter previously shown to respond specifically to activin and TGF-beta. However, Nodal is unable to induce pTlx2-Lux, a reporter specifically responsive to bone morphogenetic proteins. Recent results provide direct biochemical evidence that Nodal signaling is mediated by both activin-TGF-beta pathway Smads, Smad2 and Smad3. Further evidence has shown that the extracellular cripto protein is required for Nodal signaling, making it distinct from activin or TGF-beta signaling.
 Growth and Differentiation Factor-8 (GDF8) is also known as myostatin. GDF8 is a negative regulator of skeletal muscle mass. GDF8 is highly expressed in the developing and adult skeletal muscle. The GDF8 null mutation in transgenic mice is characterized by a marked hypertrophy and hyperplasia of the skeletal muscle (McPherron et al., Nature, 1997, 387:83-90). Similar increases in skeletal muscle mass are evident in naturally occurring mutations of GDF8 in cattle (Ashmore et al., 1974, Growth, 38:501-507; Swatland and Kieffer, J. Anim Sci., 1994, 38:752-757; McPherron and Lee, Proc. Natl. Acad. Sci. USA, 1997, 94:12457-12461; and Kambadur et al., Genome Res., 1997, 7:910-915) and, strikingly, in humans (Schuelke et al., N Engl J Med 2004; 350:2682-8). Studies have also shown that muscle wasting associated with HIV-infection in humans is accompanied by increases in GDF8 protein expression (Gonzalez-Cadavid et al., PNAS, 1998, 95:14938-43). In addition, GDF8 can modulate the production of muscle-specific enzymes (e.g., creatine kinase) and modulate myoblast cell proliferation (WO 00/43781). The GDF8 propeptide can noncovalently bind to the mature GDF8 domain dimer, inactivating its biological activity (Miyazono et al. (1988) J. Biol. Chem., 263: 6407-6415; Wakefield et al. (1988) J. Biol. Chem., 263; 7646-7654; and Brown et al. (1990) Growth Factors, 3: 35-43). Other proteins which bind to GDF8 or structurally related proteins and inhibit their biological activity include follistatin, and potentially, follistatin-related proteins (Gamer et al. (1999) Dev. Biol., 208: 222-232).
 Growth and Differentiation Factor-11 (GDF11), also known as BMP11, is a secreted protein (McPherron et al., 1999, Nat. Genet. 22: 260-264). GDF11 is expressed in the tail bud, limb bud, maxillary and mandibular arches, and dorsal root ganglia during mouse development (Nakashima et al., 1999, Mech. Dev. 80: 185-189). GDF11 plays a unique role in patterning both mesodermal and neural tissues (Gamer et al., 1999, Dev Biol., 208:222-32). GDF11 was shown to be a negative regulator of chondrogenesis and myogenesis in developing chick limb (Gamer et al., 2001, Dev Biol. 229:407-20). The expression of GDF11 in muscle also suggests its role in regulating muscle growth in a similar way to GDF8. In addition, the expression of GDF11 in brain suggests that GDF11 may also possess activities that relate to the function of the nervous system. Interestingly, GDF11 was found to inhibit neurogenesis in the olfactory epithelium (Wu et al., 2003, Neuron. 37:197-207). Hence, GDF11 may have in vitro and in vivo applications in the treatment of diseases such as muscle diseases and neurodegenerative diseases (e.g., amyotrophic lateral sclerosis).
 In certain aspects, the present invention relates to the use of certain ActRIIB polypeptides (e.g., soluble ActRIIB polypeptides) to antagonize the signaling of ActRIIB ligands generally, in any process associated with ActRIIB activity. Optionally, ActRIIB polypeptides of the invention may antagonize one or more ligands of ActRIIB receptors, such as activin A, activin B, GDF3, BMP2, BMP4, BMP7, Nodal, GDF8, and GDF11, and may therefore be useful in the treatment of additional disorders.
 Therefore, the present disclosure contemplates using ActRIIB polypeptides and antagonists of ActRIIB or ActRIIB ligands in treating or preventing diseases or conditions that are related to fatty liver disease. ActRIIB or ActRIIB ligands are involved in the regulation of many critical biological processes. Examples of such metabolic disorders or conditions related to fatty liver disease include, but are not limited to, nonalcoholic fatty liver disease, nonalcoholic steatohepatitis, alcoholic fatty liver disease, alcoholic steatohepatitis, hepatic fibrosis, cirrhosis, hypoadiponectinemia, insulin resistance, and hyperinsulinemia. These disorders and conditions are discussed below under “Exemplary Therapeutic Uses.”
 The terms used in this specification generally have their ordinary meanings in the art, within the context of this invention and in the specific context where each term is used. Certain terms are discussed below or elsewhere in the specification, to provide additional guidance to the practitioner in describing the compositions and methods of the invention and how to make and use them. The scope or meaning of any use of a term will be apparent from the specific context in which the term is used.
 “About” and “approximately” shall generally mean an acceptable degree of error for the quantity measured given the nature or precision of the measurements. Typically, exemplary degrees of error are within 20 percent (%), preferably within 10%, and more preferably within 5% of a given value or range of values.
 Alternatively, and particularly in biological systems, the terms “about” and “approximately” may mean values that are within an order of magnitude, preferably within 5-fold and more preferably within 2-fold of a given value. Numerical quantities given herein are approximate unless stated otherwise, meaning that the term “about” or “approximately” can be inferred when not expressly stated.
 The methods of the invention may include steps of comparing sequences to each other, including wild-type sequence to one or more mutants (sequence variants). Such comparisons typically comprise alignments of polymer sequences, e.g., using sequence alignment programs and/or algorithms that are well known in the art (for example, BLAST, FASTA and MEGALIGN, to name a few). The skilled artisan can readily appreciate that, in such alignments, where a mutation contains a residue insertion or deletion, the sequence alignment will introduce a “gap” (typically represented by a dash, or “A”) in the polymer sequence not containing the inserted or deleted residue.
 The term “diabetes”, as used herein, refers to non-insulin-dependent diabetes mellitus (NIDDM, also known as type II diabetes). Type I diabetes, or insulin-dependent diabetes mellitus (IDDM), is the result of an absolute deficiency of insulin, the hormone which regulates glucose utilization. Type II diabetes, or insulin-dependent diabetes (i.e., non-insulin-dependent diabetes mellitus), often occurs in the face of normal, or even elevated, levels of insulin and appears to be the result of the inability of tissues to respond appropriately to insulin. Most type II diabetics are also obese.
 “Homologous,” in all its grammatical forms and spelling variations, refers to the relationship between two proteins that possess a “common evolutionary origin,” including proteins from superfamilies in the same species of organism, as well as homologous proteins from different species of organism. Such proteins (and their encoding nucleic acids) have sequence homology, as reflected by their sequence similarity, whether in terms of percent identity or by the presence of specific residues or motifs and conserved positions.
 The term “sequence similarity,” in all its grammatical forms, refers to the degree of identity or correspondence between nucleic acid or amino acid sequences that may or may not share a common evolutionary origin.
 However, in common usage and in the instant application, the term “homologous,” when modified with an adverb such as “highly,” may refer to sequence similarity and may or may not relate to a common evolutionary origin.
2. ActRIIB Polypeptides
 In certain aspects, the invention relates to ActRIIB variant polypeptides (e.g., soluble ActRIIB polypeptides). Optionally, the fragments, functional variants, and modified forms have similar or the same biological activities of their corresponding wild-type ActRIIB polypeptides. For example, an ActRIIB variant of the invention may bind to and inhibit function of an ActRIIB ligand (e.g., activin A, activin AB, activin B, activin C, activin E, GDF3, BMP2, BMP4, BMP7, Nodal, GDF8, or GDF11). Optionally, an ActRIIB polypeptide modulates growth of tissues such as fat, muscle, bone, or cartilage. Examples of ActRIIB polypeptides include human ActRIIB precursor polypeptide (SEQ ID NO: 2), and soluble human ActRIIB polypeptides (e.g., SEQ ID NOs: 1, 5, 6, 12 and 14).
 The disclosure identifies functionally active portions and variants of ActRIIB. Applicants have ascertained that an Fc fusion protein having the sequence disclosed by Hilden et al. (Blood. 1994 Apr. 15; 83(8):2163-70), which has an Alanine at the position corresponding to amino acid 64 of SEQ ID NO: 2 (A64), has a relatively low affinity for activin and GDF-11. By contrast, the same Fc fusion protein with an Arginine at position 64 (R64) has an affinity for activin and GDF-11 in the low nanomolar to high picomolar range. Therefore, a sequence with an R64 is used as the wild-type reference sequence for human ActRIIB in this disclosure.
 Attisano et al. (Cell. 1992 Jan. 10; 68(1):97-108) showed that a deletion of the proline knot at the C-terminus of the extracellular domain of ActRIIB reduced the affinity of the receptor for activin. An ActRIIB-Fc fusion protein containing amino acids 20-119 of SEQ ID NO:2, “ActRIIB(20-119)-Fc” has reduced binding to GDF-11 and activin relative to an ActRIIB(20-134)-Fc, which includes the proline knot region and the complete juxtamembrane domain. However, an ActRIIB(20-129)-Fc protein retains similar but somewhat reduced activity relative to the wild type, even though the proline knot region is disrupted. Thus, ActRIIB extracellular domains that stop at amino acid 134, 133, 132, 131, 130 and 129 are all expected to be active, but constructs stopping at 134 or 133 may be most active. Similarly, mutations at any of residues 129-134 are not expected to alter ligand binding affinity by large margins. In support of this, mutations of P129 and P130 do not substantially decrease ligand binding. Therefore, an ActRIIB-Fc fusion protein may end as early as amino acid 109 (the final cysteine), however, forms ending at or between 109 and 119 are expected to have reduced ligand binding Amino acid 119 is poorly conserved and so is readily altered or truncated. Forms ending at 128 or later retain ligand binding activity. Forms ending at or between 119 and 127 will have an intermediate binding ability. Any of these forms may be desirable to use, depending on the clinical or experimental setting.
 At the N-terminus of ActRIIB, it is expected that a protein beginning at amino acid 29 or before will retain ligand binding activity. Amino acid 29 represents the initial cysteine. An alanine to asparagine mutation at position 24 introduces an N-linked glycosylation sequence without substantially affecting ligand binding. This confirms that mutations in the region between the signal cleavage peptide and the cysteine cross-linked region, corresponding to amino acids 20-29 are well tolerated. In particular, constructs beginning at position 20, 21, 22, 23 and 24 will retain activity, and constructs beginning at positions 25, 26, 27, 28 and 29 are also expected to retain activity.
 Taken together, an active portion of ActRIIB comprises amino acids 29-109 of SEQ ID NO:2, and constructs may, for example, begin at a residue corresponding to amino acids 20-29 and end at a position corresponding to amino acids 109-134. Other examples include constructs that begin at a position from 20-29 or 21-29 and end at a position from 119-134, 119-133 or 129-134, 129-133. Other examples include constructs that begin at a position from 20-24 (or 21-24, or 22-25) and end at a position from 109-134 (or 109-133), 119-134 (or 119-133) or 129-134 (or 129-133). Variants within these ranges are also contemplated, particularly those having at least 80%, 85%, 90%, 95% or 99% identity to the corresponding portion of SEQ ID NO:2.
 The disclosure includes the results of an analysis of composite ActRIIB structures demonstrating that the ligand binding pocket is defined by residues Y31, N33, N35, L38 through T41, E47, E50, Q53 through K55, L57, H58, Y60, S62, K74, W78 through N83, Y85, R87, A92, and E94 through F101. At these positions, it is expected that conservative mutations will be tolerated, although a K74A mutation is well-tolerated, as are R40A, K55A, F82A and mutations at position L79. R40 is a K in Xenopus , indicating that basic amino acids at this position will be tolerated. Q53 is R in bovine ActRIIB and K in Xenopus ActRIIB, and therefore amino acids including R, K, Q, N and H will be tolerated at this position. Thus, a general formula for an active ActRIIB variant protein is one that comprises amino acids 29-109, but optionally beginning at a position ranging from 20-24 or 22-25 and ending at a position ranging from 129-134, and comprising no more than 1, 2, 5, 10 or 15 conservative amino acid changes in the ligand binding pocket, and zero, one or more non-conservative alterations at positions 40, 53, 55, 74, 79 and/or 82 in the ligand binding pocket. Such a protein may retain greater than 80%, 90%, 95% or 99% sequence identity to the sequence of amino acids 29-109 of SEQ ID NO:2. Sites outside the binding pocket, at which variability may be particularly well tolerated, include the amino and carboxy termini of the extracellular domain (as noted above), and positions 42-46 and 65-73. An asparagine to alanine alteration at position 65 (N65A) actually improves ligand binding in the A64 background, and is thus expected to have no detrimental effect on ligand binding in the R64 background. This change probably eliminates glycosylation at N65 in the A64 background, thus demonstrating that a significant change in this region is likely to be tolerated. While an R64A change is poorly tolerated, R64K is well-tolerated, and thus another basic residue, such as H may be tolerated at position 64.
 ActRIIB is well-conserved across nearly all vertebrates, with large stretches of the extracellular domain conserved completely. Many of the ligands that bind to ActRIIB are also highly conserved. Accordingly, comparisons of ActRIIB sequences from various vertebrate organisms provide insights into residues that may be altered. Therefore, an active, human ActRIIB variant may include one or more amino acids at corresponding positions from the sequence of another vertebrate ActRIIB, or may include a residue that is similar to that in the human or other vertebrate sequence. The following examples illustrate this approach to defining an active ActRIIB variant. L46 is a valine in Xenopus ActRIIB, and so this position may be altered, and optionally may be altered to another hydrophobic residue, such as V, I or F, or a non-polar residue such as A. E52 is a K in Xenopus , indicating that this site may be tolerant of a wide variety of changes, including polar residues, such as E, D, K, R, H, S, T, P, G, Y and probably A. T93 is a K in Xenopus , indicating that a wide structural variation is tolerated at this position, with polar residues favored, such as S, K, R, E, D, H, G, P, G and Y. F108 is a Y in Xenopus , and therefore Y or other hydrophobic group, such as I, V or L should be tolerated. E111 is K in Xenopus , indicating that charged residues will be tolerated at this position, including D, R, K and H, as well as Q and N. R112 is K in Xenopus , indicating that basic residues are tolerated at this position, including R and H. A at position 119 is relatively poorly conserved, and appears as P in rodents and V in Xenopus , thus essentially any amino acid should be tolerated at this position.
 The disclosure demonstrates that the addition of a further N-linked glycosylation site (N—X—S/T) increases the serum half-life of an ActRIIB-Fc fusion protein, relative to the ActRIIB(R64)-Fc form. By introducing an asparagine at position 24 (A24N construct), an NXT sequence is created that confers a longer half-life. Other NX(T/S) sequences are found at 42-44 (NQS) and 65-67 (NSS), although the latter may not be efficiently glycosylated with the R at position 64. N—X—S/T sequences may be generally introduced at positions outside the ligand binding pocket. Particularly suitable sites for the introduction of non-endogenous N—X—S/T sequences include amino acids 20-29, 20-24, 22-25, 109-134, 120-134 or 129-134. N—X—S/T sequences may also be introduced into the linker between the ActRIIB sequence and the Fc or other fusion component. Such a site may be introduced with minimal effort by introducing an N in the correct position with respect to a pre-existing S or T, or by introducing an S or T at a position corresponding to a pre-existing N. Thus, desirable alterations that would create an N-linked glycosylation site are: A24N, R64N, S67N (possibly combined with an N65A alteration), E106N, R112N, G120N, E123N, P129N, A132N, R112S and R112T. Any S that is predicted to be glycosylated may be altered to a T without creating an immunogenic site, because of the protection afforded by the glycosylation. Likewise, any T that is predicted to be glycosylated may be altered to an S. Thus the alterations S67T and S44T are contemplated. Likewise, in an A24N variant, an S26T alteration may be used. Accordingly, an ActRIIB variant may include one or more additional, non-endogenous N-linked glycosylation consensus sequences.
 Position L79 may be altered to confer altered activin—myostatin (GDF-11) binding properties. L79A or L79P reduces GDF-11 binding to a greater extent than activin binding. L79E or L79D retains GDF-11 binding. Remarkably, the L79E and L79D variants have greatly reduced activin binding. In vivo experiments indicate that these non-activin receptors retain significant ability to increase muscle mass but show decreased effects on other tissues. These data demonstrate the desirability and feasibility for obtaining polypeptides with reduced effects on activin.
 The variations described may be combined in various ways. Additionally, the results of mutagenesis program described herein indicate that there are amino acid positions in ActRIIb that are often beneficial to conserve. These include position 64 (basic amino acid), position 80 (acidic or hydrophobic amino acid), position 78 (hydrophobic, and particularly tryptophan), position 37 (acidic, and particularly aspartic or glutamic acid), position 56 (basic amino acid), position 60 (hydrophobic amino acid, particularly phenylalanine or tyrosine). Thus, in each of the variants disclosed herein, the disclosure provides a framework of amino acids that may be conserved. Other positions that may be desirable to conserve are as follows: position 52 (acidic amino acid), position 55 (basic amino acid), position 81 (acidic), 98 (polar or charged, particularly E, D, R or K).
 In certain embodiments, isolated fragments of the ActRIIB polypeptides can be obtained by screening polypeptides recombinantly produced from the corresponding fragment of the nucleic acid encoding an ActRIIB polypeptide (e.g., SEQ ID NOs: 3, 4 and 15). In addition, fragments can be chemically synthesized using techniques known in the art such as conventional Merrifield solid phase f-Moc or t-Boc chemistry. The fragments can be produced (recombinantly or by chemical synthesis) and tested to identify those peptidyl fragments that can function, for example, as antagonists (inhibitors) or agonists (activators) of an ActRIIB protein or an ActRIIB ligand.
 In certain embodiments, a functional variant of the ActRIIB polypeptides has an amino acid sequence that is at least 75% identical to an amino acid sequence selected from SEQ ID NOs: 3, 4 and 10. In certain cases, the functional variant has an amino acid sequence at least 80%, 85%, 90%, 95%, 97%, 98%, 99% or 100% identical to an amino acid sequence selected from SEQ ID NOs: 1, 2, 5, 6, 12 and 14.
 In certain embodiments, the present invention contemplates making functional variants by modifying the structure of an ActRIIB polypeptide for such purposes as enhancing therapeutic efficacy, or stability (e.g., ex vivo shelf life and resistance to proteolytic degradation in vivo). Modified ActRIIB polypeptides can also be produced, for instance, by amino acid substitution, deletion, or addition. For instance, it is reasonable to expect that an isolated replacement of a leucine with an isoleucine or valine, an aspartate with a glutamate, a threonine with a serine, or a similar replacement of an amino acid with a structurally related amino acid (e.g., conservative mutations) will not have a major effect on the biological activity of the resulting molecule. Conservative replacements are those that take place within a family of amino acids that are related in their side chains. Whether a change in the amino acid sequence of an ActRIIB polypeptide results in a functional homolog can be readily determined by assessing the ability of the variant ActRIIB polypeptide to produce a response in cells in a fashion similar to the wild-type ActRIIB polypeptide, or to bind to one or more ligands, such as activin, BMP7, GDF3, BMP2, BMP4, GDF-11, or myostatin in a fashion similar to wild type.
 In certain embodiments, the present invention contemplates specific mutations of the ActRIIB polypeptides so as to alter the glycosylation of the polypeptide. Exemplary glycosylation sites in ActRIIB polypeptides are illustrated in SEQ ID NO: 2. Such mutations may be selected so as to introduce or eliminate one or more glycosylation sites, such as O-linked or N-linked glycosylation sites. Asparagine-linked glycosylation recognition sites generally comprise a tripeptide sequence, asparagine-X-threonine/serine (where “X” is any amino acid) which is specifically recognized by appropriate cellular glycosylation enzymes. The alteration may also be made by the addition of, or substitution by, one or more serine or threonine residues to the sequence of the wild-type ActRIIB polypeptide (for O-linked glycosylation sites). A variety of amino acid substitutions or deletions at one or both of the first or third amino acid positions of a glycosylation recognition site (and/or amino acid deletion at the second position) results in non-glycosylation at the modified tripeptide sequence. Another means of increasing the number of carbohydrate moieties on an ActRIIB polypeptide is by chemical or enzymatic coupling of glycosides to the ActRIIB polypeptide. Depending on the coupling mode used, the sugar(s) may be attached to (a) arginine and histidine; (b) free carboxyl groups; (c) free sulfhydryl groups such as those of cysteine; (d) free hydroxyl groups such as those of serine, threonine, or hydroxyproline; (e) aromatic residues such as those of phenylalanine, tyrosine, or tryptophan; or (f) the amide group of glutamine. These methods are described in WO 87/05330 published Sep. 11, 1987, and in Aplin and Wriston (1981) CRC Crit. Rev. Biochem., pp. 259-306, incorporated by reference herein. Removal of one or more carbohydrate moieties present on an ActRIIB polypeptide may be accomplished chemically and/or enzymatically. Chemical deglycosylation may involve, for example, exposure of the ActRIIB polypeptide to the compound trifluoromethanesulfonic acid, or an equivalent compound. This treatment results in the cleavage of most or all sugars except the linking sugar (N-acetylglucosamine or N-acetylgalactosamine), while leaving the amino acid sequence intact. Chemical deglycosylation is further described by Hakimuddin et al. (1987) Arch. Biochem. Biophys. 259:52 and by Edge et al. (1981) Anal. Biochem. 118:131. Enzymatic cleavage of carbohydrate moieties on ActRIIB polypeptides can be achieved by the use of a variety of endo- and exo-glycosidases as described by Thotakura et al. (1987) Meth. Enzymol. 138:350. The sequence of an ActRIIB polypeptide may be adjusted, as appropriate, depending on the type of expression system used, as mammalian, yeast, insect and plant cells may all introduce differing glycosylation patterns that can be affected by the amino acid sequence of the peptide. In general, ActRIIB proteins for use in humans will be expressed in a mammalian cell line that provides proper glycosylation, such as HEK293 or CHO cell lines, although other mammalian expression cell lines are expected to be useful as well.
 This disclosure further contemplates a method of generating variants, particularly sets of combinatorial variants of an ActRIIB polypeptide, including, optionally, truncation variants; pools of combinatorial mutants are especially useful for identifying functional variant sequences. The purpose of screening such combinatorial libraries may be to generate, for example, ActRIIB polypeptide variants which have altered properties, such as altered pharmacokinetics, or altered ligand binding. A variety of screening assays are provided below, and such assays may be used to evaluate variants. For example, an ActRIIB polypeptide variant may be screened for ability to bind to an ActRIIB polypeptide, to prevent binding of an ActRIIB ligand to an ActRIIB polypeptide.
 The activity of an ActRIIB polypeptide or its variants may also be tested in a cell-based or in vivo assay. For example, the effect of an ActRIIB polypeptide variant on the expression of genes involved in adipocyte differentiation or function may be assessed (e.g., adiponectin). This may, as needed, be performed in the presence of one or more recombinant ActRIIB ligand proteins, and cells may be transfected so as to produce an ActRIIB polypeptide and/or variants thereof, and optionally, an ActRIIB ligand. Likewise, an ActRIIB polypeptide may be administered to a mouse or other animal, and one or more properties of hepatocytes or adipocytes (such as adiponectin gene expression) may be assessed. Similarly, the activity of an ActRIIB polypeptide or its variants may be tested in fat cells, muscle cells, bone cells, and neuronal cells for any effect on growth of these cells, for example, by the assays as described below. Such assays are well known and routine in the art. A SMAD-responsive reporter gene may be used in such cell lines to monitor effects on downstream signaling.
 Combinatorially-derived variants can be generated which have a selective potency relative to a naturally occurring ActRIIB polypeptide. Such variant proteins, when expressed from recombinant DNA constructs, can be used in gene therapy protocols. Likewise, mutagenesis can give rise to variants which have intracellular half-lives dramatically different than the corresponding a wild-type ActRIIB polypeptide. For example, the altered protein can be rendered either more stable or less stable to proteolytic degradation or other processes which result in destruction of, or otherwise inactivation of a native ActRIIB polypeptide. Such variants, and the genes which encode them, can be utilized to alter ActRIIB polypeptide levels by modulating the half-life of the ActRIIB polypeptides. For instance, a short half-life can give rise to more transient biological effects and, when part of an inducible expression system, can allow tighter control of recombinant ActRIIB polypeptide levels within the cell.
 In certain embodiments, the ActRIIB polypeptides of the invention may further comprise post-translational modifications in addition to any that are naturally present in the ActRIIB polypeptides. Such modifications include, but are not limited to, acetylation, carboxylation, glycosylation, phosphorylation, lipidation, and acylation. As a result, the modified ActRIIB polypeptides may contain non-amino acid elements, such as polyethylene glycols, lipids, poly- or mono-saccharide, and phosphates. Effects of such non-amino acid elements on the functionality of a ActRIIB polypeptide may be tested as described herein for other ActRIIB polypeptide variants. When an ActRIIB polypeptide is produced in cells by cleaving a nascent form of the ActRIIB polypeptide, post-translational processing may also be important for correct folding and/or function of the protein. Different cells (such as CHO, HeLa, MDCK, 293, WI38, NIH-3T3 or HEK293) have specific cellular machinery and characteristic mechanisms for such post-translational activities and may be chosen to ensure the correct modification and processing of the ActRIIB polypeptides.
 In certain aspects, functional variants or modified forms of the ActRIIB polypeptides include fusion proteins having at least a portion of the ActRIIB polypeptides and one or more fusion domains. Well known examples of such fusion domains include, but are not limited to, polyhistidine, Glu-Glu, glutathione S transferase (GST), thioredoxin, protein A, protein G, an immunoglobulin heavy chain constant region (e.g., an Fc), maltose binding protein (MBP), or human serum albumin A fusion domain may be selected so as to confer a desired property. For example, some fusion domains are particularly useful for isolation of the fusion proteins by affinity chromatography. For the purpose of affinity purification, relevant matrices for affinity chromatography, such as glutathione-, amylase-, and nickel- or cobalt-conjugated resins are used. Many of such matrices are available in “kit” form, such as the Pharmacia GST purification system and the QIAexpress™ system (Qiagen) useful with (HIS 6 ) fusion partners. As another example, a fusion domain may be selected so as to facilitate detection of the ActRIIB polypeptides. Examples of such detection domains include the various fluorescent proteins (e.g., GFP) as well as “epitope tags,” which are usually short peptide sequences for which a specific antibody is available. Well known epitope tags for which specific monoclonal antibodies are readily available include FLAG, influenza virus haemagglutinin (HA), and c-myc tags. In some cases, the fusion domains have a protease cleavage site, such as for Factor Xa or Thrombin, which allows the relevant protease to partially digest the fusion proteins and thereby liberate the recombinant proteins therefrom. The liberated proteins can then be isolated from the fusion domain by subsequent chromatographic separation. In certain preferred embodiments, an ActRIIB polypeptide is fused with a domain that stabilizes the ActRIIB polypeptide in vivo (a “stabilizer” domain) By “stabilizing” is meant anything that increases serum half life, regardless of whether this is because of decreased destruction, decreased clearance by the kidney, or other pharmacokinetic effect. Fusions with the Fc portion of an immunoglobulin are known to confer desirable pharmacokinetic properties on a wide range of proteins. Likewise, fusions to human serum albumin can confer desirable properties. Other types of fusion domains that may be selected include multimerizing (e.g., dimerizing, tetramerizing) domains and functional domains (that confer an additional biological function, such as further stimulation of muscle growth).
 As a specific example, the present invention provides a fusion protein as a BMP7 antagonist which comprises an extracellular (e.g., BMP7-binding) domain fused to an Fc domain (e.g., SEQ ID NO: 13).
THTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVVV D (A)VSHE
KC K (A)VSNKALPVPIEKTISKAKGQPREPQVYTLPPSREEMTKNQVSLT
WQQGNVFSCSVMHEALH N (A)HYTQKSLSLSPGK*
 Optionally, the Fc domain has one or more mutations at residues such as Asp-265, lysine 322, and Asn-434. In certain cases, the mutant Fc domain having one or more of these mutations (e.g., Asp-265 mutation) has reduced ability of binding to the Fcγ receptor relative to a wildtype Fc domain. In other cases, the mutant Fc domain having one or more of these mutations (e.g., Asn-434 mutation) has increased ability of binding to the MHC class I-related Fc-receptor (FcRN) relative to a wildtype Fc domain.
 It is understood that different elements of the fusion proteins may be arranged in any manner that is consistent with the desired functionality. For example, an ActRIIB polypeptide may be placed C-terminal to a heterologous domain, or, alternatively, a heterologous domain may be placed C-terminal to an ActRIIB polypeptide. The ActRIIB polypeptide domain and the heterologous domain need not be adjacent in a fusion protein, and additional domains or amino acid sequences may be included C- or N-terminal to either domain or between the domains.
 In certain embodiments, the ActRIIB polypeptides of the present invention contain one or more modifications that are capable of stabilizing the ActRIIB polypeptides. For example, such modifications enhance the in vitro half life of the ActRIIB polypeptides, enhance circulatory half life of the ActRIIB polypeptides or reducing proteolytic degradation of the ActRIIB polypeptides. Such stabilizing modifications include, but are not limited to, fusion proteins (including, for example, fusion proteins comprising an ActRIIB polypeptide and a stabilizer domain), modifications of a glycosylation site (including, for example, addition of a glycosylation site to an ActRIIB polypeptide), and modifications of carbohydrate moiety (including, for example, removal of carbohydrate moieties from an ActRIIB polypeptide). In the case of fusion proteins, an ActRIIB polypeptide is fused to a stabilizer domain such as an IgG molecule (e.g., an Fc domain). As used herein, the term “stabilizer domain” not only refers to a fusion domain (e.g., Fc) as in the case of fusion proteins, but also includes nonproteinaceous modifications such as a carbohydrate moiety, or nonproteinaceous polymer, such as polyethylene glycol.
 In certain embodiments, the present invention makes available isolated and/or purified forms of the ActRIIB polypeptides, which are isolated from, or otherwise substantially free of, other proteins.
 In certain embodiments, ActRIIB polypeptides (unmodified or modified) of the invention can be produced by a variety of art-known techniques. For example, such ActRIIB polypeptides can be synthesized using standard protein chemistry techniques such as those described in Bodansky, M. Principles of Peptide Synthesis, Springer Verlag, Berlin (1993) and Grant G. A. (ed.), Synthetic Peptides: A User's Guide, W. H. Freeman and Company, New York (1992). In addition, automated peptide synthesizers are commercially available (e.g., Advanced ChemTech Model 396; Milligen/Biosearch 9600). Alternatively, the ActRIIB polypeptides, fragments or variants thereof may be recombinantly produced using various expression systems (e.g., E. coli , Chinese Hamster Ovary cells, COS cells, baculovirus) as is well known in the art (also see below). In a further embodiment, the modified or unmodified ActRIIB polypeptides may be produced by digestion of naturally occurring or recombinantly produced full-length ActRIIB polypeptides by using, for example, a protease, e.g., trypsin, thermolysin, chymotrypsin, pepsin, or paired basic amino acid converting enzyme (PACE). Computer analysis (using a commercially available software, e.g., MacVector, Omega, PCGene, Molecular Simulation, Inc.) can be used to identify proteolytic cleavage sites. Alternatively, such ActRIIB polypeptides may be produced from naturally occurring or recombinantly produced full-length ActRIIB polypeptides such as standard techniques known in the art, such as by chemical cleavage (e.g., cyanogen bromide, hydroxylamine).
3. Nucleic Acids Encoding ActRIIB Polypeptides
 In certain aspects, the invention provides isolated and/or recombinant nucleic acids encoding any of the ActRIIB polypeptides (e.g., soluble ActRIIB polypeptides), including any of the variants disclosed herein. For example, the following sequence encodes a naturally occurring human ActRIIB precursor polypeptide (SEQ ID NO: 4) (nucleotides 5-1543 of NM — 001106, 1539 bp):
 The following sequence encodes a human soluble (extracellular) ActRIIB polypeptide (SEQ ID NO: 3) (348 bp).
 The subject nucleic acids may be single-stranded or double stranded. Such nucleic acids may be DNA or RNA molecules. These nucleic acids are may be used, for example, in methods for making ActRIIB polypeptides or as direct therapeutic agents (e.g., in a gene therapy approach).
 In certain aspects, the subject nucleic acids encoding ActRIIB polypeptides are further understood to include nucleic acids that are variants of SEQ ID NO: 3. Variant nucleotide sequences include sequences that differ by one or more nucleotide substitutions, additions or deletions, such as allelic variants; and will, therefore, include coding sequences that differ from the nucleotide sequence of the coding sequence designated in SEQ ID NO: 4.
 In certain embodiments, the invention provides isolated or recombinant nucleic acid sequences that are at least 80%, 85%, 90%, 95%, 97%, 98%, 99% or 100% identical to SEQ ID NO: 3. One of ordinary skill in the art will appreciate that nucleic acid sequences complementary to SEQ ID NO: 3, and variants of SEQ ID NO: 3 are also within the scope of this invention. In further embodiments, the nucleic acid sequences of the invention can be isolated, recombinant, and/or fused with a heterologous nucleotide sequence, or in a DNA library.
 In other embodiments, nucleic acids of the invention also include nucleotide sequences that hybridize under highly stringent conditions to the nucleotide sequence designated in SEQ ID NO: 3, complement sequence of SEQ ID NO: 3, or fragments thereof. As discussed above, one of ordinary skill in the art will understand readily that appropriate stringency conditions which promote DNA hybridization can be varied. One of ordinary skill in the art will understand readily that appropriate stringency conditions which promote DNA hybridization can be varied. For example, one could perform the hybridization at 6.0×sodium chloride/sodium citrate (SSC) at about 45° C., followed by a wash of 2.0×SSC at 50° C. For example, the salt concentration in the wash step can be selected from a low stringency of about 2.0×SSC at 50° C. to a high stringency of about 0.2×SSC at 50° C. In addition, the temperature in the wash step can be increased from low stringency conditions at room temperature, about 22° C., to high stringency conditions at about 65° C. Both temperature and salt may be varied, or temperature or salt concentration may be held constant while the other variable is changed. In one embodiment, the invention provides nucleic acids which hybridize under low stringency conditions of 6×SSC at room temperature followed by a wash at 2×SSC at room temperature.
 Isolated nucleic acids which differ from the nucleic acids as set forth in SEQ ID NO: 3 due to degeneracy in the genetic code are also within the scope of the invention. For example, a number of amino acids are designated by more than one triplet. Codons that specify the same amino acid, or synonyms (for example, CAU and CAC are synonyms for histidine) may result in “silent” mutations which do not affect the amino acid sequence of the protein. However, it is expected that DNA sequence polymorphisms that do lead to changes in the amino acid sequences of the subject proteins will exist among mammalian cells. One skilled in the art will appreciate that these variations in one or more nucleotides (up to about 3-5% of the nucleotides) of the nucleic acids encoding a particular protein may exist among individuals of a given species due to natural allelic variation. Any and all such nucleotide variations and resulting amino acid polymorphisms are within the scope of this invention.
 In certain embodiments, the recombinant nucleic acids of the invention may be operably linked to one or more regulatory nucleotide sequences in an expression construct. Regulatory nucleotide sequences will generally be appropriate to the host cell used for expression. Numerous types of appropriate expression vectors and suitable regulatory sequences are known in the art for a variety of host cells. Typically, said one or more regulatory nucleotide sequences may include, but are not limited to, promoter sequences, leader or signal sequences, ribosomal binding sites, transcriptional start and termination sequences, translational start and termination sequences, and enhancer or activator sequences. Constitutive or inducible promoters as known in the art are contemplated by the invention. The promoters may be either naturally occurring promoters, or hybrid promoters that combine elements of more than one promoter. An expression construct may be present in a cell on an episome, such as a plasmid, or the expression construct may be inserted in a chromosome. In a preferred embodiment, the expression vector contains a selectable marker gene to allow the selection of transformed host cells. Selectable marker genes are well known in the art and will vary with the host cell used.
 In certain aspects of the invention, the subject nucleic acid is provided in an expression vector comprising a nucleotide sequence encoding an ActRIIB polypeptide and operably linked to at least one regulatory sequence. Regulatory sequences are art-recognized and are selected to direct expression of the ActRIIB polypeptide. Accordingly, the term regulatory sequence includes promoters, enhancers, and other expression control elements. Exemplary regulatory sequences are described in Goeddel; Gene Expression Technology: Methods in Enzymology , Academic Press, San Diego, Calif. (1990). For instance, any of a wide variety of expression control sequences that control the expression of a DNA sequence when operatively linked to it may be used in these vectors to express DNA sequences encoding an ActRIIB polypeptide. Such useful expression control sequences, include, for example, the early and late promoters of SV40, tet promoter, adenovirus or cytomegalovirus immediate early promoter, RSV promoters, the lac system, the trp system, the TAC or TRC system, T7 promoter whose expression is directed by T7 RNA polymerase, the major operator and promoter regions of phage lambda, the control regions for fd coat protein, the promoter for 3-phosphoglycerate kinase or other glycolytic enzymes, the promoters of acid phosphatase, e.g., Pho5, the promoters of the yeast α-mating factors, the polyhedron promoter of the baculovirus system and other sequences known to control the expression of genes of prokaryotic or eukaryotic cells or their viruses, and various combinations thereof. It should be understood that the design of the expression vector may depend on such factors as the choice of the host cell to be transformed and/or the type of protein desired to be expressed. Moreover, the vector's copy number, the ability to control that copy number and the expression of any other protein encoded by the vector, such as antibiotic markers, should also be considered.
 A recombinant nucleic acid of the invention can be produced by ligating the cloned gene, or a portion thereof, into a vector suitable for expression in either prokaryotic cells, eukaryotic cells (yeast, avian, insect or mammalian), or both. Expression vehicles for production of a recombinant ActRIIB polypeptide include plasmids and other vectors. For instance, suitable vectors include plasmids of the types: pBR322-derived plasmids, pEMBL-derived plasmids, pEX-derived plasmids, pBTac-derived plasmids and pUC-derived plasmids for expression in prokaryotic cells, such as E. coli.
 Some mammalian expression vectors contain both prokaryotic sequences to facilitate the propagation of the vector in bacteria, and one or more eukaryotic transcription units that are expressed in eukaryotic cells. The pcDNAI/amp, pcDNAI/neo, pRc/CMV, pSV2gpt, pSV2neo, pSV2-dhfr, pTk2, pRSVneo, pMSG, pSVT7, pko-neo and pHyg derived vectors are examples of mammalian expression vectors suitable for transfection of eukaryotic cells. Some of these vectors are modified with sequences from bacterial plasmids, such as pBR322, to facilitate replication and drug resistance selection in both prokaryotic and eukaryotic cells. Alternatively, derivatives of viruses such as the bovine papilloma virus (BPV-1), or Epstein-Barr virus (pHEBo, pREP-derived and p205) can be used for transient expression of proteins in eukaryotic cells. Examples of other viral (including retroviral) expression systems can be found below in the description of gene therapy delivery systems. The various methods employed in the preparation of the plasmids and in transformation of host organisms are well known in the art. For other suitable expression systems for both prokaryotic and eukaryotic cells, as well as general recombinant procedures, see Molecular Cloning A Laboratory Manual , 2nd Ed., ed. by Sambrook, Fritsch and Maniatis (Cold Spring Harbor Laboratory Press, 1989) Chapters 16 and 17. In some instances, it may be desirable to express the recombinant polypeptides by the use of a baculovirus expression system. Examples of such baculovirus expression systems include pVL-derived vectors (such as pVL1392, pVL1393 and pVL941), pAcUW-derived vectors (such as pAcUW1), and pBlueBac-derived vectors (such as the β-gal containing pBlueBac III).
 In a preferred embodiment, a vector will be designed for production of the subject ActRIIB polypeptides in CHO cells, such as a Pcmv-Script vector (Stratagene, La Jolla, Calif.), pcDNA4 vectors (Invitrogen, Carlsbad, Calif.) and pCI-neo vectors (Promega, Madison, Wisc.). As will be apparent, the subject gene constructs can be used to cause expression of the subject ActRIIB polypeptides in cells propagated in culture, e.g., to produce proteins, including fusion proteins or variant proteins, for purification.
 This invention also pertains to a host cell transfected with a recombinant gene including a coding sequence (e.g., SEQ ID NO: 4) for one or more of the subject ActRIIB polypeptide. The host cell may be any prokaryotic or eukaryotic cell. For example, an ActRIIB polypeptide of the invention may be expressed in bacterial cells such as E. coli , insect cells (e.g., using a baculovirus expression system), yeast, or mammalian cells. Other suitable host cells are known to those skilled in the art.
 Accordingly, the present invention further pertains to methods of producing the subject ActRIIB polypeptides. For example, a host cell transfected with an expression vector encoding an ActRIIB polypeptide can be cultured under appropriate conditions to allow expression of the ActRIIB polypeptide to occur. The ActRIIB polypeptide may be secreted and isolated from a mixture of cells and medium containing the ActRIIB polypeptide. Alternatively, the ActRIIB polypeptide may be retained cytoplasmically or in a membrane fraction and the cells harvested, lysed and the protein isolated. A cell culture includes host cells, media and other byproducts. Suitable media for cell culture are well known in the art. The subject ActRIIB polypeptides can be isolated from cell culture medium, host cells, or both, using techniques known in the art for purifying proteins, including ion-exchange chromatography, gel filtration chromatography, ultrafiltration, electrophoresis, and immunoaffinity purification with antibodies specific for particular epitopes of the ActRIIB polypeptides. In a preferred embodiment, the ActRIIB polypeptide is a fusion protein containing a domain which facilitates its purification.
 In another embodiment, a fusion gene coding for a purification leader sequence, such as a poly-(His)/enterokinase cleavage site sequence at the N-terminus of the desired portion of the recombinant ActRIIB polypeptide, can allow purification of the expressed fusion protein by affinity chromatography using a Ni 2+ metal resin. The purification leader sequence can then be subsequently removed by treatment with enterokinase to provide the purified ActRIIB polypeptide (e.g., see Hochuli et al., (1987) J. Chromatography 411:177; and Janknecht et al., PNAS USA 88:8972).
 Techniques for making fusion genes are well known. Essentially, the joining of various DNA fragments coding for different polypeptide sequences is performed in accordance with conventional techniques, employing blunt-ended or stagger-ended termini for ligation, restriction enzyme digestion to provide for appropriate termini, filling-in of cohesive ends as appropriate, alkaline phosphatase treatment to avoid undesirable joining, and enzymatic ligation. In another embodiment, the fusion gene can be synthesized by conventional techniques including automated DNA synthesizers. Alternatively, PCR amplification of gene fragments can be carried out using anchor primers which give rise to complementary overhangs between two consecutive gene fragments which can subsequently be annealed to generate a chimeric gene sequence (see, for example, Current Protocols in Molecular Biology , eds. Ausubel et al., John Wiley & Sons: 1992).
4. Antibodies and Other Antagonists
 Another aspect of the invention pertains to antibodies and other antagonists, including proteins that bind to the targets disclosed herein and nucleic acids that inhibit expression of targets disclosed herein. An antibody that is specifically reactive with an ActRIIB polypeptide (e.g., a soluble ActRIIB polypeptide) and which binds competitively with the ActRIIB polypeptide may be used as an antagonist of ActRIIB polypeptide activities. For example, by using immunogens derived from an ActRIIB polypeptide, anti-protein/anti-peptide antisera or monoclonal antibodies can be made by standard protocols (see, for example, Antibodies: A Laboratory Manual ed. by Harlow and Lane (Cold Spring Harbor Press: 1988)). A mammal, such as a mouse, a hamster or rabbit can be immunized with an immunogenic form of the ActRIIB polypeptide or ligand, an antigenic fragment which is capable of eliciting an antibody response, or a fusion protein. Techniques for conferring immunogenicity on a protein or peptide include conjugation to carriers or other techniques well known in the art. An immunogenic portion of an ActRIIB polypeptide or ligand can be administered in the presence of adjuvant. The progress of immunization can be monitored by detection of antibody titers in plasma or serum. Standard ELISA or other immunoassays can be used with the immunogen as antigen to assess the levels of antibodies.
 Following immunization of an animal with an antigenic preparation of an ActRIIB polypeptide or ligand, antisera can be obtained and, if desired, polyclonal antibodies can be isolated from the serum. To produce monoclonal antibodies, antibody-producing cells (lymphocytes) can be harvested from an immunized animal and fused by standard somatic cell fusion procedures with immortalizing cells such as myeloma cells to yield hybridoma cells. Such techniques are well known in the art, and include, for example, the hybridoma technique (originally developed by Kohler and Milstein, 1975, Nature, 256: 495-497), the human B cell hybridoma technique (Kozbar et al., 1983, Immunology Today, 4:72), and the EBV-hybridoma technique to produce human monoclonal antibodies (Cole et al., 1985, Monoclonal Antibodies and Cancer Therapy, Alan R. Liss, Inc. pp. 77-96). Hybridoma cells can be screened immunochemically for production of antibodies specifically reactive with an ActRIIB polypeptide and monoclonal antibodies isolated from a culture comprising such hybridoma cells.
 The term “antibody” as used herein is intended to include fragments thereof which are also specifically reactive with a subject ActRIIB polypeptide or ligand. Antibodies can be fragmented using conventional techniques and the fragments screened for utility in the same manner as described above for whole antibodies. For example, F(ab) 2 fragments can be generated by treating antibody with pepsin. The resulting F(ab) 2 fragment can be treated to reduce disulfide bridges to produce Fab fragments. The antibody of the present invention is further intended to include bispecific, single-chain, and chimeric and humanized molecules having affinity for an ActRIIB polypeptide conferred by at least one CDR region of the antibody. In preferred embodiments, the antibody further comprises a label attached thereto and able to be detected (e.g., the label can be a radioisotope, fluorescent compound, enzyme or enzyme co-factor).
 In certain preferred embodiments, an antibody of the invention is a monoclonal antibody, and in certain embodiments, the invention makes available methods for generating novel antibodies. For example, a method for generating a monoclonal antibody that binds specifically to an ActRIIB polypeptide or ligand may comprise administering to a mouse an amount of an immunogenic composition comprising the ActRIIB polypeptide or ligand effective to stimulate a detectable immune response, obtaining antibody-producing cells (e.g., cells from the spleen) from the mouse and fusing the antibody-producing cells with myeloma cells to obtain antibody-producing hybridomas, and testing the antibody-producing hybridomas to identify a hybridoma that produces a monocolonal antibody that binds specifically to the ActRIIB polypeptide or ligand. Once obtained, a hybridoma can be propagated in a cell culture, optionally in culture conditions where the hybridoma-derived cells produce the monoclonal antibody that binds specifically to the ActRIIB polypeptide or ligand. The monoclonal antibody may be purified from the cell culture.
 The adjective “specifically reactive with” as used in reference to an antibody is intended to mean, as is generally understood in the art, that the antibody is sufficiently selective between the antigen of interest (e.g., an ActRIIB polypeptide) and other antigens that are not of interest that the antibody is useful for, at minimum, detecting the presence of the antigen of interest in a particular type of biological sample. In certain methods employing the antibody, such as therapeutic applications, a higher degree of specificity in binding may be desirable. Monoclonal antibodies generally have a greater tendency (as compared to polyclonal antibodies) to discriminate effectively between the desired antigens and cross-reacting polypeptides. One characteristic that influences the specificity of an antibody:antigen interaction is the affinity of the antibody for the antigen. Although the desired specificity may be reached with a range of different affinities, generally preferred antibodies will have an affinity (a dissociation constant) of about 10 −6 , 10 −7 , 10 −8 , 10 −9 or less.
 In addition, the techniques used to screen antibodies in order to identify a desirable antibody may influence the properties of the antibody obtained. For example, if an antibody is to be used for binding an antigen in solution, it may be desirable to test solution binding. A variety of different techniques are available for testing interaction between antibodies and antigens to identify particularly desirable antibodies. Such techniques include ELISAs, surface plasmon resonance binding assays (e.g., the Biacore binding assay, Bia-core AB, Uppsala, Sweden), sandwich assays (e.g., the paramagnetic bead system of IGEN International, Inc., Gaithersburg, Md.), western blots, immunoprecipitation assays, and immunohistochemistry.
 In certain aspects, the disclosure provides antibodies that bind to a soluble ActRIIB polypeptide or ligand. Such antibodies may be generated much as described above, using a soluble ActRIIB polypeptide or ligand or fragment thereof as an antigen. Antibodies of this type can be used, e.g., to detect ActRIIB polypeptides in biological samples and/or to monitor soluble ActRIIB polypeptide levels in an individual. In certain cases, an antibody that specifically binds to a soluble ActRIIB polypeptide or ligand can be used to modulate activity of an ActRIIB polypeptide and/or an ActRIIB ligand, thereby treating fatty liver disease.
 Certain ligands, such as myostatin and GDF3 may be inhibited by using a polypeptide comprising a binding portion of the respective propeptide, or a variant thereof. Such propeptides may be prepared as fusion proteins, including Fc fusion proteins. Examples of suitable propeptides are disclosed in published patent applications WO 02/085306 and WO 06/002387.
 Additionally, other binding proteins, such as the so-called “traps” (e.g., follistatin, FLRG, FSTL, Cerberus and Coco), soluble type I receptors, e.g., ALK-7 may be used. Examples of such polypeptides may be found in published patent applications WO 05/115439, WO 08/109,779, WO 08/067,480, WO 07/109,686, WO 05/100563, and WO 05/025601.
 Nucleic acids, such as antisense or RNAi probes (which may include both naturally and non-naturally occurring nucleotides) may be used to inhibit expression of ActRIIB or any of the ligands discussed herein.
5. Screening Assays
 In certain aspects, the present invention relates to the use of the subject ActRIIB polypeptides (e.g., soluble ActRIIB polypeptides) to identify compounds (agents) which are agonist or antagonists of the ActRIIB polypeptides. Compounds identified through this screening can be tested in tissues such as fat, muscle, bone, cartilage, and/or neurons, to assess their ability to modulate tissue growth in vitro. Optionally, these compounds can further be tested in animal models to assess their ability to modulate tissue growth in vivo.
 There are numerous approaches to screening for therapeutic agents for modulating tissue growth by targeting the ActRIIB polypeptides. In certain embodiments, high-throughput screening of compounds can be carried out to identify agents that perturb ActRIIB-mediated effects on growth of fat, muscle, bone, cartilage, and/or neurons. In certain embodiments, the assay is carried out to screen and identify compounds that specifically inhibit or reduce binding of an ActRIIB polypeptide to its binding partner, such as an ActRIIB ligand (e.g., activin, BMP7, Nodal, GDF8, or GDF11). Alternatively, the assay can be used to identify compounds that enhance binding of an ActRIIB polypeptide to its binding protein such as an ActRIIB ligand. In a further embodiment, the compounds can be identified by their ability to interact with an ActRIIB polypeptide.
 A variety of assay formats will suffice and, in light of the present disclosure, those not expressly described herein will nevertheless be comprehended by one of ordinary skill in the art. As described herein, the test compounds (agents) of the invention may be created by any combinatorial chemical method. Alternatively, the subject compounds may be naturally occurring biomolecules synthesized in vivo or in vitro. Compounds (agents) to be tested for their ability to act as modulators of tissue growth can be produced, for example, by bacteria, yeast, plants or other organisms (e.g., natural products), produced chemically (e.g., small molecules, including peptidomimetics), or produced recombinantly. Test compounds contemplated by the present invention include non-peptidyl organic molecules, peptides, polypeptides, peptidomimetics, sugars, hormones, and nucleic acid molecules. In a specific embodiment, the test agent is a small organic molecule having a molecular weight of less than about 2,000 daltons.
 The test compounds of the invention can be provided as single, discrete entities, or provided in libraries of greater complexity, such as made by combinatorial chemistry. These libraries can comprise, for example, alcohols, alkyl halides, amines, amides, esters, aldehydes, ethers and other classes of organic compounds. Presentation of test compounds to the test system can be in either an isolated form or as mixtures of compounds, especially in initial screening steps. Optionally, the compounds may be optionally derivatized with other compounds and have derivatizing groups that facilitate isolation of the compounds. Non-limiting examples of derivatizing groups include biotin, fluorescein, digoxygenin, green fluorescent protein, isotopes, polyhistidine, magnetic beads, glutathione S transferase (GST), photoactivatible crosslinkers or any combinations thereof.
 In many drug screening programs which test libraries of compounds and natural extracts, high throughput assays are desirable in order to maximize the number of compounds surveyed in a given period of time. Assays which are performed in cell-free systems, such as may be derived with purified or semi-purified proteins, are often preferred as “primary” screens in that they can be generated to permit rapid development and relatively easy detection of an alteration in a molecular target which is mediated by a test compound. Moreover, the effects of cellular toxicity or bioavailability of the test compound can be generally ignored in the in vitro system, the assay instead being focused primarily on the effect of the drug on the molecular target as may be manifest in an alteration of binding affinity between an ActRIIB polypeptide and its binding protein (e.g., an ActRIIB ligand).
 Merely to illustrate, in an exemplary screening assay of the present invention, the compound of interest is contacted with an isolated and purified ActRIIB polypeptide which is ordinarily capable of binding to an ActRIIB ligand, as appropriate for the intention of the assay. To the mixture of the compound and ActRIIB polypeptide is then added a composition containing an ActRIIB ligand. Detection and quantification of ActRIIB/ActRIIB ligand complexes provides a means for determining the compound's efficacy at inhibiting (or potentiating) complex formation between the ActRIIB polypeptide and its binding protein. The efficacy of the compound can be assessed by generating dose response curves from data obtained using various concentrations of the test compound. Moreover, a control assay can also be performed to provide a baseline for comparison. For example, in a control assay, isolated and purified ActRIIB ligand is added to a composition containing the ActRIIB polypeptide, and the formation of ActRIIB/ActRIIB ligand complex is quantitated in the absence of the test compound. It will be understood that, in general, the order in which the reactants may be admixed can be varied, and can be admixed simultaneously. Moreover, in place of purified proteins, cellular extracts and lysates may be used to render a suitable cell-free assay system.
 Complex formation between the ActRIIB polypeptide and its binding protein may be detected by a variety of techniques. For instance, modulation of the formation of complexes can be quantitated using, for example, detectably labeled proteins such as radiolabeled (e.g., 32 P, 35 S, 14 C or 3 H), fluorescently labeled (e.g., FITC), or enzymatically labeled ActRIIB polypeptide or its binding protein, by immunoassay, or by chromatographic detection.
 In certain embodiments, the present invention contemplates the use of fluorescence polarization assays and fluorescence resonance energy transfer (FRET) assays in measuring, either directly or indirectly, the degree of interaction between an ActRIIB polypeptide and its binding protein. Further, other modes of detection, such as those based on optical waveguides (PCT Publication WO 96/26432 and U.S. Pat. No. 5,677,196), surface plasmon resonance (SPR), surface charge sensors, and surface force sensors, are compatible with many embodiments of the invention.
 Moreover, the present invention contemplates the use of an interaction trap assay, also known as the “two hybrid assay,” for identifying agents that disrupt or potentiate interaction between an ActRIIB polypeptide and its binding protein. See for example, U.S. Pat. No. 5,283,317; Zervos et al., 1993, Cell 72:223-232; Madura et al., 1993, J Biol Chem 268:12046-12054; Bartel et al., 1993, Biotechniques 14:920-924; and Iwabuchi et al., 1993, Oncogene 8:1693-1696). In a specific embodiment, the present invention contemplates the use of reverse two hybrid systems to identify compounds (e.g., small molecules or peptides) that dissociate interactions between an ActRIIB polypeptide and its binding protein. See for example, Vidal and Legrain, 1999, Nucleic Acids Res 27:919-29; Vidal and Legrain, 1999, Trends Biotechnol 17:374-81; and U.S. Pat. Nos. 5,525,490; 5,955,280; and 5,965,368.
 In certain embodiments, the subject compounds are identified by their ability to interact with an ActRIIB polypeptide of the invention. The interaction between the compound and the ActRIIB polypeptide may be covalent or non-covalent. For example, such interaction can be identified at the protein level using in vitro biochemical methods, including photo-crosslinking, radiolabeled ligand binding, and affinity chromatography (Jakoby W B et al., 1974, Methods in Enzymology 46: 1). In certain cases, the compounds may be screened in a mechanism based assay, such as an assay to detect compounds which bind to an ActRIIB polypeptide. This may include a solid phase or fluid phase binding event. Alternatively, the gene encoding an ActRIIB polypeptide can be transfected with a reporter system (e.g., β-galactosidase, luciferase, or green fluorescent protein) into a cell and screened against the library preferably by a high throughput screening or with individual members of the library. Other mechanism based binding assays may be used, for example, binding assays which detect changes in free energy. Binding assays can be performed with the target fixed to a well, bead or chip or captured by an immobilized antibody or resolved by capillary electrophoresis. The bound compounds may be detected usually using colorimetric or fluorescence or surface plasmon resonance.
 In certain aspects, the present invention provides methods and agents for regulating adiponectin signaling, insulin resistance, and obesity.
 It is understood that the screening assays of the present invention apply to not only the subject ActRIIB polypeptides and variants of the ActRIIB polypeptides, but also any test compounds including agonists and antagonist of the ActRIIB polypeptides or ActRIIB signaling. Further, these screening assays are useful for drug target verification and quality control purposes.
6. Exemplary Therapeutic Uses
 In certain embodiments, compositions (e.g., ActRIIB polypeptides) of the present invention can be used for treating or preventing a disease or condition that is associated with fatty liver disease. In certain embodiments, the present invention provides methods of treating or preventing an individual in need thereof through administering to the individual a therapeutically effective amount of an ActRIIB polypeptide as described above. These methods are particularly aimed at therapeutic and prophylactic treatments of animals, and more particularly, humans.
 As used herein, a therapeutic that “prevents” a disorder or condition refers to a compound that, in a statistical sample, reduces the occurrence of the disorder or condition in the treated sample relative to an untreated control sample, or delays the onset or reduces the severity of one or more symptoms of the disorder or condition relative to the untreated control sample. The term “treating” as used herein includes prophylaxis of the named condition or amelioration or elimination of the condition once it has been established.
 As demonstrated herein, ActRIIB-Fc promotes expression of adiponectin by white adipocytes, increases insulin sensitivity in target tissues, and inhibits hepatic steatosis. Accordingly, compositions disclosed herein may be used to treat a variety of disorders, such as nonalcoholic fatty liver disease, nonalcoholic steatohepatitis, alcoholic fatty liver disease, alcoholic steatohepatitis, hepatic fibrosis, and cirrhosis.
 In certain embodiments, compositions (e.g., soluble ActRIIB polypeptides) of the invention are used to inhibit hepatic steatosis. This effect may be coupled with an effect of promoting increased concentrations of circulating adiponectin and/or causing increased insulin sensitivity in target tissues. Blocking or antagonizing function of one or more ActRIIB ligands in vivo can effectively inhibit hepatic steatosis and thereby treat conditions in which it is unwanted. This approach is confirmed and supported by the data shown herein, whereby an ActRIIB-Fc protein was shown to increase adipocytic expression of adiponectin, increase circulating adiponectin concentrations, reduce circulating insulin concentrations consistent with increased insulin sensitivity, produce beneficial changes in the serum lipid profile, improve body composition, and prevent hepatic steatosis.
7. Pharmaceutical Compositions
 In certain embodiments, compounds (e.g., ActRIIB polypeptides) of the present invention are formulated with a pharmaceutically acceptable carrier. For example, an ActRIIB polypeptide can be administered alone or as a component of a pharmaceutical formulation (therapeutic composition). The subject compounds may be formulated for administration in any convenient way for use in human or veterinary medicine.
 In certain embodiments, the therapeutic method of the invention includes administering the composition topically, systemically, or locally as an implant or device. When administered, the therapeutic composition for use in this invention is, of course, in a pyrogen-free, physiologically acceptable form. Further, the composition may desirably be encapsulated or injected in a viscous form for delivery to a target tissue site (e.g., bone, cartilage, muscle, fat or neurons), for example, a site having a tissue damage. Topical administration may be suitable for wound healing and tissue repair. Therapeutically useful agents other than the ActRIIB polypeptides which may also optionally be included in the composition as described above, may alternatively or additionally, be administered simultaneously or sequentially with the subject compounds (e.g., ActRIIB polypeptides) in the methods of the invention.
 In certain embodiments, compositions of the present invention may include a matrix capable of delivering one or more therapeutic compounds (e.g., ActRIIB polypeptides) to a target tissue site, providing a structure for the developing tissue and optimally capable of being resorbed into the body. For example, the matrix may provide slow release of the ActRIIB polypeptides. Such matrices may be formed of materials presently in use for other implanted medical applications.
 The choice of matrix material is based on biocompatibility, biodegradability, mechanical properties, cosmetic appearance and interface properties. The particular application of the subject compositions will define the appropriate formulation. Potential matrices for the compositions may be biodegradable and chemically defined calcium sulfate, tricalciumphosphate, hydroxyapatite, polylactic acid and polyanhydrides. Other potential materials are biodegradable and biologically well defined, such as bone or dermal collagen. Further matrices are comprised of pure proteins or extracellular matrix components. Other potential matrices are non-biodegradable and chemically defined, such as sintered hydroxyapatite, bioglass, aluminates, or other ceramics. Matrices may be comprised of combinations of any of the above mentioned types of material, such as polylactic acid and hydroxyapatite or collagen and tricalciumphosphate. The bioceramics may be altered in composition, such as in calcium-aluminate-phosphate and processing to alter pore size, particle size, particle shape, and biodegradability.
 In certain embodiments, methods of the invention can be administered for orally, e.g., in the form of capsules, cachets, pills, tablets, lozenges (using a flavored basis, usually sucrose and acacia or tragacanth), powders, granules, or as a solution or a suspension in an aqueous or non-aqueous liquid, or as an oil-in-water or water-in-oil liquid emulsion, or as an elixir or syrup, or as pastilles (using an inert base, such as gelatin and glycerin, or sucrose and acacia) and/or as mouth washes and the like, each containing a predetermined amount of an agent as an active ingredient. An agent may also be administered as a bolus, electuary or paste.
 In solid dosage forms for oral administration (capsules, tablets, pills, dragees, powders, granules, and the like), one or more therapeutic compounds of the present invention may be mixed with one or more pharmaceutically acceptable carriers, such as sodium citrate or dicalcium phosphate, and/or any of the following: (1) fillers or extenders, such as starches, lactose, sucrose, glucose, mannitol, and/or silicic acid; (2) binders, such as, for example, carboxymethylcellulose, alginates, gelatin, polyvinyl pyrrolidone, sucrose, and/or acacia; (3) humectants, such as glycerol; (4) disintegrating agents, such as agar-agar, calcium carbonate, potato or tapioca starch, alginic acid, certain silicates, and sodium carbonate; (5) solution retarding agents, such as paraffin; (6) absorption accelerators, such as quaternary ammonium compounds; (7) wetting agents, such as, for example, cetyl alcohol and glycerol monostearate; (8) absorbents, such as kaolin and bentonite clay; (9) lubricants, such a talc, calcium stearate, magnesium stearate, solid polyethylene glycols, sodium lauryl sulfate, and mixtures thereof; and (10) coloring agents. In the case of capsules, tablets and pills, the pharmaceutical compositions may also comprise buffering agents. Solid compositions of a similar type may also be employed as fillers in soft and hard-filled gelatin capsules using such excipients as lactose or milk sugars, as well as high molecular weight polyethylene glycols and the like.
 Liquid dosage forms for oral administration include pharmaceutically acceptable emulsions, microemulsions, solutions, suspensions, syrups, and elixirs. In addition to the active ingredient, the liquid dosage forms may contain inert diluents commonly used in the art, such as water or other solvents, solubilizing agents and emulsifiers, such as ethyl alcohol, isopropyl alcohol, ethyl carbonate, ethyl acetate, benzyl alcohol, benzyl benzoate, propylene glycol, 1,3-butylene glycol, oils (in particular, cottonseed, groundnut, corn, germ, olive, castor, and sesame oils), glycerol, tetrahydrofuryl alcohol, polyethylene glycols and fatty acid esters of sorbitan, and mixtures thereof. Besides inert diluents, the oral compositions can also include adjuvants such as wetting agents, emulsifying and suspending agents, sweetening, flavoring, coloring, perfuming, and preservative agents.
 Suspensions, in addition to the active compounds, may contain suspending agents such as ethoxylated isostearyl alcohols, polyoxyethylene sorbitol, and sorbitan esters, microcrystalline cellulose, aluminum metahydroxide, bentonite, agar-agar and tragacanth, and mixtures thereof.
 Certain compositions disclosed herein may be administered topically, either to skin or to mucosal membranes. The topical formulations may further include one or more of the wide variety of agents known to be effective as skin or stratum corneum penetration enhancers. Examples of these are 2-pyrrolidone, N-methyl-2-pyrrolidone, dimethylacetamide, dimethylformamide, propylene glycol, methyl or isopropyl alcohol, dimethyl sulfoxide, and azone. Additional agents may further be included to make the formulation cosmetically acceptable. Examples of these are fats, waxes, oils, dyes, fragrances, preservatives, stabilizers, and surface active agents. Keratolytic agents such as those known in the art may also be included. Examples are salicylic acid and sulfur.
 Dosage forms for the topical or transdermal administration include powders, sprays, ointments, pastes, creams, lotions, gels, solutions, patches, and inhalants. The active compound may be mixed under sterile conditions with a pharmaceutically acceptable carrier, and with any preservatives, buffers, or propellants which may be required. The ointments, pastes, creams and gels may contain, in addition to a subject compound of the invention (e.g., an ActRIIB polypeptide), excipients, such as animal and vegetable fats, oils, waxes, paraffins, starch, tragacanth, cellulose derivatives, polyethylene glycols, silicones, bentonites, silicic acid, talc and zinc oxide, or mixtures thereof.
 Powders and sprays can contain, in addition to a subject compound, excipients such as lactose, talc, silicic acid, aluminum hydroxide, calcium silicates, and polyamide powder, or mixtures of these substances. Sprays can additionally contain customary propellants, such as chlorofluorohydrocarbons and volatile unsubstituted hydrocarbons, such as butane and propane.
 In certain embodiments, pharmaceutical compositions suitable for parenteral administration may comprise one or more ActRIIB polypeptides in combination with one or more pharmaceutically acceptable sterile isotonic aqueous or nonaqueous solutions, dispersions, suspensions or emulsions, or sterile powders which may be reconstituted into sterile injectable solutions or dispersions just prior to use, which may contain antioxidants, buffers, bacteriostats, solutes which render the formulation isotonic with the blood of the intended recipient or suspending or thickening agents. Examples of suitable aqueous and nonaqueous carriers which may be employed in the pharmaceutical compositions of the invention include water, ethanol, polyols (such as glycerol, propylene glycol, polyethylene glycol, and the like), and suitable mixtures thereof, vegetable oils, such as olive oil, and injectable organic esters, such as ethyl oleate. Proper fluidity can be maintained, for example, by the use of coating materials, such as lecithin, by the maintenance of the required particle size in the case of dispersions, and by the use of surfactants.
 The compositions of the invention may also contain adjuvants, such as preservatives, wetting agents, emulsifying agents and dispersing agents. Prevention of the action of microorganisms may be ensured by the inclusion of various antibacterial and antifungal agents, for example, paraben, chlorobutanol, phenol sorbic acid, and the like. It may also be desirable to include isotonic agents, such as sugars, sodium chloride, and the like into the compositions. In addition, prolonged absorption of the injectable pharmaceutical form may be brought about by the inclusion of agents which delay absorption, such as aluminum monostearate and gelatin.
 It is understood that the dosage regimen will be determined by the attending physician considering various factors which modify the action of the subject compounds of the invention (e.g., ActRIIB polypeptides). The various factors will depend upon the disease to be treated.
 In certain embodiments, the present invention also provides gene therapy for the in vivo production of ActRIIB polypeptides or other compounds disclosed herein. Such therapy would achieve its therapeutic effect by introduction of the ActRIIB polynucleotide sequences into cells or tissues having the disorders as listed above. Delivery of ActRIIB polynucleotide sequences can be achieved using a recombinant expression vector such as a chimeric virus or a colloidal dispersion system. Preferred for therapeutic delivery of ActRIIB polynucleotide sequences is the use of targeted liposomes.
 Various viral vectors which can be utilized for gene therapy as taught herein include adenovirus, herpes virus, vaccinia, or, preferably, an RNA virus such as a retrovirus. Preferably, the retroviral vector is a derivative of a murine or avian retrovirus. Examples of retroviral vectors in which a single foreign gene can be inserted include, but are not limited to: Moloney murine leukemia virus (MoMuLV), Harvey murine sarcoma virus (HaMuSV), murine mammary tumor virus (MuMTV), and Rous Sarcoma Virus (RSV). A number of additional retroviral vectors can incorporate multiple genes. All of these vectors can transfer or incorporate a gene for a selectable marker so that transduced cells can be identified and generated. Retroviral vectors can be made target-specific by attaching, for example, a sugar, a glycolipid, or a protein. Preferred targeting is accomplished by using an antibody. Those of skill in the art will recognize that specific polynucleotide sequences can be inserted into the retroviral genome or attached to a viral envelope to allow target specific delivery of the retroviral vector containing the ActRIIB polynucleotide. In one preferred embodiment, the vector is targeted to bone, cartilage, muscle or neuron cells/tissues.
 Alternatively, tissue culture cells can be directly transfected with plasmids encoding the retroviral structural genes gag, pol and env, by conventional calcium phosphate transfection. These cells are then transfected with the vector plasmid containing the genes of interest. The resulting cells release the retroviral vector into the culture medium.
 Another targeted delivery system for ActRIIB polynucleotides is a colloidal dispersion system. Colloidal dispersion systems include macromolecule complexes, nanocapsules, microspheres, beads, and lipid-based systems including oil-in-water emulsions, micelles, mixed micelles, and liposomes. The preferred colloidal system of this invention is a liposome. Liposomes are artificial membrane vesicles which are useful as delivery vehicles in vitro and in vivo. RNA, DNA and intact virions can be encapsulated within the aqueous interior and be delivered to cells in a biologically active form (see e.g., Fraley, et al., Trends Biochem. Sci., 6:77, 1981). Methods for efficient gene transfer using a liposome vehicle, are known in the art, see e.g., Mannino, et al., Biotechniques, 6:682, 1988. The composition of the liposome is usually a combination of phospholipids, usually in combination with steroids, especially cholesterol. Other phospholipids or other lipids may also be used. The physical characteristics of liposomes depend on pH, ionic strength, and the presence of divalent cations.
 Examples of lipids useful in liposome production include phosphatidyl compounds, such as phosphatidylglycerol, phosphatidylcholine, phosphatidylserine, phosphatidylethanolamine, sphingolipids, cerebrosides, and gangliosides. Illustrative phospholipids include egg phosphatidylcholine, dipalmitoylphosphatidylcholine, and distearoylphosphatidylcholine. The targeting of liposomes is also possible based on, for example, organ-specificity, cell-specificity, and organelle-specificity and is known in the art.
 The invention now being generally described, it will be more readily understood by reference to the following examples, which are included merely for purposes of illustration of certain embodiments and embodiments of the present invention, and are not intended to limit the invention.
Generation of an ActRIIB-Fc Fusion Protein
 Applicants constructed a soluble ActRIIB fusion protein that has the extracellular domain of human ActRIIB fused to a human or mouse Fc domain with a minimal linker (three glycine amino acids) in between. The constructs are referred to as ActRIIB(20-134)-hFc and ActRIIB(20-134)-mFc, respectively.
 ActRIIB(20-134)-hFc is shown below as purified from CHO cell lines (SEQ ID NO: 5)
 The ActRIIB(20-134)-hFc and ActRIIB(20-134)-mFc proteins were expressed in CHO cell lines. Three different leader sequences were considered:
(i) Honey bee mellitin (HBML):
(SEQ ID NO: 7)
(ii) Tissue Plasminogen Activator (TPA):
(SEQ ID NO: 8)
(SEQ ID NO: 9)
 The selected form employs the TPA leader and has the following unprocessed amino acid sequence:
 This polypeptide is encoded by the following nucleic acid sequence (SEQ ID NO:10):
A TGGATGCAAT GAAGAGAGGG CTCTGCTGTG TGCTGCTGCT
GTGTGGAGCA GTCTTCGTTT CGCCCGGCGC CTCTGGGCGT
GGGGAGGCTG AGACACGGGA GTGCATCTAC TACAACGCCA
ACTGGGAGCT GGAGCGCACC AACCAGAGCG GCCTGGAGCG
CTGCGAAGGC GAGCAGGACA AGCGGCTGCA CTGCTACGCC
TCCTGGCGCA ACAGCTCTGG CACCATCGAG CTCGTGAAGA
AGGGCTGCTG GCTAGATGAC TTCAACTGCT ACGATAGGCA
GGAGTGTGTG GCCACTGAGG AGAACCCCCA GGTGTACTTC
TGCTGCTGTG AAGGCAACTT CTGCAACGAG CGCTTCACTC
ATTTGCCAGA GGCTGGGGGC CCGGAAGTCA CGTACGAGCC
ACCCCCGACA GCCCCCACCG GTGGTGGAAC TCACACATGC
CCACCGTGCC CAGCACCTGA ACTCCTGGGG GGACCGTCAG
TCTTCCTCTT CCCCCCAAAA CCCAAGGACA CCCTCATGAT
CTCCCGGACC CCTGAGGTCA CATGCGTGGT GGTGGACGTG
AGCCACGAAG ACCCTGAGGT CAAGTTCAAC TGGTACGTGG
ACGGCGTGGA GGTGCATAAT GCCAAGACAA AGCCGCGGGA
GGAGCAGTAC AACAGCACGT ACCGTGTGGT CAGCGTCCTC
ACCGTCCTGC ACCAGGACTG GCTGAATGGC AAGGAGTACA
AGTGCAAGGT CTCCAACAAA GCCCTCCCAG TCCCCATCGA
GAAAACCATC TCCAAAGCCA AAGGGCAGCC CCGAGAACCA
CAGGTGTACA CCCTGCCCCC ATCCCGGGAG GAGATGACCA
AGAACCAGGT CAGCCTGACC TGCCTGGTCA AAGGCTTCTA
TCCCAGCGAC ATCGCCGTGG AGTGGGAGAG CAATGGGCAG
CCGGAGAACA ACTACAAGAC CACGCCTCCC GTGCTGGACT
CCGACGGCTC CTTCTTCCTC TATAGCAAGC TCACCGTGGA
CAAGAGCAGG TGGCAGCAGG GGAACGTCTT CTCATGCTCC
GTGATGCATG AGGCTCTGCA CAACCACTAC ACGCAGAAGA
GCCTCTCCCT GTCTCCGGGT AAATGA
 N-terminal sequencing of the CHO-cell produced material revealed a major sequence of -GRGEAE (SEQ ID NO: 11). Notably, other constructs reported in the literature begin with an -SGR . . . sequence.
 Purification could be achieved by a series of column chromatography steps, including, for example, three or more of the following, in any order: protein A chromatography, Q sepharose chromatography, phenylsepharose chromatography, size exclusion chromatography, and cation exchange chromatography. The purification could be completed with viral filtration and buffer exchange.
 ActRIIB-Fc fusion proteins were also expressed in HEK293 cells and COS cells. Although material from all cell lines and reasonable culture conditions provided protein with muscle-building activity in vivo, variability in potency was observed perhaps relating to cell line selection and/or culture conditions.
Generation of ActRIIB-Fc Mutants
 Applicants generated a series of mutations in the extracellular domain of ActRIIB and produced these mutant proteins as soluble fusion proteins between extracellular ActRIIB and an Fc domain. The background ActRIIB-Fc fusion has the sequence (Fc portion underlined) (SEQ ID NO:12):
 Various mutations, including N- and C-terminal truncations, were introduced into the background ActRIIB-Fc protein. Based on the data presented in Example 1, it is expected that these constructs, if expressed with a TPA leader, will lack the N-terminal serine. Mutations were generated in ActRIIB extracellular domain by PCR mutagenesis. After PCR, fragments were purified through a Qiagen column, digested with SfoI and AgeI and gel purified. These fragments were ligated into expression vector pAID4 (see WO2006/012627) such that upon ligation it created fusion chimera with human IgG1. Upon transformation into E. coli DH5 alpha, colonies were picked and DNAs were isolated. For murine constructs (mFc), a murine IgG2a was substituted for the human IgG1. All mutants were sequence verified.
 All of the mutants were produced in HEK293T cells by transient transfection. In summary, in a 500 ml spinner, HEK293T cells were set up at 6×10 5 cells/ml in Freestyle (Invitrogen) media in 250 ml volume and grown overnight. Next day, these cells were treated with DNA:PEI (1:1) complex at 0.5 ug/ml final DNA concentration. After 4 hrs, 250 ml media was added and cells were grown for 7 days. Conditioned media was harvested by spinning down the cells and concentrated.
 Mutants were purified using a variety of techniques, including, for example, protein A column and eluted with low pH (3.0) glycine buffer. After neutralization, these were dialyzed against PBS.
 Mutants were also produced in CHO cells by similar methodology.
 Mutants were tested in binding assays and/or bioassays. In some instances, assays were performed with conditioned medium rather than purified proteins. Variants are described, for example, in published patent applications WO 06/012627 and WO 08/097,541. Such variants may be used in the methods described herein.
Effect of Truncated Variant ActRIIB(25-131)-hFc on Hepatic Steatosis in a Mouse Model of Diet-Induced Obesity
 Nonalcoholic fatty liver disease (NAFLD) is a spectrum of increasingly common hepatic disorders widely considered to be the hepatic manifestation of metabolic syndrome and characterized by fat deposition in the liver (steatosis), often with deleterious effects. A subset of NAFLD patients develop an inflammatory condition referred to as nonalcoholic steatohepatitis (NASH), which can progress further to hepatic fibrosis, cirrhosis, and hepatocellular carcinoma (Perlemuter et al., 2007, Nat Clin Pract Endocrinol Metab 3:458-469). Applicants generated a truncated fusion protein ActRIIB(25-131)-hFc ( FIGS. 1-2 ), using the same leader and methodology as described above with respect to ActRIIB(20-134)-hFc. The mature protein purified after expression in CHO cells has the sequence shown below (SEQ ID NO: 6):
ETRECIYYNA NWELERTNQS GLERCEGEQD KRLHCYASWR
NSSGTIELVK KGCWLDDFNC YDRQECVATE ENPQVYFCCC
EGNFCNERFT HLPEAGGPEV TYEPPPT GGG THTCPPCPAP
ELLGGPSVFL FPPKPKDTLM ISRTPEVTCV VVDVSHEDPE
VKFNWYVDGV EVHNAKTKPR EEQYNSTYRV VSVLTVLHQD
WLNGKEYKCK VSNKALPAPI EKTISKAKGQ PREPQVYTLP
PSREEMTKNQ VSLTCLVKGF YPSDIAVEWE SNGQPENNYK
TTPPVLDSDG SFFLYSKLTV DKSRWQQGNV FSCSVMHEAL
 As shown below, ActRIIB(25-131)-hFc could inhibit hepatic steatosis while producing beneficial changes in closely related parameters in a mouse model of diet-induced obesity. Ten-week-old C57BL/6 mice were treated with ActRIIB(25-131)-hFc, at 10 mg/kg, s.c., or Tris-buffered-saline (TBS) vehicle twice per week for 60 days. During this period, mice had unlimited access to a diet containing 58% fat instead of the standard chow containing 4.5% fat. An additional group of mice maintained on the standard chow diet were also treated with TBS vehicle and followed as a dietary control.
 Hepatic biopsy is considered the gold standard for NAFLD diagnosis, so at study completion the liver was removed and analyzed histologically, Hepatic tissue of mice fed the high-fat diet displayed large numbers of densely packed lipid droplets, as assessed by staining with Oil Red O, whereas mice fed the standard diet showed no evidence of hepatic lipid deposits ( FIG. 3 ). Treatment with ActRIIB(25-131)-hFc almost completely prevented hepatic lipid deposition and normalized the appearance of hepatic tissue despite the high-fat diet. Thus, ActRIIB(25-131)-hFc was an effective inhibitor of hepatic steatosis caused by high-fat diet.
 ActRIIB(25-131)-hFc treatment also induced beneficial changes in endpoints that correlate strongly with hepatic steatosis in fatty liver disease of the alcoholic type as well as nonalcoholic type. Adiponectin, a fat-derived hormone whose concentration varies inversely with fat mass/obesity, exerts important insulin-sensitizing actions in target tissues (Yamauchi et al., 2001, Nat Med 7:941-946; Maeda et al., 2002, Nat Med 8:731-737; Kadowaki et al., 2005, Endocr Rev 26:439-451). Adiponectin concentrations are also important in fatty liver disease of the alcoholic type (You et al., 2009, Exp Biol Med 234:850-859). ActRIIB(25-131)-hFc treatment raised levels of adiponectin mRNA in epididymal white fat ( FIG. 4 ) as well as circulating concentrations of adiponectin ( FIG. 5 ). Importantly, these changes were accompanied in ActRIIB(25-131)-hFc-treated mice by robust decreases in circulating concentrations of insulin ( FIG. 6 ), triglycerides, free fatty acids, high-density lipoprotein (HDL), and low-density lipoprotein (LDL), leading to normalization of nearly all of these parameters. Finally, the aforementioned effects were accompanied by beneficial changes in body composition, as determined by nuclear magnetic resonance (NMR) at baseline and Day 48. Specifically, total fat mass in vehicle-treated controls under high-fat dietary conditions tripled during this 48-day period, and ActRIIB(25-131)-hFc treatment cut this increase by nearly 40%. In summary, ActRIIB(25-131)-hFc treatment under high-fat dietary conditions 1) prevented hepatic steatosis, 2) increased circulating adiponectin concentrations, 3) reduced circulating insulin concentrations, consistent with increased insulin sensitivity 4) produced beneficial changes in the serum lipid profile, and 5) improved body composition.
Effect of Truncated Variant ActRIIB(25-131)-mFc on Hepatic Steatosis in a Mouse Model of Atherogenesis
 Mice genetically deficient in the low-density lipoprotein receptor (LDLR) are a widely used experimental model of atherogenesis. When fed a diet high in fat and cholesterol, ldlr −/− mice develop hypercholesterolemia and atheromatous lesions (Breslow, 1996, Science 272:685-688). Therefore, applicants investigated the ability of ActRIIB(25-131)-mFc to ameliorate hepatic steatosis in this model when administered therapeutically (after a dietary pretreatment phase). Beginning at five months of age, male ldlr −/− mice (C57BL/6 background) were given continuous access to either a standard chow diet (Harlan Teklad Diet 2018, containing 5.8% fat) or a diet high in fat and cholesterol (Harlan Teklad Diet 94059, a cholate-free “Paigan” diet containing approximately 15.8% fat and 1.25% cholesterol). Beginning eight weeks after onset of these respective diets, mice were treated twice per week, subcutaneously, with ActRIIB(25-131)-mFc (10 mg/kg) or vehicle (Tris-buffered saline) for 16 weeks, accompanied by continued administration of the dietary regimen. Vehicle-treated wildtype mice maintained on the standard diet served as an additional control.
 There was clear evidence of hepatic steatosis in the atherogenic model. Compared to wildtype mice fed the standard diet, ldlr −/− mice fed the high-fat diet for 24 weeks displayed a marked reduction in liver tissue density, as determined by micro-computed tomography (micro-CT) ( FIG. 7 ). The reduction in overall liver density to values trending below the density of water (defined in this analysis as zero) was accompanied by liver enlargement and is indicative of hepatic steatosis. Compared to atherogenic mice treated with vehicle, those treated with ActRIIB(25-131)-mFc displayed significantly increased (improved) liver density ( FIG. 7 ), even though such mice were fed the high-fat diet for 8 weeks prior to receiving ActRIIB(25-131)-mFc. Moreover, atherogenic mice treated with ActRIIB(25-131)-mFc also exhibited normalization of body composition, glycated hemoglobin (A1C) concentrations, fat depot weights, and serum triglyceride levels to values observed in wildtype controls (data not shown).
 Taken together, these data indicate that soluble ActRIIB-Fc fusion proteins can be used as antagonists of signaling by TGF-β-family ligands to increase circulating adiponectin concentrations and improve insulin sensitivity in target tissues, thereby treating nonalcoholic fatty liver disease, alcoholic fatty liver disease, and potentially other conditions as well.
INCORPORATION BY REFERENCE
 All publications and patents mentioned herein are hereby incorporated by reference in their entirety as if each individual publication or patent was specifically and individually indicated to be incorporated by reference.
 While specific embodiments of the subject matter have been discussed, the above specification is illustrative and not restrictive. Many variations will become apparent to those skilled in the art upon review of this specification and the claims below. The full scope of the invention should be determined by reference to the claims, along with their full scope of equivalents, and the specification, along with such variations.