Eomesodermin Requires Transforming Growth Factor-β/Activin Signaling and Binds Smad2 to Activate Mesodermal Genes*

The T-box gene Eomesodermin (Eomes) is required for early embryonic mesoderm differentiation in mouse, frog (Xenopus laevis), and zebrafish, is important in late cardiac development in Xenopus, and for CD8+ T effector cell function in mouse. Eomes can ectopically activate many mesodermal genes. However, the mechanism by which Eomes activates transcription of these genes is poorly understood. We report that Eomes protein interacts with Smad2 and is capable of working in a non-cell autonomous manner via transfer of Eomes protein between adjacent embryonic cells. Blocking of Eomes protein transfer using a farnesylated red fluorescent protein (CherryF) also prevents Eomes nuclear accumulation. Transfer of Eomes protein between cells is mediated by the Eomes carboxyl terminus (456-692). A carbohydrate binding domain within the Eomes carboxyl-terminal region is sufficient for transfer and important for gene activation. We propose a novel mechanism by which Eomes helps effect a cellular response to a morphogen gradient.

capped mRNA was synthesized as previously described (9) using a Megascript T3 kit (Ambion). GSM series cDNAs were synthesized in similar splice-PCRs using mutated forward and reverse oligonucleotides in which each 18-base window encoding 6 glycines was flanked 5Ј and 3Ј with 18 bases of the appropriate wild type Eomes cDNA sequence. Deletion series cDNAs were performed similarly but with the region to be deleted missing between the two 18-base flanks of Eomes cDNA.
Xenopus Embryo Injection and Culture-Xenopus embryos were staged according to Ref. 31. Egg laying, fertilization, and embryo culture were performed as described previously (9), except that culture media contained 5 g/ml Gentamycin (Invitrogen). After de-jellying in 2% cysteine hydrochloride (pH 8.0; Sigma), fertilized eggs were placed in 1ϫ modified Barth's saline (MBS) containing 2% Ficoll (Sigma) for 6 min prior to microinjection. Synthetic mRNA was injected at the 2-to 8-cell stage into the animal pole of both blastomeres (9.2 nl/embryo; for co-immunoprecipitation (IP), 9.2 ng of Eomes or Eomes mutant mRNA) using a Drummond Nanoject II microinjector, and embryos were harvested at stage 10.5. Two hours after injection, embryos were transferred to 0.1ϫ MBS. For animal cap assays, caps were dissected in 1ϫ MBS at stage 8 and cul-tured to stage 10.5, frozen in 1ϫ MBS on dry ice, and stored at Ϫ80°C.
RNA Preparation from Xenopus laevis Animal Caps-Total RNA was isolated from whole embryos and animal caps using the phenol/NETS methods of Sambrook et al. (32) as previously described (9) with the following changes. The proteinase K digestion and LiCl precipitation steps were omitted, and the composition of the NETS homogenization buffer was 0.3 M NaCl, 1 M EDTA, 50 mM Tris (pH 7.5), 1% SDS. Cellular DNA was removed by treatment with RQ1 DNase I (Promega) for 1 h at 37°C followed by phenol extraction.
RNase Protection Assay-RNase protection assays were performed as previously described (9).
cDNA Synthesis and PCR (RT-PCR)-cDNA was synthesized for 1 h at 50°C using a ThermoScript kit (Invitrogen) according to the manufacturer's instructions using 1-2 g of total RNA as template and an oligo(dT) primer. The cDNA was diluted 5-fold with TE (pH 7.5) before PCR. PCR was performed using 2 l of 5-fold diluted cDNA template plus 0.2 l of Taq DNA polymerase (Qiagen); heated to 95°C for 3 min (1 cycle); then thermally cycled at 95°C for 30 s, 55°C for 1 min, 68°C for 1 min (30 cycles). Gene-specific primer pairs were designed based on published sequence data (GenBank). Primer sequences are shown in Table 1. All PCR products were sequenced for verification. All experiments were repeated at least three times.
Recombinant His 6 -Eomes protein was overexpressed in Escherichia coli strain SG13009 (pREP4): one single colony was inoculated and grown in 20 ml of LB broth containing 100 g/ml ampicillin and 25 g/ml kanamycin at 28°C overnight with agitation. One liter of LB containing ampicillin, kanamycin, and 0.2% glucose, was inoculated 1:50 with the overnight culture and grown 2-3 h at 28°C to A 600 ϭ 0.6. His 6 -Eomes protein overexpression was induced by addition of isopropyl 1-thio-␤-D-galactopyranoside to 2 mM and culture continued at 28°C for 4 h. Bacteria were harvested by centrifugation, and resuspended in buffer A (100 mM NaH 2 PO 4 , 10 mM Tris-HCl, 6 M guanidine hydrochloride, Invitrogen number 15502-016), pH 8.0). Cells were lysed by magnetic stirring for 1 h at room temperature. Protein extract was cleared by centrifugation at 10,000 ϫ g for 30 min at room temperature and applied to nickel-nitrilotriacetic acid resin. The resin was washed twice in Buffer C (100 mM NaH 2 PO 4 , 10 mM Tris-HCl, 8 M guanidine hydrochloride, pH 6.3) and His 6 -Eomes protein was eluted in Buffer E (100 mM NaH 2 PO 4 , 10 mM Tris-HCl, 8 M guanidine hydrochloride, pH 4.5).
Immunoprecipitation and Western Blot Analysis-Immunoprecipitation (IP) and Western blot analysis were performed as described previously (33,34), except that okadaic acid was not used in the IP buffer (34); extracts were pre-cleared 1 h (h) at 4°C with agitation using 10 l of empty Protein G-Sepharose beads (GE Healthcare, 17-0618-01); IP antibody (0.8 g) was allowed to bind overnight with 100 l of embryo extract at 4°C with agitation, then 5 l of pre-washed empty beads were added and allowed to bind 2 h at 4°C; after boiling beads in Laemmli sample buffer, before loading gels, excess IgG was removed from buffer by incubating with 10 l of empty beads for 2 h at 25°C and 1400 rpm in a Thermomixer R (Eppendorf); and the Protein G band was reduced on blots after incubation in protein A-horseradish peroxidase by washing three times for 10 min plus once for 2 h in TBST (140 mM NaCl, 10 mM Tris-HCl, pH 7.5, 0.1% Tween 20 (Bio-Rad)). Antibodies were diluted 1:1000 in TBST containing 5% milk and allowed to bind to polyvinylidene difluoride (Hybond-P, Amersham, RPN303F) Western blot membrane at 4°C overnight. Other reagents used were: Smad2 antibody (Upstate/Millipore, 07-408), anti-GFP (Sigma, G6539); recombinant protein A conjugated to horseradish peroxidase (Pierce, 32400, 1:5000); ECL plus reagent (GE Healthcare/Amersham, RPN2132).
Pre-adsorption of the Anti-NH 2 -terminal Eomes Antibody with Eomes Antigen-Eomes antigen-coupled Sepharose 4B affinity columns were prepared, and Eomes antibodies were affinity purified, by Sigma Genosys. A BSA coupling reaction was performed as for the Eomes NH 2 -terminal antigen to prepare negative control BSA-coupled Sepharose 4B beads. 0.8 g of CNBr-activated Sepharose 4B (GE Healthcare) was hydrated on a sintered glass filter by washing with 1 mM HCl, then coupling buffer (0.1 M NaHCO 3 , 0.25 M NaCl, pH 8.5). For the coupling reaction, 18 mg of BSA (Sigma Fraction V) was dissolved in coupling buffer, mixed with the gel suspension, and incubated with mixing for 2 h (h) at room temperature. After coupling of BSA to the Sepharose, any remaining active groups were quenched for 2 h at room temperature in 0.2 M glycine (Bio-Rad) (pH 8.1) using 1:10 (v/v) of gel to buffer. The BSAcoupled Sepharose 4B (BSA-Sepharose) was washed four times with coupling buffer, then with 0.1 M sodium acetate (pH 4.3) (HCl) containing 0.5 M NaCl. BSA-Sepharose was pre-equilibrated using 1ϫ phosphate-buffered saline (pH 7.7) prior to addition of the anti-Eomes antibody. One ml of a 50% slurry of BSA-Sepharose, or Eomes-Sepharose (Eomes amino acid residues 1-214; Sigma Genosys) was incubated with 5 ml of the affinity-purified anti-Eomes NH 2 -terminal antibody (1.2 mg/ml) for 4 h at 4°C. As a negative control, buffer alone was incubated with Sepharose (no added antibody). Supernatant was recovered after centrifugation of beads at 3000 ϫ g at 4°C for 2 min and used to probe Western blots.
Confocal Microscopy-Confocal microscopy of Xenopus animal pole explants was performed using a Leica DMIRE2 confocal microscope and Leica confocal software, within the Pathology Core Facility of The Children's Hospital of Philadelphia.

RESULTS
An Anti-Eomes Antibody Recognizes the Eomes Protein-To examine the molecular mechanism by which Eomes activates downstream target genes, we wished to identify candidate protein partners for Eomes. Antisera were raised against the Xenopus Eomes amino (NH 2 ) terminus (33) and in parallel, against its central DNA-binding region and carboxyl (COOH) terminus (this report) by overexpression of histidine (His 6 -) tagged Eomes protein in bacteria. The NH 2 -and COOH-terminal antisera were affinity-purified on Eomes-Sepharose 4B columns. Both of the resulting anti-Eomes antibody preparations recognized their respective bacterially produced antigens on Western blots (data not shown); and also recognized an endogenous polypeptide in Xenopus embryo extracts of the expected size for Eomes protein (76 kDa) (33) (Fig. 1A, lane 1). Pre-adsorption of the Eomes anti-NH 2 -terminal antibody with Eomes NH 2 -terminal antigen coupled to Sepharose 4B (Eomes-Sepharose) completely prevented its recognition of the 76-kDa Eomes protein band on Western blots (Fig. 1A, lane 2). Pre-adsorption with Eomes-Sepharose also prevented detection of several other polypeptides (Fig. 1A, compare lanes 1 and 2). In contrast, pre-adsorption with BSA-Sepharose failed to block recognition of Eomes but did deplete the other non-Eomes polypeptides (Fig. 1A, lane 3). Protein A-horseradish peroxidase failed to recognize any polypeptides in the absence of added primary antibody (Fig. 1A, lane 4). These data suggest that our purified anti-Eomes antisera specifically recognized the 76-kDa Eomes protein. As detection of other polypeptides was depleted with BSA, these are unrelated to Eomes.

JOURNAL OF BIOLOGICAL CHEMISTRY 2399
Eomes Protein Interacts Physically with Smad2-Eomes required both transforming growth factor-␤ signaling and phosphorylated Smad2 to activate several of its downstream target genes (data not shown). Thus Eomes protein could have cooperated with Smad2 by physically associating with phosphorylated Smad2 as a transcriptional co-activator. To test this, we overexpressed Xenopus Eomes plus Smad2 in animal caps. IP of Smad2 co-precipitated Eomes protein from caps in which both Eomes and Smad2 had been overexpressed (Fig. 1B, lanes  3 and 7). Overexpressed Eomes protein (Fig. 1B, lanes 2, 3, 6, and 7) is myc-tagged and therefore of slightly greater mass than endogenous Eomes (Fig. 1B, lanes 1 and 5; myc-Eomes migrates more slowly than expected; see below and Fig. 3L). A similar result was obtained using either of our two affinity purified antibodies, one against the Eomes NH 2 terminus (Fig. 1B, lane  3), the other against its COOH terminus (Fig. 1B, lane 7). Conversely, immunoprecipitation of Eomes using either of these two antibodies co-precipitated phosphorylated Smad2 in similarly injected caps (Fig. 1C, lanes 3 and 4), whereas a GFP antibody failed to precipitate Eomes (Fig. 1B, lanes 4 and 8) or Smad2 (Fig. 1C, lane 5). These interactions in injected embryo extracts were confirmed in uninjected Xenopus gastrulae, with immunoprecipitation of either Eomes or Smad2 able to co-precipitate the other partner (Fig. 1D, lanes 2, 5, 8, and 9). We conclude that Eomes and Smad2 form a functional transcription factor complex to activate several mesodermal genes during Xenopus embryonic development.
The COOH-terminal Domain of Eomes Binds to Agarose Polysaccharide Beads-To investigate the physical interaction between Eomes and Smad2 in more detail, we attempted to map within Eomes the location of its affinity surface for Smad2. Eomes was divided into three pieces: 1) an NH 2 -terminal domain (Met 1 -Tyr 214 , NTD); 2) a central, DNA binding domain containing the T-box (Ser 215 -Asp 455 , DBD); and 3) a COOH-terminal domain (Arg 456 -Ser 692 , CTD). mRNA encoding NTD, DBD, or CTD was microinjected into the animal pole of Xenopus embryos, embryos were frozen at stage 10.5 and subjected to co-immunoprecipitation analysis using an anti-Smad2 antibody for the immunoprecipitation, and probing the Western blot for Eomes. We found that anti-Smad2 beads bound poorly to NTD ( Fig. 2A, lane 8) and not at all to DBD (Fig. 2B, lane 7). In contrast, the beads bound robustly to CTD (Fig. 2C, lane 6). Dividing CTD in half virtually eliminated binding (Fig. 2D, lanes 6 and 7). Removing a 60-amino acid NH 2 -or COOH-terminal segment from CTD (called CC3, CN3, respectively) both retained robust binding but eliminated one antibody epitope each (Cmid1 was not recognized by the anti-CTD antibody; Fig. 2, E and F). These results suggested that there existed an affinity site in the middle region of CTD, because binding activity was prevented by halving of CTD between Asp 573 and Ala 574 , but not by removal of 60 amino acids from either end.
We therefore constructed a series of cDNAs, using wild type CC3 as a substrate, in which 21 windows of six Eomes CC3 amino acids were individually mutated. This was accomplished by substituting six glycine residues in place of the natural Eomes 6-amino acid sequence within each such window between Glu 515 and Ile 633 . The resulting "glycine-scanning" mutation clones are called GSM1 through GSM21. Each GSM mRNA was injected individually into Xenopus embryos and subjected to co-immunoprecipitation analysis in direct comparison to wild type CC3. Ten of the 21 GSMs had the same affinity for beads as wild-type CC3 peptide (GSMs 1, 2, 4 -6, 10, 15, 16, 19, and 20; Fig. 3, A-C, E, H, and J). Seven GSMs had greatly reduced affinity (GSMs 7, 8, 11-14, and 17; Fig. 3, D, F, G, and I). Two GSMs had slightly reduced (GSMs 3, 18; Fig. 3, B and I), and two slightly increased (GSMs 9, 21; Fig. 3, E and K) affinity for beads. The myc epitope tag, when fused to negative control protein DBD or GFP, failed to bind to beads (Fig. 3L).
Although full-length Eomes protein bound specifically to anti-Smad2 beads (Fig. 1), we tested the ability of NTD, DBD, and CTD to bind to anti-GFP beads. We found that CTD bound to anti-GFP beads (Fig. 6, C, lane 7, and D, lane 2) or empty protein G-agarose beads (Fig. 6D, lane 3), whereas NTD mostly failed (Fig. 6A, lane 5) and DBD failed to bind to beads (Fig. 6B, lane 6). CTD also bound to unsubstituted agarose (see below). We conclude that the Tyr 547 -Leu 612 region of CTD, spanned by GSMs 7-17 (Fig. 5A) represents a binding domain for polysaccharides.

The COOH-terminal Domain of Eomes Mediates Transfer of Eomes
Protein between Cells-It was possible that the binding of Eomes to polysaccharide beads reflected an underlying function. For example, beads might mimic the solvent-exposed surface of intra-or extracellular glycosylated protein residues. To test whether Eomes protein is capable of direct protein transfer between cells, we utilized a simple assay (Fig. 7A): we injected two cells of both two-and four-cell stage Xenopus embryos with synthetic mRNA encoding fluorescent proteins of two different colors. Embryos were cultured until mid-blastula stage 8, animal poles explanted (caps), and living caps subjected to confocal microscopy. Injections into two-cell embryos in which cytokinesis has not yet been completed would be expected to yield caps with many cells containing fluorescent proteins of both color. In contrast, for proteins unable to diffuse across or be transported across the cell plasma membrane, later stage injections should yield cells of only one or the other color.
To determine whether this assay was effective, we injected one cell of a two-cell stage Xenopus embryo with mRNA encoding a monomeric red fluorescent protein (Cherry) (35), and the second cell with histone H2B fused to GFP (H2B-GFP), at the two-cell stage, before completion of the first cytokinesis. As expected, in 10 microscopic fields of view chosen at random, many cells were found in which both fluorescent colors were present (Fig. 7B, top row). After completion of cytokinesis, two-cell embryos were injected with Cherry in one cell and GFP in the other cell: very few bi-fluorescent cells were observed (Fig. 7B, second row). In contrast, when four-cell embryos were injected; with one left-(or right-) side cell receiving Cherry, and the other side (right or left) receiving Eomes fused to GFP; many cells contained both fluorescent proteins (Fig. 7B, third row). Similarly, when the Eomes CTD (fused to Cherry) was tested opposite a monomeric form of a blue fluorescent protein (Cerulean) (36), CTD crossed the plasma membrane into adjacent cells (Fig. 7B, bottom row). Quantitation of these results is shown in Fig. 7C. We conclude that Eomes protein was translocated between adjacent cells, and that the Eomes CTD was sufficient to confer this capability.
Xbra-GFP, Eomes-NTD, and Eomes-DBD Fail to Cross the Cell Plasma Membrane-To determine whether the ability to be translocated across the plasma membrane was a general property of T-box proteins, we injected four-cell embryos with Cherry in one cell and Xbra-GFP in the other; Xbra-GFP failed to move between adjacent cells (Fig. 8B, third row), as did H2B-GFP when injected at the late two-cell stage (Fig. 8B, second  row). As a control, injection of mid-two-cell embryos with one cell receiving Cherry and the other, GFP, yielded many bi-fluorescent cells (Fig. 8B, first row). As a control for mRNA dose effects, in this experiment, embryos were injected distal to the plane of cell division (Fig. 8A). Similarly to Xbra and histone H2B, both Eomes-NTD and Eomes-DBD failed to cross the plasma membrane (Fig. 8B, fourth  and fifth rows). As expected, Cherry-CTD crossed the plasma membrane to adjacent Cerulean-injected cells (Fig. 8B, bottom row). Quantitation of these results is shown in Fig. 8C. We conclude that the ability to cross the plasma membrane is not a general property of the T-box transcription factors, and within Eomes, this property specifically mapped to the CTD. Moreover, the Eomes CTD was necessary and sufficient for protein transduction.
The Eomes Carbohydrate Binding Domain (CBD) Is Sufficient for Cell-Cell Protein Transfer-It was possible that the Eomes polysaccharide affinity surface (CBD; 547-612) could mediate Eomes protein transfer across the plasma membrane. When four-cell embryos were injected (Fig. 9A) with Cherry in one animal pole cell (left or right), and Cerulean-CBD on the other side (right or left, Fig. 9A), CBD was found in many cells along with Cherry ( Fig. 9B, third row). Eomes lacking CBD (Cerulean-⌬CBD) also crossed the plasma membrane (Fig.  9B, bottom row). Neither Cherry, GFP, nor Xbra-GFP were translocated (Fig. 9B, first and second rows). Quantitation of these results is shown in Fig. 9C. We conclude that the Eomes CBD is sufficient, but not necessary for Eomes protein transduction.
The Eomes CBD Is Important for Gene Activation-The Eomes CTD, containing the CBD, harbors a transcriptional activation domain (3). We therefore tested whether CBD conferred trans-activation capacity on Eomes by overexpressing Eomes-CBD in caps and assaying for early embryonic gene expression by RT-PCR. Although CBD itself failed to activate genes in caps, Eomes lacking CBD (Eomes-⌬CBD) was impaired in its ability to activate genes at moderate dose (0.5 ng of mRNA per cap) but  1 and 5). (GSM1, -2 lack a myc epitope tag.) B, GSM3 was slightly diminished, GSM4 binds to beads. (GSM3 lacks myc.) C, GSM5, 6 bind beads. D, GSM7, 8 failed to bind beads. E, GSM9 was slightly increased, GSM10 binds beads. F, GSM11, 12 failed to bind beads. G, GSM13, 14 failed to bind beads. H, GSM15, 16 bind beads. I, GSM17 failed to bind beads; GSM18 was diminished. GSM18 migrated slower than expected on gels. J, GSM19, 20 bind beads. K, GSM21 was slightly increased in binding to beads. L, myc-tagged DBD or GFP failed to bind to beads. Eomes DBD or GFP was overexpressed in Xenopus embryos. Extracts were left untreated (lanes 1 and  2) or immunoprecipitated (lanes 3 and 4) with an anti-Smad2 antibody and Western blotted with an anti-myc epitope antibody. (The myc tag caused anomalously slow gel migration of DBD and GFP.) not at high dose (5 ng; Fig. 10A). Both CBD and Eomes-⌬CBD also caused a high frequency of gastrulation defects when injected into the animal pole of intact embryos (Fig. 10B). We conclude that the Eomes CBD is important for Eomes gene activation. As CBD caused gastrulation defects in whole embryos, this suggested CBD interacted and interfered with important protein partners.
The Eomes CTD Activates Xnr5 but Does Not Bind DNA-Eomes-⌬CBD was capable of access to adjacent cells, which suggested other regions within CTD were also sufficient for cell transfer activity. We therefore tested CTD for its ability to activate gene transcription in caps. We found that Cherry-Eomes-CTD was capable of activating Xnr5 (Fig. 11A). As CTD contains the nuclear localization signal for Eomes (Fig. 8B, bottom  row), we asked whether CTD was capable of binding to DNAcellulose. CTD bound to DNA-cellulose beads (Fig. 11B, lanes 6  and 7) similarly to unsubstituted beads (Fig. 11B, lanes 4 and 5), and failed to bind beads to which BSA had been covalently cross-linked (Fig. 11B, lane 2). All three Eomes peptides, NTD, DBD (data not shown), and CTD, bound to heparin-agarose (positive control; Fig. 11B, lane 8). We conclude that despite its nuclear localization, CTD did not bind to DNA.
Eomes Protein Translocation Is Required for Its Nuclear Accumulation-As a transcription factor, Eomes must first enter the cell nucleus before activating gene transcription. We asked whether the transfer of Eomes protein between adjacent cells was necessary for its ability to activate genes, as measured by the nuclear accumulation of Eomes. The farnesylation signal sequence from the Xenopus H-ras protein (37) was fused to Cherry (CherryF) and overexpressed in caps. As expected, CherryF localized to the plasma membrane, and also showed some cytoplasmic staining (Fig. 12A). Cerulean-Eomes localized to nuclei (Fig. 12B). When three fluorescently tagged proteins: CherryF (ChF), Cerulean-Eomes (CerEo), and H2B-GFP, were co-expressed in caps, Cerulean-Eomes failed to accumulate in nuclei (Fig. 12D). Instead, in some cells, CherryF and Cerulean-Eomes co-localized to membrane-associated puncta (Fig. 12, C, D, and F). This suggested that attachment of lipidated CherryF to the plasma membrane was responsible for  5). B, deletion mutant D3 in CTD binds to beads; D4 failed to bind to beads. FIGURE 5. The Eomes CBD is evolutionarily conserved. A, summary of results from glycine-scanning and deletion mutation of Eomes. Glu 515 -Leu 636 is depicted in which 21 GSMs were made: red letters indicate that binding to Sepharose beads was eliminated within a 6-amino acid window; blue, diminished binding; green, increased binding. Deletions are indicated by arrows; GSMs over-or underlined; and amino acid numbers bulleted above the sequence. B, ClustalW alignment of Eomes Tyr 547 -Leu 612 (CBD) with Tbr1 and Tbet from Xenopus, zebrafish, mouse, human (GenBank accession numbers: xlEomes, P79944; zfEomes NP_57175; muEomes, O54839; huEOMES, NP_005433; xtTbr1, NP_001072587; zfTbr1, NP_001108562; muTbr1, Q64336; huTBR1, NP_006584; zfTbet, XP_001338262; muTbet, AF241242; huTBET, NP_037483.1).
preventing the nuclear accumulation of Cerulean-Eomes. We conclude that CherryF prevented Cerulean-Eomes translocation between cells, and that translocation of Cerulean-Eomes between cells was required for its nuclear localization.

DISCUSSION
Eomes Forms a Complex with Phosphorylated Smad2 to Activate Several Mesodermal Genes-Activin signaling activates mesodermal genes through phosphorylation of Smad2, which then associates with Smad4, translocates to the cell nucleus, and associates with other transcription factors (38,39). We observed that a truncated dominant negative activin receptor blocks the ability of Eomes to activate several mesodermal genes in caps (data not shown). Furthermore, a point-mutated dominant negative Smad2, which blocks endogenous Smad2 phosphorylation, also prevents Eomes from activating a number of mesodermal genes (data not shown). This suggests that Eomes requires phosphorylated Smad2 to activate these target genes, and that Eomes might activate some genes by associating physically with Smads 2 and 4 in a promoter-bound transcription factor complex. This agrees with work by others that Eomes interacts genetically with nodal in zebrafish and mouse (19,20,22) and suggests that the mechanism of this interaction is one of protein complex formation with Smad2.
The Eomes CTD Binds to Polysaccharides-Using co-immunoprecipitation analysis, we found that the Eomes protein associates physically with Smad2 (Fig. 1). To map the Eomes domain for Smad2 binding we divided Eomes into three equal fragments, NTD, DBD, and CTD, and used them in co-immunoprecipitations. Unexpectedly, we found that CTD, unlike full-length Eomes, bound to agarose beads (Fig. 6). The Eomes domain for carbohydrate binding was localized to the middle portion of the COOH-terminal third of Eomes (Fig. 2), and was further delineated by employing a glycine-scanning mutational  4 and 8). B, the Eomes DBD did not bind to beads. The blot in A was re-probed with anti-Eomes DBD antibodies. C, the Eomes CTD binds to beads. The blot in A was re-probed with anti-Eomes CTD antibodies. D, CTD binds beads without added antibody. The Eomes CTD was overexpressed in Xenopus embryos. Extracts were left untreated (lane 1) or immunoprecipitated with an anti-GFP antibody (lane 2) or no antibody (lane 3), and Western blotted with anti-Eomes COOH-terminal antibodies. approach that identified residues within the Eomes COOHterminal portion between Tyr 547 and Leu 612 that are critical for binding. These residues, when mutated to glycine (in groups of six), substantially diminished binding of an Eomes COOH-terminal peptide (CC3) to Sepharose beads (Fig. 3). Three segments important for binding were identified: 1) Tyr 547 -Lys 558 (12 amino acids); 2) Tyr 571 -Thr 594 (24 amino acids); and 3) Gly 607 -Leu 612 (6 amino acids). Two additional 6-amino acid blocks had slightly diminished binding: Gln 523 -Thr 528 and Leu 613 -Val 618 ; and two had increased binding: Ser 559 -Thr 564 and Pro 631 -Leu 636 (Fig. 3). Furthermore, deletion of 42 amino acids (Leu 553 -Thr 594 ) abolished binding of the Eomes CTD to beads (Fig. 4B). These data are summarized in Fig. 5A.
The entire Eomes CTD (Arg 456 -Ser 692 , Xenopus) is highly conserved among vertebrates (70% identical between human and Xenopus; ClustalW comparison, Fig. 5B). The Eomes CBD (Tyr 547 -Leu 612 ) within the CTD is also highly conserved (65% identical between human and Xenopus Eomes). The equivalent region of human Tbr1 or Tbet is 46 or 22% identical, respec-  tively. Comparing this domain among human, mouse, Xenopus, and zebrafish, a number of blocks of similarity are found which segregate by subfamily member (Eomes, Tbr1, and Tbet; Fig.  5B). These blocks correlate well with our glycine-scanning mutant data, such that almost the entire region spanned by GSM7-9 is conserved among species between Eomes and Tbr1. Region GSM10, dispensable for binding, lies largely within a non-evolutionarily conserved region. GSM11-14, required for binding, is conserved; and with the exception of Trp 599 and Ser 604 , GSM15 is neither required for binding nor highly conserved. Throughout this region Tbet is less well conserved than Eomes and Tbr1, with the exception of region GSM11-13, suggestive of a potential CBD in Tbet. No similarity exists between Eomes and any other carbohydrate binding domains (data not shown), and there is no apparent small, conserved motif located within the Eomes CBD region (Fig. 5B).
Eomes Protein Is Transported between Adjacent Embryonic Cells-The ability of the Eomes CTD peptide to bind polysaccharides suggested that it may normally do so intra-or extracellularly. It is unusual for transcription factors to be trafficked between cells (25). We find that full-length Eomes protein gains access to adjacent embryonic cells by crossing the plasma membrane, a barrier to which other proteins (including Xbra, histone H2B, and GFP) are impermeant (Figs. 7 and 8). The capability for intercellular protein translocation maps to the Eomes CTD ( Figs. 7 and 8); and the Eomes CBD within the CTD is sufficient for cell-cell transfer (Fig. 9).  1 and 4), Eomes-⌬CBD (lanes 2 and 5), and CBD (lanes 3 and 6). Injected or uninjected (nc, lane 7) animal poles were explanted (caps) at stage 8 and cultured until stage 10.5; and caps were frozen. Caps and control stage 10.5 whole embryos (WE, lane 8) were analyzed by RT-PCR for early embryonic genes. B, quantitation of embryonic defects in animal pole-injected, and uninjected, embryos. Embryos were scored for normalcy of gastrulation at stage 11, and for defects resulting from earlier inhibition of gastrulation at stage 23. Number of embryos (n) is shown above the histogram. Whereas CBD fails to activate early embryonic genes in caps (Fig. 10A), it is capable of interfering with normal cell movements during gastrulation (Fig. 10B). Although Eomes lacking the CBD (Eomes-⌬CBD) activates genes at high dose, at a moderate dose Eomes-⌬CBD fails to effectively activate early embryonic genes in caps (Fig. 10A). Thus at moderate doses, Eomes-⌬CBD is impaired for gene activation. Eomes-⌬CBD may be similarly impaired for cell translocation, and may only translocate between cells at high doses. CTD activates Xnr5 in caps (Fig. 11A) despite its lack of DNA-binding activity (Fig. 11B).
When direct access of Cerulean-Eomes to the plasma membrane is blocked using a farnesylated red fluorescent protein, nuclear accumulation of Eomes is also blocked (Fig. 12). This suggests that Eomes protein does not simply diffuse across the lipid bilayer (as thought to occur for HIV tat protein) (25), but rather is actively transported from one cell to the next. When CherryF and Cerulean-Eomes are co-expressed, CherryF is partially displaced from the plasma membrane (Fig. 12C), and Cerulean-Eomes is excluded from nuclei, and localizes in some cells to plasma membrane-associated puncta (Fig. 12, C and D).
(GFP-Eomes localizes to membrane puncta more readily, perhaps because of the propensity of GFP for aggregation, which may result in a reduced cell-cell transfer rate; data not shown.) This suggests CherryF and Cerulean-Eomes compete for access to the plasma membrane, reducing the rate of Cerulean-Eomes translocation and allowing visualization of Cerulean-Eomes membrane puncta just before, or during, transit. It also suggests that during or after transfer of Eomes protein between cells, Eomes protein is post-translationally modified to subsequently allow it to gain access to the cell nucleus. This putative co-translocational activation appears to regulate the nuclear import of Eomes, which itself is logically required for Eomes to activate its target genes. The Eomes CTD with its carbohydrate binding domain bears no similarity to other known peptides capable of intercellular translocation (ClustalW analysis of peptides in Ref. 25 and BLASTp analysis (40,41), data not shown (24,26).
Why should Eomes protein be regulated so as to require translocation between cells before being capable of entering cell nuclei? This novel regulation may represent a molecular mechanism for ensuring a community effect (42,43) during embryonic development. Cells that first activate the Eomes gene may wait for input from neighboring cells before making a decision to differentiate as mesoderm. In naïve cells Eomes is kept out of the nucleus and thereby prevented from activating its target genes. Cells that activate Eomes before their neighbors could directly signal their tentative decision in favor of mesoderm via translocation of Eomes protein into neighboring cells; prior translocation of Eomes is required for Eomes to activate genes. A critical threshold concentration of Eomes protein may be required within naïve cells before Eomes protein is transferred to its neighbor. Once transferred, Eomes protein would become activated for nuclear import, activate genes, and trigger cell fate commitment. This mechanism would help ensure that mesodermal cells differentiate as groups (42,43).
Future work will identify the remaining residues within the Eomes CTD that, like CBD, are also capable of protein translocation between cells. It will be important to identify the Eomes transporter, or otherwise elucidate the mechanism for Eomes translocation. Although cell to cell protein translocation is not a general property of the T-box proteins (Figs. 8 and 9), identification of an Eomes transporter could lead to elucidation of other protein substrates, perhaps including other transcription factors currently thought to behave in a cell-autonomous fashion.