BMP antagonism by Noggin is required in presumptive notochord cellsfor mammalian foregut morphogenesis

Sarah R. Fausett a, Lisa J. Brunet b, John Klingensmith a,n

a Department of Cell Biology, Duke University Medical Center, Durham, NC, United Statesb Department of Molecular & Cell Biology, University of California Berkeley, Berkeley, CA, United States

a r t i c l e i n f o

Article history:Received 3 October 2013Received in revised form21 January 2014Accepted 10 February 2014Available online 12 March 2014

Keywords:BMP signalingNogginEA/TEFForegutNotochordMouse development

a b s t r a c t

Esophageal atresia with tracheoesophageal fistula (EA/TEF) is a serious human birth defect, in which theesophagus ends before reaching the stomach, and is aberrantly connected with the trachea. Severalmouse models of EA/TEF have recently demonstrated that proper dorsal/ventral (D/V) patterning of theprimitive anterior foregut endoderm is essential for correct compartmentalization of the trachea andesophagus. Here we elucidate the pathogenic mechanisms underlying the EA/TEF that occurs in micelacking the BMP antagonist Noggin, which display correct dorsal/ventral patterning. To clarify themechanism of this malformation, we use spatiotemporal manipulation of Noggin and BMP receptor 1Aconditional alleles during foregut development. Surprisingly, we find that the expression of Noggin inthe compartmentalizing endoderm is not required to generate distinct tracheal and esophageal tubes.Instead, we show that Noggin and BMP signaling attenuation are required in the early notochord tocorrectly resolve notochord cells from the dorsal foregut endoderm, which in turn, appears to be aprerequisite for foregut compartmentalization. Collectively, our findings support an emerging model fora mechanism underlying EA/TEF in which impaired notochord resolution from the early endodermcauses the foregut to be hypo-cellular just prior to the critical period of compartmentalization. Ourfurther characterizations suggest that Noggin may regulate a cell rearrangement process that involvesreciprocal E-cadherin and Zeb1 expression in the resolving notochord cells.

& 2014 Published by Elsevier Inc.

Introduction

The trachea and esophagus are parallel tubular organs thatconnect the pharynx with the lungs and stomach, respectively.Though they are ultimately distinct organs, they both arise from asingle tube of anterior foregut endoderm, which undergoes aseries of complex morphogenetic events over the course of itsdevelopment (reviewed in Fausett and Klingensmith, 2012). Dur-ing gastrulation in the mouse, endodermal and mesodermal cellsarise from the primitive streak and migrate anteriorly along theventral aspect of the embryo as its rostro-caudal axis lengthens(Tam et al., 2007). Endodermal cells form a single epithelial celllayer ventrally, while most of the mesodermal cells lie in betweenthe endoderm and the dorsal ectoderm. At the midline, justventral to the developing neural floorplate, a narrow band of cellslies within the endodermal epithelium and expresses Brachyury(Wilkinson et al., 1990). These cells will separate from theendoderm to form the notochord, an essential signaling center

and the defining feature of the Chordate lineage (Jurand, 1974;Cleaver and Krieg, 2001).

Remarkably little is known about the cellular basis of noto-chord separation from the endoderm layer in mammals; however,the descriptive histology by Jurand (1974) provides the currentmodel. During endoderm development, the midline cells appear toevaginate and resolve from the endoderm to become the noto-chord, a distinct tissue with its own basement membrane. Follow-ing notochord resolution, the embryonic endoderm folds ventrallyto form the primitive gut tube. Next, lung buds form on the ventralsurface of the anterior foregut endoderm (FGE), heralding thecommencement of foregut compartmentalization. Finally, in aprocess that is still poorly understood, the ventral endoderm justrostral to the lung buds gives rise to the trachea while the dorsalendoderm gives rise to the esophagus. The model most consistentwith experimental evidence suggests that the lateral sides of theendoderm meet across the lumen to divide the single foregut tubein half lengthwise, creating dorsal and ventral tubes, in a caudal torostral fashion. This apparent septation process occurs concur-rently with lengthening of the entire foregut (see Fig. 1A) (Fausettand Klingensmith, 2012).

Incorrect compartmentalization of the trachea and esophagus canresult in a defect known as esophageal atresia with tracheoesophageal

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/developmentalbiology

Developmental Biology

http://dx.doi.org/10.1016/j.ydbio.2014.02.0080012-1606/& 2014 Published by Elsevier Inc.

n Corresponding author. Tel.: þ1 919 684 9402.E-mail address: john.klingensmith@duke.edu (J. Klingensmith).

Developmental Biology 391 (2014) 111–124

fistula (EA/TEF), in which the pharynx communicates with the lungs,but not the stomach. EA/TEF occurs in humans at roughly 1 in 3000live births (Torfs et al., 1995). This is a serious birth defect, becausefood and water can enter the lungs, while air enters the stomach.Despite the health concern, and the growing number of mousemodelsof EA/TEF (Fausett and Klingensmith, 2012), much remains to bediscovered regarding the normal morphogenesis of the trachea andesophagus and what goes wrong to cause EA/TEF.

Gene expression patterns prior to foregut compartmentaliza-tion presage the distinct tissue derivatives of the ventral anddorsal foregut. At E10.5 in the mouse, the ventral FGE expressesearly respiratory-tract markers including Nkx2-1, while the dorsalFGE expresses Sox2 (Que et al., 2006, 2007). Mutations disruptingdorsal/ventral (D/V) patterning of the foregut cause failure ofcompartmentalization and yield a phenotype similar to EA/TEF(Fausett and Klingensmith, 2012). When genes and signalingpathways important for ventral specification are disrupted, theentire foregut becomes esophagus-like and fails to compartmen-talize (Que et al., 2007; Harris-Johnson et al., 2009; Goss et al.,2009). By contrast, as Sox2 levels are reduced, EA/TEF often occurs,with the single foregut tube resembling a trachea (Que et al.,2007). Noggin null mutants have a very similar defect (Que et al.,2006). NOG, a secreted BMP-antagonist, is expressed in thenotochord, floorplate, and dorsal FGE. Conversely, BMP are ligandsare expressed primarily in the ventral endoderm and mesoderm(Que et al., 2006). Recently, it was shown that ventral BMPsignaling directly represses Sox2 transcription (Domyan et al.,

2011). Thus, dorsal NOG may act to antagonize BMP signalingthere, allowing dorsal Sox2 expression and maintaining correctD/V patterning.

While it is now clear that incorrect D/V patterning by geneticmisregulation can cause defective foregut compartmentalization, otherdefects may lead to EA/TEF as well. One such case is in a teratogenicmodel developed in rats. By injecting pregnant rats or mice withAdriamycin, an anthracycline antibiotic, researchers can induce a highincidence of EA/TEF that closely resembles the most common form ofhuman EA/TEF (Thompson et al., 1978; Ioannides et al., 2002). Themolecular and cellular bases of the foregut phenotype in the Adria-mycin model remain unknown; however, the foregut does not appearto be significantly mis-patterned (Ioannides et al., 2010). Instead, EA/TEF consistently occurs along with notochords that are abnormallylarge, disorganized, and sporadically attached to the dorsal foregut(Possoegel et al., 1999). Nog null mice have an identical notochorddefect, and roughly 75% have EA/TEF (Que et al., 2006; Li et al., 2007).Many theories have been put forth to explain the coincidence of thenotochord and foregut defects in these models, but there is still noconclusive evidence to show that they are linked (Fausett andKlingensmith, 2012). In 2006 and 2007, two groups published theNog null foregut and notochord phenotypes, and both suggested thatthe notochord defect might cause EA/TEF (Que et al., 2006; Li et al.,2007). A key insight was the discovery by Li et al. (2007) that not allthe cells in the mutant notochord express the notochord marker,Brachyury (T). This led them to propose a model in which theT-negative cells were actually dorsal FGE cells that ended up in the

0.001

0.01

0.1

1

10 nog id1 msx1 msx2 sox2 nkx2-1 barx1 axin2 shh

Log

rela

tive

trans

crip

t qua

ntiti

y

DA

PI

N

KX

2-1

S

OX

2

BR

E-L

AC

Z

nc

fg

fp

Wildtype Nog-/-

nc

fg fg

fp

fg

Septation model of foregut compartmentalization

E10E12

es tr

*

Fig. 1. Dorsal–ventral patterning of the anterior foregut endoderm is not disrupted in Nog� /� mutants. (A) In the septation model of foregut compartmentalization, theundivided anterior foregut at embryonic day 10 (E10) displays dorsal/ventral patterning (pink/green, respectively). Over the course of two days, the lateral edges of theendoderm meet across the lumen and divide the foregut starting caudally and progressing rostrally. (B, C) Transverse E9.5 wildtype (B) and Nog� /� (C) anterior foregutsections labeled with a BMP-response element reporter, BRE-LacZ (blue) (scale¼50 μm). (D) Relative transcript quantities of genes involved in D/V patterning of the FGE.(E, F) Immunohistochemistry showing no difference in dorsal SOX2 or ventral NKX2-1 in Nog� /� embryos (F) and wildtype littermates (E) at E10.5 in transverse section(scale¼25 μm). Abbr: fp, floorplate; nc, notochord; fg, foregut; es, esophagus; tr, trachea.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124112

notochord at the expense of the dorsal FGE during notochordresolution. They reasoned that a dorsal FGE deficient in cells mightbe too small to yield a patent esophagus after compartmentalization.

Experimental data available to date can be reconciled witheither of these models for the occurrence EA/TEF in Nog mutants.It is conceivable that NOG has two separate roles, one in notochorddevelopment and one in active compartmentalization of the of theFGE. However, it is also possible that EA/TEF arises because of thenotochord defect (Li et al., 2007). In this study we set out furthertest both hypotheses. We additionally sought to clarify the cellularbehaviors that underlie notochord resolution from the foregutendoderm. Our genetic studies pinpoint the requirement for Nogin foregut compartmentalization to be in the notochord at thetime of notochord resolution. Additionally, we present quantitativedata suggesting that a role of NOG in this domain is to facilitate acellular rearrangement event involving transcriptional down-regulation of E-cadherin (Cdh1) in notochord cells.

Results

Dorsal–ventral patterning of the foregut endoderm is not disrupted inNoggin null mutants

Nearly all mouse models of EA/TEF display aberrant D/Vpatterning of the FGE prior to compartmentalization (Fausett andKlingensmith, 2012). Since ventral BMP signaling indirectly pro-motes tracheal fates (Domyan et al., 2011), we hypothesized that aloss of the BMP antagonist NOG would lead to an inappropriateexpansion of BMP signaling, thus perturbing D/V patterning in theearly foregut. Results by Li et al. (2007) indicate that BMP signalingoccurs at low levels if at all in the early dorsal foregut, and thatloss of Noggin causes no major change in BMP signaling levels inthis domain. We investigated this issue further with a variety ofBMP signaling assays, including the use of a BMP responseelement (BRE)-driven LacZ reporter that labels cells respondingto BMP signaling (Blank et al., 2008). Consistent with the results ofLi et al. (2007), we found that the BMP signaling reporter marks agreater proportion of the foregut in Nog null mutants relative towildtype, stage-matched littermates. However, the foregut is alsosmaller in these mutants (Fig. 1B, C); therefore, we cannotconclude that BMP signaling is expanded in its dorsal extent. Wenext examined transcript levels of several targets of BMP signaling(Id1, Msx1, Msx2), as well as other presumptive targets of the D/Vpatterning network (Sox2, Barx1, Nkx2-1, Axin2) (Miyazono andMiyazawa, 2002; Tribulo et al., 2003; Woo et al., 2011; Domyanet al., 2011; Que et al., 2007; Goss et al., 2009). While levels of Nogtranscript were virtually undetectable, we found no significantchanges in expression of BMP target genes or foregut patterninggenes (Fig. 1D). Finally, we used SOX2 and NKX2-1 antibodies tolabel the dorsal and ventral domains, respectively. While thecross-sectional circumference of the Nog null FGE varies alongthe rostro-caudal axis, comparison of wildtype and mutant foregutsections of similar circumference showed no D/V patterningdefects at E10.5 (Fig. 1E, F). Together, these data support theconclusion that loss of Nog has a negligible effect on D/V pattern-ing of the foregut prior to compartmentalization.

The Noggin mutant foregut endoderm is hypocellular, but not dueto increased cell death or decreased proliferation

In order to investigate the mechanism(s) behind the occurrenceof EA/TEF in Nog null mutants in the absence of a clear patterningdefect, we examined the physical attributes of the foregut prior tocompartmentalization. Previous work by Li et al. (2007) deter-mined that the foregut endoderm appears smaller in mutants than

in wildtype at E9.5. They suggested that the dorsal FGE is reducedrelative to the ventral domain, and that it may be insufficient toyield a patent esophagus. While we were able to confirm that theFGE in the trunk and neck regions contains fewer cells at E9.5, wefound no significant difference in the relative size of the dorsal vs.ventral domains (Fig. 2A, B). By contrast, the mutant FGE is notsignificantly hypocellular at E8.5. This suggests that the foregutbecomes hypocellular in mutants between E8.5 and E9.5 (Fig. 2C).One possibility is that the reduced foregut size results from eitherdecreased proliferation or increased apoptosis, especially giventhat BMP signaling is known to promote apoptosis or cell senes-cence in other contexts (Guha et al., 2002; Buckley et al., 2003). Liet al. (2007) concluded that apoptosis is unlikely to play asignificant role in foregut size at E9.0. We extended our analysisearlier to E8.5. We quantified the levels of proliferation andapoptosis in mutant and control foreguts at E8.5 using antibodiesagainst phosphohistone H3 and cleaved caspase-3, respectively.We found that there was no significant difference in apoptosis inNog null foreguts at E8.5 compared to wildtype, and there was asignificant increase rather than decrease in proliferation (Fig. 2D).We therefore conclude that the hypocellularity in Noggin mutantsis not caused by either decreased proliferation or increasedapoptosis, and that it must be the result of a different mechanism.

Noggin expression in the foregut endoderm is not required for properforegut compartmentalization

To isolate the potential roles for Nog in notochord and foregutdevelopment, we turned to a genetic approach. Since Nog isexpressed in the neural floorplate, notochord, and dorsal FGE justprior to and during compartmentalization (Fig. 3A, B), we testedwhich of these domains is required for foregut compartmentaliza-tion and/or notochord morphogenesis. To do this, we geneticallydissected the spatial and temporal requirements for Nog. We firstsought to ascertain whether Nog expression in the dorsal foregut isrequired for correct foregut compartmentalization. To ablate Nogin this domain we employed Cre-Lox technology using Foxg1Cre

(Hebert and McConnell, 2000) in conjunction with a conditional,“floxed” allele of Nog (NogFx) (Stafford et al., 2011). Though therecombination domain for this Cre driver is quite large, it does notinclude the floorplate or the notochord; it preserves Nog expres-sion in these axial domains while ablating it in the endoderm(Fig. 3A, B, and D). To confirm successful ablation of Nog from theendoderm, we performed quantitative PCR (qPCR) on cDNA fromisolated foreguts (Fig. 3G). While this genetic manipulation pro-duced several phenotypes found in Nog nulls (exencephaly andsmall telencephalons, data not shown), it did not produce anydiscernable notochord phenotype (n¼0/11) or foregut compart-mentalization defects (n¼0/12) (Fig. 3L). We conclude that thedorsal FGE domain of Nog expression is not necessary for correctmorphogenesis of the notochord or compartmentalization of theforegut.

BMP antagonism is required in the early notochord for propernotochord and foregut morphogenesis

Having ruled out the FGE as a necessary source of NOG, weconsidered the floorplate and notochord expression domains.Given that NOG is a secreted molecule, it seemed possible thatthese nearby domains could influence notochord development,foregut compartmentalization, or both. To ablate Nog from thesedomains, while leaving the dorsal FGE domain intact, we usedShhCre, as Shh itself and the ShhCre driver are expressed innotochord and floorplate but not in the dorsal FGE at relevantstages (Figs. 3A, C, and 4A–C) (Harfe et al., 2004). Quantitative PCRand in situ hybridization confirmed that ShhCre successfully ablates

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124 113

Noggin expression in the notochord (Fig. 4B, C), but leaves it intactin the foregut endoderm (Fig. 3G). We found that this cross (ShhCre;NogFx/�) produced notochord (n¼9/9) and foregut defects (n¼10/17) identical to those seen in the Nog null (Figs. 3K and 4D, E).These data reveal that axial midline expression of Nog is requiredboth for proper notochord morphogenesis and for compartmen-talization of the common foregut into esophagus and trachea.

To supplement this finding, we employed a Cre-inducibleconstitutively active Bmpr1a (CAB) transgene targeted to theRosa26 locus (Rodriguez et al., 2010). Driving this transgene withShhCre (ShhCre;CAB) allowed us to determine whether the role ofNOG from the notochord and floorplate domains is to attenuateBMP signaling only within those domains, or whether a broaderdomain also requires BMP antagonism by axial NOG (Figs. 3E, Fand 4F, G). If the role of NOG from the notochord and floorplatewere to attenuate BMP-signaling in the foregut endoderm, forexample, we would not expect this cross to produce EA/TEF.However, this experiment also produced notochord (n¼15/15)and foregut defects (n¼27/34) very similar to those of the Nog null(Figs. 3I and 4H, I). These data imply a critical role for BMPsignaling attenuation in the axial midline. However, because ShhCre

is also expressed in the ventral foregut endoderm, it may be thatover-activation of BMP signaling in this ventral domain couldimpact the phenotype. In order to test whether activation of BMPsignaling in the FGE could cause EA/TEF, we used Foxg1Cre toinduce BMP signaling in the foregut endoderm and mesenchyme.We found that Foxg1Cre;CAB mutants to not display EA/TEF (n¼5)

(Supplemental Fig. 1). Together, these data reveal that BMPsignaling must be attenuated specifically in the floorplate, noto-chord, or both to ensure proper formation of the notochord andthe compartmentalization of the foregut.

In order to distinguish Noggin action between its notochordand floorplate domains, we turned to a tamoxifen-inducible Credriven by Shh, ShhCreER (Harfe et al., 2004). We found that underour experimental conditions, tamoxifen induces Cre-mediatedDNA recombination within 24 h of administration, and no longerinduces recombination in new Cre expression domains after 48 h(Fig. 5A, C–F). Since Shh expression in the notochord begins at E7.5in the node, but floorplate expression does not begin until E8.5, wewere able to induce recombination specifically in the notochord byadministering one dose of tamoxifen 24 h prior, at E6.5 (Fig. 5C).Additionally, by comparing the phenotypes generated with pro-gressively later administration of tamoxifen (i.e. at E6.5, E7.5, E8.5,and E9.5) in ShhCreER;Nogfx/� mutants, we tested whether or not itis possible to uncouple the notochord and foregut defects byinducing recombination after notochord resolution and beforeforegut compartmentalization.

As expected, administering tamoxifen at E9.5, and thus causingrecombination at E10.5, produced neither notochord defects(n¼0/6) nor EA/TEF (n¼0/6) (Fig. 5B, F). Tamoxifen at E8.5 causedmild notochord defects in about half of the mutant embryos(n¼10/18) (Fig. 5B and E), while tamoxifen at E7.5 caused mildto severe notochord defects in 100% of mutant embryos (n¼16/16). Strikingly, neither administration at E8.5 (n¼0/18) nor at

0

5

10

15

20

25

30

dFGE vFGE

Ave

rage

# o

f cel

ls p

er c

ross

-se

ctio

n in

the

E9.5

ant

erio

r fo

regu

t WT

MT

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

Rat

io o

f the

ave

rage

# o

f cel

ls

in th

e dF

GE

to th

e vF

GE

WT MT

0

1

2

3

4

5

6

7

dying dividing

Perc

ent o

f cel

ls in

the

E8.5

an

terio

r for

egut

WT

MT

*

*

*

Ave

rage

# o

f cel

ls p

er c

ross

-se

ctio

n in

the

E8.5

ant

erio

r FG

E

WT

MT

Fig. 2. Neither increased apoptosis nor decreased proliferation can account for the hypo-cellularity of the anterior FGE of Nog� /� mutants at E9.5. (A) The average number ofcells per cross-section in E9.5 anterior FGE is significantly lower in Nog� /� mutants compared to wildtype littermates (po0.05). (B) The ratio of the number of dorsal toventral FGE cells (as defined by ShhCre expression) is not significantly lower in Nog� /� mutants compared to wildtype littermates. (C) The average number of cells per cross-section in the E8.5 anterior FGE is not significantly lower in Nog� /� mutants compared to wildtype littermates. (D) There is no significant difference in the level ofendodermal apoptosis at E8.5 in Nog� /� mutants compared to wildtype littermates, and the level of proliferation is significantly higher (po0.05). Abbr: dFGE, dorsal foregutendoderm; vFGE, ventral foregut endoderm.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124114

fp

nc

fg

ShhCre;CABWildtype

Foxg1Cre

nc nc

fg fg

ShhCre

0.005

0.05

0.5

Log

rela

tive nog

tran

scrip

t qua

ntity

Foxg1Cre;NogFx/-ShhCre;NogFx/-

**

NogLacZ/+

fp

nc

fgfg

nc

fp

tres

Wildtype ShhCre;CAB ShhCre;NogFx/-Nog-/- Foxg1Cre;NogFx/-

fp and nc domains targeted by ShhCre

FGE domain targeted by Foxg1Cre

estr

Nog-LacZ

BMP signaling

Fig. 3. Noggin expression from the notochord and/or floorplate is both necessary and sufficient to allow proper foregut compartmentalization. (A) Cartoon showing howShhCre and Foxg1Cre target the domains of Nog expression (blue). Tissue experiencing the highest levels of BMP signaling (as judged by BRE-LacZ) is represented in orange.(B) Noggin-LacZ expression at E10.5 (scale¼50 μm). (C, D) Recombination patterns of Shh Cre (C) and Foxg1Cre (D) reported by R26RLacZ (blue) or R26ReYFP (brown) at E10.5(scale¼50 μm). (E, F) BRELacZ-reporter activity in the E9.5 foreguts of Shh Cre;CAB mutants (F) and wildtype littermates (E) (scale¼50 μm). (G) Quantitative PCR showingreduced Nog expression in the dissected E10.5 foreguts of Foxg1 Cre;NogFx/� mutants, but not Shh Cre;NoxFx/� mutants relative to wildtype littermates; (H) wildtype E12.5foregut showing a complete esophagus and trachea; (I–K) esophageal atresia (arrowheads) and tracheoesophageal fistula (arrows) in ShhCre;CAB mutants (I), Nog nullmutants (J), and ShhCre;NogFx/� mutants (K). (L) Foxg1Cre;NogFx/� mutants do not have EA/TEF (scale¼200 μm). Brown in H–L¼anti-CDH1. Abbr: fp, floorplate; nc,notochord; fg, foregut; es, esophagus; tr, trachea.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124 115

E7.5 generated EA/TEF (0/16) (Fig. 5B, D). We also note thatadministration of tamoxifen at E8.5 did not cause EA/TEF inShhCreER;CAB mutants either (n¼14). This is further evidence thatactivation of BMP signaling in the vFGE does not contribute to EA/TEF. In contrast, we were able to induce EA/TEF (n¼4/16) alongwith the notochord defects (n¼16/16) by administering tamoxifenat E6.5 (Fig. 5B, C, G, and H).

These data, together with our previous results, suggest a strongconnection between BMP antagonism in the notochord during theperiod of notochord resolution and the subsequent success offoregut compartmentalization. Furthermore, by pooling data fromeach of our genetic experiments, we determined via statisticalanalysis that the presence of severe notochord defects (4300%the size of wildtype notochords) is a strong predictor of thepresence of EA/TEF (p¼0.0047; Fig. 5B). This relationship suggeststhat larger notochords are more likely to accompany EA/TEF thansmaller ones, and is consistent with a model in which largenotochords occur at the expense of the adequate endoderm anddispose the embryo towards EA/TEF.

Characterization of notochord resolution

Because our genetic experiments strongly indicated a linkbetween the notochord defects and EA/TEF in Nog conditionalmutants, we decided to look closely at notochord resolution andhow it is disrupted in Nog null mutants. In their model, Li et al.(2007) suggested that if the notochord cells were unable to success-fully resolve from the dorsal foregut endoderm, that part of the

endoderm would be incorporated into the notochord structure. Sucha model is consistent with their observations of a reduced foregutendoderm, with only some of the notochord cells in Nog nulls beingpositive for the notochord marker Brachyury (T). We wanted to buildon their observations by testing the hypothesis that the T-negativecells in the notochord expressed the dorsal endoderm marker, SOX2.Indeed, we detected SOX2 protein in cells of the Nog null notochordat E9.5 (Fig. 6A). This result strongly suggests that some of the cells inthe mutant notochordal tissue, as defined by the basement mem-brane surrounding it, were fated to be dorsal endoderm, rather thannotochord. These data support a mechanism in which faulty resolu-tion of the notochord from the endoderm results in some presump-tive endoderm cells being sequestered along with notochord cells.

To further refine our observations, we sought to define distinctstages of wildtype notochord resolution. We stained 6–21 somiteembryos using antibodies against E-cadherin and laminin tohighlight the endoderm itself and the local tissue boundaries,respectively. Examining sections through the anterior foregutregion, we identified at least 6 distinct stages of basementmembrane morphology (Fig. 6B). These stages progress in a rostralto caudal fashion along the trunk of the embryo. During Stage 1,the basement membrane appears to interpose between thenotochord cells and endoderm cells, and by Stage 2, the basementmembrane adopts an angular arrangement around the dorsal andlateral aspects of the notochord cells. Stage 3 is characterized bythe rounding of the notochord cells, down-regulation of CDH1, anda basement membrane that begins to surround the ventral aspectof the notochord (also see Fig. 7C). At Stage 4, the basementmembrane completely separates the notochord from the dorsal

Shh

Cre

;Nog

Fx

Shh

Cre

;CA

B

WT BRE-LacZ MT BRE-LacZ

WT nc MT nc

fg

ncfp

fp

nc

fgfg

nc

fp

fgfg

nc

fp

fg

nc

fp

DAPI Brachyury Laminin

eYFP

WT Nog in situ MT Nog in situ

nc

ncnc

fp

fpnc

fg

fp fp

fg fg

E8.5 E9.5

A

H

Fig. 4. Axial domains require BMP antagonism for proper notochord development. (A) Shh Cre recombination at E8.5, reported by R26ReYFP (green). (B, C) In situ hybridizationshowing absence of Nog transcripts (purple) from the notochord at E8.5 in mutants (C) but not wildtype littermates (B). (D, E) Large notochord in which not all cells expressBrachyury seen in ShhCre;NogFx/� embryos (E) but not wildtype littermates at E9.5 (D). (F,G) BRE reporter activity (blue) showing increased BRE-LacZ expression in thenotochords of ShhCre;CAB embryos (G) but not wildtype littermates at E8.5 (F). (H, I) Large notochord in which not all cells express Brachyury seen in ShhCre;CAB embryos(I) but not wildtype littermates (H) at E9.5. In (D, E, H, I) laminin (red) marks the basement membranes and DAPI (white) marks the nuclei. Arrows indicate cells in thenotochord that do not express Brachyury (green). Arrowhead indicates example of background fluorescence. Abbr: fp, floorplate; nc, notochord; fg, foregut. Scale¼25 μm.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124116

endoderm, and during Stage 5, the notochord moves away fromthe endoderm along with the floorplate of the neural tube. AsStage 5 progresses, a trail of basement membrane material can beseen linking the notochord with the endoderm, but by Stage 6 it

has been degraded and the notochord is situated tightly againstthe floorplate.

Lastly, we wanted to confirm the published expressionpatterns of Nog and various Bmp genes (McMahon et al., 1998;

ShhCreER;

R26RLacZ

nc

nc

fp

tr

es

tam. E6.5 tam. 7.5 tam. 8.5 tam. 9.5

ShhCreER;R26RLacZ , harvest E12.5ShhCreER;NogFx/- , havest

E12.5 WT MT

Ros

tral

to E

ALe

vel o

f EA

es

tr

100% 200% 300% 400% 500%

Nog-/-

ShhCre;CAB

Tamox. E6.5

Tamox. E7.5

Tamox. E8.5

Tamox. E9.5

ShhCreER;NogFx/-

likelihood ratio test: p=0.0047

EA

/TE

Fno

EA

/TE

F

Notochord size as a percent of wildtype

fp fpfp

nc ncnc

es es es

tr tr

tr

tr

tr

es

es

Fig. 5. Large notochords in Nog� /� mutants are a good predictor of EA/TEF, and Noggin is required early in the notochord for proper foregut compartmentalization.(A) Strong R26RLacZ reporter activity (blue) in the notochord just 24 h after tamoxifen administration at E7.5 (scale¼20 μm). (B) Chart showing distribution of notochord sizesof individuals in various mutant classes against the presence of EA/TEF. Likelihood ratio test reveals that severe notochord defects are an excellent predictor of EA/TEF(p¼0.0047). (C, C0) Recombination pattern of ShhCreER at E12.5 in the notochord but not the floorplate or endoderm after a dose of tamoxifen at E6.5. (D, D0) Recombinationpattern of ShhCreER at E12.5 in the notochord and floorplate but not the endoderm after a dose of tamoxifen at E7.5. (E, E0) Recombination pattern of ShhCreER at E12.5 in thenotochord, floorplate, and ventral foregut endoderm after a dose of tamoxifen at E8.5. (F, F0) Recombination pattern of ShhCreER at E12.5 in the notochord, floorplate, andventral foregut endoderm after a dose of tamoxifen at E9.5. Blue color reflects R26RLacZ reporter activity (scale¼40 μm). (G, H) Transverse sections through the foregut atE12.5 showing esophageal atresia in ShhCreER;NogFx/� embryos that were administered tamoxifen at E6.5 (H), but not wildtype littermates (G) (scale¼40 μm). Abbr: fp,floorplate; nc, notochord; es, esophagus; tr, trachea.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124 117

Danesh et al., 2009; Arkell and Beddington, 1997; Solloway andRobertson, 1999). Importantly, genetic reduction of either Bmp4(Que et al., 2006) or Bmp7 (Li et al., 2007) suppresses the Nog nullEA/TEF phenotype. Previous work has shown that both Bmp4 andBmp7 are expressed between E8.0 and E8.5 in relevant domains.Bmp4 is expressed mainly in the ventral endoderm and mesodermduring notochord resolution (Danesh et al., 2009), with muchlower levels more dorsally. However, Bmp7 expression has beenreported in the FGE and notochordal plate (Arkell and Beddington,1997; Solloway and Robertson, 1999; Danesh et al., 2009). Welooked carefully at BMP-signaling, Nog expression, Bmp7

expression, and Brachyury expression along the length of 7-somite wildtype embryos. Brachyury is expressed in the primitivestreak as cells are undergoing EMT to accomplish gastrulation(Herrmann, 1991). It is also expressed in the node, notochordalplate, and notochord more rostrally along the midline (Fig. 6CI).Along the midline and throughout notochord resolution, Brachy-ury appears to be expressed at a roughly constant level. However,as the axial midline cells migrate rostrally from the node (Tamet al., 2007), there is a striking dynamic quality to Nog expression(Fig. 6CII). Most caudally, at these stages, Nog is expressed in thenode. Between the node and the region at which notochord

Stage 2

Stage 4 Stage 5

Stage 1

nc

Noggin in situ, 7s WT

DAPI LAMININ CDH-1

Stage 6

Stage 3

Brachyury in situ, 7s WT Bmp7 in situ, 7s WT BRE LacZ, 7s WT

DAPI LAMININ

Wildtype Nog-/-

SOX2

Brachyury

IIa IIb

IIc IId

IIe IIf

IIg IIh

IIi IIj

II

Ia Ib

Ic Id

Ie If

Ig Ih

Ii Ij

IIIa IIIb

IIIc IIId

IIIe IIIf

IIIg IIIh

IIIi IIIj

IVa IVb

IVc IVd

IVe IVf

IVg IVh

IVi IVj

IVIIII a ce g

ia ce g

ia c e g

i

a c e g

i

nc nc

nc

ncncnc

nc

nc

nc

fp

fpfp

fp

fpfp

fp

Fig. 6. Detailed characterization of notochord resolution. (A) Immunohistochemistry on serial sections showing Brachyury-negative and SOX2-positive cells in the notochordof Nog� /� mutants (scale¼20 μm). (B) 3D-rendered confocal sections of Stages 1–6 of notochord resolution (scale¼20 μm). CI–CIII) In situ hybridization (purple) showingNog, Brachyury, and Bmp7 expression in selected transverse sections (a–j) from the primitive streak to the anterior foregut in 7-somite wildtype embryos. (CIV) BRE-LacZreporter activity (blue) in selected transverse sections (a–j) from the primitive streak to the anterior foregut in 7-somite embryos. Abbr: fp, floorplate; nc, notochord; fg,foregut. Section scale bars in panel C are 20 μm. Whole mount scale bars in panel C are 500 μm.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124118

resolution is beginning, Nog transcript is nearly undetectable byin situ (Fig. 6CII:e–g). Nog is once again expressed in cells under-going resolution from the endoderm (Fig. 6CII:h–j). Interestingly,Bmp7's expression pattern is nearly identical to that of Nog

(Fig. 6CIII). BMP antagonism in the notochord by NOG, andperhaps by other BMP antagonists, is suggested by the low levelsof BMP signaling seen in the resolving notochord at later stages ofresolution (Fig. 6CIV:j). Together, these data prompted us tohypothesize that Nog and Bmp7 are part of a dynamic regulatorysystem that governs the changing behaviors of midline cells asthey migrate from the node to the trunk region and begin toresolve from the endoderm.

Noggin null notochords maintain epithelial characteristics

Previously published work, shows that CDH1 (E-cadherin) isnot expressed in the notochord after resolution (Nose andTakeichi, 1986). Since the notochord precursors are initially con-tiguous with the endodermal epithelium, we suspected thattimely loss of CDH1 from the notochord precursors would beimportant for notochord resolution from the endodermal epithe-lium. Furthermore, while the dorsal endoderm is a typical epithe-lium with both apical/basal polarity and a lumen, the notochorddoes not appear to have a lumen. We suspected that the notochordcells also lack markers of apical polarity. If such a cell rearrange-ment event were disrupted in Nog null notochords, it could causethe notochord cells to remain adhered to the endoderm cells andhelp explain the large notochord phenotype. We sought todetermine when the down-regulation of CDH1 occurs, whetherthis down-regulation is disrupted in Nog nulls, and whether otherepithelial characteristics are inappropriately maintained in Nognull notochords.

To evaluate epithelial characteristics, we observed corticalapical F-actin using phalloidin, and atypical PKCzeta (PRKCZ) byimmunofluoresence. While in wildtype notochords, there was noobvious apical domain and very little F-actin or PRKCZ, the apicaldomain in null mutants was sporadically persistent (n¼8). Strongphalloidin and anti-PRKCZ staining was observed in multiplesections along the rostro-caudal axis of the trunk, at what wasonce the apical domain of these cells (Fig. 7A).

Immunohistochemistry against CDH1 and CDH2 revealed thatwildtype notochord cells down-regulate CDH1 and up-regulateCDH2 during notochord resolution (Fig. 7B, C). While levels ofCDH-2 do not appear to be altered in Nog null mutants ascompared to wildtype littermates (Fig. 7B), CDH-1 protein wasseen to persist during Stages 3 and 4 of notochord resolution(Figs. 7C and 8A–C). To quantify the persistence of CDH1 proteinaccurately, wildtype and mutant littermates were processed andembedded together in a single block such that they were sectionedand stained side by side on the same slides. They were thenexamined by confocal microscopy using the same settings toquantify the integrated density of fluorescence. As a control wequantified DAPI (a nuclear stain) and found no significant differ-ence in fluorescence between wildtype and mutant notochords(Fig. 8A–C). At Stage 4, when the basement membrane normallyseparates the endoderm from the notochord, CDH1 fluorescencewas significantly greater in Nog nulls (Fig. 8A–C). This suggeststhat CDH1 is not down-regulated appropriately in the formingnotochord of Nog mutants. Furthermore, we were able to detect

DAPI CDH1 LAMININ

DAPI SOX9

F-actin

ncnc

nc nc

DAPI CDH2 SOX9

PRKCZ

Nog+/- Nog-/-

ncnc

Nog+/-

Nog+/-

Nog-/-

Nog-/-

Fig. 7. Epithelial characteristics persist abnormally in Nog� /� notochords afterresolution. (A) Phalloidin stain (green, upper) or anti-PRKCZ staining (green, lower)showing persistence of cortical filamentous actin and apical PRKCZ in thenotochords of Nog� /� mutants compared to wildtype littermates. SOX9 (pink)marks notochord cells (scale¼15 μm). (B) Immunohistochemistry showing thatCDH2 (red) is up-regulated in Nog� /� notochords and those of their wildtypelittermates (scale¼15 μm) DAPI (white) marks nuclei. Anti-laminin (green) marksthe basement membrane. (C) Immunohistochemistry showing that CDH1 (red)persists in Nog� /� notochords past the Stage 3 when it is down-regulated inwildtype littermates (scale¼15 μm). DAPI (white) marks nuclei. Anti-SOX9 (green)marks notochord. Abbr: nc, notochord.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124 119

persistence of Cdh1 transcript in the resolving notochord of Nogmutant embryos relative to stage-matched wildtype littermates(Fig. 8D, E).

Taken together, these data demonstrate a persistence ofepithelial characteristics in Nog mutant notochords relative tostage-matched wildtype embryos in the foregut region, perhapsindicating that disruption of epithelial characteristics must occurefficiently for successful notochord resolution

Bmp signaling may regulate Cdh1 levels during notochord formationvia Zeb1

The data presented above support a model in which normalnotochord formation involves a partial loss of epithelial integritythat allows the resolution of the notochord cells from the dorsalforegut epithelium. Since such events usually involve the expres-sion of Cdh1 repressors (Yang and Wineberg, 2008), we looked foraltered expression of known Cdh1 repressors in the resolvingnotochord of Nog� /� mutants. In situ hybridization revealed thatsnai1, snai2, and twist1 are not significantly expressed in thenotochord during resolution (Supplemental Fig. 2). The transcrip-tional repressor Zeb1 is reportedly expressed in the notochord atlow levels (Takagi et al., 1998), with notochordal expression alsorevealed by a LacZ knock-in allele of Zeb1 (Miyoshi et al., 2006).We confirmed this expression via in situ hybridization, anddetermined that while Zeb1 is expressed at low levels in wildtypenotochords during notochord resolution, Zeb1 mRNA is nearlyabsent in Nog� /� notochords (Supplemental Fig. 3A and B). We

further assayed this expression using an antibody against ZEB1protein, which also revealed low levels in wildtype notochord butnone detectable in Nog mutant notochords (Fig. 8F and G).

Work in kidney cell lines has shown that ZEB1 is a transcrip-tional repressor not only of Chd1 but also of a primary transcriptthat encodes the microRNAs miR-200a, -200b, and -429 (Eger et al.,2005). These microRNAs, in turn, target Zeb1 transcripts fordegradation (Park et al., 2008; Gregory et al., 2008). This negativefeedback loop appears to regulate Cdh1 expression and EMT/METdownstream of BMP7 and TGF-beta (Bracken et al., 2008;Samavarchi-Tehrani et al., 2010). We hypothesized that if thissystem were active during notochord resolution, given thedecrease in Zeb1, we would see persistent miR-200a/200b/429expression in Nog� /� notochords during resolution. Indeed, wedetected miR-200 primary transcripts in the endoderm, but not inthe notochords of wildtype embryos. In Nog mutants, miR-200continued to be expressed in the notochords after resolution(Supplemental Fig. 3C and D). Together, these data suggest amodel in which the ZEB1 Cdh1-regulatory system, downstreamof BMP signaling, participates in the transcriptional regulation ofCdh1 during resolution of the notochord from the dorsal foregutendoderm.

Discussion

Several mouse models of EA/TEF have demonstrated thatcorrect dorsal/ventral patterning of the anterior foregut endoderm

0

0.5

1

1.5

2

2.5

DAPI α-CDH1

Inte

grat

ed D

ensi

ty o

f Fl

uore

scen

ce N

orm

aliz

ed to

W

T C

ontr

ols

WT MT

**

Cdh1

nc nc

Nog+/- Nog-/-

ZEB1

nc nc

nt nt

Nog+/- Nog-/-

Nog-/-Nog+/-

DAPI CDH1 LAMININ

ncnc

Fig. 8. The ZEB1 Cdh1-regulatory system is altered in Nog� /� mutants during notochord resolution. (A, B) Representative images of anti-CDH1 fluorescence used to quantifyCDH1 in wildtype (A) and Noggin null (B) notochords at Stage 4. (C) Quantification of the integrated density of fluorescence of DAPI and CDH1 in Nog null mutants andwildtype littermates that were processed and imaged under identical conditions, showing increased CDH1 in mutant notochords at Stage 4 (po0.05). (D, E) In situhybridization showing persistence of Cdh1 transcript in Nog� /� notochords (E) compared to those of wildtype littermates (D). (F, G) Immunohistochemistry showing thatZEB1 expression is reduced in the neural tube and notochord of Nog null mutants (G) as compared to wildtype littermates (F). Dashed lines show the boundaries of theneural tube and notochord tissues (scale¼15 μm). Abbr: nc, notochord.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124120

is essential for correct compartmentalization of the trachea andesophagus. Using spatiotemporal genetic dissection of the require-ments for the various domains of Nog expression, we find thatNoggin is dispensable in the endoderm itself, but is required in theearly notochord. We further show that this action of Nogginresults in an essential attenuation of BMP signaling in the resol-ving notochord. Our conclusions lend further support to the viewthat the EA/TEF defects in Noggin arise as a consequence of toofew cells in the endoderm for compartmentalization (Li et al.,2007). Finally, we present mechanistic evidence suggesting thatBMP antagonism by notochordal NOG allows a cell rearrangementevent involving Cdh1 down-regulation in the notochord cells asthey resolve from the foregut endoderm.

Loss of ventral BMP signaling or other ventral patterning genes/pathways results in loss of the ventral domain of the foregutendoderm, and consequently, failure of compartmentalization(Harris-Johnson et al., 2009; Goss et al., 2009; Minoo et al.,1999; Li et al., 2008; Domyan et al., 2011). Conversely, reducedlevels of SOX2 from the dorsal domain appear to ventralize theendoderm and induce EA/TEF (Que et al., 2007). Furthermore,these genes and pathways antagonize one another such that SOX2expression is low in the ventral domain and NKX2-1 is absentfrom the dorsal domain (Que et al., 2007; Domyan et al., 2011).Given that Nog is expressed dorsally, it was possible that loss ofNog could induce ventralization. Our data, however, support thesurprising conclusion that dorsal/ventral patterning in Nog� /�

mutants is not disrupted. The result is even more strikingconsidering that in nearly every model of EA/TEF in which D/Vpatterning has been assessed, the phenotype has been attributedto mis-patterning. In light of this finding, we turned to thealternative hypothesis that the smaller foregut endoderm preventssuccessful compartmentalization (Li et al., 2007).

Li et al. (2007) observed that the Nog� /� dorsal FGE appearedreduced compared to the ventral domain. However, our quantifi-cation revealed a smaller foregut endoderm overall. The discre-pancy may be dependent on the specific rostro-caudal regionobserved. Our quantification included the entire region of theforegut thought to give rise to the trachea and esophagus, from thelung buds to the fourth pharyngeal arch at E9.5. Importantly, wefound no difference in the size of the foregut endoderm at E8.5,indicating that the deficiency arises between E8.5 and E9.5. Fewpossibilities exist that could account for such tissue loss. First,rates of cell proliferation or apoptosis could be altered. Second,endoderm cells could be lost to the notochord during resolution.We directly addressed both possibilities. We found no significantchange in the amount of apoptosis at E8.5, but rather, significantlymore cell proliferation. The increase in proliferation was puzzling,but could have been a response to tissue loss during this timeperiod. To address the second possibility we looked for SOX2-expressing cells in the Nog� /� notochord, and found that they doindeed exist, though SOX2 is never seen in wildtype notochords.We therefore conclude that the endoderm is reduced due toimpaired notochord resolution.

The results of our genetic experiments also support a strongconnection between the notochord defect and EA/TEF in Nog� /�

mutants. While it was possible that Nog expressed in and aroundthe foregut during foregut compartmentalization was importantfor that process, we were surprised to find that these spatio-temporal domains of Nog expression were completely dispensablefor compartmentalization. Additionally, we discovered using theconstitutively active BMPR1A (CAB) in conjunction with ShhCre,Foxg1Cre, and ShhCreER that it is the notochord cells themselves thatrequire BMP antagonism during notochord resolution in order forforegut compartmentalization to occur correctly. Despite ourefforts, we were not able to uncouple the notochord and foregutdefects by removing NOG or activating BMP signaling after

notochord resolution; we succeeded in inducing EA/TEF only byremoving NOG or activating BMP signaling in the notochord priorto resolution. Furthermore, we observed that the most severedefects (largest notochords) seemed to correlate with the presenceof EA/TEF, and statistically showed that a large notochord is anexcellent predictor of EA/TEF.

Noggin is known to be expressed in the early notochord(McMahon et al., 1998), but detailed analysis has not beenreported. Bmp4 and Bmp7 are also known to be expressed duringnotochord resolution. However, while Bmp7 expression has beenreported in the notochordal plate, Bmp4 expression has only beenreported in the endoderm and mesoderm (Danesh et al., 2009;Arkell and Beddington, 1997; Solloway and Robertson, 1999).Therefore, we looked carefully at BMP signaling and the expres-sion of Nog and Bmp7 during migration of the midline cells fromthe primitive streak and resolution of those cells from theendoderm. Two striking phenomena were observed. First, Nogexpression is dynamic in the midline cells, and correlates withactive notochord morphogenesis. Second, Bmp7 is expressed in anearly identical fashion as Nog in and around the notochord. Theseexpression patterns suggest that BMP signaling and BMP antagon-ism act in balance to dynamically regulate the morphogenesis ofthe notochord at this stage. Importantly, reducing the genedosages of Bmp4 or Bmp7 can rescue the Nog null notochorddefects as well as EA/TEF (Que et al., 2006; Li et al., 2007). It wouldbe interesting to test whether removing Bmp7 after notochordresolution can still rescue EA/TEF in Noggin null mutants using aninducible Cre and a floxed allele of Bmp7.

Since the process of resolution must involve the notochord cellsdetaching from their endoderm neighbors, and since BMP7 isknown to promote epithelial characteristics in several contexts, wehypothesized that NOG was allowing the notochord cells to losesome of their epithelial characteristics in order to resolve from theendoderm. Indeed, we found that the epithelial markers apicalF-actin, PRKCZ, and CDH1 are absent from the notochord cells afterresolution. However, these markers persist in Nog� /� notochordsat least until E9.5. Furthermore, careful quantification of anti-CDH1 fluorescence confirmed significantly higher levels of CDH1in Stage 4 Nog� /� notochords compared to wildtype. Together,these data support a model in which BMP antagonism in thenotochord by NOG is important for reducing epithelial character-istics that prevent proper notochord resolution. In the future, itwould be very interesting to develop mouse lines that would allowgenetic reduction of epithelial characteristics in the notochords ofNog� /� mutants to rescue these phenotypes. Conversely, geneti-cally enhancing epithelial characteristics in wildtype notochordsmay induce both large notochords and EA/TEF. We note that CDH1was also reported to be expressed in the notochord of embryos inthe Adriamycin model for EA/TEF (Hajduk et al., 2012). It remainsunclear whether there is a molecular connection between theeffects of Adriamycin in embryonic organogenesis and theincreased BMP signaling resulting from loss of Noggin. It seemslikely that the two molecular perturbations converge on somecommon molecular effectors of notochord development.

In order to discover factors that may be downstream of BMPantagonism and upstream of Cdh1, we chose to look at theexpression of known Cdh1 repressors. While we found that snai1,snai2, and twist1 are not expressed at the right time or place, Zeb1 isboth expressed in wildtype notochords and abnormally reduced inNog� /� notochords. ZEB1 is a transcriptional repressor of Cdh1 aswell as of the miR-200a/200b/429 transcript (Bracken et al., 2008).The mir-200a/200b/429 transcript, in turn, generates mature micro-RNAs that target Zeb1 transcripts for degradation. Thus, ZEB1 andthe miR-200-family microRNAs negatively regulate one another, andexpression of one over the other acts as a switch that is capable ofturning Cdh1 expression on or off (Bracken et al., 2008).

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124 121

Furthermore, this paradigm is known to be downstream of BMP7 iniPSCs and a mouse model of renal injury, such that BMP7 drivesCdh1 expression and other epithelial characteristics (Samavarchi-Tehrani et al., 2010; Ziesberg et al., 2003). Our expression datastrongly reflect the known roles of these genes. Cdh1 and the miR-200a/200b/429 transcripts are expressed highly in the endodermalepithelium, and are absent from the notochord cells during and afterresolution as other epithelial characteristics are lost. Conversely,Zeb1 is not highly expressed in the endoderm, but turns on in thenotochord during resolution. In Nog� /� mutants this arrangement isdisrupted and both Cdh1 and the mir-200a/200b/429 transcript areexpressed abnormally in the notochord cells while Zeb1 is absent. Inlight of these results, we hypothesized that Zeb1 null mutants mighthave also notochord defects. We examined histological sections ofE10.5 Zeb1 null embryos provided by Dr. Yujiro Higashi and Dr.Przemko Tylzanowski (Takagi et al., 1998). In the limited number ofsamples examined (n¼5) only one embryo had an abnormally largenotochord (data not shown). The relative lack of notochord defectswe observed in Zeb1 null mice could be due to either functionalredundancies, or to incomplete penetrance. In any case, it would bevery interesting to look more closely at the notochord in Zeb1mutants and in Zeb1/Zeb2 double knockouts, as well as to developnew genetic tools that target the miR-200 family. Of note, Zeb1expression is also altered in ventral neural tube in Nog� /� mutants.It is unclear whether this results in any functional or physicaldefects, and it would be interesting to test whether levels of Cdh1and Cdh2 are altered in the floorplate of Nog nulls.

In conclusion, this study has revealed that EA/TEF can arise by amechanism other than that of failed D/V patterning of the foregutendoderm. We have presented evidence that supports a model,previously put forth by Li et al. (2007), in which faulty notochordresolution causes foregut hypocellularity prior to compartmenta-lization, and thus causes EA/TEF. We have established that noto-chord resolution from the foregut endoderm involves geneexpression changes that reflect a loss of epithelial characteristicsin presumptive notochord cells, perhaps driven by the ZEB/miR-200 regulatory system downstream of BMP7. The development ofnew genetic tools should allow further exploration into themechanistic details of notochord resolution and its impact oncompartmentalization of anterior foregut.

Methods

Generation of mutant embryos and genotyping

Except for lines used with Foxg1Cre, which were backcrossed 5–6 generations to 129/SvEv, all mice were bred on undefinedoutbred backgrounds. All genotyping was performed on dissectedtissues (tails, limbs, or extraembryonic membranes) according topublished protocols (except NogDel, see below) and wildtypelittermates were used as controls (Supplemental Table 1).

Previously published alleles used in this study: BRE-Hspa1a-lacz reporter strain (BRE-LacZ) (Blank et al., 2008), Gt(ROSA)26Sortm1Sor (R26R-LacZ) (Soriano, 1999), Gt(ROSA)26Sortm1(EYFP)Cos

(R26R-eYFP) (Srinivas et al., 2001), Shhtm1(EGFP/cre)Cjt (ShhCre)(Harfe et al., 2004), Shhtm2(cre/ERT2)Cjt(ShhCreER) (Harfe et al., 2004),Foxg1tm1(cre)Skm (Foxg1Cre) (Hebert and McConnell, 2000),Nogtm1.1Rmh(NogFx) (Stafford et al., 2011), NogLacZ (McMahon et al.,1998), Rosa26CAG-loxpstoploxp-caBmprIa (CAB) (Rodriguez et al., 2010),Zeb1tm2Yhi (Takagi et al., 1998).

The null allele of Noggin used in this study (NogDel) wasgenerated in our colony by recombining the NogFx allele inthe germline using Alpltm1(cre)Nagy (Lomeli et al., 2000). Excisionof the single exon of the Nog gene was confirmed using primers

recognizing the 50 and 30 regions of the targeting constructflanking loxP sites (Supplemental Table 1).

To induce cre-mediated recombination using the ShhCreER allele,pregnant dams were administered tamoxifen (Sigmas) dissolvedin corn oil by oral gavage at doses between 0.08 and 0.1 mg/g bodyweight.

Immunofluorescent and immunohistochemical staining

With the exception of the ZEB1 antibody, which was applied tounfixed tissue, embryos were fixed 30 min to 1 h at roomtemperature in 4% paraformaldehyde. For sectioning, embryoswere processed through a sucrose gradient and embedded in 1:315% Sucrose:OCT (Tissue-Teks). Tissue was sectioned at 6–10 μmand then stained using standard protocols.

Primary antibodies used: goat anti-SOX2 (Neuromics), mouseanti-NKX2-1 (ThermoScientific), rabbit anti-GFP (Abcam), mouseanti-phosphohistone H3 (Cell Signaling Technologies), rabbit anti-cleaved caspase 3 (Cell Signaling Technologies), rat anti-CDH1(Invitrogen), rabbit anti-laminin (Lightner et al., 1989), goat anti-Brachyury (Santa Cruz), rabbit anti-PRKCZ (Santa Cruz), anti-ZEB1(E-20:sc-10572, Santa Cruz), rat anti-CDH2 (MNCD2, Developmen-tal Studies Hybridoma Bank), goat anti-ZEB1(E-20) (sc-10572,Santa Cruz). Filamentous actin was visualized using AlexaFluors-488 conjugated phalloidin at 1:200. Nuclei were visualizedusing DAPI.

For colorimetric visualization of antibodies, an HRP-conjugatedsecondary was used and subsequently detected with a solution of3,30-diaminobenzidine and nickel ammonium sulfate.

RNA isolation and quantitative PCR

Foreguts (between the 4th pharyngeal pouch and lung buds)were dissected from E10.5 Noggin nulls and wildtype littermates,and stored at �80 1C. Two to five foreguts were pooled for eachbiological replicate. RNA was extracted using a standard TRIzols

protocol. Reverse transcription was carried out using the iScriptcDNA synthesis kit from BioRad. Quantitative PCR was performedusing SYBRGreen (Applied Biosystems) and an Applied BiosystemsStepOnePlus Real Time RT-PCR system. Primers are listed inSupplemental Table 1.

H&E and Xgal staining

H&E staining and Xgal staining were conducted using standardprotocols.

Cell counting assays and statistics

Serial sections (6 μm) of E8.5 or 9.5 foreguts were stained withanti-cleaved caspase 3, anti-phosphohistone H3, and DAPI. Stainedslides were imaged and images were analyzed using ImageJ.Subsets of labeled cells in the foregut are reported as a percentof the total cells counted in the foregut between the 4th phar-yngeal pouch and division of the lung bud at E9.5 or between the2nd pharyngeal pouch and the anterior intestinal portal at E8.5.Values are reported as mean7s.e.m and were compared using aStudent's T-test. Differences were considered significant atpr0.05.

In situ hybridization and cloning of Zeb1 and miR-200 in situ probes

A 947 bp fragment of exon 6 from the Zeb1 mRNA was isolatedfrom E8.5 ICR (Harlan) cDNA. Similarly, a 3 kb fragment spanningmost of the miR-200b/mir-200a/mir-429 unprocessed transcriptwas isolated from E8.5 ICR (Harlan) cDNA. These fragments were

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124122

each cloned into a pGEMs-T easy vector from Promega. All otherconstructs used to generate riboprobes were previously published:Bmp7 (Ozkaynak et al., 1991), Brachyury (Herrmann, 1991), Noggin(McMahon et al., 1998), Cdh1 (Xu et al., 2002).

To detect transcripts in situ, embryos were fixed overnight at4 1C in 4% PFA then dehydrated to 100% methanol for storage.Hybridization was performed as previously published (Belo et al.,1997), except that the hybridization buffer contained 5% dextransulfate. After whole-mount in situ hybridization, embryos wereembedded for frozen sectioning.

Notochord comparative measurements and statistics

Serial sections of various mutant classes at E12.5 were used fornotochord measurement. The cross-sectional area of the noto-chord, in each section from the rostral limit of the esophagus tothe branching of the main bronchi, was measured using ImageJ.The average cross-sectional area for each embryo was compared tothat of a wildtype littermate that had been processed identically,and reported as a percent of wildtype.

A likelihood ratio test was performed using JMP8s.

Microscopy and CDH1 quantification with statistics

All microscopy and imaging was carried out either on a Zeiss 710inverted confocal, a Zeiss AxioPlan 2, or a Leica MZ12 stereoscope.

For quantification of anti-CDH1 fluorescence, mutant embryoswere frozen embedded alongside a wildtype littermate so thatthey were sectioned, stained, and imaged together. Slides werestained with anti-CDH1, anti-laminin (to highlight tissue bound-aries), and DAPI. Sections of paired littermates' Stage 4 notochordswere imaged on the confocal microscope and images wereanalyzed using ImageJ. The fluorescence was quantified withinthe notochord, using laminin to mark the boundaries, and usingthe “integrated density” measurement in ImageJ. Pairs of embryoswere re-scaled to the control embryo for comparison using theformula xnew¼(x�xmin/xmin�xmax) where xmin and xmax are takenfrom the control embryo. Normalized values were reported asmean7sem and compared using Student's T-test. pr0.05 wasconsidered significant. Quantification of DAPI staining using thesame method is presented for comparison.

Acknowledgments

We are grateful to members of the Klingensmith and Capel Labsfor discussion and technical assistance, particularly Kathy Carm-ody. Dr. Harold Erikson provided the anti-laminin antibody. Wealso acknowledge the Duke University Light Microscopy Core. TheBMP reporter mouse strain was kindly provided by Dr. LiefOxenberg. The CAB mouse and the NogFx strains were generouslyprovided by Dr. Fan Wang and Dr. Richard Harland, respectively.Dr. Przemko Tylzanowski and Dr. Yujiro Higashi provided Zeb1 nullembryo samples. This work was supported by grants from theNational Institutes of Health to Dr. John Klingensmith.

Appendix A. Supplementary materials

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.ydbio.2014.02.008.

References

Arkell, R., Beddington, R.S., 1997. BMP-7 influences pattern and growth of thedeveloping hindbrain of mouse embryos. Development 124 (1), 1–12.

Belo, J.A., Bouwmeester, T., Leyns, L., Kertesz, N., Gallo, M., Follettie, M., De Robertis,E.M., 1997. Cerberus-like is a secreted factor with neutralizing activityexpressed in the anterior primitive endoderm of the mouse gastrula. Mech.Dev. 68, 45–57.

Blank, U., Seto, M.L., Adams, D.C., Wojchowski, D.M., Karolak, M.J., Oxburgh, L., 2008. Anin vivo reporter of BMP signaling in organogenesis reveals targets in the developingkidney. BMC Dev. Biol. 8, 86, http://dx.doi.org/10.1186/1471-213X-8-86.

Bracken, C.P., Gregory, P.A., Kolesnikoff, N., Bert, A.G., Wang, J., Shannon, M.F.,Goodall, G.J., 2008. A double-negative feedback loop between ZEB1-SIP1 andthe microRNA-200 family regulates epithelial–mesenchymal transition. CancerRes. 68, 7846.

Buckley, S., Shi, W., Dirscoll, B., Ferrario, A., Anderson, K., Warburton, D., 2003.BMP4 signaling induces senescence and modulates the oncogenic phenotype ofA549 lung adenocarcinoma cells. Am. J. Physiol. Lung Cell Mol. Physiol. 286,L81–L86, http://dx.doi.org/10.1152/ajplung.00160.2003.

Cleaver, O., Krieg, P.A., 2001. Notochord patterning of the endoderm. Dev. Biol. 234(1), 1–12.

Danesh, S.M., Villasenor, A., Chong, D., Soukup, C., Cleaver, O., 2009. BMP and BMPreceptor expression during murine organogenesis. Gene Expr. Patterns 9 (5),255–265.

Domyan, E.T., Ferretti, E., Throckmorton, K., Mishina, Y., Nicolis, S.K., Sun, X., 2011.Signaling through BMP receptors promotes respiratory identity in the foregutvia repression of Sox2. Development 138, 971–981.

Eger, A., Aigner, K., Sonderegger, S., Dampier, B., Oheler, S., Schreiber, M., Berx, G.,Cano, A., Beug, H., Foisner, R., 2005. Delta-EF1 is a transcriptional repressor ofE-cadherin and regulates epithelial plasticity in breast cancer cells. Oncogene24, 2375–2385.

Fausett, S.R., Klingensmith, J., 2012. Compartmentalization of the foregut tube:developmental origins of the trachea and esophagus. WIREs Dev. Biol. 1,184–202, http://dx.doi.org/10.1002/wdev.12.

Goss, A.M., Tian, Y., Tsukiyama, T., Cohen, E.D., Zhou, D., Lu, M.M., Yamaguchi, T.P.,Morrisey, E.E., 2009. Wnt2/2b and beta-catenin signaling are necessary andsufficient to specify lung progenitors in the foregut. Dev. Cell. 17, 290–298.

Gregory, P.A., Bert, A.G., Paterson, E.L., Barry, S.C., Tsykin, A., Farshid, G., Vadas, M.A.,Khew-Goodal, Y., Goodall, G.J., 2008. The miR-200 family and miR-205 regulateepithelial to mesechymal transition by targeting ZEB1 and SIP1. Nat. Cell Biol.10, 593–601.

Guha, U., Gornes, W.A., Kobayashi, T., Pestell, R.G., Kessler, J.A., 2002. In vivoevidence that BMP signaling is necessary for apoptosis in the mouse limb.Dev. Biol. 249, 108–120.

Hajduk, P., May, A., Puri, P., Murphy, P., 2012. The effect of adriamycin exposure onthe notochord of mouse embryos. Birth Defects Res. B Dev. Reprod. Toxicol. 95(2), 175–183.

Harfe, B.D., Scherz, P.J., Nissim, S., Tian, H., McMahon, A.P., Tabin, C.J., 2004.Evidence for an expansion-based temporal Shh gradient in specifying verte-brate digit identities. Cell 118 (4), 517–528.

Harris-Johnson, K.S., Domyan, E.T., Vezina, C.M., Sun, X., 2009. Beta-cateninpromotes respiratory progenitor identity in mouse foregut. Proc. Natl. Acad.Sci. USA 106, 16287–16292.

Hebert, J.M., McConnell, S.K., 2000. Targeting of cre to the foxg1 (BF-1) locusmediates loxp recombination in the telencephalon and other developing headstructures. Dev. Biol. 222, 296–306.

Herrmann, B.G., 1991. Expression pattern of the Brachyury gene in whole-mountTWis/Wis mutant embryos. Development 113, 913–917.

Ioannides, A.S., Chaudhry, B., Henderson, D.J., Spitz, L., Copp, A.J., 2002. Dorsoven-tral patterning in oesophageal atresia with tracheo-oesophageal fistula: evi-dence from a new mouse model. J. Pediatr. Surg. 37, 185–191.

Ioannides, A.S., Massa, V., Ferraro, E., Cecconi, F., Spitz, L., Henderson, D.J., Copp, A.J.,2010. Foregut separation and tracheo-oesophageal malformations: the role oftracheal outgrowth, dorso-ventral patterning, and programmed cell death. Dev.Biol. 337, 351–362.

Jurand, A., 1974. Some aspects of the development of the notochord in mouseembryos. J. Embryol. Exp. Morphol. 32, 1–33.

Li, Y., Gordon, J., Manley, N.R., Litingtung, Y., Chiang, C., 2008. BMP4 is required fortracheal formation: a novel mouse model for tracheal agenesis. Dev. Biol. 322,145–155.

Li, Y., Litingtung, Y., Dijke, P.T., Chiang, C., 2007. Aberrant BMP signaling andnotochord delamination in the pathogenesis of esophageal atresia. Dev. Dyn.236, 746–754.

Lightner, V.A., Gumbowski, F., Bigner, D.D., Erickson, H.P., 1989. Tenascin/hexab-rachion in human skin: biochemical identifcation and localization by light andelectron microscopy. J. Cell Biol. 108, 2483–2493.

Lomeli, H., Ramos-Mejia, V., Gertsenstein, M., Lobe, C.G., Nagy, A., 2000. Targetedinsertion of Cre recombinase into the TNAP gene: excision in primordial germcells. Genesis 26 (2), 116–117.

McMahon, J.A., Takada, S., Zimmerman, L.B., Fan, C.M., Harland, R.M., McMahon, A.P., 1998. Noggin-mediated antagonism of BMP signaling is required for growthand patterning of the neural tube and somite. Genes Dev. 12 (10), 1438–1452.

Minoo, P., Su, G., Drum, H., Bringas, P., Kimura, S., 1999. Defects in tracheosophagealand lung morphogenesis in Nkx2.1(�/�) mouse embryos. Dev. Biol. 209,60–71.

Miyoshi, T., Maruhashi, M., Van De Putte, T., Kondoh, H., Huylebroeck, D., Higashi, Y.,2006. Complementary expression pattern of Zfhx1 genes Sip1 and deltaEF1 inthe mouse embryo and their genetic interaction revealed by compoundmutants. Dev. Dyn. 235 (7), 1941–1952.

Miyazono, K., Miyazawa, K., 2002. id: a target of BMP signaling. Sci. Signal. 151, pe40.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124 123

Nose, A., Takeichi, M., 1986. A novel cadherin cell adhesion molecule: its expressionpatterns associated with implantation and organogenesis of mouse embryos. J.Cell Biol. 103 (6), 2649–2658.

Ozkaynak, E., Schnegelsberg, P.N., Oppermann, H., 1991. Murine osteogenic protein (OP-1): high levels of mRNA in kidney. Biochem. Biophys. Res. Commun. 179, 116–123.

Park, S.M., Gaur, A.B., Lengyel, E., Peter, M.E., 2008. The miR-200 family determinesthe epithelial phenotype of cancer cells by targeting the E-cadherin repressorsZEB1 and ZEB2. Genes Dev. 22, 894–907.

Possoegel, A.K., Diez-Pardo, J.A., Morales, C., Tovar, J.A., 1999. Notochord involve-ment in experimental esophageal atresia. Pediatr. Surg. Int. 15, 201–205.

Que, J., Choi, M., Ziel, J.W., Klingensmith, J., Hogan, B.L.M., 2006. Morphogenesis ofthe trachea and esophagus: current players and new roles for Noggin andBMPs. Differentiation 74, 422–437.

Que, J., Okubo, T., Goldenring, J.R., Nam, K., Kurotani, R., Morrisey, E.E., Taranova, O.,Pevny, L.H., Hogan, B.L.M., 2007. Multiple dose-dependent roles for Sox2 in thepatterning and differentiation of the anterior foregut endoderm. Development134, 2521–2531.

Rodriguez, P., Da Silva, S., Oxburgh, L., Wang, F., Hogan, B.L.M., Que, J., 2010. BMPsignaling in the development of the mouse esophagus and forestomach.Development 137, 4171–4176.

Samavarchi-Tehrani, P., Golipour, A., David, L., Sung, H., Beyer, T.A., Datti, A.,Woltjen, K., Nagy, A., Wrana, J.L., 2010. Functional genomics reveals a BMP-driven mesenchymal-to-epithelial transition in the initiation of somatic cellreprogramming. Cell Stem Cell 7 (1), 64–77.

Solloway, M.J., Robertson, E.J., 1999. Early embryonic lethality in Bmp5;Bmp7double mutant mice suggests functional redundancy within the 60A subgroup.Development 126 (8), 1753–1768.

Soriano, P., 1999. Generalized lacZ expression with the ROSA26 Cre reporter strain.Nat. Genet. 21 (1), 70–71.

Srinivas, S., Watanabe, T., Lin, C.S., William, C.M., Tanabe, Y., Jessell, T.M., Costantini,F., 2001. Cre reporter strains produced by targeted insertion of EYFP and ECFPinto the ROSA26 locus. BMC Dev. Biol. 1 (1), 4.

Stafford, D.A., Brunet, L., Khokha, M.K., Economides, A.N., Harland, R.M., 2011.Cooperative activity of Noggin and gremlin 1 in axial skeleton development.Development 138 (5), 1005–1014.

Takagi, T., Moribe, H., Kondoh, H., Higashi, Y., 1998. DeltaEF1, a zinc finger andhomeodomain transcription factor, is required for skeleton patterning inmultiple lineages. Development 125 (1), 21–31.

Tam, P.P.L., Khoo, P.L., Lewis, S.L., Bildsoe, H., Wong, N., Tsang, T.E., Gad, J.M., Robb, L.,2007. Sequential allocation and global pattern of movement of the definitiveendoderm in the mouse embryo during gastrulation. Development 134,251–260.

Thompson, D.J., Molello, J.A., Strebing, R.J., Dyke, I.L., 1978. Teratogenicity ofadriamycin and daunomycin in the rat and rabbit. Teratology 17, 151–158.

Torfs, C.P., Curry, C.J.R., Bateson, T.F., 1995. Population-based study of tracheoeso-phageal fistula and esophageal atresia. Teratology 52, 220–232.

Tribulo, C., Aybar, M.J., Nguyen, V.H., Mullins, M.C., Mayor, R., 2003. Regulation ofmsx genes by a bmp gradient is essential for neural crest specification.Development 130, 6441–6452.

Wilkinson, D.G., Bhatt, S., Herrmann, B.G., 1990. Expression pattern of the T geneand its role in mesoderm formation. Nature 343, 657–659.

Woo, J., Miletich, I., Kim, B.M., Sharpe, P.T., Shivdasani, R.A., 2011. Barx1-mediatedinhibition of Wnt signaling in the mouse thoracic foregut controls tracheo-esophageal septation and epithelial differentiation. PLoS One 6 (7), e22493,http://dx.doi.org/10.1371/journal.pone.0022493.

Xu, L., Overbeek, P.A., Reneker, L.W., 2002. Systematic analysis of E-, N-,and P-cadherin expression in mouse eye development. Exp. Eye Res. 74,753–760.

Yang, J., Wineberg, R.A., 2008. Epithelial–mesenchymal transition: at the crossroadsof development and tumor metastasis. Dev. Cell 14 (6), 818–829.

Ziesberg, M., Hanai, J., Sugimoto, H., Mammoto, T., Charytan, D., Strutz, F., Kalluri, R.,2003. BMP-7 counteracts TGF-beta1-induced epithelial-to-mesenchymal tran-sition and reverses chronic renal injury. Nat. Med. 9 (7), 964–968.

S.R. Fausett et al. / Developmental Biology 391 (2014) 111–124124