Pax3 and Zic1 drive induction and differentiationof multipotent, migratory, and functional neuralcrest in Xenopus embryosCécile Mileta,b, Frédérique Maczkowiaka,b, Daniel D. Rochea,b, and Anne Hélène Monsoro-Burqa,b,1

aInstitut Curie, Centre National de la Recherche Scientifique Unité Mixte de Recherche 3347, Institut National de la Santé et de la Recherche Médicale U1021,F-91405 Orsay Cedex, France; and bBiology Department, Université Paris Sud-11, F-91405 Orsay Cedex, France

Edited by Nicole M. Le Douarin, Centre National de la Recherche Scientifique, Gif-sur-Yvette, France, and approved February 14, 2013 (received for reviewNovember 6, 2012)

Defining which key factors control commitment of an embryoniclineage among a myriad of candidates is a longstanding challenge indevelopmental biology and an essential prerequisite for developingstem cell-based therapies. Commitment implies that the induced cellsnot only express early lineagemarkers but further undergoanauton-omous differentiation into the lineage. The embryonic neural crestgenerates a highly diverse array of derivatives, including melano-cytes, neurons, glia, cartilage, mesenchyme, and bone. A complexgene regulatory network has recently classifiedgenes involved in themany steps of neural crest induction, specification, migration, anddifferentiation. However, which factor or combination of factors issufficient to trigger full commitment of this multipotent lineageremains unknown. Here, we show that, in contrast to other potentialcombinations of candidate factors, coactivating transcription factorsPax3 and Zic1 not only initiate neural crest specification from variousearly embryonic lineages in Xenopus and chicken embryos but alsotrigger full neural crest determination. These two factors are suffi-cient to drive migration and differentiation of several neural crestderivatives in minimal culture conditions in vitro or ectopic locationsin vivo. After transplantation, the induced cells migrate to and in-tegrate into normal neural crest craniofacial target territories, indi-cating an efficient spatial recognition in vivo. Thus, Pax3 and Zic1cooperate and execute a transcriptional switch sufficient to activatefull multipotent neural crest development and differentiation.

neural crest developmental program | ectoderm to neural cresttranscriptional switch

The neural crest, a transient embryonic cell population,develops into an amazing array of derivatives, including pe-

ripheral nervous system, pigment cells, cartilage, mesenchyme,and bone (1). During neural development, definitive neural crest(NC) induction is preceded by formation of a neural border ter-ritory between the neural plate and the nonneural ectoderm. Thisregion is initiated by transcription factor TFAP2-α (AP2a, tran-scription factor activating enhancer binding protein 2 alpha) andreinforced by Hairy2, Msx1, and AP2a itself along with secretedbone morphogenetic protein (BMP) antagonists. In addition,Pax3/Pax7, Gbx2, and Zic1 are also essential for neural borderspecification (2–8). In turn, these transcription factors cooperateto activate the NC specifiers snail2 (snai2), soxE (sox8, 9, 10), andfoxd3 in the ectoderm (reviewed in ref. 9). Although each neuralborder specifier is necessary for NC formation in vivo, none ofthese factors alone is sufficient to initiate NC induction in theectoderm (3, 7, 8). Addition of a secreted BMP antagonist, a Wntsignal, or another transcription factor is needed to activate earlyNC specifiers (reviewed in ref. 10). In particular, Pax3 and Zic1can synergize and initiate expression of early NC markers inblastula ectoderm (3, 5, 7–11). However, it remains unknownwhich of these factors are sufficient for switching on definitive NClineage commitment. These factors should be sufficient to drivean autonomous and complete NC-like developmental program,including migration, multipotency, and differentiation. To tackle

the issue, we have focused on the early neural border regulators(ap2, hairy2, msx1, pax3, and zic1) acting during gastrulation. Toaddress which combination is sufficient to elicit NC development,we took advantage of the ability of pluripotent prospective ecto-derm cells of the Xenopus blastula (also named the animal cap) todevelop autonomously in culture. These early ectodermal cellsrespond to embryonic patterning cues and thus, provide an effec-tive system for studying pluripotent cells. Amajor challenge in stemcell biology and especially, craniofacial biology will be to build newtissues in vitro for repair of injuries. As such, understanding whichof the myriad factors involved in NC specification are sufficient todrive the process is amajor goal.We find here that the transcriptionfactors Pax3 and Zic1 are the best combination to provide all of thenecessary information to trigger multipotent NC specification inthe absence of additional inducers. We challenged both migrationand differentiation of the induced cells in ectopic locations in vivo,long-term culture in vitro, and orthotopic backgrafting in vivo.

ResultsNeural border specifiers are essential transcription factors thatinitiate NC specification in vivo. However, by itself, none of thesefactors induces efficient NC development from early pluripotentprospective ectoderm. Using the main neural border regulators,which act early upstream of the NC gene regulatory network (e.g.,Pax3, Zic1, Msx1, Hairy2, and AP2), we tested the various possiblepairs to identify which combinations would be sufficient for drivingthe next key steps of NC development [e.g., NC induction (snail2expression) and cell delamination in animal cap ectoderm explants](Fig. 1 and SI Experimental Procedures). We observed that Pax3activation promoted modest snail2 expression and cell delami-nation. This activity was strongly enhanced by the coactivation ofMsx1 or Zic1, whereas other pairs of factors failed to induce snail2and delamination (Fig. 2A). Moreover, the Pax3/Zic1 combinationpromoted the highest snail2 induction, efficient explant attach-ment, and a massive cell delamination from the explants. In addi-tion, in vivo, Pax3/Zic1 also activated NC induction in the pro-spective ventral ectoderm in X. laevis and the extra embryonicectoderm in chicken embryos (Fig. S1). We have, therefore, de-cided to further characterize the NC-inducing activity of Pax3/Zic1.NC specification involves a clear temporal sequence of gene

activation starting at late gastrula–early neurula stage (stages11.5–12) (12) for the earliest NC specifiers (snail1, sox8, sox9,and myc) followed by genes activated at early–midneurula stage

Author contributions: C.M. andA.H.M.-B. designed research; C.M., F.M., D.D.R., and A.H.M.-B.performed research; F.M. contributed new reagents/analytic tools; C.M., F.M., and A.H.M.-B.analyzed data; and C.M. and A.H.M.-B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: anne-helene.monsoro-burq@curie.fr.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1219124110/-/DCSupplemental.

5528–5533 | PNAS | April 2, 2013 | vol. 110 | no. 14 www.pnas.org/cgi/doi/10.1073/pnas.1219124110

Dow

nloa

ded

by g

uest

on

June

12,

202

0

(stages 12.5–13; snail2 and foxd3) and finally, genes activatedduring later neurulation (stage 14+; sox10) (13, 14).We, therefore,

first explored whether Pax3/Zic1 recapitulated the normal pro-gression of gene activation using inducible constructs activatedat early gastrulation (stage 10). Inducible Pax3 and Zic1 displaysimilar NC inducing activity as the WT forms (Fig. S2) (1, 11, 15,16). Indeed, a group of early genes (snail1, sox8, and myc) wasactivated during early neurulation (between stages 12 and 15);a group of intermediate genes, including foxd3 and sox9, wasexpressed at midneurula stage 15. Finally, snail2 and sox10 wereinduced at a later neurula stage (Fig. 2B). Thus, the sequentialgene activation observed in vivo was clearly reproduced in vitro,albeit with slight delay.Because early ectoderm is competent to form NC only until late

gastrula/early neurula stages in vivo, we then tested the devel-opmental window of ectodermal competence to respond to Pax3/Zic1 in vitro (17). To this end, we activated Pax3 andZic1 at gastrulathrough neurula stages for assay at the late neurula stage (Fig. 2Band Fig. S3A). Similar to the normal competence for ectoderm, weobserved efficient NC specifiers’ induction only when the activationwas done before neurulation (stage 11.5). We also asked whetherPax3/Zic1 might alter the specification of an already committedtissue. When Pax3 and Zic1 were activated in already committedmesoderm (stage 10.5 dorsal, dorsal–lateral, or ventral mesoderm),the mesodermal markermyod was maintained, and activation of theNC marker snail2 was not observed (Fig. S3 B and C).We conclude that the Pax3/Zic1 combination faithfully activates

the schedule of NC specification in pluripotent prospective ecto-derm. This result raised the question of whether the tissue con-tinued NC-like development, including epithelial-to-mesenchymaltransition (EMT), migration, and terminal differentiation intomultiple types of derivatives.NC cells undergo an EMT that allows dispersal from their

birthplace in the dorsal neural tube and migration to their finalpositions. If Pax3/Zic1 drives NC specification, one would expectthat these cells would display the molecular signature of cells un-dergoing an EMT. In particular, a cadherin switch occurs beforeEMT: e-cadherin is expressed in cohesive epithelial/neuroepithelialcells, whereas n-cadherin is expressed in migrating NC in quail,mouse, and frog (18, 19). On Pax3 or Zic1 induction, e-cadherin

Fig. 1. General experimental design. X. laevis embryos were injected at thetwo-cell stage into both blastomeres with eitherWT or inducible pax3 and zic1[e.g., dexamethasone activable glucocorticoid receptor (GR) fusions]. Embryoswere grown until blastula stage 9 when blastocoele roof ectoderm was cut(about a 20-cell-wide square). Explants were further grown in 1/3 MMR or 3/4NAM without growth factor supplements until gastrula or neurula stageequivalent 10–18 (12, 15). Control sibling embryos served as a reference toevaluate developmental stages. Dexamethasone was added at late blastula–early gastrula stage 10 unless otherwise mentioned to activate Pax3GR andZic1GR in the explants. At neurula stage 18, the explants were processed (forRT-PCR, Western blot, or luciferase assay), put on fibronectin-coated plates forvideomicroscopy, or backgrafted into the cranial NC territory of a stage 18uninjected host siblingafter ablating either a part or all of thehost NC.Graftedembryos were analyzed either at migration stages 22–25 or tadpole stage 41for differentiation. Targeted ectopic injections are described below.

Fig. 2. Pax3/Zic1 is the best combination to induce NC specification in vitro with a time schedule reflecting the steps of induction in vivo. (A) Comparison ofneural border specifiers’ combinations in NC induction and emigration from ectoderm. Neural border specifiers, which expand the NC domain in vivo, weretested for initiating NC specification (indicated by snail2 induction) and delamination in “animal cap” prospective ectoderm explants. Delamination wasanalyzed after plating the explants onto a fibronectin substratum. Ten different combinations of the main five neural border specifiers (pax3, zic1, msx1, ap2,and hairy2) were tested. Results were scored as follows: percent of explants showing delamination (an average of 15 explants per condition was analyzed);relative levels of snail2 induction (from 24 explants per condition) compared with themaximal induction observed using quantitative RT-PCR: −, value ≤ 25%; +, 25%< value ≤ 50%; ++, 50% < value ≤ 75%; +++, value ≥ 75%. Single injections were analyzed for pax3 and hairy2, because they had not been described forsnail2 induction in ectoderm explants before. att, attachement; NA, not analyzed. (B) RT-PCR analysis was done after induction and lysis at various time pointsduring gastrulation and neurulation for the following NC specifiers: snail2, foxd3, sox8, 9, and 10, myc, and snail1. When induction was done at stage 10 andlysis at increasing developmental time points during gastrulation and neurulation (stages 11, 12, 15, and 18), appearance of NC specifiers followed the se-quential appearance described in vivo. Similarly, when induction was done at various times during gastrulation (stages 10, 10.5, and 11.5) and lysis was doneat stage 18, responsiveness drastically decreased, indicating the same stage limit in ectoderm competence as described in vivo. Lane 1, uninjected wholeembryo; lane 2, − reverse transcriptase (RT) control; lane 3, uninjected ectoderm; induction/lysis, stage of dexamethasone addition/stage of analysis.

Milet et al. PNAS | April 2, 2013 | vol. 110 | no. 14 | 5529

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

12,

202

0

expression was decreased, whereas Pax3 activated n-cadherin ex-pression. Pax3/Zic1 coexpression potentiated these two effects(Fig. 3 A–D). This clear-cut cadherin switch was, thus, consistentwith EMT initiation in the induced ectoderm in vitro in registerwith the time when NC undergoes EMT in vivo. To validate afunctional EMT in vitro, we compared the ability of NC cells andGFP-labeled Pax3/Zic1-induced ectoderm to undergo EMT andmigrate on fibronectin (18).When uninduced animal caps attachedand spread on fibronectin, these control cells did not emigrate fromthe explant. In contrast, both NC and Pax3/Zic1-induced ectodermefficiently attached, spread, and underwent EMT (Fig. 3 B–G).Individual cells detached from the explant, produced numerousprotrusions, and actively migrated, with a slightly slower overallvelocity for Pax3/Zic1-induced ectoderm compared with NC cells(Fig. 3H and Movies S1, S2, and S3). Altogether, these in vitroresults showed that Pax3/Zic1 activation is sufficient to induceEMT and migration in early prospective ectoderm.Because Pax3/Zic1 cells display the molecular signature of

premigratory NC and EMT and undergo migration in vitro, weasked if they were capable of migration in vivo. We grafted suchcells, unilaterally, into an unlabeled host after the time when en-dogenous NC induction had been completed (Fig. 1). Thus, weensured that the potential NC induction only depended on the invitro activation of Pax3 and Zic1. Control explants healed into thehost ectoderm and did not migrate (Fig. 3I). Pax3/Zic1-inducedectoderm efficiently migrated in host embryos and followed nor-mal NCmigration paths around the eye and along themandibular,

hyoid, and branchial arches (Fig. 3J). Ectoderm forced to expressPax3 only also migrated, albeit far less efficiently (only 30% of thegrafted explants migrated) (Fig. 3K). In conclusion, when back-grafted in vivo, after the time when NC can be induced, the Pax3/Zic1-induced ectodermal cells migrate along the normal NCstreams of migration, indicating a fine recognition of the spatialcues that guide NC migration in the embryo.The molecular signatures and cell migratory behaviors suggested

that Pax3/Zic1 cells were equivalent to the NC cells developing invivo. However, the ultimate test for NC cell commitment is theirability to differentiate into diverse cell types. We, therefore, asked ifthe Pax3/Zic1-induced ectoderm could undergo autonomous NC-type terminal differentiation in vitro. Strikingly, Pax3/Zic1 express-ing ectodermal explants frequently developed some pigmented andstellate cells reminiscent of NC-derived melanocytes (in 44% of theexplants), whereas the controls formed only ciliated ectoderm likeuninjected animal caps (Fig. 4 A and B and Fig. S4A). Occasionally,a few melanocytes differentiated within Pax3 alone-induced ex-plants (14% of the explants) (Fig. S4A). Importantly, Pax3/Zic1-induced ectoderm expressed mitf, a key molecular marker for ter-minal melanocyte differentiation (Fig. 4 and Fig. S4) (20, 21).Because a major function of NC cells in vivo is to differentiate

into the chondrocytes that build the skeletal elements of the faceand jaw, we asked if Pax3/Zic1 expressing ectoderm ultimatelyformed chondrocytes. We found that the explants expressed sox9and runx2, which indicate chondrocyte and bone commitment,respectively (Fig. 4C) (22, 23). In addition, the NC differentiates

Fig. 3. Pax3/Zic1-induced ectoderm displays cadherin switch and migratory activity in vitro and migrates in vivo. (A) n- and e-cadherin expression was analyzedby RT-PCR on explants induced at gastrula stage 10 and lysed at neurula stage 18. Low e-cadherin and high n-cadherin mark the Pax3/Zic1-induced ectoderm.Lane 1, uninjected whole embryo; lane 2, −RT control; lanes 3 and 7, uninjected ectoderm; −Dex, ethanol-treated Pax3GR/Zic1GR-injected ectoderm; +Dex,dexamethasone-treated Pax3GR/Zic1GR-injected ectoderm. (B–H) Using histone2b-GFP mRNA coinjections, we plated either Pax3GR/Zic1GR/GFP-injected ec-toderm (B and E; uninduced controls; C and F; induced explants) or GFP-labeled NC (D and G) at stage 18 on fibronectin-coated plates. NC and Pax3/Zic1-induced(+Dex) cells attached, spread, and exhibited EMT. E-cadherin was prominent at cell junctions in uninduced explants, whereas actin staining showed numerousprotrusions in both NC and induced explants. Videomicroscopy (H) indicated that individual cells actively migrated outside of the Pax3/Zic1-induced ectoderm,albeit slightly slower than control NC cells (t test: P < 0.0001; error bars: SEM). Uninduced cells (−Dex) did not migrate. (Scale bars: 100 μm.) (I–K) When graftedinto the cranial NC territory, the control (−Dex) explants integrated the ectoderm and remained at the graft site (I and I′; 0% migration, n = 24, white arrow),whereas the induced (+Dex) ectoderm actively migrated along the normal NC migration paths a few hours postgrafting (J and J′; 77%, n = 70, red arrows) whenhost NC was both present and fully ablated. Pax3GR-only injected grafts exhibited some but less-efficient migration (K). (Scale bars: 500 μm.)

5530 | www.pnas.org/cgi/doi/10.1073/pnas.1219124110 Milet et al.

Dow

nloa

ded

by g

uest

on

June

12,

202

0

into neurons and glia of the peripheral nervous system. We foundthat the Pax3/Zic1 explants expressed the neuronal markers c-retand phox2b, which mark autonomous neurons of the sympatheticand enteric nervous systems as well as the glial markers sox10 andfoxd3 (Fig. 4C) (24–30). Similar results were observed with bothWT and activable pax3 and zic1 injections (Fig. 4C). In contrast,when Zic1 alone was activated, only neural tissue was formed,indicated by robust ncam activation, but no other differentiationmarker tested was found (31). The few Pax3 alone-inducedexplants that formed pigment were picked up for RT-PCR: theyalso expressed ncam, sox9 (Fig. 4C), sox10, and mitf (Fig. S4B),which is in line with Pax3 being a direct mitf activator duringmelanocyte differentiation (32).To test if the induced cells could form morphologically differ-

entiated neurons, we used a defined medium enriched to allowneuronal growth and survival after the initial induction (Fig. 4 D–I). Control NC formed numerous neurofilament-positive neuritesin both saline and supplemented medium (Fig. 4 F and I). Pax3/Zic1-induced ectoderm cells efficiently formed neuritis only in thesupplemented medium (Fig. 4 E and H), whereas control cells didnot form neurites (Fig. 4 D and G). Hence, the ectoderm inducedby Pax3 and Zic1 has the potential to differentiate into neurons.

Because Msx1 coinjected with Pax3 was the second best combi-nation to activate early NC development from animal caps (Fig.2A), we have then tested if Msx1 would promote neuronal dif-ferentiation in Pax3- or Pax3/Zic1-induced ectoderm. How-ever, coexpression of Msx1 did not improve neuron formationin minimal culture conditions (Fig. S5).We then tested the differentiation of cells induced ectopically

in vivo using injections targeted to either the prospective ventralectoderm or the prospective endoderm of 16-cell stage Xenopusembryos (33). The induced cells formed a pigmented tissue masseither ventrally or within the endoderm (Fig. 5 A–C and Fig. S6);they also expressed ectopic sox9 (Fig. 5M–O), sox10 (Fig. 5 J–L),neural tubulin (Fig. 5 D–F), tyrosine hydroxylase (TH) (Fig. 5G–I),and low levels of runx2 (Fig. S6 P–R). These ectopic stainingsindicated the formation of multiple NC-like derivatives, suchas melanocytes, TH-positive peripheral neurons, and cartilage,in ectopic locations, including the prospective endoderm in vivo.Finally, because the Pax3/Zic1 cells induced in vitro and

backgrafted in vivo specifically followed endogenous NC migra-tion paths, we have challenged the ability of the induced ecto-derm to form differentiated tissues in host embryos. To favordifferentiation from the graft, the whole cranial NC was ablatedon one side of the unlabeled host before grafting. NC removalresulted in a dramatic failure of craniofacial development thatcould be rescued by backgrafting premigrating GFP-labeled NCcells taken at neurulation stages 17–18 (Fig. S7). The controlgraft (i.e., injected Pax3/Zic1/GFP but uninduced) formed skinepithelium along with WT blastocoele roof ectoderm (Fig. 6Aand Fig. S8). In contrast, after Pax3/Zic1 activation in vitro, wenoted long distance migration of the grafted Pax3/Zic1/GFP-

Fig. 4. Pax3/Zic1-induced ectoderm differentiates into multiple NC deriva-tives in vitro. (A–C) We have grown the control (A) and induced (B) explantsin 3/4 NAM or 1/3 MMR without any supplements (except for gentamicin) forseveral days (6–8 d at 15 °C), from late neurula stage 18 to swimming tad-pole stage 41 (differentiation stage). Melanocytes differentiated into theinduced (B) but not the control (A) explants. (C) RT-PCR analysis showed thatmarkers for various NC derivatives were expressed when both WT and in-ducible pax3 and zic1 were coinjected. Lanes 1 and 7, uninjected wholeembryo; lanes 2 and 8, −RT control; lanes 3 and 9, uninjected explant; −Dex,ethanol-treated Pax3GR/Zic1GR-injected explants; +Dex, dexamethasone-treated Pax3GR/Zic1GR-injected explants. (D–I) Using histone2b-GFP mRNAcoinjections, we plated control ectoderm (D and G), Pax3GR/Zic1GR-inducedectoderm (E and H), or GFP-labeled NC (F and I) on fibronectin-coated platesafter the initial induction in 3/4 NAM in vitro. The cells were grown either in3/4 NAM for 3–4 d or switched to Neurobasal/B27 medium after 1 d on fi-bronectin. In 3/4 NAM, only NC formed neurites (F; red, antineurofilamentimmunostaining). When Neurobasal/B27 medium was added, both NC andPax3GR/Zic1GR-induced cells formed neurites (H and I), whereas the unin-duced cells did not (G).

Fig. 5. Pax3 and Zic1 ectopic coactivation induces NC-like differentiation inprospective ventral ectoderm and endoderm. (A–C) Pax3GR/Zic1GR/β-galmRNAs coinjections were targeted to the prospective ventral epidermis (BInset; blastomere V1.1) or the prospective endoderm (C Inset; vegetal V2.2/3dblastomere) in 16-cell stage blastulas. Dexamethasone activation was per-formed at the 32-cell stage. Stage 41-injected tadpoles exhibit ectopic mel-anocytes in the ventral ectoderm (B) and endoderm (C), respectively (whitearrows) compared with control siblings (A). (D–O) After β-gal staining tolocalize the injected area (red), these tadpoles were stained for variousNC markers (WISH; purple-blue staining) and sectioned. In tadpoles injectedinto both the prospective ventral epidermis and the prospective endoderm,we observed ectopic staining for neural tubulin (E and F), TH (H and I), sox10(K and L), and sox9 (N and O) compared with control uninjected embryos (D,G, J, and M). (Scale bars: A–C, 1 mm; D–O, 100 μm.)

Milet et al. PNAS | April 2, 2013 | vol. 110 | no. 14 | 5531

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

12,

202

0

labeled cells either dorsally or to the craniofacial areas of thehost, and in the best cases, we noted restoration of the generalhead morphology on the grafted side at swimming tadpole stage(Fig. 6 C and F). Moreover, the GFP-positive cells differentiatedinto pigmented and stellate melanocytes under the ectoderm andwere also found in deeper locations around the eye, along theoptic nerve, and in the head mesenchyme (Fig. 6B and Fig. S8).They formed GFP-positive fibroblasts in the branchial archesmesenchyme, well-shaped sox9+/GFP+ Meckel’s cartilage, andother cartilage elements on the grafted side (Fig. 6 D–G),whereas contralateral cartilages were formed by sox9+/GFP−host cells (Fig. 6H). Occasionally, at the rhombencephalon level,GFP-labeled cells were found adjacent to craniofacial musclesand ganglia on the grafted side, but we could not clearly identifythose cells as muscle or glial precursors.In conclusion, the cells committed to a NC fate by Pax3 and

Zic1 activation, integrated several NC destinations in the cra-niofacial area in vivo, and underwent terminal differentiationinto multiple bona fide NC-type derivatives.

DiscussionThe mechanisms of NC induction and development have beenextensively studied. Wnt, BMP, and FGF secreted by the superfi-cial ectoderm and paraxial/intermediate mesoderm activatea network of genes at the neural plate border, which then specifiesthe NC (3, 34–36). It was shown that AP2, Pax3, Msx1, and Zic1are key transcription factors that initiate expression of the earliestNC markers after neural border formation (3, 5, 7, 8). In partic-ular, Pax3 and Zic1 cooperate with Wnt signals, whereas AP2cooperates with BMP antagonists to initiate snail2 expression

in the prospective ectoderm (3, 5, 7). Despite these numerousstudies, the key question of exactly which cell autonomous factorsmay be sufficient to commit ectoderm to a multipotent NC fateand drive its complete development until differentiation remainedelusive. Because individual neural border regulators fail to activatede novo NC development from early prospective ectoderm, welooked for a combinatorial activity of pairs of neural borderspecifiers, which would be sufficient to drive full NC developmentfrom uncommitted ectoderm. It was shown previously that coin-jection of the two neural border specifiers Pax3 and Zic1 activatesectopic expression of early premigratory NC markers in frogventral ectoderm in vivo (5, 7). Positive cross-regulation activatesPax3 or Zic1 expression when Zic1 or Pax3 are activated, re-spectively, inWnt-treated animal caps (5).Moreover, the balance ofactivity of these two factors is critical for a choice between NC,placode, and hatching gland fates (11). Whether other combina-tions of factors were playing a similar role remained unknown.Altogether, our results indicate that activating the two factors Pax3and Zic1 efficiently commits early cells—whether they are pro-spective ectoderm or endoderm—to multipotent and functionalNC. This coactivation triggers NC specifiers’ expression in thenormal sequence and schedule followed by EMT, migration, anddefinitive NC-type differentiation both in vitro and in vivo. Al-though the WT NC autonomously forms neurons in minimal con-ditions, the Pax3/Zic1-induced cells formed neurites only insupplemented medium. In addition, the induced cells undergoa random migration in vitro, which was described for isolated NCcells devoid of a chemoattractant source, but directional migrationin vivo, indicating appropriate recognition of guidance cues (18, 37).Our data, thus, provide an experimental paradigm to elucidate thefine-tuning of the NC gene regulatory network as well as developassays for differentiating NC derivatives from stem cells in vitro.During ectoderm patterning in vivo, balanced levels of Pax3

and Zic1 are essential to control the choice between NC, pla-code, and other cell fates at the neural border (7, 11). WhenPax3/Zic1 levels are appropriate, a full NC development pro-gram is, thus, triggered by these two transcription factors, similarto activation of the skeletal muscle program by MyoD and car-diac development by a combination of Nkx2.5, GATA4, andmyocardin (38). Within the general context of understandingadult cell reprogramming and controlling pluripotent cell dif-ferentiation into chosen lineages (39), the NC commitmenttriggered by Pax3 and Zic1 provides a framework to study NCand craniofacial development from stem cells. It also suggeststhat coregulation of these two factors is an essential node in theNC gene regulatory network that may have allowed NC forma-tion during vertebrate evolution.

Experimental ProceduresXenopus Embryo Manipulation and Explant Culture. X. laevis embryos wereobtained and staged using standard procedures (12, 15). H2b-GFPmRNA wascoinjected as a lineage tracer for migration and grafting assays, and β-galwas used as a lineage tracer for ectopic injection assays. mRNAs were syn-thesized using the mMESSAGE mMACHINE kit (Ambion). The plasmids usedas templates were pax3, zic1, pax3-GR, zic1-GR, β-gal, and H2bGFP (TablesS1, S2, and S3). For animal cap experiments, the two blastomeres of stage 2embryos were injected at the animal pole. The blastocoele roof ectodermwas dissected at stage 9 and aged in 3/4 Normal Amphibian Medium (NAM)or 1/3 Mark’s Modified Ringer (MMR) until the sibling controls reached stage18 or 41. For grafting, the cranial NC of stages 17–18 host embryos waspartially or totally ablated on one side. A piece of animal cap explant grownto stage 18 was carefully rinsed and immediately put in place of the ablatedNC. Operated embryos were then grown until stages 22–41.

Chicken Embryos Electroporation. Gallus gallus embryos were incubated untillate blastula/early gastrula stage (2–4 Hamburger–Hamilton stage). Theywere then placed in the electroporation chamber as described in ref. 40. Theplasmid solution containing Xenopus Pax3 and Zic1 constructs was thenelectroporated. The embryos were reincubated overnight on albumin, fixed,and processed for in situ hybridization (adapted from ref. 41).

Fig. 6. Pax3/Zic1-induced ectoderm differentiates into multiple cranial NCderivatives in vivo. Embryos were grafted orthotopically as previously de-scribed (Figs. 1 and 3) and grown until stage 41. Although control GFP+ cells(i.e., ethanol-treated Pax3GR/Zic1GR/H2bGFP-injected cells) integrated theskin (A), induced GFP+ cells (i.e., dexamethasone-treated Pax3GR/Zic1GR/H2bGFP-injected cells) formed melanocytes (B Inset; note adjacent blackmelanosome and GFP+ nucleus) and migrated into deeper (thus out of focus)locations dorsally, around the eye, and into the branchial arches (B, stage 45;C, stage 35). Scheme of a stage-41 tadpole head in transverse section (D),indicating the location of the three NC-type derivatives found in operatedembryos (shown in B and E–H). Transverse head sections processed with sox9in situ hybridization and anti-GFP immunostaining analysis show GFP+fibroblasts in the maxillary mesenchyme (D, 2 and E) and GFP+ Meckel’scartilage (D, 3, F, and G) on the grafted side but only sox9+/GFP− cartilage onthe contralateral control side (H). (Scale bars: A–C, 500 μm; E–H, 100 μm.)

5532 | www.pnas.org/cgi/doi/10.1073/pnas.1219124110 Milet et al.

Dow

nloa

ded

by g

uest

on

June

12,

202

0

In Vitro Cell Culture and Videomicroscopy. For in vitro migration assay, NCexplants of stages 17–18 control embryos and injected animal cap ectodermgrown until stage 18 were plated on fibronectin (10 μg/mL on plastic) in six-well plates, cultured, and imaged in 3/4 NAM (18). Individual cell trackingwas performed using the Manual Tracking Image J plugin (http://rsbweb.nih.gov/ij/plugins/track/track.html) (42). Statistical analysis by two-way ANOVAindicated an extremely significant difference of migration overall velocitybetween control NC and induced ectoderm explants (P < 0.0001). For in vitroneuronal differentiation, similar NC or explants were plated on fibronectin-coated coverslips (100 μg/mL) in six-well plates and cultured 24 h in 3/4 NAM.Then, 2/3 of the NAM was removed, and an equivalent volume of Neuro-basal medium supplemented with antibiotics and B-27 supplement wasadded (#21103 and #17504; Gibco). Neuronal differentiation occurred within2–3 additional d of culture for NC and 3–4 d for Pax3/Zic1-induced ectoderm.

In Situ Hybridization and Immunochemistry. After β-gal staining, whole mountin situ hybridization on ectopically injected whole embryos was performed(41); then, embryos were embedded in Albumin/Gelatin mix for vibratomesectioning. Grafted embryos were first embedded in paraffin for sectioningand then stained by in situ hybridization using DIGoxigenin-labeled probes(http://geisha.arizona.edu). Anti-GFP immunostaining was performed afterin situ hybridization. For in vitro explants staining, NC explants (taken fromstages 17–18 control embryos) and injected ectoderm explants grown untilstage 18 were plated on fibronectin (10 mg/mL) in six-well plates and cul-tured in 3/4 NAM (18). Explants were fixed in 3.7% formaldehyde/96.3% PBS(vol/vol) and stained with phalloidin, anti–E-Cadherin, or antineurofilamentP200 antibodies (Tables S1, S2, and S3).

Pharmacological Treatments. Inducible Pax3 and Zic1 (Pax3-GR and Zic1-GR)were activated by dexamethasone (16). Controls included pax3GR/zic1-GR–injected embryos grown in 0.2% ethanol and noninjected embryos grownin dexamethasone.

RT-PCR. Semiquantitative and quantitative RT-PCR (15) included the follow-ing controls (Figs. 2B, 3A, and 4C): whole embryo (lane 1), embryo withoutreverse-transcriptase treatment (lane 2), and uninjected caps (lane 3). Elon-gation factor 1a or Ornithine Decarboxylase served as a loading control, andmuscle actin controlled for mesoderm contamination (not shown). Detailsand primer pairs are presented in Tables S1, S2, and S3.

ACKNOWLEDGMENTS. The authors thank Tatjana Sauka-Spengler for help inchick electroporation experiments. We also thank Eric Theveneau for helpfuldiscussions and Karen Liu, Richard Harland, and John Wallingford for theircritical reading of the manuscript. We thank J.-P. St-Jeannet, R. Tsien, andM. Perron for the gift of various reagents. The authors thank the PICT@OrsayImaging Facility and the Animal Facility of the Institut Curie. C.M. wasa postdoctoral fellow of Region Ile de France (Domaine d’IntérêtMajeur StemPole), Universite Paris Sud-11 (Attaché Temporaire d’Enseignement et deRecherche), and Agence Nationale de la Recherche. D.D.R. was a Region Ile-de-France (DIM Stem Pole) and Fondation de France fellow. This work wasfunded by the Université Paris Sud-11 (Attractivite 2011), the Centre de laRecherche Scientifique (Action Thématique et Incitative sur Programme), As-sociation pour la Recherche contre le Cancer Grant SFI20101201882, Liguecontre le Cancer, and Agence Nationale de la Recherche (Agence Nationalede la Recherche Programme Blanc; A.H.M.-B.).

1. Le Douarin ML, Kalcheim C (1999) The Neural Crest (Cambridge Univ Press, Cam-bridge, United Kingdom).

2. Li B, Kuriyama S, Moreno M, Mayor R (2009) The posteriorizing gene Gbx2 is a directtarget of Wnt signalling and the earliest factor in neural crest induction. De-velopment 136(19):3267–3278.

3. de Crozé N, Maczkowiak F, Monsoro-Burq AH (2011) Reiterative AP2a activity controlssequential steps in the neural crest gene regulatory network. Proc Natl Acad Sci USA108(1):155–160.

4. Basch ML, Bronner-Fraser M, García-Castro MI (2006) Specification of the neural crestoccurs during gastrulation and requires Pax7. Nature 441(7090):218–222.

5. Sato T, Sasai N, Sasai Y (2005) Neural crest determination by co-activation of Pax3 andZic1 genes in Xenopus ectoderm. Development 132(10):2355–2363.

6. Nichane M, et al. (2008) Hairy2-Id3 interactions play an essential role in Xenopusneural crest progenitor specification. Dev Biol 322(2):355–367.

7. Monsoro-Burq AH, Wang E, Harland R (2005) Msx1 and Pax3 cooperate to mediateFGF8 and WNT signals during Xenopus neural crest induction. Dev Cell 8(2):167–178.

8. Luo T, Lee YH, Saint-Jeannet JP, Sargent TD (2003) Induction of neural crest in Xen-opus by transcription factor AP2alpha. Proc Natl Acad Sci USA 100(2):532–537.

9. Milet C, Monsoro-Burq AH (2012) Neural crest induction at the neural plate border invertebrates. Dev Biol 366(1):22–33.

10. Pegoraro C, Monsoro-Burq AH (2013) Signaling and transcriptional regulation inneural crest specification and migration: lessons from xenopus embryos. WIREs DevBiol 2:247–259.

11. Hong CS, Saint-Jeannet JP (2007) The activity of Pax3 and Zic1 regulates three distinctcell fates at the neural plate border. Mol Biol Cell 18(6):2192–2202.

12. Nieuwkoop PD, Faber J (1994) Normal Table of Xenopus laevis (Daudin) (Garland,New York), 3rd Ed.

13. Aybar MJ, Nieto MA, Mayor R (2003) Snail precedes slug in the genetic cascade re-quired for the specification and migration of the Xenopus neural crest. Development130(3):483–494.

14. O’Donnell M, Hong CS, Huang X, Delnicki RJ, Saint-Jeannet JP (2006) Functionalanalysis of Sox8 during neural crest development in Xenopus. Development 133(19):3817–3826.

15. Sive HL, Grainger RM, Harland RM (2000) Early Development of Xenopus laevis: ALaboratory Manual (Cold Spring Harbor Laboratory Press, Plainview, NY).

16. Kolm PJ, Sive HL (1995) Efficient hormone-inducible protein function in Xenopuslaevis. Dev Biol 171(1):267–272.

17. Mancilla A, Mayor R (1996) Neural crest formation in Xenopus laevis: Mechanisms ofXslug induction. Dev Biol 177(2):580–589.

18. Theveneau E, et al. (2010) Collective chemotaxis requires contact-dependent cellpolarity. Dev Cell 19(1):39–53.

19. Wheelock MJ, Shintani Y, Maeda M, Fukumoto Y, Johnson KR (2008) Cadherinswitching. J Cell Sci 121(Pt 6):727–735.

20. Hornyak TJ, Hayes DJ, Chiu LY, Ziff EB (2001) Transcription factors in melanocytedevelopment: Distinct roles for Pax-3 and Mitf. Mech Dev 101(1-2):47–59.

21. Kumasaka M, Sato S, Yajima I, Yamamoto H (2003) Isolation and developmentalexpression of tyrosinase family genes in Xenopus laevis. Pigment Cell Res 16(5):455–462.

22. Mori-Akiyama Y, Akiyama H, Rowitch DH, de Crombrugghe B (2003) Sox9 is requiredfor determination of the chondrogenic cell lineage in the cranial neural crest. ProcNatl Acad Sci USA 100(16):9360–9365.

23. Bialek P, et al. (2004) A twist code determines the onset of osteoblast differentiation.

Dev Cell 6(3):423–435.24. Pattyn A, Morin X, Cremer H, Goridis C, Brunet JF (1999) The homeobox gene Phox2b

is essential for the development of autonomic neural crest derivatives. Nature399(6734):366–370.

25. Young HM, et al. (1998) A single rostrocaudal colonization of the rodent intestine by

enteric neuron precursors is revealed by the expression of Phox2b, Ret, and p75 andby explants grown under the kidney capsule or in organ culture. Dev Biol 202(1):67–84.

26. Thomas AJ, Erickson CA (2009) FOXD3 regulates the lineage switch between neuralcrest-derived glial cells and pigment cells by repressing MITF through a non-canonicalmechanism. Development 136(11):1849–1858.

27. Paratore C, Goerich DE, Suter U, Wegner M, Sommer L (2001) Survival and glial fateacquisition of neural crest cells are regulated by an interplay between the tran-

scription factor Sox10 and extrinsic combinatorial signaling. Development 128(20):3949–3961.

28. Kos R, Reedy MV, Johnson RL, Erickson CA (2001) The winged-helix transcription

factor FoxD3 is important for establishing the neural crest lineage and repressingmelanogenesis in avian embryos. Development 128(8):1467–1479.

29. Lang D, et al. (2000) Pax3 is required for enteric ganglia formation and functions withSox10 to modulate expression of c-ret. J Clin Invest 106(8):963–971.

30. Kuhlbrodt K, Herbarth B, Sock E, Hermans-Borgmeyer I, Wegner M (1998) Sox10,

a novel transcriptional modulator in glial cells. J Neurosci 18(1):237–250.31. Balak K, Jacobson M, Sunshine J, Rutishauser U (1987) Neural cell adhesion molecule

expression in Xenopus embryos. Dev Biol 119(2):540–550.32. Corry GN, Underhill DA (2005) Pax3 target gene recognition occurs through distinct

modes that are differentially affected by disease-associated mutations. Pigment Cell

Res 18(6):427–438.33. Moody SA (1987) Fates of the blastomeres of the 32-cell-stage Xenopus embryo. Dev

Biol 122(2):300–319.34. Monsoro-Burq AH, Fletcher RB, Harland RM (2003) Neural crest induction by paraxial

mesoderm in Xenopus embryos requires FGF signals. Development 130(14):3111–

3124.35. García-Castro MI, Marcelle C, Bronner-Fraser M (2002) Ectodermal Wnt function as

a neural crest inducer. Science 297(5582):848–851.36. Betancur P, Bronner-Fraser M, Sauka-Spengler T (2010) Assembling neural crest reg-

ulatory circuits into a gene regulatory network. Annu Rev Cell Dev Biol 26:581–603.37. Theveneau E, Mayor R (2011) Collective cell migration of the cephalic neural crest:

The art of integrating information. Genesis 49(4):164–176.38. Sachinidis A, et al. (2003) Cardiac specific differentiation of mouse embryonic stem

cells. Cardiovasc Res 58(2):278–291.39. Yamanaka S, Blau HM (2010) Nuclear reprogramming to a pluripotent state by three

approaches. Nature 465(7299):704–712.40. Strobl-Mazzulla PH, Sauka-Spengler T, Bronner-Fraser M (2010) Histone demethylase

JmjD2A regulates neural crest specification. Dev Cell 19(3):460–468.41. Monsoro-Burq AH (2007) A rapid protocol for whole-mount in situ hybridization on

Xenopus embryos. CSH Protoc 2007(Aug 1):pdb prot4809.42. Abramoff MD, Magalhaes PJ, Ram SJ (2004) Image processing with ImageJ.

Biophotonics Int 11(7):36–42.

Milet et al. PNAS | April 2, 2013 | vol. 110 | no. 14 | 5533

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

12,

202

0