RESEARCH ARTICLE

Nutritional Endoderm in a Direct DevelopingFrog: A Potential Parallel to the Evolution ofthe Amniote EggDaniel R. Buchholz,1† Srikanth Singamsetty,2 Uma Karadge,2 Sean Williamson,2 Carrie E. Langer,2

and Richard P. Elinson2*

The egg of the direct-developing frog, Eleutherodactylus coqui, has 20! the volume as that of the modelamphibian, Xenopus laevis. Increased egg size led to the origin of nutritional endoderm, a novel cell typethat provides nutrition but does not differentiate into digestive tract tissues. As the E. coqui endodermdevelops, a distinct boundary exists between differentiating intestinal cells and large yolky cells, whichpersists even when yolk platelets are depleted. The yolky cells do not become tissues of the digestive tractand are lost, as shown by histology and lineage tracing. EcSox17, an endodermal transcriptional factor, didnot distinguish these two cell types, however. When cleavage of the yolky cells was inhibited, embryogenesiscontinued, indicating that some degree of incomplete cleavage can be tolerated. The presence ofcellularized nutritional endoderm in E. coqui may parallel changes that occurred in the evolution of theamniote egg 360 million years ago. Developmental Dynamics 236:1259–1272, 2007. © 2007 Wiley-Liss, Inc.

Key words: Eleutherodactylus coqui; direct development; endoderm; yolk; Sox17; lineage tracing; amniote egg; amnioteevolution

Accepted 14 March 2007

INTRODUCTIONA key innovation in the evolution ofthe terrestrial vertebrates was theamniote egg. The amniote egg gener-ates not only the embryo, but also aset of extraembryonic membranesthat support development on land. Inaddition to these membranes, the am-niote egg is characterized by a largeamount of yolk and by meroblasticcleavage, which does not divide theyolk into cells. In contrast to the largeyolky egg of turtles, snakes, lizards,and birds, extant amphibians haveeggs that are usually 1–5 mm in diam-

eter, and all have complete or holo-blastic cleavage. The amphibian modeis considered basal for terrestrial ver-tebrates, because egg size, cleavage,and gastrulation in amphibians issimilar to lungfish and sturgeon,among others (Elinson, 1989; Collazoet al., 1994). This situation raises thequestion as to how a meroblasticcleaving egg evolved from a holoblas-tic cleaving one.

This evolution occurred approxi-mately 360 million years ago, so it isunlikely that traces of these changesin soft, embryonic tissues will be

found in the fossil record. A source ofinformation for possible scenarios is toexamine amphibians with large yolkyeggs to see how increased yolk affectsearly development. One such amphib-ian, available for experimental analy-sis, is the Puerto Rican tree frog, Eleu-therodactylus coqui. Its egg is 3.5 mmin diameter, 20! the volume of theegg of the frog model Xenopus laevis.

The large egg size in E. coqui isassociated with differences in molecu-lar organization of early embryonicpatterning as compared with X. laevis.The RNA for EcVegT, the orthologue

1Section on Molecular Morphogenesis, LGRD, NICHD, NIH, Bethesda, Maryland2Department of Biological Sciences, Duquesne University, Pittsburgh, PennsylvaniaGrant sponsor: the National Science Foundation; Grant number: IBN-0080028; Grant number: IOB-0343403.†Dr. Buchholz’s present address is Department of Biological Sciences, University of Cincinnati, Cincinnati, OH 45221.*Correspondence to: Richard P. Elinson, Department of Biological Sciences, Duquesne University, Pittsburgh, PA 15282.E-mail: elinson@duq.edu

DOI 10.1002/dvdy.21153Published online 18 April 2007 in Wiley InterScience (www.interscience.wiley.com).

DEVELOPMENTAL DYNAMICS 236:1259–1272, 2007

© 2007 Wiley-Liss, Inc.

of the X. laevis endo–mesodermal de-terminant, is concentrated in the ani-mal cytoplasm of the oocyte, ratherthan localized to the vegetal cortex(Beckham et al., 2003). Mesoderm-in-ducing activity is restricted to mar-ginal surface in the E. coqui early gas-trula and is absent from the deeper,yolky cells and the vegetal pole (Ni-nomiya et al., 2001). These differencesled us to propose a different fate mapfor E. coqui, in which ectodermal andmesodermal tissues are displaced abovethe equator, near the animal pole, com-pared with X. laevis (Ninomiya et al.,2001; Elinson and Beckham, 2002). Theprospective endoderm is greatly en-larged.

In the investigation reported here,we examined the development of theendoderm in E. coqui. We present ev-idence that most of the yolk-rich re-gion of the early embryo serves only anutritional role and does not contrib-ute to differentiated tissues or organsof the digestive system. This conditionmay parallel the one that occurred inthe evolution of the amniote egg.

RESULTSGross Morphology ofIntestine DevelopmentThe egg of E. coqui has approximately20! the volume of the egg of X. laevis.The increased volume is due to yolk,required to support the nonfeedingembryo to a free-living frog. Cleavage

Fig. 1.

Fig. 2.

c,d.

Fig. 3.

1260 BUCHHOLZ ET AL.

and gastrulation are similar to am-phibians with smaller eggs, but theembryo develops directly to a frogwithout passing through a tadpolestage (see Fig. 10 for selected stages).The frog hatches from its jelly capsuleapproximately 3 weeks after fertiliza-tion. It still has a large amount ofyolk, which enables it to grow for 1–2weeks before eating.

As the gut initially forms, the largemass of yolky cells, the yolkyendoderm, is connected to the mouthand the anus by thin tubes. This sim-ple arrangement persists until TS 9.At this stage, the embryo looks quitefrog-like with big eyes, limbs, and dig-its, atop the yolk (Fig. 10c). At TS 9,the yolky endoderm turns by 90 de-grees, so the original anterior–poste-rior axis becomes the right–left axis(Fig. 1a). Consequently, the anteriortube bends to the right to enter theyolky endoderm, and the posteriortube exits from left of the midline. Be-ginning at TS 10, the anterior tubebecomes translucent, and differentia-tion into stomach, liver, and lungs isapparent (Fig. 1b).

Over the next few days, representedby stages TS 10–13, the yolkyendoderm becomes longer, withgreater elongation on the ventral sidethan on the dorsal side. To accommo-date this elongation, the anterior andposterior ends of the yolky endoderm

are folded dorsally and in toward themidline. Due to these folds, there is ananterior and a posterior lobe, with amore pronounced fold separating theposterior lobe from the rest of theyolky endoderm (Fig. 1c). The yolkyendoderm is like a thick sash or belt.Its outer surface is pressed againstthe ventral and lateral sides of thecoelom. It is folded in on itself dor-sally, so that from a dorsal aspect,there is an anterior lobe and a poste-rior lobe with the stomach, liver, andheart in between (Fig. 1c).

To achieve the final morphology, thegut has bent and folded (Fig. 1d,e).The folding is topologically compli-cated, because the elongation of thegut is far less on the side attached tothe dorsal mesentery. The entry of theanterior tube into the yolky endodermand the exit of the posterior tube re-main relatively close together, whilethe opposite side of the yolkyendoderm elongates more. The foldingthen is more like the folding of a fan,rather than the complete coils presentin the gut of the developing X. laevisembryo (Chalmers and Slack, 1998).This complex coiling exists in mosttadpoles, as they are herbivorous, andlittle if any trace of the tadpole-typegut is present in the nonfeeding E.coqui embryo.

As the amount of yolk decreases af-ter hatching (TS 15), the anterior and

posterior lobes diminish and disap-pear, and the internal luminal surfacebecomes smooth. Consequently, thegut of the hatched froglet has an un-usual appearance because of the pres-ence of large yolky cells in the futureintestinal region (Fig. 1f,g). The dorsalsurface of the yolky endoderm, at-tached to the mesentery, is translu-cent like the differentiated tissues ofthe stomach anteriorly and the hind-gut posteriorly, but it joins laterally toa large mass of yolky cells (Fig. 1g).The presence of this dorsal translu-cent tissue is highlighted by the largeamount of green bile pigments, re-leased from the gall bladder. As yolk isused, the yolk mass shrinks, and thegut becomes a continuous tube of di-gestive tissues (Fig. 1h). The cloacaoften is swollen with wastes.

The morphology of the developinggut suggests that the large yolky cellsare distinct from the cells forming therest of the intestine. We consider thispossibility further by histology, lin-eage tracing, and expression of anendodermal marker.

Prehatching IntestinalHistologyBefore hatching, the anterior and pos-terior tubular structures and theyolky endoderm as described aboveare distinct in histology (Fig. 2). The

Fig. 1. Gross morphology of digestive tract development. a: TS 9 embryo. The ventral skin and body wall of this TS 9 embryo were removed to seethe orientation of the gut. The posterior tube exits from the left side of the yolk-rich region and runs posteriorly to the anus. This left exit is caused bya 90-degree rotation of the yolk-rich region. b: TS 10 gut. This isolated TS 10 gut, viewed from ventral side, has liver (L) and a stomach (S), which entersthe yolk-rich region to the embryo’s right (left in this ventral view). There is a fold to produce a posterior lobe (P) in the yolk-rich region on the embryo’sleft, from which the posterior tube exits. c: TS 12 gut. This isolated TS 12 gut viewed from the dorsal side shows the fold to produce a posterior lobeon the embryo’s left (left in this dorsal view as well) from which the posterior tube exits. The liver and stomach are in the top center, and the anteriortube from the stomach enters an anterior lobe (A), which has folded less than the posterior one. d,e: Early TS 15 gut viewed from dorsal (d) andposterior (e) aspects. The posterior lobe is deeply folded, and the developing organs sit between the posterior and anterior lobes of the yolk-rich region.f: Post-TS 15 gut. This isolated gut was dissected from a froglet and extended to reveal the relationship between the different areas. The two pink liverlobes and the green gall bladder are connected to the anterior tube at the base of the stomach. The anterior tube enters the yolk-rich region, and theposterior tube exits from it. g: Post-TS 15 gut. The froglet was dissected and the yolk-rich region was displaced to the frog’s left to reveal therelationships. The gall bladder has released a large amount of green bile pigments. These highlight the translucent tissue on the dorsal surface of theyolk-rich region, which is connected to the anterior and posterior tubes. The bile accumulates in the cloaca (C). h: Gut from a late prefeeding froglet.A small amount of yolk (Y) remains in the intestine of this isolated gut. The cloaca is filled with brown bile pigments and yolk debris. Scale bars " 0.5mm. b–e and g are the same scale.

Fig. 2. Histology of the prehatching digestive tract. a: TS 14 embryo, removed from the egg capsule, with a diagram of the intestine superimposed.The vertical line indicates the plane of the cross-section in the next panel. b: A cross-section at level shown in (a) indicates a large amount of yolk takingup the majority of the peritoneal volume. Boxes indicate areas photographed at higher magnification in the next panels. c: Cross-section of the anteriortube reveals undifferentiated epithelium and a small lumen. d: Cross-section of the posterior tube shows similar histology as in (c). Note yolk platelets(magenta) in anterior and posterior tubes. Scale bar " 1 mm in b, 10 #m in c,d.

Fig. 3. Intestinal remodeling after hatching. a: TS 15 froglet with tail nearly gone. b: Cross-section through the whole body shows less yolk and moreintestinal loops than seen in Figure 2b. c: The anterior tube has multiply folded epithelium and an absence of yolk in the epithelial cells. d: The posteriortube also lacks yolk in the epithelial cells but has not undergone epithelial remodeling. In both anterior and posterior tubes, muscle (outer ring of cells)has differentiated from the connective tissue (cells between muscle and epithelium). Scale bar " 1 mm in b, 10 #m in c,d.

FROG NUTRITIONAL ENDODERM 1261

anterior and posterior tubes are simi-lar in that there are two cell layers, anouter layer one to two cells thick ofundifferentiated connective tissuewith very little cytoplasm and an in-ner columnar epithelial layer facingthe small lumen (Fig. 2c,d). The co-lumnar layer seems to be a singlelayer of cells with unaligned nuclei,but it is possible that there are multi-ple layers where not all cells are incontact with the basal lamina, pre-sumed to be present between the con-nective tissue and epithelial cells.Yolk is still present in the epithelialcells at this stage (Fig. 2c,d). Mitoticfigures are often seen.

The yolky endoderm fills the major-ity of the abdominal cavity (Fig. 2b).Nuclei are widely dispersed through-out this tissue, and cell boundaries areoften not readily observable in histo-logical sections. This appearance re-flects the difficulty in fixation of thisyolk-rich tissue, as in some prepara-tions, individual cells are clearly visu-alized (e.g., Fig. 11). The cellular na-ture of the yolky endoderm wasconfirmed by direct observation of dis-sected, living embryos at TS 3 (neu-rula) and TS 6 (gill circulation). Inaddition, when placed into 100%Ca2$-free Steinberg’s solution with0.5 mM ethyleneglycoltetraacetic acid(EGTA), the yolky endoderm dissoci-ated into a large pile of cells. Theyolky endoderm is surrounded by themesentery, a squamous layer of con-nective tissue, distinct from the undif-ferentiated connective tissue aroundthe anterior and posterior tubes.

Intestinal RemodelingHistological remodeling of the intes-tine begins when most of the yolk isstill present (Fig. 3a,b). Between TS14 and TS 15, the anterior tubechanges from a simple tube to themore complex structure of the adult(Fig. 3c). The epithelium becomesmultiply folded, and the presumptivemuscle around the outside and con-nective tissue cells in between themuscle and epithelium become dis-tinct. In the posterior tube, muscleand connective tissue are visible, butthe epithelial in-folding seen in theanterior tube has not yet occurred(Fig. 3d).

Beginning at TS 15, the process of

replacing the yolky endoderm with in-testinal epithelium begins, and thisreplacement can be considered a sec-ond phase of intestinal remodeling.Intestinal epithelium first developsbetween the anterior and posteriortubes on the dorsal surface of theyolky endoderm tube, starting fromboth ends (Fig. 4a). Simultaneously,there is a ventrally directed expansionof the epithelium along the walls ofthe yolky endoderm tube, so that thesurface area of epithelium expandsand yolky endoderm decreases (Fig.4b–d). The epithelial expansion alongdorsal surface connects from bothends before the ventral expansion is50% complete. Thus, the intestinal ep-ithelium proper gradually replacesthe yolky endoderm. Simultaneouswith this second phase of histologicalremodeling, the number of yolk plate-lets in the yolky endoderm cells de-creases. However, yolk platelets actu-ally disappear before many of theyolky endoderm cells are replaced.These changes leave two distinctkinds of cells facing the lumen: theintestinal epithelium and the now de-pleted yolky endoderm cells, with onlya few cells having a few yolk platelets(Fig. 4e,f).

Fate of the Yolky EndodermCellsThe histology described above raisedthe question of whether the yolkyendoderm cells contribute to the epi-thelium proper. Before TS 15, yolk ispresent in all intestinal endodermcells, that is, the yolky endoderm aswell as the epithelial cells of the ante-rior and posterior tubes. An exclusivehistological distinction develops be-tween yolky endoderm and adult in-testinal regions, as the cells of the an-terior and posterior tubes graduallylose yolk. By the time intestinal re-modeling is complete and yolkyendoderm replacement has begun atTS 15, no yolk platelets exist in anycell except the yolky endoderm. Thepresence of intracellular yolk plateletsis a unique identifier of yolkyendoderm. In addition, even when de-pleted of yolk and reduced in size nearborder regions, the shape of nuclei andstaining characteristics of the yolkyendoderm are distinct from the epi-thelium proper (Fig. 4e,f).

The histological distinctiveness ofthe yolky endoderm suggests that itmight not become part of the epithe-lium. To address whether the depletedyolky endoderm cells die by apopto-is, terminal deoxynucleotidyl trans-ferase–mediated dUTP nick end-la-beling (TUNEL) assay was carried outat different stages of intestine remod-eling. This assay failed to detect celldeath, even when yolk was completelydepleted (data not shown), suggestingapoptotic cells are not commonenough to detect or the yolkyendoderm cells die by some othermechanism. Evidence for an alternatedeath mechanism comes from serialsections of intestine, when most of theyolky endoderm is depleted. At thistime, it is common to see cells appar-ently being sloughed into the lumen(Fig. 5). The two examples shown aretaken from a large number of suchobservations seen exclusively in thedepleted yolky endoderm regions. Inaddition, close examination of thesloughed cells indicated that in everycase, there were multiple nuclei insideeach cell, suggesting that each yolkyendoderm cell is multinucleate. Fur-ther evidence of sloughing is that atthis stage, the posterior intestinal lu-men is filled with nuclei, an occur-rence not seen at other stages.

Fate MappingHistological examination suggeststhat the yolky endoderm does not be-come differentiated intestinal tissue.To test this suggestion further, largevegetal cells of morulae, with approx-imately 40–60 cells (Fig. 6), were in-jected with the lineage tracer fluores-cein dextran amine (FDA), and theirfate was followed. Embryos wereraised to different stages, when theirdigestive tract was dissected out andexamined for FDA fluorescence. Of the50 animals, 13 were prehatching (TS9–TS 15), 13 within the first weekposthatching, 15 within the secondweek posthatching, 8 within the thirdweek, and 1 at 27 days posthatching.

Injection of FDA into vegetal cellslabeled only the large, yolky cells ofthe endoderm and not differentiatedgut tissues in 43 of 50 cases. In sixcases, there were specks of FDA in theliver as well as labeled yolkyendoderm. The remaining froglet was

1262 BUCHHOLZ ET AL.

the oldest at 27 days posthatch, andwill be described later. The labeledyolky endoderm was sharply delin-eated from the more dorsal endoder-mal tissue, which was unlabeled (Fig.7a). As the amount of yolk decreased,the region of labeling decreased, with-

Fig. 4.

Fig. 5. Sloughing of depleted yolky endoderm cells. The top three panels show serial sections ofa large yolky endoderm cell appearing to come out of the cell layer and into the lumen. The bottomthree panels are another example. Multiple profiles of nuclei indicate that the sloughed cells maybe multinucleated. Scale bar " 10 #m in all panels.

Fig. 4. Loss of yolky endoderm. a–d: Cross-sections through the small intestine at differentstages of yolk resorption. This series is fromEleutherodactylus planirostris, and similar ob-servations for E. coqui indicate this pattern ofyolk resorption may be general for the entiregenus. a: After hatching, much of the peritonealcavity is still filled with yolk, and the dorsalmargin of the entire small intestine is soon re-placed with epithelium. b: As the epitheliumreplaces the yolky endoderm in a dorsal to ven-tral direction around the tube, the thickness ofthe yolky endoderm decreases as well as theintestine diameter. Note that the epithelial layer ismultiply folded as it progresses ventrally. c: As themultiply folded epithelium continues to expand,most yolk platelets have been resorbed, but thedepleted yolky endoderm cells are still present. d:Near the end of resorption, only a few remainingendoderm cells have yolk, whereas the majority ofthe surface is now intestinal epithelium. Figurese,f, from E. coqui, show that the boundary be-tween intestinal epithelium and yolky endodermcells is sharp. Note that these two cell typeshave different cell staining characteristics andnuclear morphology. e: At TS 14, the yolkyendodermal cells have many yolk platelets. f:The boundary remains distinct even with nearcomplete yolk resorption. Scale bar " 0.5 mmin a–d; 10 #m in e,f.

FROG NUTRITIONAL ENDODERM 1263

COLOR

out label appearing in the more differ-entiated intestinal tissue (Fig. 7b). Inthe third week posthatching, the lastof the label was found in the intestine(Fig. 7c), and a few froglets near theend of the third week had only a fewfluorescent specks remaining. At thistime, label was also found in the clo-aca (Fig. 7c) and in the mesonephros(Fig. 7d). The label in the cloaca wasdull and diffuse and associated withthe wastes accumulated there (Fig.7c). The only froglet that survivedlonger (27 days) had no yolky cells leftand no label in any part of the gut, butits mesonephros was labeled.

At the 40- to 60-cell stage, there is acap of smaller animal cells as well as afew intermediate-sized cells at theboundary between the small animalcells and the huge vegetal ones (Fig.6). To ensure that descendants of cellslabeled at the 40- to 60-cell stage re-tained label, FDA was injected intoanimal and intermediate cells. Whenthe embryos were raised to frogletsand dissected, label was found inmany different tissues including spi-nal cord, vertebrae, and intestine (Fig.7e–h). These results show that the la-

bel persisted for at least 2 weeks post-hatching. Although we have not com-pletely fate mapped the embryo, thepresence of label in differentiated in-testinal tissue shows that these tis-sues arise from more animal positions.

Labeling in Pronephros andMesonephrosApart from yolky endoderm, the onlyorgans labeled in the entire body ofthe froglets, injected with FDA intothe large vegetal cells, were the pro-nephros or mesonephros. Over half ofthe embryos at TS 9 to TS 15 showeddistinct labeling in pronephros (7 of13). Frogs after TS 15 showed no pro-nephric labeling, corresponding to theeventual disappearance of this struc-ture. Disappearance of dye in prone-phros was coupled with the gradualappearance of label in mesonephros.There was no mesonephric labeling inthe TS 9 (n " 2) and TS 11 (n " 2)embryos, but label started appearingweakly as early as TS 13 (n " 4). AtTS 15, two of five embryos had weaklabeling in mesonephros, but labelingwas strong in froglets up to 3 weeks

posthatching (36 of 37), correspondingto the period when most of the yolkyendoderm disappears (Fig. 7d). Thisfinding includes the oldest froglet at27 days that did not have label in itsgut. These results suggest that theFDA was accumulating, due to clear-ance from the blood after the utiliza-tion of the labeled yolky endoderm.

Expression of EcSox17In an attempt to follow the moleculardevelopment of the endoderm in E.coqui, we chose the transcription fac-tor Sox17. Based on expression pat-tern, overexpression, and deficiencyphenotypes, Sox17 is a key regulatorof endoderm development in X. laevis,mouse, and zebrafish (Hudson et al.,1997; Stainier, 2002; Shivdasani,2002; Lewis and Tam, 2006; Dickin-son et al., 2006). Sox17 in X. laevis is amember of a proposed core autoregu-latory network required for endoderm(Sinner at al., 2006).

Our EcSox17 cDNA (accession no.EF397004) consists of a 36-nucleotide(nt) 5%-untranslated region (UTR),1,143-nt open reading frame (ORF),

Fig. 6. Eleutherodactylus coqui morula. a: The E. coqui morula has a cap of small cells at the animal pole (A) and very large vegetal cells. There isa distinct boundary between the small animal cells and the larger vegetal ones. The animal pole is tilted forward in this photograph, so the full extentof the vegetal cells cannot be appreciated. b: This drawing of an &40-cell morula shows the relationship between the animal cap of small cells andthe very large vegetal cells. A few intermediate sized cells are forming at the boundary. The vegetal furrows are indistinct but extend to the vegetalpole. The inset shows a &48 cell X. laevis morula, at the same scale as the E. coqui morula in (a) and (b). The size disparity between animal and vegetalcells is clearly less than in E. coqui. Reproduced with permission from Nieuwkoop and Faber (1956).

1264 BUCHHOLZ ET AL.

and 363-nt 3%-UTR, ending in a polyAtail. Comparing to Sox 17’s in X. lae-vis, the nucleotide sequence of theORF is 68% identical to XSox17'1,53% identical to XSox17'2, and 48%identical to XSox17(, and the aminoacid sequence identities are 69%, 66%,and 45%, respectively. The HMG Box,nucleotides 211–417, is 82% identicalto the XSox17'1 HMG Box domain.The EcSox17 HMG domain is virtu-ally identical at the amino acid level toother vertebrate Sox17 HMG do-mains, e.g., 89% with zebrafish, 95%with human, and 97% with XSox17'1.

In situs of EcSox17 RNA in gastru-lae showed nice staining vegetal to theblastoporal lip, but not in the rest ofthe vegetal region, as expected (Fig.8a). When gastrulae were cut sagit-tally, staining was found on the invo-luted cells, which will line the arch-enteron (Fig. 8c). The large yolky cells,now directly exposed to the probe, stilldid not stain.

Although these patterns suggestedthat EcSox17 is expressed only at thesurface of the embryo near the blas-toporal lip, in situs of Sox17 in yolk-rich regions of X. laevis can give false-negative reactions (D’Souza et al.,2003). To test whether EcSox17 RNAis present in the yolk-rich region, E.coqui gastrulae, equivalent to Nieu-wkoop and Faber stage 10, were dis-sected into dorsal and ventral mar-ginal zones and the yolky core (Fig.9a,b). RNA was isolated from pooledpieces, and examined for the presenceof EcSox17 and EcBra RNA by reversetranscriptase-polymerase chain reac-tion (RT-PCR). EcBra is expressed inthe prospective mesoderm as in othervertebrate embryos (Ninomiya et al.,2001), and EcBra expression in themarginal zone was confirmed (Fig. 9).Expression was absent in the vegetalcore in two cases and present at a lowlevel in one case, similar to our previ-ous report (Ninomiya et al., 2001). Incontrast, EcSox17 RNA was found inboth the marginal zones and in thevegetal core (Fig. 9). Neither EcSox17nor EcBra RNA was detected in sam-ples without reverse transcription.Detection of EcSox17 RNA by RT-PCRsuggests that EcSox17 RNA cannot beused to distinguish the endodermforming near the blastoporal lip fromthat forming from the yolk-rich vege-tal core.

The EcSox17 RNA, detected in thevegetal core, is likely in part maternalin origin. EcSox17 RNA was found inRNA from early cleaving embryos andfrom ovarian oocytes by RT-PCR (datanot shown). Also, in our re-cloning ofEcSox17 cDNA from Hawaiian E. co-qui, ovarian RNA was used as men-tioned in the Experimental Proce-dures section.

Effect of Vegetal CleavageInhibitionThe yolky endoderm cells appear to beplaying mainly a nutritional role. Thisfinding raises the question whethercleavage of these cells is necessary fordevelopment to proceed. To answerthis question, we used the cell cycleregulator c-mos, which can arrestblastomeres in X. laevis embryos atmetaphase (Sagata et al., 1989; Yewet al., 1991; Masui, 2000). Accord-ingly, one or more blastomeres of E.coqui cleaving embryos were injectedwith c-mos RNA. Although the mitoticstage was not checked, large areas ofvegetal surface remained uncleaved 1day later, when embryos were blastu-lae. When cleavage of a large vegetalblastomere at the 60- to 100-cell stagewas inhibited, the embryo succeededin gastrulation. The blastopore closed,excluding a large amount of yolk (Fig.10a,b). Despite this loss, patterning ofthe embryo’s body was normal as in-dicated by a head with eyes and thefore- and hindlimb buds (Fig. 10b,c).Embryos with reduced yolk could de-velop normally to become smallerthan normal prefeeding frogs (Fig.10d).

When vegetal cells of embryos a fewdivisions later were injected with c-mos RNA, the embryos internalizedall of the yolky cells at gastrulationand continued development. Largeuncleaved blastomeres were presentamong the cells of the gut (compareFig. 11a vs. 11b). These included caseswhere the uncleaved blastomere ex-tended ventrally from the floor of thearchenteron (Fig. 11c). This position isexpected from fate mapping, becausethe vegetal surface of the cleaving em-bryo ends up as the archenteric floor.

We succeeded in raising a few mos-injected embryos to a time when thecontrol froglets had used up most oftheir yolk. Five of six had a large

amount of amorphous debris in theirgut lumen (Fig. 11d), with debrispushed into the stomach and the hind-gut. This appearance indicates poorutilization of the yolk.

DISCUSSIONThe treatment of yolk is fundamen-tally different between the amphibi-ans and the amniotes, such as reptilesand birds. In amphibians, yolk in theform of yolk platelets is incorporatedinto cells. These cells later become tis-sues of the embryo. In amniotes, thelarge mass of yolk is not incorporatedinto cells. The yolk is broken downand moved into the developing embryoby means of the blood vessels of theyolk sac, which surrounds the un-cleaved yolk. The evidence that wepresent here suggests that an inter-mediate state is present in the directdeveloping frog E. coqui. The yolk iscompartmentalized into cells, butthose cells do not contribute to thetissues of the froglet.

Evidence for NutritionalEndoderm in E. coquiSeveral lines of evidence suggest thatthe yolky endoderm cells serve only anutritional role. The gut of thehatched froglet consists of two grosslydifferent tissues. These are the trans-lucent tissues, forming the differenti-ating stomach and hindgut, and theyolky endoderm. The junction be-tween these two tissues, especiallyalong the dorsal side, is distinct, andthis sharp distinction is emphasizedby histological examination. One pos-sibility is that the junction is whereyolky endoderm cells differentiateinto functional intestinal cells, buthistology suggests otherwise. Evenwhen depleted of yolk, the formerlyyolky cells at the junction remainlarge and undifferentiated. Sloughingof these cells into the lumen is oftendetected.

Fate mapping provides further evi-dence for loss of the yolky cells, ratherthan differentiation. The yolky cellswere strongly labeled by injection ofFDA into large vegetal cells. If theyolky cells later differentiate, labelshould be found in the differentiatedgut tissues, as the mass of yolky cellsdecreased. This transition did not oc-cur. Rather, with the disappearance of

FROG NUTRITIONAL ENDODERM 1265

Fig. 7.

Fig. 8.

Fig. 10.

Fig. 11.

the yolky cells, label was detected inmesonephros and cloaca. Mesonephriclabeling indicates that FDA was beingcleared from blood. This finding mayreflect uptake of FDA into circulationupon death of the labeled cell. Alter-natively, the FDA may be taken intocirculation along with yolk-derivedmetabolites, and then filtered out inthe mesonephros. The strong labeling

of the mesonephros indicates that theFDA is trapped there.

There are alternative explanationsfor the cloacal labeling as well.Hatched froglets use yolk for 1–2weeks before they start eating. Duringthis time, the cloaca is often swollenwith waste. One possibility for thepresence of FDA in the wastes is thata small amount of FDA is moved from

the mesonephros through the meso-nephric duct to the cloaca. A secondpossibility is that FDA is included inthe wastes as the labeled yolky cellsbreak down. A resolution betweenthese alternatives may come by ob-serving the path of FDA, injected di-rectly into blood. The histological im-ages of cells, apparently entering thelumen of the gut, along with detectionof FDA in the cloacal wastes stronglysuggests that the fate of the largeyolky cells is death.

We attempted to show death di-rectly by TUNEL labeling, but no la-beled cell nuclei were observed. Apo-ptotic epithelial cell death in theremodeling intestine of X. laevis is as-sociated with an increase in thyroidhormone, either added to the rearingwater or endogenous thyroid hormoneduring natural metamorphosis and isdetectable above background levelsfor only 2–3 days (Ishizuya-Oka,1996). In the case E. coqui, loss ofyolky endoderm cells begins after thethyroid hormone peak (see below) andoccurs gradually over a period of 2–3weeks, where there may not be a burstof cell death that is detectable abovethe background levels. Also, changesin nuclear histology associated with

Fig. 9. EcSox17 reverse transcriptase-polymerase chain reaction (RT-PCR) of gastrula. a: Earlygastrula embryos were dissected into three pieces: dorsal marginal zone (D), ventral marginal zone(V), and vegetal core (C) as indicated in the diagrams of a sagittal section and a ventral view. b: Thisphotograph of a dissected embryo shows three pieces, corresponding to the diagram of the vegetalview. Note the dorsal lip of the blastopore on the piece of dorsal marginal zone. c: RT-PCR wascarried out on RNA isolated from pooled samples of the three pieces: dorsal, ventral, and core.EcSox17 RNA was detected in all pieces. EcBra RNA, a mesodermal marker, was detected in thedorsal and ventral pieces, but not in the core. EcL8 RNA, which codes for a ribosomal protein wasused as a loading control. No amplification occurred in the absence of reverse transcription ()RT).

Fig. 7. Fate mapping. a: This gut, isolated from a froglet injected with fluorescein dextran amine (FDA) into a vegetal cell, still has a lot of yolk. Thelarge yolky cells fluoresce green while the more differentiated gut tissues lack fluorescence and appear orange. b: There is less yolk in this gut. Againthe large yolky cells fluoresce green and the rest of the tissues show little fluorescence. The lack of fluorescence in the differentiated tissues indicatesthat green yolky cells do not subsequently become intestinal tissues. c: This gut has very little yolk left, visible as the intense green stain on the rightside. The more diffuse green stain on the left is material within the cloaca, which will be defecated. d: This mesonephros is from the same froglet asFigure 7c. There is a lot of FDA present, which is likely due to the attempt of the froglet to excrete the dye, which was filtered from the blood. Thismesonephric staining was generally not found in embryos, sacrificed with more yolk present. e: This spinal cord was isolated from a froglet that wasinjected with FDA into a cell near the animal pole. Label is present on one side only. f: This froglet was also injected with FDA into an animal cell, andlabel is present in vertebrae. g,h: Phase (g) and fluorescent confocal (h) views of the junction between the stomach and the intestine in a froglet thatwas injected with FDA into an intermediate sized cell. Labeled cells are present in the intestine but not the stomach. Scale bar " 0.5 mm in a–c,300 #m in e, 150 #m in f,g,h.

Fig. 8. EcSox17 in situ hybridization of gastrula. a: Cells vegetal to the blastoporal lip (B) express EcSox17. b: No signal is present with a sense probe.c: To reveal internal expression, gastrulae were bisected, before in situ hybridization. With an antisense probe (left), the large mass of yolky cells isunstained, and staining is present along the involuted cells (arrowhead) that will line the archenteron. There is no staining of these cells with the senseprobe (right).

Fig. 10. Vegetal cleavage inhibition at 60- to 100-cell stage. Mos RNA was injected into a large vegetal blastomere at the 60- to 100-cell stage, andthe embryos were raised. a: A mos-injected embryo (bottom) is compared with a normal TS5 embryo (top). The mos-injected embryo failed toincorporate all of the yolk. Nonetheless, the embryo has a normal anterior–posterior axis as indicated by the presence of all forelimbs and hindlimbs.b: A normal TS 6 embryo (left) is shown next to a mos-injected embryo (bottom). The injected embryo has a normal head and limbs, but lacks mostof its yolk (Y). c: A normal TS 9 embryo (top) is compared with a mos-injected embryo (bottom), which has a small supply of yolk. d: A froglet (bottom),derived from a mos-injected embryo that had lost a lot of yolk, is much smaller than its normal sibling. The mos-injected embryo has not finishedabsorbing its tail. Scale bars " 1 mm.

Fig. 11. Vegetal cleavage inhibition at the &200-cell stage. Mos RNA was injected into a large vegetal blastomere a few divisions later than in Figure10. a: A section through a control TS 12 embryo shows many large endodermal cells. The spaces are artifacts. b: A section through a mos-injectedembryo shows the presence of several huge, uncleaved vegetal cells in the endoderm. c: Another example of a mos-injected embryo shows severalhuge, uncleaved cells, including one that extends to the floor of the archenteron. Note the limb buds. d: This section shows the gut of a mos-injectedembryo that developed to a froglet. There is a large mass of debris (D) in the intestinal lumen, as well as wastes (W) in the cloaca. Mesonephros (M).Scale bar " 0.5 mm in a–c, 0.25 mm in d.

FROG NUTRITIONAL ENDODERM 1267

apoptosis in X. laevis, including chro-matin condensation and nuclearbreak up (Ishizuya-Oka and Ueda,1996), are not observed in E. coqui.Even when the yolky endoderm cellsare being extruded into the lumen, thenuclear structure is intact, lackingany histological indications of apopto-sis. Therefore, death of the yolkyendoderm cells appears not to be byapoptosis and occurs after extrusioninto the lumen.

Comparison BetweenE. coqui and X. laevisYolk platelets are present throughoutthe cytoplasm of the fertilized X. lae-vis egg (Danilchik and Gerhart, 1987;Danilchik and Denegre, 1991). Allcells inherit yolk, which is metabo-lized within the embryonic cells. Inbirds, this metabolism within cells oc-curs, but in addition, the uncleavedyolk is broken down, moved into theyolk sac endoderm, and distributed bymeans of the blood to the growing em-bryo (Bellairs, 1964; Yoshizaki et al.,2004). This second utilization of yolklikely occurs in X. laevis, although thisuse does not appear to be documented.The X. laevis endodermal cells areyolk-rich, and this yolk is consumedbefore initiation of feeding by the tad-pole. Some of the metabolites fromthese endodermal yolky cells likelyenter circulation to supply nutrition tothe rest of the embryo. A simple test ofthis possibility is to see whether thedry mass of the embryo, minus thegut, increases as the yolk is used be-tween stages NF 41 and 46.

When fate mapping is done, all re-gions of the endoderm in a X. laevisneurula contribute cells to the differ-entiated tadpole intestine (Chalmersand Slack, 2000). This result suggeststhat a nutritional endoderm, as wehave documented for E. coqui, doesnot exist in X. laevis. The fate map-ping, however, is not sufficiently de-tailed to determine the fate of eachindividual cell. It is possible that afraction of the X. laevis endodermalcells die after depletion of their yolk,and these cells would be equivalent tothe nutritional endoderm in E. coqui.Alternatively, the X. laevis embryomust generate a very long coiled gutin a short amount of time (Chalmersand Slack, 1998), so all of the endoder-

mal cells may use their yolk supply forrapid division and differentiation.

Expression of EcSox17

Zygotic expression of X. laevis Sox17depends on the maternally expressedtranscription factor VegT and Nodalligands, activated by VegT (Yasuo andLemaire, 1999; Xanthos et al., 2001).RNA for EcVegT, the E. coqui ortho-logue, is greatly enriched near the an-imal pole of the E. coqui oocyte andcleaving embryo, unlike the localiza-tion of VegT RNA to the vegetal cortexof the X. laevis oocyte (Beckham et al.,2003). Nodal signaling in the E. coquiembryo, as assayed by mesoderm in-ducing activity, is restricted to thesurface cells of the early gastrula (Ni-nomiya et al., 2001). It is not presentat the vegetal pole or in the deep cellsextending from the vegetal pole to theblastocoele floor.

We, therefore, expected EcSox17 ex-pression to be limited to the equatorialsurface and to be absent from the veg-etal pole and the vegetal core. Thisexpected absence of EcVegT expres-sion would correspond to the nutri-tional endoderm, whose fate is to de-generate rather than to differentiateas intestinal tissues. Although in situpatterns of EcSox17 RNA fit these ex-pectations, RT-PCR of isolated vegetalcores demonstrated the presence ofEcSox17 RNA there. This result raisesquestions of the origin and function ofEcSox17 RNA in the vegetal core.With respect to origin, some of theEcSox17 RNA is maternal, an expres-sion not present in X. laevis (Hudsonet al., 1997). One possibility is thatzygotic signaling is very inefficientwhen cells are so big and yolky, so thelarge vegetal cells in E. coqui may notbe able to rely on a zygotic source ofany necessary RNAs. This speculationthen begs the question as to why anyRNA is necessary for cells, whichserve only a nutritional function andare scheduled to die. Nonetheless,these cells survive for more than amonth, and they must deliver metab-olites of yolk to the embryo. Furtheranalysis of gene expression of the nu-tritional endoderm would help in afurther understanding of the nature ofthese cells.

Relationship BetweenThyroid Hormone andIntestinal DevelopmentIn all frogs studied to date, thyroidhormone (TH) is required for intestineremodeling from the larval form to theadult version (Shi, 2000). Our analy-sis of gut development in E. coqui sug-gests biphasic intestinal developmentwith respect to TH. In X. laevis, intes-tinal remodeling occurs in a sequenceof events preceded by limb outgrowthand followed by tail resorption. All ofthese events, as well as other organtransformations, require TH, whichreaches a peak in blood concentrationwhen the tail begins resorbing. In E.coqui, intestinal remodeling of the an-terior part from a simple tube to themultiply folded adult form is completeby TS 15 before hatching. This remod-eling occurs in the same sequence asin Xenopus, that is, limbs, intestine,and then tail. At this point, most ofthe yolky endoderm is still present,and then over the next 6–8 days afterhatching and tail resorption, the yolkyendoderm is replaced by an epithe-lium in what can be considered a sec-ond phase of postembryonic intestinaldevelopment.

Unlike in X. laevis and other frogs,the role played by TH in intestinaldevelopment in E. coqui is unclear.Direct measurements of TH across de-velopment have not been carried out,but the beginning of endogenous THproduction is predicted to be aroundTS 11, just after the thyroid glandforms (Jennings and Hanken, 1998)and when TH-dependency begins(Callery and Elinson, 2000; Callery etal., 2001). The peak histological activ-ity of the thyroid gland and beginningof tail resorption occurs at TS 13 (Jen-nings and Hanken, 1998, Townsendand Stewart, 1985), suggesting thepeak of endogenous TH titer at thisstage. However, peak TH levels mayoccur after hatching because TH re-ceptor beta (TR() expression, which ispresumably correlated with TH titersas in X. laevis (Shi and Brown, 1993;Shi, 1994), continues to increase afterhatching (Callery and Elinson, 2000).Thus, TH is likely present during pre-hatching and potentially involved inintestinal remodeling of the anteriortube, and the high TR( expressionposthatching suggests high TH levels

1268 BUCHHOLZ ET AL.

even after hatching during yolkyendoderm replacement. Future hor-mone measurements and endocrineexperiments are required to deter-mine the role of TH in E. coqui intes-tinal development.

Why Is Holoblastic CleavageConserved in Amphibians?The amniote egg was a key innovationin the evolution of terrestrial verte-brates, the tetrapods. Eggs of extantamphibians and amniotes differ inthree main ways. First, the amnioteegg generates a set of extraembryonicmembranes, the amnion, chorion, al-lantois, and yolk sac, which the am-phibian egg does not. Second, themaximum size of an amphibian egg isapproximately 1 cm in diameter,while eggs of turtles, lizards, snakes,and birds, are generally much larger.Third, amphibian eggs divide holo-blastically, while amniote eggs, exceptfor the small eggs of marsupial andplacental mammals, divide meroblas-tically.

Many hypotheses have been pro-posed for the origin of the extraembry-onic membranes. The two hypothesesunder most active current consider-ation are that the membranes wereoriginally used to permit terrestrialdevelopment and that the membraneswere originally used to mediate ma-ternal exchange for an embryo devel-oping within the mother’s body. Vigor-ous attempts have been made to usephylogenetic methods to decide be-tween these hypotheses, but these at-tempts have not produced a resolution(Laurin and Reisz, 1997; Wilkinson etal., 2002; Laurin, 2005). On the otherhand, there is general consensus onbiological and phylogenetic groundsthat the eggs of the original tetrapodswere much like those of extant am-phibians and cleaved holoblastically(Romer, 1957; Goin and Goin, 1962;Carroll, 1969, 1970; Elinson, 1987,1989; Collazo et al., 1994; Packardand Seymour, 1997; Arendt andNubler-Jung, 1999; Elinson and Beck-ham, 2002; Chea et al., 2005). A tran-sition to meroblastic cleavage wouldbe one of the conditions, along withprovision for gas exchange, which per-mitted the great increase in yolk con-tent and egg size, found in the am-niotes.

Based on phylogenetic analysis,Collazo et al. (1994) proposed that atransition from holoblastic to mero-blastic cleavage occurred five times inthe chordates. The only instance of ameroblastic to holoblastic transition iswithin mammals. All amphibianshave holoblastic cleavage (Elinson,1987). Chipman et al. (1999) calledcleavage in Hyperolius puncticulatus“pseudo-meroblastic,” because theyolk-rich vegetal cells appeared to beso large. Their observations werebased on fixed embryos, however, inwhich cell boundaries among yolk-richcells may be obscured. It would beworth examining living embryos of H.puncticulatus and related Hyperoli-idae with larger eggs to see if thereare any exceptions to the holoblasticrule. The conservation of amphibianholoblastic cleavage raises the ques-tion as to why there has been no sec-ond viable origin of meroblastic cleav-age among tetrapods in the past 360million years.

Given the presence of nutritionalendoderm in E. coqui, we askedwhether cleavage of the yolky vegetalregion of the egg is necessary. Inhibi-tion of early vegetal cleavage led to afailure to incorporate all of the yolk atgastrulation. Nonetheless, the em-bryos closed the blastopore and hadnormal patterning of the anterior–posterior axis. When vegetal cleavagewas inhibited a few divisions later, allof the yolky cells, including large un-cleaved ones, were internalized dur-ing gastrulation. In X. laevis, yolkyvegetal cells actively participate in therearrangements during gastrulationin a movement called vegetal rotation(Winklbauer and Schurfield, 1999;Ibrahim and Winklbauer, 2001). It isnot known whether vegetal rotationoccurs in E. coqui, but it is unlikely tobe required, given the closure of theblastopore in E. coqui embryos withhuge cleavage-inhibited cells. Themorphology of gastrulation is differ-ent between amphibians and am-niotes, and Arendt and Nubler-Jung(1999) hypothesized that differencesin gastrulation movements evolved inconjunction with the presence of anuncleaved yolk mass. Nonetheless,our results indicate that the amphib-ian pattern of gastrulation move-ments does not require complete

cleavage of the vegetal region intosmall cells.

In amniotes, the yolk is envelopedby the yolk sac after gastrulation. InE. coqui, the yolky cells are second-arily enveloped by the body wall,which brings musculature from thesomites to the ventral midline (Elin-son and Fang, 1998). Although themovement of the body wall in E. coquiis superficially similar to that of theyolk sac, it is not homologous. Thebody wall includes somitic mesodermrather than the splanchnic mesodermof the lateral plate, and it does notinclude the endoderm. Nonetheless,the similarity suggests that the inde-pendent evolution of an envelopinglayer can occur, and it could be used tosurround an uncleaved yolk mass.Neither the amphibian gastrulationpattern nor the lack of a yolk envelop-ing layer appears to preclude a tran-sition to meroblastic cleavage.

We managed to raise a few embryoswith internalized large yolk cells tofroglets, and these animals had alarge amount of debris in the lumen ofthe gut. The debris spread from theintestine into the stomach. These dif-ficulties in using the yolk from thelarge uncleaved cells suggest that themembranes of the yolky endodermcells in amphibians may function inyolk metabolism, either its breakdownor its transport. Perhaps this functionwould have to be acquired by anothertissue, such as the amniote’s yolk sac,before cleavage of the yolky regioncould be abandoned.

EXPERIMENTALPROCEDURESAnimals and EmbryosAdult E. coqui were collected inPuerto Rico under permits issued bythe Departmento de Recursos Natura-les y Ambientales and in Hawaii un-der Injurious Wildlife Export Permitsissued by the Department of Land andNatural Resources. A reproductivecolony was maintained at DuquesneUniversity, and adults and embryoswere treated according to protocolsapproved by the Institutional AnimalCare and Use Committee.

All embryos were obtained fromnatural matings within our laboratorycolony (Elinson et al., 1990). After the

FROG NUTRITIONAL ENDODERM 1269

fertilized eggs were laid, they were re-moved from the guarding male andkept in a moist chamber. When em-bryos reached a desired stage, theywere submerged in 20% Steinberg’ssolution (12 mM NaCl, 0.134 mM KCl,0.068 mM Ca(NO3)2, 0.166 mMMgSO4, 1 mM Tris HCl, pH &6.5) toallow the jelly capsule to swell. Thejelly capsule, consisting of three jellylayers, and the fertilization envelopewere removed using watchmaker’sforceps, depending on the use of theembryos. While it is relatively easy todo this on embryos that have com-pleted neurulation, other treatmentswere used for earlier embryos withtighter fertilization envelopes. Afterremoving the outer and middle jellylayers of the earlier embryos with for-ceps, the inner jelly layer was dis-solved with 2.5% cysteine, pH 8.0, bygently swirling for 5–8 min. Embryoswere thoroughly rinsed and thenraised in 20% Steinberg’s solution. Toweaken the fertilization envelope, de-jellied embryos were further treatedwith 2.5% cysteine, 2% papain, and2% '-chymotrypsin, pH 8.0, for 10–20sec (Hennen, 1973). The embryos werewashed with 20% Steinberg’s solutionand transferred to 1% bovine serumalbumen in 100% Steinberg’s solution.The fertilization envelope was re-moved carefully with forceps.

The stage of development was de-termined according to the table byTownsend and Stewart (1985). Eachstage is denoted “TS number,” rangingfrom TS 1, the cleaving embryo, to TS15, the hatched froglet. The hatchedfroglet has a large mass of yolky cells,and the yolk is used over the next 1–2weeks. Developing guts after hatchingwere compared based on the amountof yolk remaining.

In addition to E. coqui embryos, afew embryos of Eleutherodactylus pla-nirostris, collected in Tallahassee,Florida, were examined. This speciesis closely related to E. coqui. Theyhave the same nesting ecology andprogress through the same develop-mental stages.

HistologyEmbryos or isolated guts were fixed inSmith’s fixative (Rugh, 1962) or 4%paraformaldehyde in phosphate buff-ered saline, pH 7.4 (PFA). The former

was generally used for younger em-bryos with a lot of yolk, whereas thelatter was used for isolated later gutswith reduced yolk. Embryos wereanesthetized in 0.05% Tricaine meth-anesulfonate (MS222), buffered to pH7.3 by the addition of solid Na2PO4,and guts were dissected in 200%Steinberg’s solution. For fixation us-ing Smith’s fixative, equal parts of twostock solutions were mixed fresh foruse, and embryos were fixed in thedark. Smith’s Stock A is 1% potassiumdichromate (K2Cr2O7), and Smith’sStock B is a 5% acetic acid, 7.4% form-aldehyde. After fixation, the embryoswere rinsed with water and stored inbuffered 3.7% formaldehyde at roomtemperature. For fixation using PFA,guts were fixed for 2 hr and thenstored in 70% ethanol at room temper-ature or 100% ethanol at )20°C.Fixed guts were then dehydrated andinfiltrated with paraffin in an Excel-sior ES tissue processor (Thermo Elec-tron Corp.) before embedding in par-affin blocks. They were sectioned at 8#m and dried onto Probe-On Plus mi-croscope slides (Fisher). For histologi-cal staining, sections on slides wererehydrated, stained with hematoxylinand eosin, dehydrated, and thenmounted with Permount (Fisher).Digital pictures were taken using aMicroprocessor digital camera (Q-Im-aging) on an Olympus compound mi-croscope.

Lineage TracingAfter removal of their outer and middlejelly layers, embryos with 40–60 cellswere partially submerged in 5% Ficollprepared in 100% Steinberg’s solution.One large vegetal blastomere was in-jected with 9.2 nl of 50 mg/ml FDA(10,000 molecular weight, anionic, ly-sine fixable, Molecular Probes), using aDrummond Nanoject Variable Auto-matic Injector. Injected embryos werecultured individually in 5% Ficoll in100% Steinberg’s solution with 50#g/ml gentamicin overnight and thenplaced in 20% Steinberg’s solution,which was changed regularly.

When the embryos or frogletsreached the desired stage, they wereanesthetized in 1% MS 222 in 100%Steinberg’s solution for 4–5 sec anddissected in 200% Steinberg’s solu-tion. They were fixed in MEMFA (100

mM MOPS, 2 mM EGTA, 1 mMMgSO4, 3.7% formaldehyde, pH 7.4;Harland, 1991) and stored at 4°C. Thefixed frogs were dissected further toseparate different parts of the body,for example, foregut, hindgut, liver,mesonephros, heart, pronephros (ifpresent), vertebrae, spinal cord, brain,eyes, forelimbs, and hindlimbs. Eachseparated part or organ of the frogbody was mounted on a glass slide in adrop of anti-fade buffered glycerol (25mg of n-propyl gallate, 5 ml of PBS, 5ml of glycerol, pH 8.0–8.6). Themounted organs were covered with acover glass, which was gently presseddown, and the slides were kept at 4°C.For controls, organs from uninjectedfroglets were processed in the sameway.

The specimens were examined us-ing a Nikon Microphot-SA microscopewith fluorescence filters for fluores-cein isothiocyanate (FITC). Confocalimaging was done using a Leica TCSSP2 Confocal Microscope. The wave-length was set at FITC WIDE and spe-cial FITC GREY, and the lasers were488 Ar/ArKr set at minimum.

Cloning EcSox17 cDNARNA was extracted from neurula andlimb bud stages (TS 3–4) with Trizol(Invitrogen) and used to preparecDNA. Degenerate primers, based onconserved Sox17 sequences betweenX. laevis, zebrafish, mouse, and hu-man, were used to amplify a 261-bpfragment by PCR. This fragment wascloned into a pGEM T-Easy vector. Ithad high sequence identity to Sox17’sin other species, so it was extended by3%- and 5%-rapid amplification of cDNAends to yield a consensus sequence of1,565 nt, which contains the completeORF for a 380 amino acid protein. OurEcSox17 clones were derived from E.coqui, collected in Puerto Rico, butduring the course of these experi-ments, our source of E. coqui changedfrom Puerto Rico to Hawaii. We re-cloned EcSox17 from ovary of Hawai-ian E. coqui by PCR, using exact prim-ers, and sequencing confirmed thatthe two populations had the same Ec-Sox17.

In Situ HybridizationDejellied embryos were fixed inMEMFA. After 3 hr, the fertilization

1270 BUCHHOLZ ET AL.

envelope was removed carefully withforceps. Bisection of embryos, if re-quired, was done at this point with arazor blade. The embryos were fixedfor a further 2 hr, washed twice in100% ethanol, and stored at )20°C in100% ethanol until further use.

EcSox17 cDNA was cloned into apGEM-T EASY vector (Promega). Theplasmid was cut with NcoI and tran-scribed with SP6 to produce the anti-sense probe, labeled with digoxigenin.The plasmid was cut with SpeI andtranscribed with T7 to produce thesense probe. In situ hybridization wascarried out with hydrolyzed probes ac-cording to standard procedures (Har-land, 1991; Nath et al., 2005), whichincluded a 20-min treatment of thefixed embryos with 10 #g/ml protein-ase K. Hybridization was visualizedby using BM Purple AP substrate(Roche) for alkaline phosphatase.

Dissection of Embryos andRT-PCRGastrulae were dissected into differ-ent pieces for RT-PCR. After removalof the jelly layers and fertilization en-velope, the embryo flattened into theshape of a fat pancake. The embryowas positioned with a hair loop so thatits vegetal half was up and the dorsalblastoporal lip was visible. Using finetungsten needles, two incisions ofequal length, extending toward thecenter of the embryo were made onopposite sides lateral to the dorsal lip(Fig. 9). The inner ends of the inci-sions were joined by cutting in a semi-circular manner on either side of thecentral vegetal core. At the end of dis-section, a semicircular dorsal piecewith the dorsal lip, a semicircular ven-tral piece, and a central core of theearly gastrula remained (Fig. 9). RNAwas extracted from 8–10 dissectedpieces or 4–5 whole embryos withTrizol.

cDNA for PCR was reverse tran-scribed from RNA using the M-MLVenzyme and oligo dT. Primers forEcSox17 were as follows: Forward5%-CTACAGCTTGCCTACACCTGA-3%and Reverse 5%-CATCCCATCATAC-TCTGTGCA-3%. These primers am-plify a 201-bp fragment. EcL8, a ribo-somal protein gene that is expressedthroughout development, was used aloading control. Primers for EcL8

were published previously (Calleryand Elinson, 2000), as were primersfor the mesodermal marker EcBra(Ninomiya et al., 2001).

Cleavage InhibitionThe plasmid pX64, with the X. laevisc-mos coding region in pSP64 (Pro-mega), was provided by Nick Dues-bery, Van Andel Institute, Grand Rap-ids, MI. The pX64 plasmid was cutwith BglI and transcribed with SP6,using the mMessage mMachine Kit(Ambion, Austin, TX). Cleaving E. co-qui embryos were placed in 20% Stein-berg’s solution, and the outer jelly lay-ers were removed with forceps. One ormore blastomeres near the vegetalpole were injected with 9.2–36.8 nl of200 ng/#l c-mos RNA using a Drum-mond Nanoject Injector, and the em-bryos were allowed to develop.

ACKNOWLEDGMENTSWe thank Kimberly Nath for technicalassistance, Dr. Nick Duesbery (VanAndel Institute, Grand Rapids, MI)for the c-mos plasmid, Brian Storz forthe gift of E. planirostris, and RobertHogarth for the drawing of the E. co-qui cleaving embryo.

REFERENCES

Arendt D, Nubler-Jung K. 1999. Rearrang-ing gastrulation in the name of yolk: evo-lution of gastrulation in yolk-rich am-niote eggs. Mech Dev 81:3–22.

Beckham YM, Nath K, Elinson RP. 2003.Localization of RNAs in oocytes of Eleu-therodactylus coqui, a direct developingfrog, differs from Xenopus laevis. EvolDev 5:562–571.

Bellairs R. 1964. Biological aspects of theyolk of the hen’s egg. Adv Morphog 4:217–272.

Callery EM, Elinson RP. 2000. Thyroidhormone-dependent metamorphosis in adirect developing frog. Proc Natl AcadSci U S A 97:2615–2620.

Callery EM, Fang H, Elinson RP. 2001.Frogs without polliwogs: evolution ofanuran direct development. Bioessays 23:233–241.

Carroll RL. 1969. Problems of the origin ofreptiles. Biol Rev 44:393–432.

Carroll RL. 1970. Quantitative aspects ofthe amphibian-reptilian transition.Forma Functio 3:165–178.

Chalmers AD, Slack JMW. 1998. Develop-ment of the gut in Xenopus laevis. DevDyn 212:509–521.

Chalmers AD, Slack JMW. 2000. TheXenopus tadpole gut: fate maps and

morphogenetic movements. Development127:381–392.

Chea HK, Wright CV, Swalla BJ. 2005.Nodal signaling and the evolution of deu-terostome gastrulation. Dev Dyn 234:269–278.

Chipman AD, Haas A, Khaner O. 1999.Variations in anuran embryogene-sis: yolk-rich embryos of Hyperoliuspuncticulatus (Hyperoliidae). Evol Dev1:49–61.

Collazo A, Bolker JA, Keller R. 1994. Aphylogenetic perspective on teleost gas-trulation. Amer Nat 144:133–152.

Danilchik MV, Gerhart JC. 1987. Differen-tiation of the animal–vegetal axis in Xe-nopus laevis oocytes. I. Polarizedintracellular translocation of plateletsestablishes the yolk gradient. Dev Biol122:101–112.

Danilchik MV, Denegre JM. 1991. Deepcytoplasmic rearrangements duringearly development in Xenopus laevis. De-velopment 111:845–856.

Dickinson K, Leonard J, Baker JC. 2006.Genomic profiling of Mixer and Sox17 (targets during Xenopus endoderm devel-opment. Dev Dyn 235:368–381.

D’Souza A, Lee M, Taverner N, Mason J,Carruthers S, Smith JC, Amaya E, Pa-palopulu N, Zorn AM. 2003. Molecularcomponents of the endoderm specifica-tion pathway in Xenopus tropicalis. DevDyn 226:118–127.

Elinson RP. 1987. Change in developmen-tal patterns: embryos of amphibianswith large eggs. In: Raff RA, Raff EC,editors. Development as an evolutionaryprocess. NY: Alan R. Liss, Inc. p 1–21.

Elinson RP. 1989. Egg evolution. In: WakeDB, Roth G, editors. Complex organis-mal functions: integration and evolutionin vertebrates. Chichester, England:John Wiley & Sons. p 251–262.

Elinson RP, Beckham Y. 2002. Develop-ment in frogs with large eggs and theorigin of amniotes. Zoology 105:105–117.

Elinson RP, Fang H. 1998. Secondary cov-erage of the yolk sac by the body wall inthe direct developing frog, Eleutherodac-tylus coqui: an unusual process for am-phibian embryos. Dev Genes Evol 208:457–466.

Elinson RP, del Pino EM, Townsend DS,Cuesta FC, Eichhorn P. 1990. A practicalguide to the developmental biology ofterrestrial-breeding frogs. Biol Bull 179:163–177.

Goin OB, Goin CJ. 1962. Amphibian eggsand the montane environment. Evolu-tion 16:364–371.

Harland RM. 1991. In situ hybridization:an improved whole-mount method forXenopus embryos. Methods Cell Biol 36:685–695.

Hennen S. 1973. Competence tests of earlyamphibian gastrula tissue containingnuclei of one species (Rana palustris)and cytoplasm of another (Rana pipiens).J Embryol Exp Morphol 29:529–238.

Hudson C, Clements D, Friday RV, Stott D,Woodland HR. 1997. Xsox17' and -( me-diate endoderm formation in Xenopus.Cell 91:397–405.

FROG NUTRITIONAL ENDODERM 1271

Ibrahim H, Winklbauer R. 2001. Mecha-nisms of mesendoderm internalization inthe Xenopus gastrula: lessons from theventral side. Dev Biol 240:108–122.

Ishizuya-Oka A. 1996. Apoptosis of larvalcells during amphibian metamorphosis.Microsc Res Tech 34:228–235.

Ishizuya-Oka A, Ueda S. 1996. Apoptosisand cell proliferation in the Xenopussmall intestine during metamorphosis.Cell Tissue Res 286:467–476.

Jennings DH, Hanken J. 1998. Mechanis-tic basis of life history evolution in anu-ran amphibians: thyroid gland develop-ment in the direct-developing frog,Eleutherodactylus coqui. Gen Comp En-docrinol 111:225–232.

Laurin M. 2005. Embryo retention, charac-ter optimization, and the origin of theextra-embryonic membranes of the am-niotic egg. J Nat Hist 39:3151–3161.

Laurin M, Reisz RR. 1997. A new perspec-tive on tetrapod phylogeny. In: MartinKLM, Sumida SS, editors. Amniote ori-gins: completing the transition to land.San Diego: Academic Press. p 9–59.

Lewis SL, Tam PPL. 2006. Definitiveendoderm of the mouse embryo: forma-tion, cell fates, and morphogenetic func-tion. Dev Dyn 235:2315–2329.

Masui Y. 2000. The elusive cytostatic fac-tor in the animal egg. Nat Rev Mol CellBiol 1:228–232.

Nath K, Boorech JL, Beckham YM, BurnsMM, Elinson RP. 2005. Status of RNAs,localized in Xenopus laevis oocytes, inthe frogs Rana pipiens and Eleuthero-dactylus coqui. J Exp Zool (Mol DevEvol) 304B:28–39.

Nieuwkoop PD, Faber J. 1956. Normal ta-ble of Xenopus laevis (Daudin). Amster-dam: North-Holland Publishing Co.243 p.

Ninomiya H, Zhang Q, Elinson RP. 2001.Mesoderm formation in Eleutherodacty-lus coqui: body patterning in a frog witha large egg. Dev Biol 236:109–123.

Packard MJ, Seymour RS. 1997. Evolutionof the amniote egg. In: Martin KLM, Su-mida SS, editors. Amniote origins: com-pleting the transition to land. San Diego:Academic Press. p 265–290.

Romer AS. 1957. Origin of the amniote egg.Scientific Monthly 85:57–63.

Rugh R. 1962. Experimental embryology.3rd ed. Minneapolis: Burgess PublishingCo. 501 p.

Sagata N, Watanabe N, Vande Woude GF,Ikawa Y. 1989. The c-mos proto-onco-gene product is a cytostatic factor re-sponsible for the meiotic arrest in verte-brate eggs. Nature 342:512–518.

Shi YB. 1994. Molecular biology of amphib-ian metamorphosis. Trends EndocrinolMetab 5:14–20.

Shi YB. 2000. Amphibian metamorphosis:from morphology to molecular biology.New York: Wiley-Liss. 288 p.

Shi YB, Brown DD. 1993. The earliestchanges in gene expression in tadpoleintestine induced by thyroid hormone.J Biol Chem. 268:20312–20317.

Shivdasani RA. 2002. Molecular regulationof vertebrate early endoderm develop-ment. Dev Biol 249:191–203.

Sinner D, Kirilenko P, Rankin S, Wei E,Howard L, Kofron M, Heasman J, Wood-land HR, Zorn AM. 2006. Global analysis

of the transcriptional network control-ling Xenopus endoderm formation. De-velopment 133:1955–1966.

Stainier DY. 2002. A glimpse into the mo-lecular entrails of endoderm formation.Genes Dev 16:893–907.

Townsend DS, Stewart MM. 1985. Directdevelopment in Eleutherodactylus coqui(Anura: Leptodactylidae); a staging ta-ble. Copeia 1985:423–436.

Wilkinson M, Richardson MK, Gower DJ,Oommen OV. 2002. Extended embryo re-tention, caecilian oviparity and amnioteorigins. J Nat Hist 36:2185–2198.

Winklbauer R, Schurfield M. 1999. Vegetalrotation, a new gastrulation movementinvolved in the internalization of the me-soderm and endoderm in Xenopus. De-velopment 126:3703–3713.

Xanthos JB, Kofron M, Wylie C, HeasmanJ. 2001. Maternal VegT is the initiatorof a molecular network specifyingendoderm in Xenopus laevis. Develop-ment 128:167–180.

Yasuo H, Lemaire P. 1999. A two-stepmodel for the fate determination of pre-sumptive endodermal blastomeres in Xe-nopus embryos. Curr Biol 9:8869–8879.

Yew N, Oskarsson M, Daar I, Blair DG,Vande Woude GF. 1991. mos gene trans-forming efficiencies correlate with oocytematuration and cytostatic factor activi-ties. Mol Cell Biol 11:604–610.

Yoshizaki N, Soga M, Ito Y, Mao KM, Sul-tana F, Yonezawa S. 2004. Two-step con-sumption of yolk granules during the de-velopment of quail embryos. Dev GrowthDiffer 46:229–238.

1272 BUCHHOLZ ET AL.