Follistatin is a developmentally regulated cytokine in neural differentiation.

Activin acts mitogenically on P19 cells as well as being inhibitory of the differentiation of retinoic acid-treated P19 cells and some neuroblastoma cell lines. Here, we show some lines of evidence that follistatin, an activin-binding protein, is also involved in neural differentiation. Counteracting the activity of activin, addition of follistatin suppresses the anchorage-independent growth of P19 cells in soft agar and stimulates neurite outgrowth of a neuroblastoma cell line, IMR-32 cells. While activin does not seem to be expressed significantly, follistatin is demonstrated in the conditioned medium of these cells. Furthermore, the expression of follistatin in P19 cells is subject to dynamic fluctuations in response to retinoic acid treatment. These neural cells may produce follistatin in a cell stage-specific manner in order to interact with exogenously derived activin.

Activin, a member of the transforming growth factor+ family peptides, was first characterized from gonadal fluids as a stimulator of follicle-stimulating hormone (FSH)' release from anterior pituitary gland (1, 2). Since erythroid differentiation factor was shown to be identical to activin-A (3, 4), diverse biological roles of activin outside the reproductive system have been extensively studied. The most intriguing finding may be the role of activin in embryogenesis. In amphibian systems, activin is supposed to be responsible for mesodermal induction (5-7). Activin induces different regions of dorsoventral tissues in a dose-dependent manner as well as anterior structures in axial pattern formation (8,9). Thus, activin is involved in morphogenesis of early development. It should also be noted that activin acts distinctly from RA in various aspects of cell differentiation. In amphibian develop-* This work was supported by grants from the Ministry of Education, Science and Culture, Japan, and from the Foundation for Promotion of Cancer Research, Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The abbreviations used are: FSH, follicle stimulating hormone; RA, retinoic acid; FBS, fetal bovine serum; TBS, Tris-buffered saline. ment, activin and RA differentially activate separate homeo box genes (10). In cultured mammalian cells, activin inhibits RA-induced differentiation of P19 cells and some types of neuroblastoma cells (11,12). Activin can also inhibit the RAinduced secretion of alkaline phosphatase from osteoblastic cells.' These observations could be explained by the existence of a RA-regulated mediator which interacts with activin.
The follistatin, a protein that can bind with activin and inhibit its activity (13, 14) appears a good candidate for it. Recent studies demonstrating follistatin transcripts in various tissues (15-la), such as kidney or brain, may indicate that follistatin has a significant role in extragonadal tissues. Here, we report some observations which suggest that follistatin is developmentally regulated to interact with activin in neural differentiation.

EXPERIMENTAL PROCEDURES
Cell Culture"P19 cells were obtained from Dr. M. W. McBurney (University of Ottawa) (19). IMR-32 cells were obtained from the Japanese Cancer Research Resources Bank (20). They were usually maintained in a modification of Eagle's minimum essential medium (GIBCO) supplemented with 10% fetal bovine serum (FBS) (Cell Culture Laboratories) and 10 pg/ml gentamycin sulfate (Shering-Plough) at 37 "C in a humidified atmosphere of 5% COz.
Reagents-Recombinant activin-A was produced using Chinese hamster ovary cells as previously described (4). Follistatin was purified from porcine ovaries as previously described (13).
Follistatin Treatment of IMR-32 Cells-IMR-32 cells (104/ml) were plated on 35-mm dishes in 2 ml of a-medium containing 8% FBS. After 24 h, cells were added with 0.3 p~ RA and increasing amounts of follistatin (0, 0.03,0.3,3, 30 nM), 0.3 p~ RA, and 1 nM inhibin, 0.3 p~ RA, and 3 nM follistatin for the first 2 days and 1 nM activin for the last 3 days or no agents. After 5 days, phase-contrast photomicrographs were taken, and the cell number was counted by hemocytometer. In another experiment, cells were treated with follistatin (0, 0.03, 0.3, 3, or 30 nM) alone for 14 days. In this case, medium was changed at day 7 when cells were detached with Ca2+-free phosphatebuffered saline and replated. The concentration of follistatin is expressed in molar units by calculating the molecular mass of follistatin as 35 kDa (21). Cell numbers were statistically analyzed by Duncan's multiple range test.
Colony Formation Assay in Soft Agar-Anchorage-independent cell growth was measured in soft agar in 96-well plates as previously described (3) with some modifications. First, a 50-p1 basal layer consisting of a-medium, 10% FBS, and 0.67% agar was prepared. Next, the middle layer containing 100/well of single cells in 50 pl of a-medium, 10% FBS, and 0.3% agar was overlaid. The top layer consisted of 100 p1 of a-medium containing porcine follistatin at the indicated concentrations. Colonies were scored at day 9 after plating by counting aggregates whose diameter was more than 125 pm. Phasecontrast photomicrographs were simultaneously taken for some samples.
Northern Blotting-P19 cells were treated with 0.3 p~ RA as previously described (19). In brief, cells were cultured in bacterial dishes in the presence of 0.3 p M RA for the first 4 days with a medium change at day 2 and then transferred to tissue culture dishes. 10 pg of poly(A)+ RNA isolated from cells at the indicated times was electrophoresed on agarose gel (1.0%) under denaturing conditions and blotted to a nylon membrane. Mouse activin receptor cDNA probe was produced by reverse transcription-polymerase chain reaction according to the published sequence (22). The sense primer, 888-901 position) in the polymerase chain reaction reaction to generate a putative 237-base pair cDNA fragment. The filter was hybridized with mouse follistatin cDNA probe: activin-receptor cDNA probe, and &actin, respectively. Immunoand Ligand Blottings-Immuno-and ligand blot analyses were done as previously described (13)' with some modifications. P19 cells (106/ml) were incubated in bacterial dishes with or without addition of 0.3 p M RA for 48 h. 10 ml of conditioned medium was collected, added with 1 g of sulfate-Cellulofine (Seikagaku Kogyo), and stirred gently for 12 h at 4 'C. The particles were washed with 0.35 M NaCl, 20 mM Tris-HC1 (pH 7.5) and eluted with 4 ml of 1.5 M NaCl, 20 mM Tris-HC1 (pH 7.5). The elute was dialyzed against 0.15 M NaCl, 20 mM Tris-HC1 (pH 7.5) and concentrated with dialysis and lyophilized. Proteins resolved by sodium dodecyl sulfate-polyacrylamide (10%) electrophoresis were electroblotted onto an Immobilon membranes (Millipore). For immunoblotting, the blots were first blocked for 1 h with 5% skim milk in Tris-buffered saline (TBS) and then incubated with rabbit anti-human follistatin polyclonal antibody (23) for 2 h at room temperature, washed with TBS, incubated with affinity-purified goat antibody to rabbit IgG conjugated with horseradish peroxidase (Sigma) at a 1:lOOO dilution in 1% skim milk in TBS for 2 h at room temperature, and washed with 0.1% Triton X-100 in TBS for 30 min. The proteins were developed with 4-chloro-1-naphthol. As a control, a sample of 10 ml of a-medium containing 10% FBS was treated in the same manner. In the case of ligand blotting, the blots were incubated with Iz5I-activin for 2 h at room temperature, washed with TBS containing 0.1% Triton X-100 for 1 h, and autoradiography was developed for 4 days. 100 ng of purified porcine follistatin was also blotted as a control. IMR-32 cells (105/ml) were grown in 100-mm culture dishes for 5 days to a confluent state. Immuno-and ligand blottings were performed in the same manner using 50 ml of the conditioned medium and the same volume of a-medium containing 10% FBS. Exposure for the autoradiograph was 14 days. Iodination of activin was performed by the chloramine-T method as described elsewhere (24). The specific activity of 1251-activin was about 25,000 cpm/ng.

Among various neural cell lines investigated (IMR-32, GOTO,
NB-1, SK-N-SH, and PC12), IMR-32 cells were found to differentiate in response to porcine follistatin. When cells were treated with 0.3 PM RA alone, the degree of morphological differentiation was not so striking (Fig. lA, a). However, addition of more than 3 nM follistatin markedly stimulated the cellular aggregation and neurite outgrowth (Fig. lA, b). r. . They were electrophoresed on 1.0% agarose gel under denaturing conditions, blotted to nylon membrane, and hybridized with mouse follistatin, mouse activin receptor, and @-actin cDNA probes.
The effect was completely and reversibly abolished by the addition of an equimolar activin (not shown). Follistatin, by itself, also stimulated the neural differentiation in a longerterm culture (Fig. lA, c and d). Although neural differentiation markers for IMR-32 cells were not available, the differentiation-promoting effect was clearly reflected on the cell number. Follistatin significantly inhibited the growth of the cells in a dose-dependent manner either under the presence or absence of RA (Fig. lB, a and b). Inhibin, by contrast, had no effects on the growth and differentiation of IMR-32 cells (Fig. lB, a).

Follistatin Suppresses the Growth of P19 Cells in a Soft Agar
Assay-Our preliminary results showed that follistatin had little effect on the RA-induced neural differentiation of P19 cells. However, follistatin dose dependently suppressed the anchorage-independent growth of P19 cells in soft agar, though the effective concentration was much higher than that employed on the differentiation of IMR-32 cells (Fig. 2).
The Expression of Follistatin and Activin Receptor mRNAs in PI9 Cells-Northern blotting showed an oscillating change of follistatin mRNA during the course of P19 differentiation  follistatin (lane 7). c, ligand blots of the conditioned medium of IMR-32 cells (lane 8) and a-medium + 10% FBS ( l a n e 9 ) . (Fig. 3). The early response to RA was so rapid, which may suggest that the follistatin gene is a RA-regulated gene. After a transient peak of 24 h, the mRNA disappeared within 48 h and became significant again in the later stage. This characteristic pattern was reproducible. By contrast, the activin receptor mRNA was increased not rapidly but gradually at a later stage. The Production of Follistatin in the Conditioned Medium of P19 and IMR-32 Cells-Next, we demonstrated the presence of follistatin at the protein level. Fig. 4a shows the existence of molecules immunoreactive to follistatin in the medium of undifferentiated P19 cells and a drastic decrease after treatment with RA. Two forms of follistatin (32 and 35 kDa) are known to be produced by alternative splicing (21), and they are further modified with glycosylation (13, 25). The pattern of bands derived from the conditioned medium of PI9 cells is similar to that of pituitary, where the 35-kDa band is more predominant (14). Ligand blotting using lZ5I-activin also showed a consistent result, although the sensitivity of detection was much higher than immunoblotting (Fig. 2b). Immunoblotting using a polyclonal antibody which recognizes the common peptide sequence between human and mouse follistatin could not detect the band in IMR-32 cells (data not shown). However, ligand blotting by long-exposure autoradiography showed that the follistatin band in IMR-32 cells was stronger than that of the control serum (Fig. 4c). Thus, the expression of follistatin in IMR-32 cells seemed to be much lower than P19 cells.

DISCUSSION
We and others have observed a series of activin effects on neural cells. Activin acts as a potent mitogen on PI9 cells in an extracellular matrix-dependent manner (11,26). Activin is also inhibitory on the neural differentiation of RA-treated P19 cells and some neuroblastoma cell lines (11,12). Furthermore, activin acts as a survival factor for other types of neuronal cells (27). These observations suggest that activin plays significant roles in the nervous system. In pituitary, the secretion of FSH is controlled by activin and its related proteins. The FSH stimulatory effect of activin is neutralized by follistatin as an activin binding protein (14). Recent studies have also suggested the presence of follistatin transcripts in nervous system (16). Therefore, the analysis of follistatin kinetics is indispensable to the understanding of the activin actions in the nervous system.
We first examined the effect of exogenously added porcine follistatin on a variety of neural cell lines. As expected from the observations in pituitary, the effect of follistatin was almost contrary to the activin actions. Follistatin, by itself or synergistically with RA, promoted the differentiation of IMR-32 cells (Fig. 1). It also inhibited the anchorage-independent growth of P19 cells in soft agar but at rather high concentrations (Fig. 2). Although RA-treated P19 cells and IMR-32 cells are different neuronal cell types, these observations suggest that the proliferation and differentiation of these neural cells are controned by both activin and follistatin. Follistatin exogenously added to IMR-32 cells may have changed the subtle balance of activin and follistatin and caused cells to turn toward differentiation. The specific balance of activin and follistatin may also be required to keep P I 9 cells in a proliferating state.
Then, where is the origin of activin and follistatin? First, the endogenous production of activin in both P19 and IMR-32 cells has not been detected by either bioassay as erythroid differentiation factor activity or Northern blot analy~is.~ Our previous observations showed that a low serum condition made P19 cells aggregated and differentiated, but the effect was counteracted by the addition of activin (11). Therefore, we suppose that the majority of activin activity should be derived from FBS in this culture system.
By contrast, we could demonstrate the unique expression M. Hashimoto and Y. Eto, unpublished data. pattern of follistatin during the neural differentiation of PI9 cells. Northern blot analysis has suggested that the mRNA expression of follistatin is dynamically controlled by RA while that of activin receptor was gradually increased (Fig. 3).
Immuno-and ligand blotting5 demonstrated the presence of follistatin in the conditioned medium of both P19 and IMR-32 cells (Fig. 4). Although the physiological meaning of the early transient rise of follistatin mRNA after RA treatment is not clear, the expression of follistatin protein was downregulated at the early stage of differentiation. These observations, indicating that these neural cells produce follistatin and activin receptor which interact with the exogenously derived activin in a developmentally regulated manner, seem to be in line with the observations in -gonadal tissues. It has been confirmed that granulosa cells in ovary produce follistatin (28), whereas the origin of activin is still obscure. It is intriguing if a similar mechanism of differentiation is employed in these two different systems. The physiological role of follistatin also remains to be disclosed. Recently, follistatin was characterized as a heparin binding protein, suggesting that it regulates activin action by associating with proteoglycan on the cell surface (29). Further investigations of the follistatin functions will be required to clarify the divergent effects of activin.