SEARCH

SEARCH BY CITATION

Keywords:

  • Tgif1;
  • TG-interacting factor;
  • nodal, VegT;
  • holoprosencephaly (HPE);
  • maternal mRNA

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

In Xenopus, the maternal transcription factor VegT is necessary and sufficient to initiate the expression of nodal-related genes, which are central to many aspects of early development. However, little is known about regulation of VegT activity. Using maternal loss-of-function experiments, we show that the maternal homeoprotein, Tgif1, antagonizes VegT and plays a central role in anteroposterior patterning by negatively regulating a subset of nodal-related genes. Depletion of Tgif1 causes the anteriorization of embryos and the up-regulation of nodal paralogues nr5 and nr6. Furthermore, Tgif1 inhibits activation of nr5 by VegT in a manner that requires a C-terminal Sin3 corepressor-interacting domain. Tgif1 has been implicated in the transcriptional corepression of transforming growth factor-beta (TGFβ) and retinoid signaling. However, we show that Tgif1 does not inhibit these pathways in early development. These results identify an essential role for Tgif1 in the control of nodal expression and provide insight into Tgif1 function and mechanisms controlling VegT activity. Developmental Dynamics 237:2862–2873, 2008. © 2008 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Nodal proteins are secreted growth factors, related to TGFβ, that are essential for germ layer induction and patterning in vertebrate embryos. Nodal activates receptor serine/threonine kinases, resulting in the phosphorylation and nuclear translocation of Smad2/3 transcription factors (reviewed in Shen,2007). These proteins form complexes with other transcription factors, such as FoxH1 or Mixer, to regulate expression of target genes, which include Nodal itself and the Nodal antagonists, Lefty and Cerberus-like. The interplay of these self-propagating and self-limiting outcomes of Nodal signaling establishes temporally and spatially graded Nodal activity, the regulation of which is required for the coordinate induction and patterning of the mesoderm and endoderm germ layers (reviewed in Shen,2007). Because Nodal expression and activity are highly dynamic, precise control over the initial levels and pattern of its expression is essential for normal development.

The initiation of nodal-related (nr) expression and signaling has been well characterized in Xenopus embryos, in contrast to in other organisms. There are six nodal paralogues in Xenopus (Chea et al.,2005), which might be considered problematic for studying their regulation. However, two of the genes, nr5 and nr6, lack the autoinduction by Smad2/3 signaling that is typical of nodal genes (Takahashi et al.,2000), making this arrangement ideal for investigating the onset of nodal expression. Instead of autoinduction, nr5 and nr6 are dependent on the maternal transcription factors VegT and β-catenin for their expression (Takahashi et al.,2000; Xanthos et al.,2002; Hilton et al.,2003). Thus, the activation of these genes can be studied independently of Nodal signaling itself. Expression of nr5 and nr6 begins at the mid-blastula transition exclusively in vegetal cells, and becomes rapidly down-regulated before gastrulation (Takahashi et al.,2000), suggesting that spatiotemporal control of these genes is critical for normal development.

Previous studies showed that the Xenopus zinc-finger transcription factor Zic2 is a repressor of nodal-related gene expression in early development (Houston and Wylie,2005). In humans, mutations in ZIC2 cause holoprosencephaly (HPE; Brown et al.,1998), one of the most common forms of brain malformation. HPE is characterized by failure of the developing forebrain hemispheres to segregate, and is linked to abnormal development of the ventral midline in the brain (reviewed in Cohen,2004). Mutations in the Shh pathway (SHH, PTC, DISPA, MEGALIN) are the predominant genetic cause of HPE, but other proteins not linked to Shh are also implicated, including the transcription factors SIX3 and TGIF1, as well as ZIC2 (reviewed in Cohen,2004). The roles of these factors in the etiology of HPE is, however, unclear, because their genetic manipulation in mouse embryos does not exactly recapitulate all the features of human HPE (Pasquier et al.,2000; Lagutin et al.,2003; Shen and Walsh,2005; Bartholin et al., 2006; Jin et al.,2006; Mar and Hoodless,2006). Given our characterization of Zic2 in relation to Nodal signaling in early frog development, and because one of these other HPE candidates, Tgif1, had been implicated Nodal signaling (Wotton et al.,1999a; Gripp et al., 2000), we tested the hypothesis that Tgif1 might also control Nodal signaling in Xenopus embryos.

Tgif1 is a three amino-acid loop extension (TALE) homeodomain protein, first discovered as a repressor of retinoic acid-mediated transcription (Bertolino et al.,1995) and later identified as a corepressor that bound activated Smad2 (Wotton et al.,1999a). Tgif1 uses multiple mechanisms to repress target genes, including recruitment of C-terminal binding protein (CtBP), direct binding of histone deacetylases (HDACs; Wotton et al.,1999b) and recruitment of the Sin3 corepessor complex. This complex associates with the highly conserved C-terminal domain of Tgif1 (Wotton et al.,2001), whereas CtBP and HDACs bind to N-terminal and central regions respectively (Wotton et al.,1999b). Experiments on cultured cells have identified several potential functions at the molecular level, including inhibition of TGFβ signaling (Wotton et al.,1999a,b), inhibition of retinoid signaling (Bartholin et al.,2006) and regulation of the cell cycle (Mar and Hoodless,2006). Interestingly, Mar and Hoodless (2006) noted that Tgif1 null mouse embryos had a low incidence of laterality defects, which further suggests an interaction with Nodal signaling, a pathway widely implicated in establishing left–right asymmetry (reviewed in Tabin,2006).

In this study, we show that the Xenopus homologue of TGIF1 is expressed maternally and is critically required to repress early nodal-related gene expression. We also show that Tgif1 inhibits the activity of VegT, through a mechanism requiring the C-terminus of Tgif1. The data presented here provide insight into the mechanisms that regulate Nodal activity in embryos, and are the first to suggest a molecular function for Tgif1 during early vertebrate development.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Identification of Xenopus TG-interacting Factor Homologues

Previous studies showed that maternal Xenopus Zic2 represses nodal-related gene expression (Houston and Wylie,2005). To determine whether homologues of other genes implicated in human HPE might also control nodal-related expression, we looked for expression of these homologues in Xenopus oocytes. Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of oocyte RNA showed that tgif1 (tg-interacting factor 1) was expressed in Xenopus oocytes, whereas other major HPE-associated genes, shh, six3, or ptch1, were not expressed (data not shown). We used human TGIF1 to query Xenopus EST databases, and we identified clones encoding complete Tgif1 proteins from both X. laevis (accession no. BC044016; Klein et al., 2002) and X. tropicalis (accession no. BC064716). A composite protein sequence, generated from overlapping ESTs, had been reported for Xenopus Tgif1 (Hyman et al.,2003) and the X. laevis EST encoded an identical protein. The predicted X. laevis protein is 96% identical to X. tropicalis Tgif1, and shows 73% and 71% identity to the human and mouse proteins, respectively (Fig. 1A). The Xenopus proteins are 272 amino acids in length and contain conserved regions including a homeodomain, a putative binding motif for CtBP, a putative nuclear localization signal and a short Tgif1-specific motif (Hyman et al.,2003; Fig. 1A). The C-termini of Tgif1 homologues are also nearly identical across species. This C-terminal region mediates transcriptional repression through recruitment of the Sin3 corepressor complex (Wotton et al.,2001) and includes a known ERK phosphorylation site, thought to control Tgif1 stability (Lo et al.,2001). Residues mutated in human HPE (Gripp et al.,2000) are also conserved in the Xenopus proteins (Fig. 1A). We obtained the full-length X. laevis clone and used this cDNA in expression and functional studies.

thumbnail image

Figure 1. Sequence and expression of Xenopus tgif1. A: Alignment of human (TGIF1), mouse (Tgif1), Xenopus tropicalis (xtr Tgif1), and Xenopus laevis (xla Tgif1) translated amino acid sequences. Identical residues are in black, similar residues are in grey. Boxes above the sequences mark the locations of a CtBP-binding motif, a nuclear-localization sequence, a Tgif-specific motif and an ERK phosphorylation site. The homeodomain is underlined with a dashed line. Arrowheads mark the sites of HPE mutations. Amino acid (a.a.) numbers are indicated for the X. laevis protein. B: Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of tgif1 and tgif2 expression in animal (An) and vegetal (Vg) oocyte halves, compared with whole oocytes (Oo) and stage 10.5 embryos (10.5). The −RT indicates a stage 10.5 sample processed in the absence of reverse transcriptase. C: Quantitative real-time PCR (QPCR) analysis of tgif1 and tgif2 expression during early development. Samples were normalized to odc and values displayed as a percentage relative to an uninjected oocyte sample (relative expression %). D–I: in situ hybridization of tgif1. D: Stage 10.5 (top row) and stage 13 (bottom row) embryos. Both rows show an animal pole view. In the bottom row, anterior is to the left. E: Stage 18, anterior to the left. F,G: Stage 24 (F) and stage 36 (G). H,I: Sections through stage 30 embryos, hindbrain level section (H), spinal cord level section (I). Anterior is to the left. no, notochord.

Download figure to PowerPoint

Vertebrates have multiple Tgif paralogues, including Tgif2, whose expression is similar to Tgif1 in mouse (Jin et al.,2005; Shen and Walsh,2005). Tgif-related genes are also present on the sex chromosomes in mammals (Tgiflx and Tgifly; Blanco-Arias et al.,2002). We identified two tgif2 sequences in Xenopus laevis, likely representing allo/pseudoalleles, and one tgif2 gene in X. tropicalis (accession nos. BC081024, BC081037, CT030554, respectively, not shown). We did not identify any ESTs homologous to the sex chromosome-specific Tgif genes. Synteny analysis using Metazome indicated that the X. tropicalistgif1 and tgif2 loci correspond with the mammalian Tgif genes (data not shown).

Expression of tgif1 in Normal Development

We detected expression of tgif1 by RT-PCR at all stages analyzed, from egg through the swimming tadpole stage (not shown). Notably, tgif1 transcripts are present in stage VI oocytes, and are evenly distributed between animal and vegetal hemispheres (Fig. 1B, and in situ data not shown), suggesting that the RNA is not localized. Expression of tgif2 was also detected maternally, in a similar distribution as that found for tgif1 (Fig. 1B). Comparison of tgif1 and tgif2 message levels over a series of developmental stages showed that tgif1 transcripts decline in abundance after fertilization, whereas tgif2 levels showed a marked increase during the gastrula stages (Fig. 1C).

Spatial analysis of tgif1 expression identified expression in animal cap ectoderm (Fig. 1D, top row) and surrounding the blastopore. Expression in two stripes flanking the notochord plate in late gastrula embryos was also evident (Fig. 1D, bottom row). By the neurula stages, tgif1 is expressed in the nervous system, particularly in the anterior and posterior regions of the neural plate, with some lateral expression along the periphery of the neural folds (Fig. 1E). Tail bud embryos express high levels of tgif1 in the nervous system, branchial arches, pronephros, ear vesicle, and in the eye (Fig. 1F,G). Sectioning of these embryos revealed that in the nervous system, tgif1 is expressed in the dorsal neural tube, predominantly in the mid- and hindbrain (Fig. 1H,I). These data show that tgif1 is maternally expressed in Xenopus, and that its embryonic expression pattern is consistent with roles in early development, as well as in neural development and organogenesis.

Antisense Inhibition of Maternal tgif1 Results in Excess Anterior Development

To elucidate the function of tgif1 in early embryonic development, we used antisense oligos to deplete the maternal stores of tgif1 mRNA. DNA-based antisense oligonucleotides (oligos) cause the degradation of complementary mRNAs in Xenopus oocytes through an endogenous RNase-H activity (Dash et al.,1987). We screened several antisense oligos specifically targeting tgif1, but not tgif2, and identified two that were effective in degrading tgif1 mRNA. Figure 2A shows a dose-dependent reduction of tgif1 mRNA levels in oocytes injected with 3.0 or 6.0 ng of oligo tgif1-as2mp. Quantitative real-time PCR (QPCR) analysis of the same cDNA samples found that tgif1 was reduced to 23% and 6% of control levels, at the 3.0 ng and 6.0 ng doses respectively. The oligos used are not complementary to tgif2 mRNA, and levels of tgif2 were unaffected in these experiments (data not shown). A 3.0-ng dose of tgif1-as2mp was used in the following experiments because higher doses resulted in fewer surviving embryos. Another DNA oligo targeting a different sequence in the mRNA gave similar results (not shown). We also designed a morpholino oligo complementary to the start codon to block tgif1 translation.

thumbnail image

Figure 2. Depletion of maternal tgif1 anteriorizes embryos. A: Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of tgif1 mRNA levels in uninjected (Un) and tgif1-depleted oocytes, injected with 3.0 ng or 6.0 ng of oligo tgif1-as2mp. Oocytes were analyzed after overnight progesterone (Pg) treatment. Values from separate Quantitative real-time PCR (QPCR) runs on the same cDNA samples are shown below each marker as a percentage of the uninjected sample, normalized to odc. B: QPCR analysis of tgif1 mRNA levels in uninjected (Un) and tgif1-depleted oocytes and embryos (tgif1-; 3 ng tgif1-as2mp). Samples were normalized to odc and values displayed as a percentage relative to the uninjected oocyte sample (relative expression %). C–H: Phenotypes of representative control (C, E, G, Uninjected), 3 ng of tgif1-as2mp-injected (D, F, tgif1-) and 30 ng of tgif1-MO-injected embryos (H) tail bud stage embryos. Embryos in C,D and E,F are from separate host-transfer experiments. Arrows indicate enlarged head and anterior endoderm (D), and a secondary head (F). I: RT-PCR of late patterning markers in uninjected (Un) and tgif1-depleted (tgif1-) stage 30 embryos, obtained by fertilization of tgif1-as2mp-injected oocytes (3.0 ng) by the host-transfer method.

Download figure to PowerPoint

We used the host-transfer method (Heasman et al.,1991) to assess the developmental effects of depleting maternal tgif1 in oocytes. This procedure is necessary because cultured oocytes cannot be fertilized in vitro directly, but rather must acquire jelly coats and other surface modifications to become competent for fertilization and normal development (Heasman et al.,1991). Control and antisense oligo-injected oocytes were matured in vitro using progesterone, vital dye stained, and transferred into the body cavity of egg-laying females. The eggs were recovered, fertilized normally in vitro, and sorted. Analysis of tgif1 mRNA levels in depleted embryos over a series of stages in early development showed that tgif1 accumulates to normal zygotic levels by mid-gastrulation (stage 11). This result suggests that effects of tgif1 loss-of-function are due primarily to loss of maternal tgif1.

We examined tail bud stage tgif1-depleted embryos for phenotypic abnormalities. These embryos were shorter than controls and had swollen anterior endoderm, truncated tails and expanded cement glands (Fig. 2C–F; Table 1). We saw small secondary heads with cement glands in approximately 25% of surviving, affected tgif1-depleted embryos (Table 1, series 1; Fig. 2E,F). Similar phenotypes were seen in embryos derived from oocytes injected with 15 and 30 ng of an MO against tgif1 (Table 1, series 2; Fig. 2G,H). Injection of the same morpholino doses into fertilized eggs had no effect on development (data not shown), further suggesting that it is the maternal store of tgif1 that is relevant for anteroposterior patterning. To provide molecular evidence for anteriorization, we analyzed tailbud stage tgif1-depleted embryos by RT-PCR (Fig. 2I). We found elevated levels of anterior markers: wnt1, a hindbrain marker, and hex, a marker of liver diverticulum at this stage, along with concomitant decreased levels of ventroposterior markers: hoxb9, a spinal cord marker and globin (〈 T4 globin), a blood island marker (Fig. 2I).

Table 1. Tgif1 Regulates Anteroposterior Patterning in a Nodal-Dependent Fashiona
 nExogastrulaExpanded Anterior 2° headsNormal% Anteriorized % 2° heads
  • a

    Values represent proportion of embryos (stage 28), out of the total (n), that displayed the indicated phenotypes. Numbers in italics on the second row indicate the proportion of the anteriorized embryos with duplicated heads containing cement glands. The data for series 1 were pooled from three host-transfer experiments using oocytes from three different donor females. tgif1 mRNA was injected into oocytes in the rescue experiments. In series 2 and 3, data for each were pooled from two host-transfer experiments, using oocytes from two donor females. The κ2 analysis showed that depleted and rescued samples differ significantly from each other and from the controls, P ≤ 0.001.

Series 1     
Uninjected503/500/5047/500
   N/A N/A
3 ng tgif1-as2mp345/3424/345/3471
   6/24 25
3 ng as2mp + 50 pg tgif1 RNA222/227/2213/2232
   1/7 14
Series 2     
Uninjected200/200/200/200
   N/A N/A
15 ng tgif1-MO100/106/104/1060
   0/6 0
30 ng tgif1-MO180/1812/186/1867
   1/12 8
Series 3     
Uninjected300/300/3030/300
   N/A N/A
3 ng tgif1-as2mp162/1610/164/1663
   3/10 30
3 ng as2mp + 50 pg cerS RNA173/176/178/1741
   0/6 0

We controlled for the specificity of the antisense oligos through rescue experiments, in which tgif1 mRNA was injected back into depleted oocytes. Oligo-injected oocytes were incubated for 24 hr to allow for oligo turnover, and were then injected with tgif1 mRNA (50 pg). Oocytes were transferred into host females and fertilized as above. Phenotypic examination at the tail bud stage showed that reintroduction of tgif1 mRNA could rescue the anteriorization phenotype (Table 1, series 1, also Fig. 4). In summary, our phenotypic analyses show that depletion of tgif1 results in an expansion of anterior development, which is reversed by reintroduction of tgif1 mRNA.

thumbnail image

Figure 4. Rescue of maternal tgif1 depletion by injected tgif1 mRNA. A–E: Phenotypes of representative control (C, Uninjected), tgif1-depleted (D, tgif1-), rescued (E, tgif1-;+RNA), and 50 pg tgif1 RNA-injected (F) embryos, at stage 32. E: Quantitative real-time polymerase chain reaction analysis of nr5, gsc, xbra, and xnr3 expression at stage 10. Relative expression values are shown on the y-axis; samples were normalized to odc and values expressed as a percentage of uninjected control stage 10 embryos (100%).

Download figure to PowerPoint

Because tgif1-depleted embryos showed abnormal embryonic body axis patterning, we analyzed their early development to identify the more proximal defects in these embryos. Depleted embryos developed normally up to the gastrula stage, but were slower than control embryos in progressing through gastrulation. Embryos deficient in tgif1 displayed open blastopores with deeply recessed yolk plugs (Fig. 3A,B), a phenotype which oftentimes indicates hyper-dorsalization. The majority (> 75%) of tgif1-depleted embryos eventually completed gastrulation 3–4 hr later than controls; the remainder exogastrulated and did not survive past the neurula stages (Table 1, series 1).

thumbnail image

Figure 3. Depletion of maternal tgif1 causes gastrulation defects and abnormal nodal-related gene expression. A,B: Phenotypes of control uninjected (A) and tgif1-depleted (B) stage 13 embryos. C: Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of early mesendoderm markers in uninjected (Un) and tgif1-depleted (tgif1-) stage 11 embryos, obtained by fertilization of as2mp-injected oocytes (3.0 ng) by the host-transfer method. Values from separate quantitative real-time PCR (QPCR) runs on the same cDNA samples are shown below each marker as a percentage of the uninjected sample, normalized to odc. D: QPCR analysis of nodal-related genes in uninjected (Un) and tgif1-depleted (tgif1-) stage 9.5 and 10.5 embryos, obtained by fertilization of as2mp-injected oocytes (3.0 ng) by the host-transfer method. Samples were normalized to odc and values displayed as a percentage relative to the uninjected stage 9.5 sample (relative expression %).

Download figure to PowerPoint

We next assayed molecular markers of dorsoventral and germ layer patterning at the gastrula stage (stage 11) by RT-PCR. Compared with sibling embryos, tgif1-depleted embryos had elevated levels of dorsal markers such as goosecoid (gsc), and siamois (sia), as well as increased levels of the early endodermal nodal-related genes, nr5 (Fig. 3C) and nr6 (not shown). Representative mesendodermal (xbra, wnt8, mix.2, sox17α) markers were also increased to a lesser degree (Fig. 3C) in tgif1-depleted embryos. Markers of ectoderm (lhx5, zic1) were unchanged compared with controls (Fig. 3C), as were a subset of genes known to suppress TGFβ signaling in the animal hemisphere, foxi1 and sox3 (Suri et al.,2005; Zhang et al.,2003; data not shown). Thus, the increases in mesendodermal gene expression suggest misregulation of these genes in situ, rather than alterations in overall germ layer patterning.

Because nr5 and nr6 were prominently increased in tgfi1-depleted embryos, we examined the expression of the other nodal-related genes in pre- and early gastrula stage embryos (stage 9 and 10.5) to identify any dynamic changes in their expression (Fig. 3D). Interestingly, although nr5/6 were up-regulated at both stages, the other nrs showed differing patterns of expression. The genes nr1 and nr3 were both little affected in pregastrula embryos but up-regulated by early-to-mid-gastrulation (Fig. 3D). Of interest, nr2 and nr4 were down-regulated before gastrulation, but were normally expressed or slightly down by mid-gastrulation, suggesting that Tgif1 may positively influence their expression. These changes in gene expression are specific to deficiency in tgif1, because normal development and near-normal levels of representative markers, nr5, nr3, gsc, and xbra, were restored in tgif1-depleted embryos injected with tgif1 mRNA (Fig. 4).

The elevated levels of nr5 in tgif1-depleted embryos suggested that Nodal signaling might be involved in causing anteriorization in these embryos. Expression of nr5 is initiated coincident with, or immediately after MBT (Takahashi et al.,2000), and Nodal signaling is known to contribute to the expression of genes elevated in tgif1-depleted embryos, such as gsc and sia (Nishita et al.,2000). Therefore, we determined if blocking Nodal activity could reduce the severity of tgif1-deficiency. We injected a low dose of cerberus-short (cerS) mRNA (50 pg), a Nodal antagonist (Piccolo et al.,1999), into tgif1-depleted embryos. In controls, this dose did not significantly alter anteroposterior patterning on its own. When low-dose cerS was injected into tgif1-depleted embryos, the overall proportion of normal embryos increased (Table 1, series 3). It has been shown that overexpression of cerS in frog embryos increases nr5 expression (Kofron et al.,2004; Houston and Wylie,2005; Takahashi et al.,2006), even though overall Nodal activity is inhibited (Piccolo et al.,1999; Agius et al.,2000). These results support the idea that the phenotypic consequences of tgif1 depletion are due to mis-regulated nodal expression and signaling.

Repression of nodal-related Genes 5 and 6 by Tgif1 Depends on VegT and Is Regionally Specific

Expression of nr5 is activated directly by VegT and β-catenin (Hilton et al.,2003), and is repressed by maternal transcription factors, including Zic2, Tcf3, Sox3, and FoxH1 (Houston et al.,2002; Zhang et al.,2003; Kofron et al.,2004; Houston and Wylie,2005), in addition to Tgif1 (this study). To distinguish whether Tgif1 antagonizes the VegT or β-catenin aspect of nr5 regulation, we generated embryos deficient in both Tgif1 and VegT. We depleted tgif1 mRNA as above, and we used a previously characterized morpholino oligo to inhibit VegT function (Heasman et al.,2001). We analyzed expression of the VegT target nr5, and of the Wnt/ β-catenin target gene, sia, at the gastrula stage by QPCR. Depletion of maternal VegT alone reduced nr5 expression to approximately 25% of control levels (Fig. 4A, left panel) but did not greatly affect sia levels (Fig. 4A, right panel). Both genes were elevated upon depletion of tgif1 (Fig. 4A). In embryos depleted of both tgif1 and VegT, nr5 levels were similar to that seen with VegT depletion alone, suggesting that Tgif1 represses nr5 in a VegT-dependent context. In contrast, in tgif1/VegT double depleted embryos, sia expression remained at control levels (Fig. 5A, right panel), suggesting that factors downstream of VegT are responsible for boosting sia expression in tgif1-depleted embryos. Thus, Tgif1 is likely to function as an antagonist of VegT activity, rather than of Wnt/β-catenin activity in Xenopus embryos.

thumbnail image

Figure 5. Tgif represses nr5/6 expression in vegetal cells. A: Quantitative real-time polymerase chain reaction (QPCR) analysis of nr5 expression in embryos depleted of tgif1, VegT, or both. Relative expression values; samples were normalized to odc and values expressed as a percentage of uninjected control stage 10.5 embryos. tgif1-, embryos injected with 3.0 ng of tgif1-as2mp as oocytes, VegT-, embryos injected with 18.0 ng of vegt-MO as oocytes. B: Reverse transcriptase-polymerase chain reaction (RT-PCR) of mesendodermal markers in uninjected and tgif1-depleted explants. Cap, animal caps; Eq, equatorial/marginal zone region; Vg, vegetal mass/endodermal mass. Explants were cut at stage 9, and cultured to stage 11. Each sample contained five explants or two intact embryos. C–D′: In situ hybridization of nr5/6 expression in control uninjected (B, Un) and tgif1-depleted embryos (C,D) at stage 10, obtained by fertilization of tgif1-as2mp-injected oocytes (3.0 ng) by the host-transfer method. Dorsal is toward the top.

Download figure to PowerPoint

In line with these results, spatial analysis of gene expression in tgif1-depleted embryos showed that nr5 was only up-regulated in vegetal cells. We analyzed explants of prospective ectoderm (animal caps), mesoderm (equatorial regions), and endoderm (vegetal masses) from either control or tgif1-depleted embryos by real-time RT-PCR (Fig. 5B). The expression of nr5 was increased vegetally, reflecting the appropriate spatial distributions of this gene. gsc and sia were increased both vegetally and equatorially, likely due to their potential to be up-regulated by TGFβ signaling (Nishita et al.,2000). These results were confirmed by in situ hybridization analysis of nr5/6 expression (Fig. 5C,D), which showed that these genes were increased only in vegetal cells of tgif1-deficient embryos. Thus, Tgif1 is only required to repress nr5/6 in cells that normally express those genes, rather than being required as a global repressor of nodal expression.

Tgif1 Inhibits VegT-Mediated Activation of nr5

To further test the hypothesis that Tgif1 inhibits VegT activity, we performed coexpression experiments in animal caps. Injection of caps with VegT alone caused induction of nr5, whereas this expression was inhibited by coinjection of tgif1 (Fig. 6A). This was also true for nr6 (data not shown), but we focused on nr5 for simplicity. Injection of tgif1 partially inhibited the activation other markers by VegT, such as xbra and wnt8, but did not block dorsal mesoderm or endoderm markers, gsc and sox17α (Fig. 6A). These genes are also targets of VegT, (Xanthos et al.,2002) suggesting that Tgif1 does not equally repress all VegT targets. Because Tgif1 is known to repress transcriptional activation by recruiting several different corepressor complexes, we used deletion and site-directed mutagenesis to identify which potential mechanisms are required for the inhibition of VegT activity by Tgif1. The C-terminus of Tgif1 is highly similar across vertebrate species (Fig. 1A; Hyman et al.,2003), and is thought to mediate transcriptional repression through recruitment of the Sin3/HDAC complex (Wotton et al.,2001). We, therefore, made a C-terminal truncation mutant of Tgif1 (Tgif1ΔC) and tested this construct in animal cap assays. A similar deletion in human TGIF1 nearly eliminated Sin3 binding in cell line assays (Wotton et al.,2001). In caps coexpressing tgif1ΔC and vegt, induction of nr5 was unaffected (Fig. 6B) compared with caps expressing vegt alone, suggesting that the Tgif1 C-terminus is required for the regulation of nr5 expression by VegT.

thumbnail image

Figure 6. The tgif1 C-terminus is required to inhibit VegT activation of nr5. A: Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of gene expression in uninjected (Un), vegt-injected (1.0 ng) and vegt+tgif1-injected (1.0 ng each) stage 10.5 animal caps. We, whole stage 10.5 embryo. B: RT-PCR analysis of nr5 expression in uninjected (Un) caps and in caps overexpressing vegt, vegt+tgif1, vegt+tgif1ΔC or tgif1ΔC alone (1.0 ng each). We, whole stage 10.5 embryo. C: RT-PCR analysis of nr5 expression in uninjected (Un), vegt-injected and vegt+ wild-type (tgif1) or mutant tgif1-injected stage 10.5 animal caps (1.0 ng each). We, whole stage 10.5 embryo. C′: Immunoblotting of embryos injected with tgif1, tgif1ΔC, tgif1-S28C, and tgif1-T151A. The left panel shows a representative blot using anti-TGIF1 N-terminal mAb; the right panel shows a representative blot using anti-TGIF1 C-terminal mAb. Tubulin was used as a loading control in the right panel. D: RT-PCR analysis of nr5 expression in uninfected (Un), vegt-injected and vegt+tgif1, and vegt+tgif1+cerS-injected stage 10.5 animal caps (1.0 ng each). We, whole stage 10.5 embryo. D′: RT-PCR analysis of nr5 and nr1 expression in uninjected (Un) and cerS-injected stage 10.5 embryos.

Download figure to PowerPoint

We next tested point mutations in Tgif1 that have been identified in cases of human HPE, S28C and T151A (Gripp et al.,2000). The S28C mutation is defective in CtBP recruitment (Melhuish et al.,2000; El-Jaick et al., 2007) as well as Smad2- and retinoid-dependent transcriptional repression (Gripp et al.,2000; El-Jaick et al.,2007). The T151A mutation reduced the ability to bind activated Smad2 (Gripp et al.,2000) but a later study showed normal binding to Smad3, as well as normal transcriptional repression function in cell line assays (El-Jaick et al.,2007). Xenopus Tgif1 has a serine at position 151 rather than a threonine, but we kept the mammalian nomenclature in our constructs for consistency. In contrast to tgif1ΔC construct, tgif1-S28C and tgif1-T151A both retained the wild-type ability to inhibit nr5 induction by VegT in animal cap assays (Fig. 6C). Immunoblotting of sibling embryos using cross-reacting antibodies against human TGIF1 confirmed that wild-type and mutant Tgif1 proteins were equivalently produced in embryos (Fig. 6C′).

Because the S28C and T151A constructs were defective in Smad-dependent corepression and putatively defective in Smad2 binding respectively, we concluded that the ability to antagonize VegT was independent of Smad-associated corepression. To confirm this result, we coexpressed vegt and tgif1 in the presence or absence of the Nodal antagonist, cerS, and then analyzed nr5 expression in animal caps. Tgif1 was able to repress nr5 regardless of the presence or absence of injected cerS (Fig. 6D), showing that ongoing Nodal signals are not required for Tgif1 to inhibit VegT function. Embryos injected with cerS were analyzed separately to confirm that nr5 was not blocked under these conditions. Indeed, consistent with our observations and with previous reports (Kofron et al.,2004; Houston and Wylie,2005; Takahashi et al.,2006), cerS did not inhibit, but rather up-regulated, nr5 (Fig. 6D′). Overall, these results indicate that corepressor proteins recruited by the Tgif1 C-terminus, likely including the Sin3 complex, are involved in attenuating VegT function.

Tgif1 Does Not Inhibit Retinoid Signaling or TGFβ Signaling in Animal Cap Cells

Our functional analysis of Tgif1 function in early Xenopus development indicated a role in the inhibition of VegT, either directly or indirectly. In cultured cells and in adult mouse tissues, Tgif1 can repress both Smad2 and retinoid signaling, although whether Tgif1 regulates these pathways during embryonic development is not known. We used animal cap assays to determine whether either of these pathways could account for the regulation of VegT activity and nodal expression by Tgif1. We first determined the extent to which Tgif1 interacts with retinoic acid (RA) signaling by testing the ability of Tgif1 to block RA-mediated posteriorization of experimentally induced neural tissue. Animal caps expressing tgif1, with or without the bone morphogenetic protein (BMP) antagonist noggin to induce neural tissue, were cultured in the presence or absence of all-trans retinoic acid (RA). We then assayed markers of anteroposterior neural patterning at the neural groove stage (Fig. 7A). As expected, noggin expression alone induced anterior markers (NCAM, six3), and the addition of RA caused the expression of posterior markers, such as the spinal cord marker, hoxb9. Tgif1 did not alter either neural induction by noggin, or neural posteriorization by RA (Fig. 7A). Additionally, Tgif1 alone did not induce neural markers, suggesting that Tgif1 also does not block BMP/Smad1 signaling in this context, an activity that causes neural induction in animal cap assays.

thumbnail image

Figure 7. Tgif1 does not regulate retinoid or Smad2 signaling in animal caps. A: Reverse transcriptase-polymerase chain reaction (RT-PCR) of neural markers in stage 22 caps injected with 200 pg of noggin mRNA (nog), 500 pg of tgif1, or both (nog+tgif1), cultured alone or in the presence of 1 μM all-trans retinoic acid (+ATRA). B: RT-PCR analysis of xbra expression in uninjected (Un) and tgif1-injected stage 11 animal caps treated with 10 ng/ml activin (act). Note weak xbra expression in caps injected with tgif1. We, whole stage 11 embryo. C: RT-PCR analysis of xbra and mix.2 expression in uninjected (Un), tgif1-injected (tgif1), smad2-HA-injected and tgif1+smad2-HA-injected animal caps. We, whole stage 11 embryo. D: Immunoblotting of overexpressed Smad2-HA and Tgif1 levels in sibling whole embryos from the experiment in C. E–F′; Examples of Tgif1 overexpression phenotype (F–F′) compared with an uninjected control embryo (E).

Download figure to PowerPoint

We next determined the ability of Tgif1 to inhibit Smad2-dependent signaling, using mesoderm induction assays in animal caps. Control or tgif1-injected caps were cultured in the presence or absence of recombinant activin protein, to stimulate Smad2 activation, and mesoderm was assayed by detection of xbra in RT-PCR assays. In these experiments, Tgif1 was unable to block induction of xbra by activin (Fig. 7B). To test the role of Tgif1 in Smad2 signaling more directly, we coexpressed Tgif1 with Smad2 (HA-smad2) in animal caps. In line with our results using activin, Tgif1 failed to block Smad2-mediated activation of xbra or mix.2, another well-characterized direct transcriptional target of Smad2 (Fig. 7C). Because the injected tgif1 had no effect in these experiments, we confirmed the presence of the exogenous proteins by extracting total protein from sibling whole embryos and immunoblotting for Smad2-HA and Tgif1 (Fig. 7D). In other experiments, we inferred that tgif1 was active because sibling embryos injected with a high dose of tgif1 mRNA (500 pg to 1 ng) developed a characteristic phenotype (Fig. 7E–F′), typified by truncated posterior tissues and open blastopores. This gain-of-function phenotype is likely the result of either a cell movement defect or a deficiency in posterior mesoderm development, and is currently under investigation. Overall, these data indicate that antagonism of RA and Smad signaling are not the major activities of Tgif1 in early Xenopus development. Taken together with the results from the VegT experiments, these results suggest the interesting possibility that Tgif1 may be a more general regulator of gene expression in early development, and not simply a modulator of other pathways.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Using primarily maternal loss-of-function experiments, we have shown a role for maternal Tgif1 in establishing proper levels of nodal-related genes in Xenopus. Additionally, we provide gain- and loss-of-function evidence that this regulation occurs through the regulation of VegT activity, but not through the regulation of Smads or retinoids, widely characterized functions of Tgif proteins. These findings thus detail novel mechanisms for the control of gene expression by Tgif1.

We show that depletion of maternal tgif1 mRNAS results in overexpression of the early endodermally expressed nodal paralogues, nr5/6. Tgif1-depleted embryos develop hyper-anteriorized features, the formation of which could be blocked by injection of a Nodal antagonist, demonstrating that this phenotype results from overproduction of Nodal proteins. Nodal-related proteins are highly potent signaling molecules. Femtogram doses of injected nr5 mRNA are capable of causing developmental defects (Kofron et al.,2004), suggesting that nr5 and nr6 likely require tight regulation. Interestingly, Takahashi et al. (2006) reported that nr5 and nr6 are amplified in the genome, and they hypothesized that gene amplification may be necessary to overcome these repression mechanisms to achieve levels needed for mesoderm induction. The narrow window of nr5/6 expression before gastrulation, and the observation that several different maternal transcription factors in addition to Tgif1, including Tcf3, Sox3, FoxH1, and Zic2, are also required to repress nr5 and nr6 (Houston et al.,2002; Zhang et al.,2003; Kofron et al.,2004; Houston and Wylie,2005), support this idea. Of interest, we find that nrs2 and 4 are slightly decreased in tgif1-depleted embryos, in contrast to the increased expression of the other nodal-related genes. One possibility is that Tgif1 acts in a context-dependent manner to either repress or enhance gene expression. The differential roles and varied expression of the various Nodal proteins in Xenopus development are not well understood and this deficiency represents an important unsolved problem.

Mechanistically, our evidence suggests that Tgif1 regulates nr5/6 by antagonizing their activation by VegT. First, other VegT target genes (e.g., sox17α, gsc) are up-regulated in tgif1-depleted embryos. Some of the genes up-regulated by Tgif1 depletion, such as sia and nr3, are primarily regulated by Wnt signaling, but excess TGFβ/Nodal signaling, presumably downstream of VegT, is known to enhance their expression (Nishita et al.,2000). We showed that sia was not up-regulated in embryos deficient in both VegT and Tgif1, indicating that Tgif1 likely does not repress endogenous Wnt responses. Second, we showed that Tgif1 could block the activation of nr5 by injected VegT in animal cap assays. This result is interesting because Zic2 is also required to inhibit nr5 but does not inhibit VegT (Houston and Wylie,2005), pointing to possible mechanistic differences between these proteins in the regulation of nr5 genes.

Tgif1 proteins have multiple potential mechanisms of transcriptional repression, including CtBP recruitment, direct recruitment of HDAC1, and indirect recruitment of HDACs by Sin3 (Wotton et al.,1999b). Our mutational analysis of the Xenopus Tgif1 protein implicated the C-terminus of Tgif1 in the inhibition of VegT activity, the region that contains the Sin3-interacting motif. This result was somewhat surprising, because it demonstrated that Tgif1 does not use all possible transcriptional repression mechanisms equivalently, or quantitatively, in the context of VegT regulation. Future work will be needed to determine whether the Sin3 complex is indeed involved in the antagonism of VegT in vivo or whether other proteins are required. Tgif1 could potentially inhibit VegT by binding directly to it and recruiting corepressors, or by binding DNA on its own and creating a repressive chromatin environment. Two pieces of evidence suggest the latter. First, our animal cap assays showed that some VegT targets, such as sox17α and gsc, are not inhibited by Tgif1 overexpression (Fig. 6A), arguing against a global role for Tgif1 in the repression of VegT. Second, excess Tgif1 does not greatly inhibit endogenous nr5 expression (Fig. 4E), which might be the case if Tgif1 simply bound and inhibited available VegT. Although the in vivo mechanistic roles of Tgif1 in transcriptional repression are unclear, Tgif proteins are related to the Irx family of chromatin insulators, also TALE-homeodomain transcription factors, suggesting the possibility that Tgifs may also regulate chromatin structure around their target promoters (Mukherjee and Bürglin,2007).

Tgif1 is thought to have two main functions at the molecular level, including transcriptional corepression of Smad2 signaling (Wotton et al.,1999a,b), and inhibition of retinoid signaling (Bertolino et al.,1995; Bartholin et al.,2006). We did not find evidence for either of these activities using gain- and loss-of-function assays in early Xenopus embryos. Our maternal mRNA depletion experiments targeted only the initial function of Tgif1 in the embryo and do not necessarily argue against a role for Tgif1 in regulating aspects of Smad or RA signaling in other contexts. It is likely that Smad and RA inhibition by Tgif1 plays important roles either in later development or in differentiated cell types. In further support of this idea, recent functional analysis of the Tgif1-related protein, Tgif2, in Xenopus (Spagnoli and Brivanlou,2008) identified an important role for zygotic Tgif2 in pancreatic development. However, their data point to BMP inhibition as an important function of Tgif2 in this context. We did not find any evidence for BMP inhibition using Tgif1 gain- or loss-of-function, and studies in cell lines did not identify significant regulation of BMP signaling by Tgif1 or Tgif2 (Wotton et al.,1999a; Melhuish et al., 2001). Our expression analyses of tgif1 and tgif2 in frogs found that tgif1 was higher maternally and declined zygotically, whereas the reverse was true for tgif2. Because BMP signaling is not initiated until after zygotic transcription begins, frog Tgif2 may have evolved as a more potent inhibitor of BMP signaling. Spagnoli and Brivanlou (2008) also showed that Tgif2 could block activin signaling to some extent, resulting in inhibition of activin-induced wnt8 expression. Interestingly, and similar to our results, xbra and other Nodal targets were not inhibited (Spagnoli and Brivanlou,2008), suggesting that like Tgif1, Tgif2 is not a global antagonist of TGFβ signaling, but acts in a promoter-specific context. It will be of interest in future experiments to determine the extent that maternal Tgif2 also regulates nodal-related gene expression.

It is currently unclear to what extent the role of Tgif1 in early development is conserved. Recently, however, Bartholin et al. (2008) showed that maternal Tgif1 is required for normal trophoblast differentiation and placenta formation in mouse. Of interest, these authors provide evidence that Smad and retinoid signaling are unlikely to account for the placental defects in maternal Tgif1-null animals, echoing our results in frog. It is thus intriguing to speculate that murine Tgif1 regulates the activity of VegT homologues, such as Eomesodermin, which has a well-known role in trophoblast differentiation.

EXPERIMENTAL PROCEDURES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Plasmids

A 1.8 kilobase (kb) Xenopus laevis tgif1 cDNA was obtained commercially (Open Biosystems; clone ID# 5571016, accession no. BC044016) in the vector pCMV-SPORT6 (Invitrogen). For over-expression studies, the tgif1 coding region (CDS) was sub-cloned into pCS2+ using BamHI and AvrII sites flanking the CDS, ligated to BamHI/XbaI-cut vector. Site-directed mutagenesis was carried out using the Quickchange kit (Stratagene) according to the manufacturer's instructions. Mutagenic primers were designed using the PrimerX Web application (bioinformatics.org/primerx/). A carboxy-terminal deletion of the last 120 amino acids (Tgif1ΔC) was made by digestion with SacI and PstI, blunt end generation with T4 polymerase and intramolecular re-ligation.

Oocytes and Embryos

Ovary was removed from mature females by laparotomy, dissected into 1.0-cm segments, and stored in modified oocyte culture medium at 18°C (OCM; 70% L-15, 0.04% bovine serum albumin, 1 mM Glutamax (Invitrogen), 1.0 μg/ml gentimicin, pH 7.6–7.8; adapted from Heasman et al.,1991). Individual oocytes for microinjection were isolated by manual defolliculation and cultured at 18°C in OCM. Oocyte maturation was induced by the addition of 2.0 μM progesterone overnight. Ovulated eggs were collected from females induced with human chorionic gonadotropin (hCG), and were fertilized using macerated testis tissue. Embryos were reared in 0.1× MMR (1× MMR: 0.1 M NaCl, 1.8 mM KCl, 2.0 mM CaCl2, 1.0 mM MgCl2, 15.0 mM HEPES, pH 7.6) and for most uses, embryonic jelly coats were removed using cysteine (2% l-cysteine; Sigma, nonanimal source) in 0.1× MMR, pH 7.8).

Antisense Oligos and Host-Transfer

The oligodeoxynucleotide (oligo) used for tgif1 depletion was tgif1-as2mp; C*A*T*CCTCTGTCTCACTG*C*C*T. Modified oligos containing three phosphorothioate linkages (*) on the 5′ and 3′ termini were dissolved in nuclease-free H2O to 1.0 mM and stored at −80°C. MOs against vegt and tgif1 were obtained from Gene-Tools: vegt-MO, 5′ CCCGACAGCAGTTTCTCATTCCAGC 3′ (Heasman et al.,2001); tgif1-MO, 5′ AGCCTTTCTTGGCTTTCATTGTAGC 3′ (new). MOs were brought to 1.0 mM in nuclease-free water and used as described in the text. Manually defolliculated stage VI oocytes were injected vegetally at doses indicated in the text and maintained in OCM at 18° C for 24 to 48 hr before maturation with progesterone. Oocytes were stimulated to mature approximately 10–12 hr before implantation into laying host females essentially as described (Heasman et al.,1991).

Microinjection

Template DNAs for microinjection were prepared by linearization with appropriate restriction enzymes and cleaned by Proteinase K digestion, followed by phenol extraction. Capped RNA synthesis was carried out using the mMessage mMachine transcription kits (Ambion/Applied Biosystems) according to the manufacturer's protocol. RNA was recovered using lithium chloride precipitation and dissolved in nuclease-free water. RNA was injected at doses of 50 pg-to-1.0 ng into two- to four-cell embryos in Ficoll solution (2% Ficoll, 0.5× MMR, pH 7.8). A total of 2–10 nl of RNA solution was delivered per 1-sec injection at a pressure of 18–22 PSI, using an air-driven injection system (PLI-100; Harvard Apparatus). tgif1, cerS, noggin (a gift from R. M. Harland), and HA-smad2 (a gift from J. Gurdon), were all in the pCS2+ vector, and their mRNAs were prepared by digestion with NotI and transcription with SP6.

Whole-Mount In Situ Hybridization

Digoxigenin-labeled RNA probes were synthesized from linearized templates, followed by transcription with the appropriate polymerases: tgif1-pCMV-SPORT6, NotI/SP6 (sense), SalI/T7 (antisense); nr5 and nr6 probes (gifts from M. Asashima), NotI/T7. RNA probes were synthesized as described (Sive et al.,2000), using enzymes and reaction buffers from Promega. DNA templates were removed by DNase I treatment and RNAs were precipitated and re-suspended in 30 μl of nuclease-free H2O. Probes were used at 1.0 μg/ml in hybridization buffer. Because the RNAs are coexpressed, nr5 and nr6 probes were mixed together to increase the overall signal.

Whole-mount in situ hybridization was performed as described (Sive et al.,2000), with the exception that the triethanolamine and acetic anhydride steps were replaced with a single 20-min treatment of 0.1% active DEPC in phosphate buffered saline/0.1%Tween 20, adapted from the method of Braissant and Wahil (1998). Color was detected using BM Purple (Roche Applied Science) and the reaction was stopped by fixation in Bouin's solution. Samples were washed multiple times in 70% ethanol/10 mM Tris, pH 8.0 and bleached as described (Sive et al.,2000).

RT-PCR

Embryos for RT-PCR analysis were frozen on dry ice and stored at −80°C. Samples were homogenized in 200 μl per embryo of RNA lysis buffer, DNase I treated, and subjected to RT-PCR as described (Houston et al.,2002; Houston and Wylie,2005). For QPCR, PCR was run on an ABI Prism 7000 (Applied Biosystems) using a 2× SYBR green master mix (Applied Biosystems). Data were normalized to a passive reference dye (ROX) and to levels of ornithine decarboxylase (odc) and quantified against a standard curve of serial diluted control embryo cDNA. The figures show the results of representative QPCR runs of individual samples, which were repeated at least three times to show reproducibility of the results. The primer sequences and full RT-PCR protocols are available upon request.

Immunoblotting

Embryos were frozen at the desired stages and immunoblotting was performed as described (Yokota et al.,2003). The following antibodies and dilutions were used: anti–α-tubulin (1:10,000, mAb clone DM1A, Sigma), anti-HA-HRP (1:500; Roche Applied Science) and anti-TGIF1 (1:250, mAb clones H-1 and 946C1a, Santa Cruz Biotechnology). The anti-human TGIF1 antibodies cross-react with overexpressed but not endogenous Tgif1 in Xenopus embryos. mAb H-1 recognizes the TGIF1 C-terminus; mAb 946C1a recognizes the N-terminus.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

The authors thank Drs. M. Asashima, J. Gurdon, and R.M. Harland for sending reagents. We also thank members of the Houston lab and Drs. D. Slusarski, D. Weeks, and J. Weiner for critical reading of the manuscript. T.C.K. received an Iowa Research Experiences for Undergraduates (IREU) award and D.W.H. was funded by The University of Iowa and from The Roy J. Carver Charitable Trust.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES