Role of TSC-22 during early embryogenesis in Xenopus laevis

Authors

  • Akiko Hashiguchi,

    1. Department of Biological Sciences, Graduate School of Science, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8654,
    Search for more papers by this author
  • Koji Okabayashi,

    1. International Cooperative Research Project (ICORP), Japan Science and Technology Corporation (JST) and
    Search for more papers by this author
  • Makoto Asashima

    Corresponding author
    1. International Cooperative Research Project (ICORP), Japan Science and Technology Corporation (JST) and
    2. Department of Life Sciences (Biology), Graduate School of Arts and Sciences, The University of Tokyo, 3-8-1 Komaba, Meguro-ku, Tokyo 153-8902, Japan.
    Search for more papers by this author

*Author to whom all Correspondence should be addressed.
Email: asashi@bio.c.u-tokyo.ac.jp.

Abstract

Transforming growth factor-β1-stimulated clone 22 (TSC-22) encodes a leucine zipper-containing protein that is highly conserved. During mouse embryogenesis, TSC-22 is expressed at the site of epithelial–mesenchymal interaction. Here, we isolated Xenopus laevis TSC-22 (XTSC-22) and analyzed its function in early development. XTSC-22 mRNA was first detected in the ectoderm of late blastulae. Translational knockdown using XTSC-22 antisense morpholino oligonucleotides (XTSC-22-MO) caused a severe delay in blastopore closure in gastrulating embryos. This was not due to mesoderm induction or convergent-extension, as confirmed by whole-mount in situ hybridization and animal cap assay. Cell lineage tracing revealed that migration of ectoderm cells toward blastopore was disrupted in XTSC-22-depleted embryos, and these embryos had a marked increase in the number of dividing cells. In contrast, cell division was suppressed in XTSC-22 mRNA-injected embryos. Co-injection of XTSC-22-MO and mRNA encoding p27Xic1, which inhibits cell cycle promotion by binding cyclin/Cdk complexes, reversed aberrant cell division. This was accompanied by rescue of the delay in blastopore closure and cell migration. These results indicate that XTSC-22 is required for cell movement during gastrulation though cell cycle regulation.

Introduction

Transforming growth factor-β stimulated clone-22 (TSC-22) encodes a leucine zipper-containing protein that is highly conserved between many species. It was first isolated as a target gene of TGF-β1 in mouse osteoblastic cells (Shibanuma et al. 1992). Transcription of TSC-22 is upregulated by a variety of growth factors, including follicle-stimulating hormone (Hamil & Hall 1994), fibroblast growth factor-2 (Kawa-uchi et al. 1995) and progesterone (Kester et al. 1997). Embryonic development requires the complex interaction of different growth factors to generate morphogenetic fields (Hogan 1999), which must require the fine coordination of multiple extracellular signals by specific regulatory factors. One such factor is Drosophila bunched, a homologue of TSC-22, which has been shown to set the boundary between bone morphogenetic protein (BMP)- and epidermal growth factor (EGF)-induced cell fates (Dobens et al. 1997). During mouse embryogenesis, TSC-22 is expressed at sites of epithelial–mesenchymal interactions (Dohrmann et al. 1999).

TSC-22 is also a potential tumor suppressor gene. Human TSC-22 was re-isolated in a screen of anticancer drug responsive genes (Kawamata et al. 1998), and was shown to downregulate cell proliferation, and induce apoptosis in human salivary gland cells (Nakashiro et al. 1998; Hino et al. 2000). In addition, TSC-22 is upregulated in growth-inhibited breast cancer cells (Kester et al. 1997) and inactivated in human brain tumors (Shostak et al. 2003). TSC-22 inhibits cell growth by upregulating expression of the p21 gene, a cyclin-dependent kinase (Cdk) inhibitor which inhibits Cdk by binding cyclin/Cdk complexes (Gupta et al. 2003). Modulation of the cell cycle is closely related to differentiation. In Xenopus, promotion of cell cycle by overexpression of cyclin A2 and Cdk2 inhibits differentiation of skin and primary neurons (Richard-Parpaillon et al. 2004). Depletion of the cell cycle inhibitor p27Xic1 also impairs neuronal differentiation (Carruthers et al. 2003; Vernon et al. 2003). Abnormal cell division also affects morphogenesis. Increased cell division induced by depletion of Wee1 and Wee2, tyrosine kinases that inhibit Cdc2, causes a delay in gastrulation and anterior–posterior embryo elongation, respectively (Leise 3rd & Mueller 2004; Murakami et al. 2004).

Here, we have isolated Xenopus laevis TSC-22 (XTSC-22) to clarify its role during early embryogenesis. Using specific antisense morpholino oligonucleotides (MO) to block XTSC-22 translation, we show that XTSC-22 is required for cell movement in gastrulation via regulation of the cell cycle.

Materials and Methods

Embryo manipulation

Eggs were obtained from adult Xenopus laevis injected with 300 units of human chorionic gonadotrophin (Gestron; Denka Seiyaku, Kawasaki, Japan). Fertilized embryos were dejellied with 4% cysteine hydrochloride in Steinberg's solution and allowed to develop in 10% Steinberg's solution. Embryos were staged according to Nieuwkoop & Faber (1956).

Cloning of XTSC-22

Polymerase chain reaction (PCR) was carried out using a Xenopus tailbud (Stage 32) cDNA. Primers were designed based on expressed sequence tag (EST) sequences that showed homology with mouse TSC-22: forward 5'-GCAGCTATGGATCTTGGGGTC-3', reverse 5'-GTGTGAGTAACTGTGTTGAGG-3'. An amplified 0.31 kb fragment was cloned and used as a probe to screen a Xenopus gastrula (Stage 10/12) cDNA library.

Microinjections of embryos

Capped mRNAs were made by in vitro transcription using an SP6 mMASSAGE mMACHINE kit (Ambion, Austin, TX, USA). For in vitro transcription, pCS2+-XTSC-22, pSP64T-BMP-4 (Nishimatsu et al. 1992), pCS2+-Xnr-1 (Takahashi et al. 2000), and pCS2+-β-gal were used. Antisense MO were obtained from Gene Tools (Philomath, OR, USA) LLC: TSC-22-MO, 5'-AGCTGACAGACCCCAAGATCCATAG-3'.

For animal cap assay, morphological analysis, whole-mount in situ hybridization and immunohistochemistry, in vitro synthesized RNAs or MO were injected into both blastomeres of 2-cell-stage embryos. Cell lineage tracing was performed according to Murakami et al. (2004) with some modifications. Briefly, RNA or MO was injected into both cells of 2-cell-stage embryos. β-galactosidase (β-gal) mRNA was then injected into one of the B1 blastomeres at the 32-cell-stage. The migration pattern was determined by β-gal staining at stage 12.5. β-gal was visualized by Red-gal (Research Organics, Cleveland, OH, USA).

Growth factors and antibody

Human recombinant TGF-β1 and basic fibroblast growth factor (FGF) were purchased from Wako Pure Chemical Industries (Osaka, Japan). Human recombinant activin A was a kind gift from Dr Yuzuru Eto of Ajinomoto (Tokyo, Japan). These factors were used in the following concentrations: 100 ng/mL TGF-β1, 5 or 10 ng/mL activin A, 100 ng/mL bFGF. Anti-phospho-histone H3 antibody was purchased from Upstate (Charlottesville, VA, USA).

Reverse transcriptase (RT)-PCR Analysis

Total RNA was extracted with an ISOGEN kit (Nippongene, Tokyo, Japan) from embryos or animal caps. First-strand cDNA was synthesized from 1 µg RNA with an Oligo-(dT) primer using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA). A negative control excluding enzyme was included for each sample. A total of 3% of the cDNA obtained was used for each PCR reaction. Primers used were as follows: XTSC-22, forward 5'-CGGTTACAGTGGACAGTAGC-3', reverse 5'- GTGTGAGTAACTGTGTTGAGG-3'; ODC, forward 5'-GTCAATGATGGAGTGTATGGATC-3', reverse 5'-TCCATTCCGTCTCCTGAGCAC − 3'; Xbra, forward 5'-GGATCGTTATCACCTCTG-3', reverse 5'-GTGTAGTCTGTAGCAGCA-3'; Chordin, forward 5'-AACTGCCAGGACTGGATGGT-3', reverse 5'-GGCAGGATTTAGAGTTGCTTC-3'.

TUNEL assay

The whole-mount terminal deoxyribonucleotidyl transferase-mediated dUTP–digoxigenin nick end-labeling (TUNEL) assay was as described by Hensey & Gautier (1997). Embryos were fixed for 2 h in MEMFA (100 mm MOPS, pH 7.4, 2 mm EGTA, 4% formaldehyde) and dehydrated in methanol. The assay was conducted with 150 units/mL of terminal deoxynucleotidyltransferase (Invitrogen) and 0.5 µM digoxygenin-dUTP (Roche). Visualization was obtained using alkaline phosphatase-conjugated anti-digoxygenin antibody (Roche). Staining was carried out with BM purple (Roche).

Whole-mount in situ hybridization and immunohistochemistry

For whole-mount in situ hybridization, the protocol of Harland (1991) was used with some modifications, including the use of BM purple (Roche Diagnostics, Basel, Switzerland) as substrate and without RNase treatment. An XTSC-22 probe was prepared by linearizing pBluescriptII(SK +)-XTSC-22 with NotI and transcribing with T7 RNA polymerase.

Anti-phosopho-histone H3 staining was performed as described by Nechiporuk & Keating (2002), using a 1:100 dilution of the antibody.

Results

Isolation and sequence analysis of Xenopus TSC-22 cDNA

Approximately 600 000 clones from a Xenopus gastrula library were screened with PCR-generated probes. We obtained a clone with an insert of approximately 1.9 kb which encoded an open reading frame (ORF) of 411 nucleotides, giving a putative protein of 137 amino acids. There was a stop codon at the fifth amino acid upstream of the first methionine. The protein shared 79% identity with mouse TSC-22, with conservation of the leucine zipper and adjacent N-terminal region, which has been designated the TSC-box (Fig. 1a). It lacked the first 10 amino acids of mouse TSC-22. A database search revealed highly similar amino acid sequences in Xenopus tropicalis, a close relative of X. laevis (Fig. 1b). These features identify the isolated cDNA as a Xenopus ortholog of TSC-22 (accession no. AB191720).

Figure 1.

Sequence of Xenopus TSC-22. (a) Structure of XTSC-22 protein. LZ, leucine zipper; RD1 and RD2, repression domain 1 and 2. (b) Comparison of TSC-22 proteins from Xenopus laevis (accession no. AB191720), Xenopus tropicalis EST (AL776759), mouse (X62940) and human (U35048). White letters on black background indicate amino acid residues that are conserved across species. Note that the leucine zipper and TSC-box is 100% identical to mammalian TSC-22.

Pattern of expression of XTSC-22 in the developing embryo

Reverse transcription-PCR did not detect any maternal XTSC-22 component, but XTSC-22 transcripts were clearly detectable at late blastula (stage 9) and expression was maintained to the tadpole stage (Fig. 2a). Whole-mount in situ hybridization with XTSC-22 probes showed heavy staining of the entire ectodermal region at the blastula stage (Fig. 2b,c). During neurulation, staining was weakly concentrated in the neural plate, but was not evident in the rest of the embryo (Fig. 2d). This expression was localized mainly in the region of the eyes, and brain (Fig. 2e,f). Additional expression was detected in the dorsal epidermis.

Figure 2.

Expression pattern of TSC-22 (a) Temporal expression pattern of XTSC-22 mRNA detected by reverse transcription-polymerase chain reaction (RT-PCR). Transcripts were first detected at late blastula (stage 9) and persisted until tadpole stage (stage 35). Ornitine decarboxylase (ODC) expression serves as the quantitative control. RT-PCR was carried out to avoid contamination with genomic DNA. (b–f) Spatial expression pattern of XTSC-22 determined by whole-mount in situ hybridization. At stage 10 (b,c), the ectoderm was stained. (c) Sagittal section of the stage-10 embryo in (b). The black lines indicate the blastopore. At neurula stage (d,e), transcripts were observed in the neural region including eyes and brain. This localization was clear at stage 35 (f). Bars, 0.3 mm (g) Induction of XTSC-22 transcription by growth factors: there was no change in XTSC-22 expression with any treatment. WE, whole embryo.

Induction of the XTSC-22 gene

Mammalian TSC-22 was originally identified as a TGF-β1-responsive gene, and it may be up- or downregulated by several growth factors that play important roles in development. We tested the effects of a number of growth factors on XTSC-22 expression. Animal cap explants were treated with TGF-β1, activin or bFGF for three hours, or mRNA for BMP-4 or Xnr-1 were injected into 2-cell-stage embryos and animal caps were excised. Consistent with the results of whole-mount in situ hybridization, XTSC-22 transcripts were detected in untreated animal caps. None of these treatments had any effect on XTSC-22 mRNA levels, suggesting that the level of expression of XTSC-22 is strictly regulated in animal cap explants (Fig. 2g).

Overexpression of XTSC-22 leads to apoptosis

To determine the role of XTSC-22 in early development, we overexpressed XTSC-22 by injecting embryos with XTSC-22 mRNA and allowed to develop. As gastrulation proceeded, cells were excluded into the vitelline space in these embryos (Fig. 3a,b), and there was also accumulation of excluded cells into the blastocoel (data not shown). In Xenopus development, expulsion of cells from embryos is an indicator of apoptosis. TUNEL staining confirmed that these cells had undergone apoptosis (Fig. 3c,d).

Figure 3.

Overexpression of XTSC-22 resulted in apoptosis (a,b) Morphology of control embryos at stage 11 (a) and XTSC-22 mRNA-injected embryos (b). Expulsion of cells at animal pole was observed in XTSC-22 injected embryos. (c,d) Terminal deoxyribonucleotidyl transferase-mediated dUTP–digoxigenin nick end-labeling (TUNEL) staining of control embryos (c) and XTSC-22 injected embryos (d). TUNEL-positive cells were seen in XTSC-22-injected embryos (d). Bars, 0.5 mm.

XTSC-22 depletion affects epiboly

To investigate the function of XTSC-22 in more detail, we designed an antisense MO directed against XTSC-22 (XTSC-22-MO). MO have been shown to prevent translation of their target mRNAs by binding to sequences near the initiation methionine (Heasman et al. 2000). Injection of XTSC-22-MO caused gastrulation defects. By stage 11, it was evident that embryos injected with MO had a delay in blastopore closure (Fig. 4a). This gastrulation defect caused by XTSC-22 depletion could be rescued by co-injection of XTSC-22 mRNA, where the XTSC-22-MO binding sequence had been slightly modified so that translation is not prevented by the MO (Fig. 4a,b).

Figure 4.

XTSC-22 depletion affects gastulation movements (a) Morphology of XTSC-22-morpholino oligonucleotides (MO)-injected embryos at stage 11. XTSC-22 depletion led to delayed gastrulation. This phenotype was rescued by co-injection of XTSC-22 mRNA. (b) Frequency of phenotypes seen in experiment shown in (a). (c) Cell lineage tracing of XTSC-22-MO-injected embryos at stage 12. Epiboly was perturbed in XTSC-22-depleted embryos. Co-injection of RNA with XTSC-22-MO partially rescued the perturbation. (d) Animal cap assay using explants excised from uninjected embryos and XTSC-22-MO-injected embryos. Explants treated with activin elongated whether the explants were injected with the MO or not. (e) Expression of mesodermal gene in control embryos and XTSC-22-MO-injected embryos. No difference in expression of Xbra and Chordin was observed. (f) Expression of mesodermal gene in activin-treated animal caps from uninjected embryos and XTSC-22-MO-injected embryos. Activin treatment induced Xbra and Chordin in both uninjected and XTSC-22-MO injected animal caps.

During gastrulation, the ectoderm expands to cover the entire embryo, a process called epiboly, which is essential for the segregation of the three germ layers. Because TSC-22 transcripts accumulated in the ectoderm of gastrulae (Fig. 2b,c), we investigated epiboly in XTSC-22-depleted embryos by cell lineage tracing (Murakami et al. 2004). At stage 12, cells derived from one of the B1 blastomeres migrated toward the blastopore to form a narrow column (Fig. 4c). In contrast, in MO-injected embryos these cells formed a broad, belt-like band, indicating that epiboly was impaired in XTSC-22-depleted embryos. Co-injection of XTSC-22 RNA could partially reverse the effect of XTSC-22 depletion. In these embryos, B1 cell progeny migrated to form a line (Fig. 4c).

Gastrulation movements include both epiboly of ectoderm and convergent-extension of mesodermal tissue. Animal cap cells explanted at the blastula stage (stage 8.5) can be induced to form mesodermal tissue by activin. Induction of mesoderm causes an elongation of the explants. This process is thought to imitate convergent-extension movements (Keller & Danilchik 1988). Therefore, we examined the effect of XTSC-22 depletion on convergent-extension using animal cap explants. Treatment with activin-induced control animal caps to elongate. Such elongation was also observed in XTSC-22 depleted animal caps (Fig. 4d).

Whole-mount in situ hybridization showed that the gastrulation delay was not due to inhibition of mesodermal gene induction. We found that Xbra was expressed around the blastopore and chordin was expressed in the region of the Spemann's organizer, regardless of MO injection (Fig. 4e). RT-PCR analysis performed with activin-treated animal caps further confirmed that XTSC-22-depletion did not affect mesodermal gene induction (Fig. 4f).

XTSC-22 depletion promotes cell division in gastulating embryos

The gastrulation process requires tight regulation of cell migration, mesodermal gene transcription and cell division. Recent studies have shown the importance of cell cycle regulation in gastrulation. In Xenopus embryos, cell divisions (cell cycles 2–12) are rapid until they reach late blastula stage (stage 9). These cycles consist of only the S and M phases. After midblastula transition (MBT), the cycles come to include gap phases (Newport & Kirschner 1982; Kimelman et al. 1987). Transition to a normal cell cycle requires phosphorylation of the Wee1 tyrosine kinase, which inhibits the CyclinB/Cdc2 complex by phospholylating Cdc2. Translational knockdown of Wee1 causes cell cycle promotion and gastrulation delay (Murakami et al. 2004). Mammalian TSC-22 has been shown to inhibit cell proliferation in various carcinoma cell lines (Kester et al. 1997; Kawamata et al. 1998; Nakashiro et al. 1998; Shostak et al. 2003), and Gupta et al. (2003) proved that this growth inhibition is in part due to its ability to increase expression levels of the p21 gene, a Cdk inhibitor that inhibits Cdk by binding to a complex of cyclins and Cdk.

Taken together, these suggest that XTSC-22-depleted embryos might contain an increased number of dividing cells. Therefore, we examined whether XTSC-22 depletion affects cell division using antiphospho-histone H3 antibody (αPH3), a mitotic marker. Embryos injected with XTSC-22-MO showed a marked increase in the number of dividing cells (Fig. 5a). This increase could be reversed by co-injection of XTSC-22 RNA. In addition, αPH3 staining of embryos injected with XTSC-22 RNA alone showed a strong reduction in the number of mitotic cells (Fig. 5a).

Figure 5.

XTSC-22 depletion affects gastrulation movements through cell cycle promotion (a) αPH3 staining of XTSC-22-MO injected embryos. The number of αPH3-positive cells increased in XTSC-22-depleted embryos. This increase was reversed by co-injection of XTSC-22 mRNA. The number of αPH3-positive cells decreased when XTSC-22 mRNA alone was injected. (b) αPH3 staining of embryos injected with XTSC-22-MO and p27Xic1 mRNA. Single injection of p27Xic1 mRNA led to a decrease in the number of αPH3-positive cells. The increase in number of dividing cells by XTSC-22 depletion was reversed by co-injection of p27Xic1 mRNA. (c) Morphology of embryos injected with XTSC-22-MO and p27Xic1 mRNA. p27Xic1 alone did not affect gastrulation. Delay of blastopore closure seen in XTSC-22-depleted embryos was rescued by co-injection of p27Xic1 mRNA. (d) Frequency of phenotypes seen in experiment shown in (c). (e) Cell lineage tracing of rescued embryos. Disruption of epiboly movement in XTSC-22-depleted embryos was rescued in embryos injected with XTSC-22-MO and p27Xic1 mRNA.

Co-injection of p27Xic1 cell cycle inhibitor rescues gastrulation defects

The results of αPH3 staining suggested that the gastrulation delay in XTSC-22-depleted embryos arose from abnormal cell division in the ectoderm. Based on this hypothesis, we attempted a rescue experiment with p27Xic1, the only CDK inhibitor found in Xenopus (Su et al. 1995). Single injection of p27Xic1 mRNA inhibited cell division, and led to a decrease in the number of αPH3-positive cells. When co-injected with XTSC-22-MO, p27Xic1 suppressed the cell cycle promotion caused by XTSC-22-depletion (Fig. 5b). This was accompanied by a rescue of the delay of blastopore closure (Fig. 5c,d). We next investigated whether p27Xic1 also rescued the epiboly movement in the ectoderm. Cell migration was traced as described above. In XTSC-22-MO-injected embryos, B1 progeny were widely distributed through the embryo. In contrast, cells derived from the B1 blastomere in p27Xic1 mRNA co-injected embryos migrated to converge on the dorsal midline, similar to uninjected embryos (Fig. 5e). These results confirmed that the phenotype seen in XTSC-22 depleted embryos was due to abnormal cell division.

Discussion

We have shown that XTSC-22, a Xenopus ortholog of TSC-22, is crucial for correct cell movement during gastrulation, especially in epiboly, through regulation of the cell cycle. Expression of XTSC-22 was initiated just before gastrulation, and was not affected by treatment with any of a number of growth factors including TGF-β1, suggesting that expression of this gene must be maintained at very strict levels (Fig. 2). Consistent with this, perturbation of the XTSC-22 levels by overexpression by amounts as low as 30–40 pg in the animal pole caused extensive apoptosis (Fig. 3). Taken together, these suggested that XTSC-22 may play some role in the ectoderm of gastrula embryos.

Translational knockdown of XTSC-22 resulted in delayed gastrulation (Fig. 4a). In these embryos, ectodermal cells did not migrate properly (Fig. 4c). However, other important driving forces of gastrulation, such as convergent-extension and mesoderm induction, were not affected (Fig. 4d–f). αPH3 staining showed that the rate of cell division was increased in XTSC-22-depleted embryos and decreased in embryos in which XTSC-22 was overexpressed (Fig. 5a), which is consistent with the fact that mammalian TSC-22 can inhibit cell growth in carcinoma cell lines (Nakashiro et al. 1998; Gupta et al. 2003). Rescue experiments performed with the cell cycle inhibitor p27Xic1 confirmed that the gastrulation delay depends on accelerated cell division (Figs 5b,c). We may therefore conclude that XTSC-22 is required for gastrulation through its ability to suppress cell cycle progression.

Abnormal cell division is well known to affect morphogenetic movements. Wee1 tyrosine kinase is an inhibitor of Cdc2 and is expressed in cleavage- and blastula-stage embryos. Knockdown of Wee1 tyrosine kinase causes cell cycle progression and leads to a delay in gastrulation (Murakami et al. 2004). In these embryos, both epiboly and convergent-extension were disrupted. In contrast, XTSC-22 depletion only affected epiboly (Fig. 4c). This result could be explained by the observed accumulation of TSC-22 in the ectoderm of gastrulae (Fig. 2b,c). Tissue-specific regulation of cell division is an important mechanism in morphogenesis. In Xenopus, cell division occurs uniformly until the late blastula stage. Once gastrulation begins, there are no mitotic cells in the involuted dorsal axial mesoderm, which undergoes convergent-extension movement (Saka & Smith 2001). Suppression of cell division in this region is mediated by the Wee2 tyrosine kinase, a postgastrula form of Wee1, which is specifically expressed in presumptive notochord (Leise 3rd & Mueller 2002). Knockdown of Wee2 blocked convergent-extension and resulted in insufficient axis elongation without affecting gastrulation (Leise 3rd & Mueller 2004). Our results suggest that XTSC-22 plays a similar role in the ectoderm of the developing Xenopus embryo, where it also regulates the cell cycle to control morphogenetic movements of that tissue.

Mammalian TSC-22 regulates the cell cycle by increasing p21 mRNA levels (Gupta et al. 2003). However, no p21 gene has been reported in X. laevis to date. The only Cdk inhibitor in Xenopus is p27Xic1. p27 is closely related to p21 but is not regulated transcriptionally. If TSC-22 functions as a transcription factor, it may have a different target gene, such as cyclin or Cdk which regulate the cell cycle and have unique spatial expression patterns during development (Vernon & Philpott 2003).

Acknowledgements

This work was supported by ICORP (International Cooperative Research Program) of the Japan Science and Technology Agency and by Grants-in-Aid for Scientific Research from Ministry of Education, Science, Sports and Culture of Japan. We thank Dr Yuzuru Eto of the Central Research Laboratories of Ajinomoto for providing human recombinant activin A.

Ancillary