Activin/Inhibin

First printed in R&D Systems' 1998 Catalog.

Overview

Inhibin is a molecule that was first described over 60 years ago.1-4 At that time, it was known that experimental destruction of rat seminiferous tubules resulted in hypertrophy of the pituitary gland. This hypertrophy was suspected to be due to the lack of a circulating factor derived from the seminiferous tubules (testes). As part of a feedback loop, hypertrophied pituitary cells were believed to be overproducing a factor that would normally induce the testes to produce a pituitary-targeted molecule. When it was noted that the development of hypertrophied cells following castration could be blocked or inhibited by the administration of an aqueous testicular extract, this water soluble substance was named inhibin.1 Subsequent studies showed that 1) the molecule being over-expressed by the anterior pituitary was follicle-stimulating hormone (FSH), and 2) FSH secretion could be both inhibited and promoted (activated) by two closely related molecules termed inhibin and activin, respectively.2, 3, 5, 6

Currently, inhibin and activin are still believed to be physiologically relevant in the regulation of pituitary gonadotrophin (FSH and LH) secretion.7 Now, however, a number of aspects of the inhibin story are undergoing reevaluation, including the physiological sources of activin and inhibin7, 8 and the exact form(s) the bioactive molecules may take.9 Furthermore, there is now some uncertainty regarding past measurements of circulating levels of activin and inhibin, making previous correlations with endocrine cycles difficult to interpret.10-14 Finally, the receptors for activin (inhibin receptors have not yet been identified) are now reported to bind a number of non-activin/inhibin molecules.15 These considerations show that the activin/inhibin system is far more complex than previously imagined and make the interpretation of older observations problematic.

Structural Information

Both activin and inhibin are members of the TGF-beta superfamily of molecules.3, 4, 16 As such, these molecules are synthesized as large precursors that contain a signal sequence, a pro-domain of variable size, and a mature C-terminal segment that ranges from 110 to 140 amino acids (aa) in length. Within the mature segment, there are seven cysteines that are considered invariant across the superfamily. The basic structure for both activin and inhibin is a disulfide-linked dimer. This dimer is composed of one or two fundamental subunit types; an alpha chain encoded by a distinct gene and a beta chain, one of three basic types (in the human), each encoded by a separate gene. Beta-beta (beta-beta) dimers are homodimeric only in the most general sense, since 1) there may be products of different beta genes involved, and 2) each gene can give rise to multiple cleavage forms that may differ in biological activity.3, 9, 17-19 The activin molecule is sometimes described as being a homodimer, but this is because the dimer is composed of two, same-gene beta-subunit types.4, 17, 18 Inhibin, in contrast, is often described as being a heterodimer, composed of a beta-type subunit linked to some form of alpha-type subunit.4, 17, 18 In any case, the "form" of the subunits is what is important, because various combinations of subunit cleavage products (for both alpha- and beta-subunits) are known to form dimers, circulate, and show bioactivity.17, 19

To date, four beta-type chains for mammals have been described.18, 20-22 Designated betaA, betaB, betaC and betaE (betaD has been found only in Xenopus),23 three of the chains have been identified in human (A, B, C) and all four in mouse.22 Although each chain shows remarkable conservation of structure, each chain is also the product of a separate gene.3, 19 In humans, the prepropeptides for beta-chains A, B, and C consist of 425, 407, and 352 aa residues, respectively.18, 20 The differences in length are due to the size of the pro-regions, as the mature segments are 116, 115, and 116 aa residues, respectively. Within the mature segments, there is 64% aa residue identity, betaA to betaB, and approximately 50% aa residue identity betaC to betaA and betaB.20 Mouse to human, each counterpart chain is approximately 95% conserved at the aa residue level.18, 21 Finally, in mice where four chains have now been identified, it appears that beta-chains A and B comprise one subset, while chains C and E (plus Xenopus D) constitute a second subset.22

Activin, consisting of only beta-type chains, is named for the presence of those chains. That is, activin A is a homodimer with two betaA chains; activin B is a homodimer with two betaB chains; and activin AB is a heterodimer with an A and B chain. Although subunit C can theoretically link with subunits A and B, little is known about its contribution to activin formation. All mature 115/116 aa residue beta-chains are always unglycosylated, and each is about 14 kDa in size.18 Thus, any activin dimer composed of fully processed beta-proforms (called betaA) will be approximately 24 to 28 kDa in size.9, 18 However, possibly due to variable proteolytic processing of the beta-subunit prosequence (called probetaA) at the surface of secreting cells, beta-precursors can also be released that form dimers with other beta-precursors or mature (betaA) forms. This can give rise to a number of activin molecules, which, in the case of the betaA subunit only, will yield a 110 kDa homodimer composed of two probetaAs, a 70 kDa heterodimer composed of one probetaA and one betaA, and a 24 kDa dimer representing two fully processed betaA chains.9 In addition, while the mature peptide is not glycosylated, the pro-segment may be monoglycosylated.9 Since it is estimated that the addition of a sugar moiety will likely increase the MW of activin by 3 kDa,11 it becomes clear that there are potentially a large number of activin size species that can circulate at any one time. Whether all forms of activin are bioactive is unknown. Data suggests, however, that the presence of at least one mature (14 kDa) beta-chain is sufficient to confer bioactivity.9 Cells known to express the betaA chain include fibroblasts,24, 25 endothelial cells,26 hepatocytes,27 vascular smooth muscle cells, macrophages, and cuboidal epithelium,28, 29 keratinocytes,25, 30 osteoclasts,31 bone marrow monocytes,32 prostatic epithelium,33 neurons,34 chondrocytes and osteoblasts,35 Leydig cells,3, 36 Sertoli cells,3, 37 ovarian granulosa cells,3, 36 and possibly, anterior pituitary gonadotrophs.38 betaB subunits are known to be expressed in anterior pituitary gonadotrophs,39 Sertoli cells,3, 37 Leydig cells,3, 36 testicular interstitial cells,37 ovarian granulosa cells,3, 36 and keratinocytes.34 Expression of the betaC chain has not been well described and currently appears to be limited to liver, ovaries and testes, and primary spermatocytes.19, 20, 40

The alpha-chain (or subunit) is unique to inhibin molecules. All inhibin molecules, by definition, possess one alpha-chain and one beta-chain. The inhibin name (i.e., inhibin A, B or C) derives from the general form of the beta-chain present. Each alpha-subunit is synthesized as a 53 kDa, 366 aa residue precursor containing an 18 aa residue signal sequence, a 43 aa residue pro-region (pro-alpha), and a 305 aa residue "mature" segment, which itself is divided into a 171 aa residue N-terminal segment (alphaN) and a mature 134 aa residue C-terminal segment (alphaC).9, 18 Unlike the mature beta-chain, the mature alphaC fragment has sites for glycosylation, and usually shows mono-, if not di-, glycosylation. The molecule shows differential proteolytic processing, creating variants with different combinations of segments. For example, inhibin A (an alpha-chain dimerized with a betaA-chain) can exist as either a 105 kDa pro-alpha/alphaN/alphaC:probetaA/betaA dimer, a 75 kDa alphaC:probetaA dimer, a 56 kDa alphaN/alphaC:betaA dimer, or a 32 kDa alphaC:betaA dimer.9, 17 In addition, various monomeric forms of the alpha-subunit also exist. Forms reported to circulate include 26 kDa and 29 kDa mono- and diglycosylated pro-alpha/alphaC and 20 kDa monoglycosylated alphaC.9, 17 As with activin, the number of alpha-betaA,B,C combinations for inhibin is quite large. Given that partially-processed subunit dimers may, or may not, be bioactive, difficulties can arise when correlating physiological activity with circulating inhibin levels. Overall, alphaC segments are less than 30% identical at the aa residue level to mature betaA and betaB chains.16 Cells known to express a-subunits include ovarian granulosa cells,3, 36 Sertoli cells,3, 37 testicular interstitial cells,37 Leydig cells,3, 37 prostatic epithelial cells,33 and anterior pituitary gonadotrophs.39

Receptors

The receptors for activin are transmembrane serine/threonine kinases whose pattern parallels that seen for other members of the TGF-beta superfamily. That is, there is a type II ligand-binding receptor, a type I signal-transducing receptor, and possibly a type III ligand-presenting receptor.13, 15, 41-43

Type II Receptors:

In humans, there are two type II receptors known for activin. These are termed activin receptor type II (ACTRII)44 and activin receptor type IIB (ACTRIIB).45 ActRII is a 70 kDa, 513 aa residue transmembrane glycoprotein that has a 116 aa residue extracellular region, a 26 aa residue transmembrane segment, and a 352 aa residue cytoplasmic domain. As with other TGF-beta type II receptors, ActRII's extracellular region contains a cysteine-rich domain with 10 conserved cysteines. In addition, the cytoplasmic region contains a 285 aa residue kinase domain characteristic of serine/threonine kinases.44 The mouse homolog of human ActRII has also been cloned and is found to be 99% identical to the human receptor at the aa residue level. The Kd for activin A binding to ActRII is reported to be 180-360 pM,44, 46 with activin B binding with slightly less affinity.46 Inhibin A will also bind to ActRII, but with 10-fold less affinity.46 At 512 aa residue, human ActRIIB is almost identical in size to human ActRII. However, even though ActRIIB shows the same structural features as ActRII, it is only 69% identical at the aa residue level.45 As with ActRII, ActRIIB demonstrates almost complete identity to its mouse ActRIIB counterpart. However, for reasons not understood, the mouse ActRIIB gene gives rise to multiple receptor isoforms, a phenomenon not reported for humans.47 Four mouse splice forms are known and result from splice events in both the extracellular and cytoplasmic regions. Within the four mouse forms (ActRIIB1-4), ActRIIB2 is considered the homolog of the single human receptor. In mice, activin A binds to the ActRIIB isoforms with Kds ranging from 100 to 400 pM, with the B2 Kd reported to be 100 pM. In Balb/C 3T3 fibroblasts, all four B isoforms have been shown to be expressed, with the B2 isoform accounting for 58%, the B4 isoform (Kd = 380 pM) for 36%, the B3 for 4%, and the B1 for 2% of the total activin B mRNA.47 However, in the same cell type, levels of ActRII mRNA were reported to be 10-fold more abundant than the total of all of the ActRIIB mRNAs.47

Type I Receptors:

As with the type II receptor, there are two forms of the type I receptor, ActRI and ActRIB. Human ActRI, also called SKR1 (serine/threonine kinase receptor 1) or Alk-2 (activin receptor-like kinase-2), is a 65 kDa, 509 aa residue transmembrane glycoprotein comprised of a 21 aa residue signal sequence, a 102 aa residue extracellular region, a 23 aa residue transmembrane segment, and a 363 aa residue cytoplasmic tail.48-50 Like the type II receptor, the type I receptor has a cysteine-rich extracellular domain and a serine-threonine kinase domain. Unlike the type II receptor, the type I receptor shows two cytoplasmic modifications. First, it lacks a serine/threonine-rich extension distal to its kinase domain, and second, it possesses a glycine-serine (GS) rich segment between the transmembrane segment and kinase domain. This GS region is suggested to be phosphorylated by a ligand-activated type II receptor, an event that subsequently leads to type I receptor-mediated downstream intracellular signaling.15 Mouse ActRI has also been cloned, and found to be 98% identical to human ActRI.48, 51 Relative to ActRII, ActRI has less than 15% aa residue identity in the extracellular segment, and only 33% aa residue identity in the cytoplasmic region.49, 51 ActRI will not directly bind activin, but will form a noncovalent heteromeric signalling complex with either ActRII or ActRIIB.48, 49

There is a second human type I receptor known for activin. Also known as ActRIB (or SKR2 or Alk-4), this transmembrane molecule is 505 aa residues long, demonstrating a potential for up to four alternatively spliced isoforms.52, 53 When compared to ActRI, this receptor has less than 20% aa residue identity in its extracellular region, and only 67% identity in its cytoplasmic kinase domain.52 It would appear that the 33% kinase difference is important, as ActRI and IB are reported to perform very different functions. For example, in cell proliferation assays, activin A/ActRII/ActRIB complexes will inhibit cell proliferation, while activin A/ActRII/ActRI complexes show no such activity.52

Type III Receptors:

Although a type III receptor is known for TGF-beta,16, 41 no equivalent receptor has yet been identified for the activin family. Nevertheless, there is some suggestion that a high molecular weight (145 kDa) type III receptor may participate in activin receptor-ligand interactions.26, 43

Two points should be made relative to the activin receptor system. First, no receptor for inhibin has yet been characterized. Until this happens, the molecular biology of inhibin must remain something of a black-box. Second, the receptors for activin are known to bind a number of other TGF-beta superfamily members. For example, ActRII will bind GDF-555 and BMP-7 (bone morphogenetic protein-7),56 while ActRIIB will bind GDF-5 (growth/differentiation factor-5)55 and BMP-2.57 Although ActRI does not bind ligand directly, it will form a heteromeric complex with BMP-7/ActRII, and BMP-2/BMPRII.15, 56, 58 Thus, a number of ligand-receptor combinations (and confusions) are possible within the activin receptor system.

Cells known to express ActRII include oocytes and spermatids,59 endothelial cells,26 fibroblasts,47 and neurons.59 ActRIIB expression parallels that for ActRII, being found in neurons,59 Leydig cells,59 fibroblasts,47 trophoblasts,7 and endothelial cells.26 In the brain, both ActRII and IIB are ubiquitous and almost always co-expressed. However, the region of the basal ganglia (motor coordination) shows ActII but not ActIIB expression.59 Cells known to express ActRI include fibroblasts and vascular endothelial cells,49, 60 hepatocytes and keratinocytes.49 ActRIB has been identified in fibroblasts, endothelial cells and hepatocytes.53, 60

Binding Proteins

Activin and inhibin rarely circulate freely. In both blood and extravascular fluid, these dimers are complexed to one of two binding proteins: 725 kDa alpha2-macroglobulin (alpha2M) or 32-40 kDa Follistatin (FS). alpha2M is a noncovalent tetramer of two 370 kDa disulfide-linked homodimers.61 Known to bind both activin and inhibin, alpha2M circulates in blood at concentrations of approximately 3 mg/mL, levels suggested to provide abundant opportunity for activin binding.61 When bound to alpha2M, neither activin bioactivity nor immunoreactivity are reported to be affected.13, 61 The biological significance of this binding is not understood, but it is suggested to play a role in either delivery or clearance of activins and inhibins.13, 61 This suggested role is similar to that proposed for alpha2M binding of TGF-beta and numerous other growth factors.

Follistatin, by contrast, is believed to have a significant effect on both inhibin and activin availability and activity.62 Synthesized as one of two alternatively spliced forms, this glycoprotein is known to bind both activin and inhibin through their beta-chains with a Kd of 50 pM.63-65 When complexed with activin/inhibin, follistatin is reported to effectively neutralize all ligand bioactivity and immunoreactivity.8, 13, 66, 67 It is estimated that FS levels in the blood are in the range of 10 ng/mL.66 Given that activin A levels are variously reported to be in the pg/mL, if not the ng/mL range,8, 10, 68 it seems clear that FS binding could have a material impact on activin function.8, 66 Again, whether this takes the form of a storage/presenting molecule7, 8, 65 or a "removal" molecule65 is unknown.

Biological Functions

Functionally, the activins/inhibins have been implicated in a number of processes. These include embryogenesis, osteogenesis, hematopoiesis, and reproductive physiology. Within these systems, the idea that activin is "activating" should not be carried too far. Although activin will induce the proliferation of cells such as CFU-E32 and fibroblasts,69 it will also inhibit the proliferation of hepatocytes27 and induce apoptosis in plasmacytoma cells.70

During development, activin is perhaps best described as acting locally and at many sites during mammalian embryogenesis.71 The betaA subunit is known to be expressed (and presumably acts) in heart, brain, bone marrow and skeletal muscle, while the betaB subunit is associated with brain and gonads. Activin has been proposed to be a mesoderm-inducing factor. However, knockout mice have failed to demonstrate this conclusively.71, 72 One of the more interesting observations in knockout mice suggests that activin may not be signaling through its type II receptor(s) in the embryo, recalling again the lack of specificity in the activin system.71, 73 In this regard, the importance of the activin receptors is seen in mice that lack the ActRIIB receptor (whatever the ligand). These animals show marked cardiovascular positional abnormalities, some of which mimic complex human disorders.74

Activin has also been implicated in bone remodeling. During bone formation, a temporal interaction between FS and activin A may be involved in the development and conversion of cartilage to bone. In concert with the BMPs, activin A may also contribute to intramembranous ossification.35

Blood cell formation also may be influenced by both inhibin A and activin A. Inhibin A has been shown to suppress BFU-E and CFU-GEMM colony expansion, while activin A is reported to promote BFU-E and CFU-GEMM expansion.75, 76 The activin effect is believed to be indirect, being mediated by activated monocytes and T cells. The overall role for activin is considered to be secondary, not primary (i.e., activin fine-tunes rather than drives the expansion of hematopoietic progenitors).

The best known role for activin/inhibin involves the modulation of the female reproductive cycle. At the beginning of the menstrual cycle, anterior pituitary gonadotrophic cells release FSH. This FSH travels to the ovary where it binds to immature oocytes, initiating their maturation. Such maturing oocytes or follicles make estrogen, which gives negative feedback to the pituitary to inhibit further FSH release. After about 14 days, one oocyte matures and releases a final burst of estrogen that causes the pituitary gonadotrophs to secrete both FSH and LH (luteinizing hormone). This "spike" of LH/FSH induces follicle rupture, leading to a decline in estrogen production. Cells in the ovary that contributed to earlier follicle maturation and now remain following ovulation develop into a structure called the corpus luteum. This structure becomes a center for the synthesis of progesterone, a molecule that both inhibits pituitary LH production, and promotes the growth of the uterus. In the absence of fertilization, the corpus luteum involutes, leading to a period of menses with low circulating estrogen and progesterone levels. The lack of these two hormones is a positive signal for new FSH release and the start of a new cycle. Although estrogen is the primary stimulus for the "LH spike", gonad-produced inhibin is believed to limit the amount of FSH released during this spike, thus fine-tuning the response.4, 7 In addition, during the luteal or post-ovulation period, ovarian-derived inhibin levels rise, helping to suppress pituitary FSH release during this time.7 Then when inhibin levels fall late in the menstrual cycle, an impediment to FSH production is removed. The above must be viewed as naively simplistic, since there are multiple sites of activin and inhibin production (including the pituitary39), and multiple combinations of growth factors that affect production at local levels.4, 7 The scenario is further complicated by the large number of circulating activin/inhibin forms,77 as well as the fundamental question of whether there is actually any circulating bioactive activin/inhibin in blood.8, 12 Given FS blood levels and affinity for activin, it is unclear if the activin/inhibin system is more an endocrine, or an autocrine/paracrine system. Along these lines, it should be noted that during the last trimester of pregnancy, activin A levels reach a reported level of 1-6 ng/mL. At these levels, the binding-capacity of FS is likely to be exceeded, making activin available in a true endocrine manner.78

References

  1. McCullagh, D.R. (1932) Science 76:19.
  2. Vale, W. et al. (1988) Rec. Prog. Horm. Res. 44:1.
  3. Mayo, K.E. (1994) Trends Endocrinol. Metab. 5:407.
  4. Vale, W. et al. (1990) "The Inhibin/Activin Family of Hormones and Growth Factors." In: Peptide Growth Factors and their Receptors. Sporn, M., Roberts, A., eds., Berlin, Springer-Verlag. pp. 211-248.
  5. Ling, N. et al. (1986) Nature 321:779.
  6. Vale, W. et al. (1986) Nature 321:776.
  7. Halvorson, L.M. and A.H. DeCherney (1996) Fertil. Steril. 65:459.
  8. Muttukrishna, S. et al. (1996) J. Clin. Endocrinol. Metab. 81:3328.
  9. Mason, A.J. et al. (1996) Mol. Endocrinol. 10:1055.
  10. Groome, N.P. et al. (1996) J. Clin. Endocrinol. Metab. 81:1401.
  11. Robertson, D. et al. (1996) J. Clin. Endocrinol. Metab. 81:669.
  12. Lambert-Messerlian, G.M. et al. (1995) J. Clin. Endocrinol. Metab. 80:3043.
  13. Krummen, L.A. et al. (1993) Endocrinology 132:431.
  14. Schneyer, A.L. et al. (1992) J. Clin. Endocrinol. Metab. 74:1320.
  15. Hill, C.S. (1996) Cell. Signal. 8:533.
  16. Kingsley, D.M. (1994) Genes Dev. 8:133.
  17. Knight, P.G. (1996) Frontiers Neuroendocrinol. 17:476.
  18. Mason, A.J. et al. (1987) "Human Inhibin and Activin: Structure and Recombinant Expression in Mammalian Cells." In: Inhibin: Non-Steroidal Regulation of Follicle-Stimulating Hormone Secretion. Burger, H. de Kretser, D., Findlay, J., & M. Igarashi, eds., New York, Raven Press, 42:77.
  19. Fang, J. et al. (1997) Biochem. Biophys. Res. Commun. 231:655.
  20. Hotten, G. et al. (1995) Biochem. Biophys. Res. Commun. 206:608.
  21. Schmitt, J. et al. (1996) Genomics. 32:358.
  22. Fang, J. et al. (1996) Biochem. Biophys. Res. Commun. 228:669.
  23. Oda, S. et al. (1995) Biochem. Biophys. Res. Commun. 210:581.
  24. Hubner, G. et al. (1996) Dev. Biol. 173:490.
  25. Hubner, G. and S. Werner (1996) Exp. Cell Res. 228:106.
  26. McCarthy, S.A. and R. Bicknell (1993) J. Biol. Chem. 268:23066.
  27. Yasuda, H. et al. (1993) J. Clin. Invest. 92:1491.
  28. Matsuse, T. et al. (1995) Am. J. Respir. Cell. Mol. 13:17.
  29. Matsuse, T. et al. (1996) Am. J. Pathol. 148:707.
  30. Seishima, M. et al. (1996) FEBS Lett. 398:120.
  31. Hosoi, T. et al. (1996) Gerontology 42 (Suppl 1):20.
  32. Shao, L. et al. (1992) Exp. Hematol. 20:1235.
  33. Ying, S-Y. et al. (1997) Life Sci. 60:397.
  34. Tretter, Y. et al. (1996) NeuroReport 7:1819.
  35. Funaba, M. et al. (1996) Endocrinology 137:4250.
  36. Rabinovici, J. et al. (1991) J. Clin. Endocrinol. Metab. 73:1141.
  37. Roberts, V. et al. (1989) Endocrinology :2350.
  38. DeMura, R. et al. (1996) Endocrine J. 43:429.
  39. Roberts, V.J. et al. (1992) Neuroendocrinology 56:214.
  40. Loveland, K.L. et al. (1996) J. Mol. Endocrinol. 17:61.
  41. Attisano, L. et al. (1994) Biochim. Biophys. Acta 1222:71.
  42. ten Dijke, P. et al. (1996) Curr. Opin. Cell Biol. 8:139.
  43. McCarthy, S.A. and R. Bicknell (1993) J. Biol. Chem. 269:3909.
  44. Donaldson, C.J. et al. (1992) Biochem. Biophys. Res. Commun. 184:310.
  45. Hilden, K. et al. (1994) Blood 83:2163.
  46. Mathews, L.S. and W.W. Vale (1991) Cell 65:973.
  47. Attisano, L. et al. (1992) Cell 68:97.
  48. Attisano, L. et al. (1993) Cell 75:671.
  49. Matsuzaki, K. et al. (1993) J. Biol. Chem. 268:12719.
  50. ten Dijke, P. et al. (1993) Oncogene. 8:2879.
  51. Ebner, R. et al. (1993) Science 260:1344.
  52. Carcamo, J. et al. (1994) Mol. Cell. Biol. 14:3810.
  53. Xu, J. et al. (1994) Proc. Natl. Acad. Sci. USA 91:7957.
  54. Lebrun, J.J. and W.W. Vale (1997) Mol. Cell. Biol. 17:1682.
  55. Nishitoh, H. et al. (1996) J. Biol. Chem. 271:21345.
  56. Yamashita, H. et al. (1995) J. Cell Biol. 130:217.
  57. Hoodless, P.A. et al. (1996) Cell 85:489.
  58. ten Dijke, P. et al. (1994) J. Biol. Chem. 269:16985.
  59. Cameron, V.A. et al. (1994) Endocrinology 134:799.
  60. ten Dijke, P. et al. (1994) Science 264:101.
  61. Vaughan, J.M. and W.W. Vale (1993) Endocrinology 132:2038.
  62. Shimonaka, M. et al. (1991) Endocrinology 128:3313.
  63. Shimasaki, S. et al. (1988) Proc. Natl. Acad. Sci. USA 85:4218.
  64. Inouye, S. et al. (1991) Endocrinology 129:815.
  65. Schneyer, A.L. et al. (1994) Endocrinology 135:667.
  66. Khoury, R.H. et al. (1995) J. Clin. Endocrinol. Metab. 80:1361.
  67. de Winter, J.P. et al. (1996) Mol. Cell. Endocrinol. 116:105.
  68. Muttukrishna, S. et al. (1997) Lancet 349:1285.
  69. Sakurai, T. et al. (1994) J. Biol. Chem. 269:14118.
  70. Sternberg, D. et al. (1995) Growth Factors 12:277.
  71. Ying, S-Y. et al. (1997) Proc. Soc. Exp. Biol. Med. 214:114.
  72. St.-Jacques, B. and A.P. McMahon (1996) Curr. Opin. Genet. Dev. 6:439.
  73. Matzuk, M.M. et al. (1995) Nature 374:356.
  74. Oh, S.P. & E. Li (1997) Genes Dev. 11:1812.
  75. Hangoc, G. et al. (1992) Exp. Hematol. 20:1243.
  76. Broxmeyer, H.E. et al. (1988) Proc. Natl. Acad. Sci. USA 85:9052.
  77. Robertson, D.M. et al. (1997) J. Clin. Endocrinol. Metab. 82:889.
  78. Woodruff, T.K. et al. (1997) J. Endocrinol. 152:167.