869Development 125, 869-877 (1998)Printed in Great Britain © The Company of Biologists Limited 1998DEV3754

FGF1 patterns the optic vesicle by directing the placement of the neural retina

domain

Jeanette Hyer, Tatsuo Mima and Takashi Mikawa*

Department of Cell Biology and Anatomy, Cornell University Medical College, 1300 York Avenue, New York, NY 10021, USA*Author for correspondence (e-mail: tmikaw@mail.med.cornell.edu)

Accepted 18 December 1997: published on WWW 4 February 1998

Patterning of the bipotential retinal primordia (the opticvesicles) into neural retina and retinal pigmentedepithelium depends on its interaction with overlayingsurface ectoderm. The surface ectoderm expresses FGFsand the optic vesicles express FGF receptors. PreviousFGF-expression data and in vitro analyses support thehypothesis that FGF signaling plays a significant role inpatterning the optic vesicle. To test this hypothesis in vivowe removed surface ectoderm, a rich source of FGFs. Thisablation generated retinas in which neural and pigmentedcell phenotypes were co-mingled. Two in vivo protocolswere used to replace FGF secretion by surface ectoderm:(1) implantation of FGF-secreting fibroblasts, and (2)injection of replication-incompetent FGF retroviralexpression vectors. The retinas in such embryos exhibited

segregated neural and pigmented epithelial domains. Theneural retina domains were always close to a source of FGFsecretion. These results indicate that, in the absense ofsurface ectoderm, cells of the optic vesicles display bothneural and pigmented retinal phenotypes, and thatpositional cues provided by FGF organize the bipotentialoptic vesicle into specific neural retina and pigmentedepithelium domains. We conclude that FGF can mimic oneof the earliest functions of surface ectoderm during eyedevelopment, namely the demarcation of neural retinafrom pigmented epithelium.

Key words: Fgf, Neural retina, Optic vesicle, Eye development,Retrovirus, Vertebrate

SUMMARY

ndg, itrect.

lderscttivez,

ng toeof inhenein).ye

E.hat;

INTRODUCTION

Development of the vertebrate eye depends on the coordininteraction of three distinct tissues: neuroepithelium, overlyisurface ectoderm and mesenchyme. In one of the earliest sof eye formation, anterior neuroepithelium of the neural tugives rise to the optic vesicles (OVs) by evaginating bilatera(Fig. 1). Each OV extends until it contacts the surface ectod(SE). After contact, the distal portion of the OV invaginatthus initiating formation of the bilayered optic cup. The innlayer of the optic cup forms the multiple layers of the neuretina (NR), while the outer layer remains a simple cuboiepithelium, which differentiates as retinal pigmenteepithelium (RPE). Simultaneously, the SE, at its point contact with the OV thickens into a placode that will first fora vesicle and later differentiate into the lens.

Although different in form and function, the NR and RPderive from common progenitors in the OV. Differenexperimental approaches have demonstrated that single cethe OV are bipotential, able to form both NR and RP(Mikami, 1939; Dragomirov, 1937; Stroeva, 1960; Dorri1938; Lopashov, 1963). Thus, it has been postulated patterning or segregation of the OV into separate NR and Rdomains must arise from local inductive events (Mikami, 193Detwiler and VanDyke, 1954; Dragomirov, 1937). On

atedngtepsbelly

ermeserraldaldofm

Et

lls ofE

s,thatPE9;e

hypothesis states that tight apposition between the distal tip athe SE may be necessary for exclusion of RPE-inducinmesenchymal cells (Lopashov and Stroeva, 1961). Howeveris also possible that SE or SE plus mesenchyme exerts a diinstructive role, specifying tissue-specific domains in the OVAlternatively, information defining these tissue patterns coube transmitted by local contact between inner and outer layof the optic cup, thereby establishing the NR; when contabetween these two layers is prevented, the inner prospecNR differentiate as RPE (Orts-Llorca and Genis-Galve1960).

One approach that has proven to be successful in clarifyithe complex steps underlying embryonic morphogenesis isdefine the molecular components involved in the inductivreactions during the morphogenetic process. A number peptide growth factors have been shown to be key playerssuch reactions. That growth factors play such a role in tdeveloping eye is supported by the fact that at least omember of every major growth factor family is expressed the early embryonic eye (McAvoy and Chamberlain, 1990However, of all the growth factors found in ocular tissue onlFGF family members and their receptors are known to bexpressed prior to or during the association of OV with the SBoth FGF1 and FGF2 are expressed in the SE at the time tit makes contact with the OV (DeIongh and McAvoy, 1993

870

ally

s aE

hets,

ar)cle

s onith

%enal

gifth00. onnd

essegift0ti-and

ueupheFs. thee,ticE

sed inveedF

ly

ofFig.).fesith

ig.lls;ted,ig.

J. Hyer, T. Mima and T. Mikawa

E 1.5 E 2.0 E 4.0

Fig. 1.Schematic representation of eye development during chickenembryogenesis from optic vesicle stage to formation of the optic cup;purple, the presumptive retinal pigmented epithelium; yellow, thepresumptive neural retina; blue, the surface ectoderm; white, the lensplacode and lens.

Pittack et al., 1997). A cognate receptor for these FGFGFR1, is found throughout the OV (Wanaka et al., 199FGFR2, a second potential receptor, colocalizes with FGFin the early optic cup (Tcheng et al., 1994).

A number of studies have addressed the role of FGF signain retinal development. In the differentiation of the NR, FGFhas been shown to elevate cellular opsin levels (Hicks aCourtois, 1988) and FGF2 increases the number of opspositive cells (Hicks and Courtois, 1992). Significantly, FGFare able to transform the differentiated RPE into NR (Guillemand Cepko, 1992; Park and Hollenberg, 1991). These data hled to the suggestion that FGF might be the true inducer of NR. However, when FGF2 is removed from the SE, the OVstill patterned into NR and RPE domains (Pittack et al., 199thus, another component(s) of the SE may be responsibleOV patterning. Previous work on the commitment andifferentiation of the NR lineage has been based on in viassays or on transdifferentiation of RPE after optic cup staof development. There have been no experiments focusingearly steps of OV development in vivo. In the present study,in vivo assay has been established for examining the influeof FGF on patterning of the OV. Evidence is presented thatthe absence of SE, optic vesicle cells can initiate tdifferentiation program into both neural retina and pigmentepithelium, and that FGF alone, mimicking surface ectodercan drive the cells to organize into two distinct retinal domain

MATERIALS AND METHODS

Virus constructs and productionConstruction of the viral vectors and propagation of the recombinviruses have been described elsewhere (Mima et al., 1995; Itoh e1996). Biological activities of the FGF virus, the FGFR1 virus anthe antisense FGFR virus have been used previously to assess thof FGF signaling in cellular differentiation and proliferation withinsomitic mesoderm, limb mesenchyme, cardiogenic mesoderm coronary vasculature as described in Itoh et al. (1996), Mima et(1995) and Mikawa (1995).

MicroinjectionFertile White Leghorn chicken eggs (Truslow Farms, NJ) weincubated at 38°C until Hamburger-Hamilton (HH) stage 10 (1somites, ~1.5 days). A small hole was made in the egg shell andextraembryonic membranes overlying the embryo removed. Withglass capillary, approximately 1 nl of virus solution, at a concentratof 108 virions/ml, was pressure-injected using a picospritzer (GeneValve Co., NJ) into the selected area. Eggs were sealed with Paraand incubated until indicated stages.

ImplantationThe surface ectoderm surrounding the optic vesicle was surgicremoved at stage 10 using 1 µl of Nile blue sulfate (Sigma, 1.5% inwater). As described in Yang and Niswander (1995), this causeslight blistering of the ectoderm and facilitates the removal of Swithout disturbing the underlying neuroectoderm. Upon removal, tSE was replaced by FGF-producing fibroblasts. D17 fibroblaspreviously infected with a retrovirus encoding FGF1 and β-galactosidase (β-gal), were plated on polyester membranes (Costcut to 0.45×0.70 mm. Membranes were placed onto the optic vesiat various sites. The average embedded membrane held ~102 cells and1-10 ng of FGF was secreted from 106 infected fibroblasts per day(Itoh et al., 1996), as assayed by immunoblot and dosage studiePC12 neurite outgrowth (Togari et al., 1985). Eggs were sealed wParafilm and reincubated until the indicated stage.

Fs,1).R1

ling1ndin-sotavethe is7); fordtroges on annce, inheedm,s.

antt al.,d

e role

and al.

re0

the a

ionralfilm

ImmunohistochemistryEmbryos were removed from eggs, fixed overnight at 4°C in 2paraformaldehyde, stained with X-gal at 37°C overnight and thprocessed for cryo- or paraffin sectioning. Hu mouse monoclonantibody (Kostyk et al., 1996; Marusich et al., 1993), a generous from Dr Henri Furneaux (SKI, New York) and Dr Michael Marusic(Univ. of Oregon, OR) was used on cryosections at a dilution of 1:5Anti-β-galactosidase rabbit polyclonal antibody (Cappel) was used10 µm cryosections at a dilution of 1:500. Secondary anti-mouse aanti-rabbit antibodies conjugated to indicated fluorochrom(American Qualux) were used at a dilution of 1:500. RA4 moumonoclonal antibody (McLoon and Barnes, 1989), a generous from Dr Steven McLoon (Univ. of Minneapolis, MN) was used on 1µm paraffin section at a dilution of 1:500, together with goat anmouse secondary conjugated to alkaline phosphatase (Sigma) BCIP/NBT (Calbiochem) as chromogen.

RESULTS

The NR and PE are patterned prior to optic cupformationIn vitro, RPE is competent to transdifferentiate into NR tissin response to FGF. Paradoxically, shortly after optic cformation, the NR itself begins to secrete FGFs in vivo, yet tRPE continues to develop as it would in the absence of FGThese results lead to the hypothesis that, in the optic cup,juxtaposed NR inhibits the RPE from adopting a NR cell fatimplying that NR and RPE domains are established in the opcup stage of development. Alternatively, in the optic cup, RPmight not see or respond to FGF secreted from the juxtapoNR. We asked whether the RPE is capable of forming NRthe context of an intact optic cup by introducing a constitutiFGF signal into the forming optic cup. This was accomplishby using a replication-defective retrovirus to express FGligand or the FGFR gene with a reporter β-gal. To control fornon-specific effects of viral infection, a vector expressing onβ-gal was used in parallel experiments.

Viruses were microinjected into the presumptive RPE zonethe OV at E1.5 (HH stage 9) and the eyes examined at E7 (2A). Staining with X-gal identified the infected cells (Fig. 2B,CInfection with β-gal virus (Fig. 2B) gave rise to blue sectors oX-gal staining in the eye; FGF1-producing virus, which producectopic sources of FGF within the RPE, gave rise to eyes wsectors lacking pigmentation around sites of viral infection (F2C). β-gal virus had no apparent effect on the fate of these cesections showed that, when the presumptive RPE was infecpigmented cells were unchanged in number or distribution (F

871FGF patterns the optic vesicle

d

A B C

IH

D

GF

E

iNR

NR

ig. 2. Effect of RPE-supplied FGF1 on transdifferentiation in intact cup. (A) Diagram ofperation; FGF-encoding retrovirus is added to presumptive RPE. (B) Whole-mount X-gal stainingf an E7 embryo after infection of presumptive RPE with control virus, encoding β-gal alone.C) Whole-mount X-gal staining after infection of presumptive RPE with FGF-encoding virus.D-G) All cryosections are through optic cups of E7 embryos and double-stained for Hu (red) and-gal (green). (D,E) Section through optic cup of embryo infected with control virus in theresumptive RPE (D) or NR (E). (F) Section through optic cup of embryo as in (C), iNR denotes

he induced NR. (G) Section through optic cup of embryo infected in the mesoderm posterior to theresumptive RPE. (H) Induced NR, stained for RA4, arrow points towards stained axons of retinalanglion cells. As this section captures the anterior of the optic cup, the true NR does not have anyanglion cells and therefore is not stained. (I) RA4 staining showing emergence of RA4-positiveells; X-gal staining identifies the FGF virus infected cells. NR, the neural retina; PE, the retinaligmented epithelium.

2D). Retroviral infection of thepresumptive NR with the β-galvector also had no apparent effect(Fig. 2E) on neuronal numbers ordistribution. Sections through FGF-infected eyes, however, revealedthat zones lacking pigment (Fig. 2C)consisted of a double-layered NR(NR and iNR) as evidenced bylocalization of Hu protein (Fig. 2F),an RNA-binding protein expressedin neurogenic cells prior to overtdifferentiation. Hu protein wasnever seen in the RPE or in any non-neuronal tissues (Fig. 2D), makingit an ideal marker for identifyingcells of the neuronal lineage withinthe developing retina.

Infection of presumptive NR withthe FGF-producing virus did notinduce the adjacent RPE todifferentiate into NR (data notshown). In contrast, infection of themesoderm underlying thepresumptive RPE domain with FGFvirus transformed the RPE into NR-expressing Hu (Fig. 2G). Infectionof the RPE with FGFR-encodingvirus did not alter the developmentalfate of the RPE and the tissue wasidentical to control tissue infectedwith the β-gal vector (data notshown). Using RA4 as a marker ofdifferentiated cell types, such asganglion cell bodies and axonalprocesses, we examined the induceNR. RA4 immunoreactivity wasstrongly positive in the induced NR(Fig. 2H), but was reduced in thesame tissue when in close proximityto the RPE (Fig. 2I).

We conclude that the RPE iscompetent to form NR in thepresence of endogenous NRduring normal development.However, the RPE does notrespond to FGF present in the NReven when supplied with highlevels of receptor. These data donot support the hypothesis thatinhibition of NR differentiation inthe RPE is caused by endogenousNR since RPE can be transformedinto NR when FGF is suppliedfrom the underlying mesoderm.Since this second source of FGFcould be mimicking the SE, wehypothesized that FGF comingfrom the SE might direct theformation of the NR. This impliesthat, by the optic cup stage of

Foo((βptpggcp

872

e a

ateyeiondnrer

ed

J. Hyer, T. Mima and T. Mikawa

A B

C D

E F

NR

PE

Fig. 3.Effect of ablating surface ectoderm on eye morphogenesis. (A) Whole-mount view of a day 7 embryo, showing overall brain and eyeformation. (B) An embryo in which the surface ectoderm was removed at E1.5, showing dramatic and specific reduction of the eye (arrow).(C) Section through a normal optic cup of an E7 embryo, showing the correct relationship between columnar NR and epithelial RPE tissues.(D) Section through embryo treated as in B, showing a pigmented tissue in a columnar array (arrowheads, position of Hu-positive cells as inF). (E) Hu staining in a normal retina stains various cell types in the neural retinal layer (arrow); the RPE is not stained. (F) Hu staining oftissue section in D where Hu-positive cells (arrowheads) are present within the pigmented retinal layer. OT, the optic tectum; NR, the neuralretina; PE, the retinal pigmented epithelium. Bars represent 1.0 mm.

development, delineation of NR and RPE domains haalready been established.

The surface ectoderm is required for optic vesiclepatterningTo test the hypothesis that SE determines the location ofNR, we removed the SE at E1.5, before it makes contact wthe OV (Fig. 3A). The resulting embryos were examined at EE5 and E7. Examination in whole mount of the operatembryos revealed that the effects of ablation were restricteeye development; overall facial and brain morphogenesis

ve

theith4,

edd towas

indistinguishable from unoperated embryos (Fig. 3B,C). In thabsence of the SE, the OV in most cases developed intopigmented vesicle (86% of embryos, n=14). These remnantswere severely reduced in size and had no lens, indicating ththe SE had been successfully ablated. Sections through the remnant revealed that such eyes lacked the normal organizatof RPE and NR tissues (Fig. 3D) and failed to form bilayereoptic cups (Fig. 3E). Importantly, the cells of OV displayed aindeterminate phenotype. Retinal cells in these embryos wepigmented but, unlike the normal RPE, they were no longearranged as a simple cuboidal epithelium. Instead, they form

873FGF patterns the optic vesicle

he of

antsVle-

ls,erg.ngne in

d/or

the thickened retinal tissue typical of NR differentiation (F3D). Hu+ cells were sequestered within this pigmented thickened retinal layer (Fig. 3F). Therefore, in the absencethe SE, the OV had lost patterning information but the capabto initiate the differentiation into NR and RPE was preserve

FGF organizes the optic vesicle into NR and RPEdomainsSE could be playing three major roles in this early stage of development: (1) regulating the proliferation of cells forming tretina, (2) inducing optic cup formation or (3) organizing RPand NR domains. To determine what role FGF1 is playing,introduced a new source of FGF1 ligand in the SE-ablated(Fig. 4A). FGF virus was injected into the mesoderm posteto the OV, adjacent to presumptive RPE. The infected cells co-expressed FGF1 and β-gal were identified by X-gal staining(data not shown). The resulting embryos were examined at

An optic cup was never formed. The OV exhibited clea

stage 10

B

E

A

F

C

PE

PE

lumen

lumen

Fig. 4. Effect ofmesodermally-located FGF1on NR differentiation.(A) Diagram of operation;FGF-encoding retrovirus isadded to the mesoderm.(B-D) Examples of treatedembryos at E5, the extent ofeye remnant previously incontact with the virallyinfected cells is indicated bydashed lines. (E) Sectionthrough the eye remnant ofone treated embryo,showing the transition frompigmented to unpigmentedzones, “lumen” signifies theenduring primary opticcavity, since these eyeremnants did not form aneye cup or true vitreouscavities (arrowheads,position of Hu-positive cellsas in F). (F) Hu staining ofsame section showingemergence of Hu-positiveNR cells (arrowheads). NR,the neural retina; PE, theretinal pigmentedepithelium. Bars represent0.2 mm.

ig.and of

ilityd.

eyeheE

we eyeriorthat

E5.rly

delineated pigmented and non-pigmented domains, with tnon-pigmented domains nearest the mesodermal sourceexogenous FGF (55%, n=9) (Fig. 4B-D). In controlexperiments, where virus encoding only β-gal was injected,these effects were never seen. Sections through eye remnin embryos infected with the FGF virus showed that the Ohad not formed a bilayered optic cup but remained as a singlayered vesicle. β-gal protein was not found in NR tissue. Thepigmented region had a cuboidal pigmented layer of celcharacteristic of RPE. The adjacent non-pigmented layexhibited a columnar arrangement of cells typical of NR (Fi4E); at a distance from the transition point, strong Hu stainiwas observed (Fig. 4F), indicating that the unpigmented zohad been committed as NR. Hu staining was never robustareas adjacent to pigmentation.

FGF patterns through short-range interactionsWe removed the SE and placed an FGF signal at the distal an

ablate SE inject

D

NR

NR

874

ed

liedo

of

4t

hator

J. Hyer, T. Mima and T. Mikawa

medial positions of the OV (Fig. 5A). This procedure providan FGF source that more closely mimicked the normal releasFGF ligand from SE by being spatially removed from the OV.also eliminated the possibility that FGF-producing mesodermcells might have been secreting unknown active factconsequent to infection. Fibroblasts co-expressing FGF1 anβ-gal were plated onto polyester membranes and these implaonto the exposed OV. Cells expressing only β-gal were used ascontrols for non-specific effects of implantation. Whole-mouX-gal staining of embryos revealed that β-gal-positive cellsremained localized to the site of implantation (data not showControl implants of cells expressing only β-gal did not rescue thedisarray in eye formation that results after SE ablation (100%embryos, n=8) (Fig. 5B). Implantation of FGF-producing cells asingle sites on the exposed OVs gave rise to eyes that were a

stage 10

B

E F

H

A

CFig. 5. Effect ofexogenously suppliedFGF on NRdifferentiation.(A) Diagram of operation;FGF-expressing cells areimplanted at distal tip ofnaked optic vesicle. (B) Incontrol embryos, the opticvesicle forms a uniformlypigmented vesicle.(C,D) Examples of treatedembryos at E7, asteriskindicates the site ofimplanted cells.(E-G) Examples ofembryos treated with twoimplants so that entireoptic vesicle is convertedto a NR cell fate. (H) RA4staining of a normal opticcup, arrow indicates thedirection of cell migrationand asterisk indicatesRA4 staining of cellbodies, lumen is thevitreal cavity. (I) RA4staining of sectioned eyeremnant of G, arrow ofcell migration points inopposite direction, sincethe NR did not form anoptic cup, lumen is theenduring primary opticcavity, asterisk indicatesRA4 staining of cellbodies. NR, the neuralretina; PE, the retinalpigmented epithelium.Bars represent 0.2 mm.

ede of Ital

orsd nted

nt

n).

oftgain

divided by a sharp boundary into pigmented and non-pigmentzones (41% of embryos, n=22) (Fig. 5C,D). In all cases, the non-pigmented zone was restricted to the site where FGF was suppto the OV. Importantly, when cells were placed at two sites sthat FGF was supplied to both anterior and posterior regionsthe OV, all pigmentation was lost (59% of embryos, n=22) (Fig.5E-G). In the expanded non-pigmented layer, RAimmunoreactivity could be detected (Fig. 5H,I), indicating thathe non-pigmented zone had differentiated as NR.

DISCUSSION

FGF1 organizes the NR domain in the OVIn vitro studies, using explant assays, have demonstrated tthe single cells of the OV are poised to differentiate as RPE

implantablate SE

G

I

D

875FGF patterns the optic vesicle

her.GFedg toWeior,ysGFly

ueas theellr

ethat

.,eh,onay

wollyct-

siss

t

l.,:

nalne2sed in

wild-type -SE -SE+FGF -SE+2X FGF

Fig. 6 FGF signal refines NR field todistal tip of OV. Schematicpresentation of OV as a circle. Duringnormal development, the entire OV iscompetent to differentiate as NR, butFGF coming from the SE ensures thatonly the distal tip will do so (yellow).Although the RPE has potential todifferentiate as NR (purple andyellow), it does not see FGF signalingand will therefore slowly lose thispotential. When the SE is removed,no NR domain is established and theOV is unpatterned; the RPE and NR lose their distinct identities. By adding back the FGF signal, the pattern is re-established with the NRdomain in closest proximity with the new source of FGF. We show that the portion of the OV out of reach of the new FGF signal will lose thepotential to realize a NR fate. That the entire OV is competent to form NR is demonstrated by adding two sources of FGF, which specifies theentire OV as NR.

NR and are not fixed at the early OV stage (Lopashov, 196During in vivo development, the OV is divided into RPE anNR domains at an early stage. Here we show that removingSE causes a loss of the RPE/NR boundary. The OV becodisorganized, and individual cells of the OV exhibit bopigmentation phenotype of the RPE and the tissue organizaphenotype of the NR. This randomization of the OV reminiscent of prior results with explants, which demonstratthat the OV can initiate differentiate into both NR and RPwithout SE (Lopashov, 1963 p.11-25). Our results do nsupport the concept that SE is the inducer of the NR, ratthey reinforce in vitro findings that all of the cells of the Oare bipotential and able to form both RPE and NR withoinstructive signals from the SE.

Although FGF2 has been demonstrated to induce a neurophenotype in cultured PC12 cells (Rydel and Greene, 198FGF2 has only been shown to influence the differentiationselected cell types of the NR in the developing vertebrate (Zhao et al., 1996; Hicks and Courtois, 1992). A series recent experiments support the notion that FGF is not inducer of NR in vivo. McFarlane et al. (1996) transfecteddominant negative FGF receptor into the early eye field of Xenopusembryo and found that the transgenic cells could fudifferentiate. Pittack et al. (1997) applied FGF2-neutralizinantibodies to organ culture explants of embryonic chicken Oand found that NR and RPE formed distinctly but that the Ndid not express differentiated markers. The ability of the Oto spontaneously differentiate into NR tissue in explacultures (Lopashov, 1963, p12) leads to the conclusion thatthe time the OV is formed, an inducer of the NR is nnecessary. One view considers the process of induction duembryonic development to be the result of cumulative tissinteractions arising from several tissues (Jacobson, 1966)the OV, induction of the NR by SE may be viewed as only tlast step of a sequential series of inductions, one of whmight involve FGF signaling. Although SE may produce othfactors in addition to the FGFs that participate in patterning OV, our data indicate that FGF alone substitutes for torganizational function of the SE.

In our experiments, two different approaches producidentical results. The OV is always organized such that the domain is closest to the FGF signal. When the SE is remoand the mesoderm underneath the OV produces FGF, theis organized such that the NR domain is oriented towards

3).d themesthtionisedEotherVut

nal7),

ofeyeofan athellygVsRVnt, byotringue. Inheicherthehis

edNRved OVthe

FGF source in the mesoderm. RPE always formed on the otside of the OV away from the FGF-producing mesodermWhen the SE is removed and exogenous cells expressing Fare placed on the OV, again the OV differentiates in an orientfashion with NR closest to the FGF source. Implantinmembranes carrying FGF-secreting fibroblasts enabled usassess FGF responsiveness in different regions of the OV. found no variations in FGF responsiveness between anterposterior or lateral sectors of the OV; the NR domain alwadeveloped nearest the source of the FGF. If two sources of Fwere applied to the OV, the RPE domain was lost completeand the entire OV differentiated as NR.

An interesting observation from the current study is that Hand RA4, the markers of the NR, are always observed in arseparated from pigmentation and are not expressed nearNR/RPE boundary. The results suggest that critical NR cnumbers and/or NR-intrinsic factors are required fodetermination (Hu) and differentiation (RA4) of the NR. Thpatterns of Hu and RA4 expression supports the hypothesis an intrinsic retinal growth factor is needed for thedetermination and differentiation of the NR (Altshuer et al1993; Watanabe and Raff, 1992; Sparrow et al., 1990; R1992), and might also indicate the presence of some inhibitiof NR by PE in that transition zone. Such transition zone mcorrespond to the presumptive ciliary body.

The cells of the OV respond to FGF1 in a short-range paracrine fashionIn the above experiments, complete loss of RPE required timplanted FGF sources. This short-range FGF signaling furecapitulated the expected action of the SE. Thus, the contadependent nature of NR patterning during morphogeneinsures that NR forms at the distal tip of the OV. Inductionprior to OV formation most likely endow the cells of the OVwith the potential to form NR. FGF signaling reinforces thapotential in a specific region of the OV, its distal tip. FGF fromthe SE is deposited in the extracellular matrix (Jeanny et a1987), and limitations in diffusion of FGF ligand requires OVSE contact to transmit requisite developmental cues.

The viral construct that we use to express FGF has a sigsequence derived from FGF4 upstream of the FGF1 ge(Forough et al., 1993). It is still uncertain how FGF1 and FGFare secreted from cells, as they lack signal peptides. When uto infect the tissue of the OV, the RPE cells respond to FGF

876

n-

the

en.

irnds

rsl

of

.

e

an

ly

b

t.

nt

e

t

J. Hyer, T. Mima and T. Mikawa

a paracrine fashion; they themselves are not infected respond to infected neighbors which secrete FGFs. Tparacrine response is always local and restricted to immediate area of infection, again indicative of a short-ransignal.

The RPE domain is competent to respond to FGFbut may not see the signal in embryoThere is a substantial literature demonstrating ttransdifferentiative properties of RPE in vitro (reviewed Guillemot and Cepko, 1992; Park and Hollenberg, 1991). asked a different question: what conditions define tdifferential commitment of RPE and NR in vivo? Our resulindicate that, when high levels of FGFRs are activated in viRPE does not respond any differently to endogenous sourcof FGF. These results suggest that differences in RPE andcommitment are not based on FGF receptor levels. The limitcomponent seems to be the FGF ligand itself. This is surprissince prior experiments have shown that, when NR and Rtissues are prevented from contacting one another placement of an obstruction between the two cell populatioRPE developed into NR nearest the obstruction (Orts-Llorcaal., 1960), suggesting the presence of a local inhibitor.

Molecular mechanism of NR domain determinationCells of the OV are not committed to a NR and PE fate eain development. Plasticity of these cells has been demonstrwell after optic cup formation with amphibian tissue. Ichicken embryonic optic cups, plasticity of cell fate is seen to embryonic day 3 (Dorris, 1938). Even after the NR becomfixed in its fate, RPE cells remain plastic (Lopashov, 196Coulombre and Coulombre, 1965; Zhao et al., 1995).

In this present report, we show that the NR domainreinforced at the distal tip of the OV by FGF signaling. Opossible mechanism for specifying the NR domain is throuthe differential expression of genes between RPE and domains. FGF signaling might specify the NR domain driving the expression of the specific transcription factofound in the OV, either inducing or inhibiting the expressioof the appropriate factors in the NR. While still at the OV stagthe presumptive NR expresses Msx-2 (Monaghan et al., 19Chx-10 (Liu et al., 1994), Rx (Mathers et al., 1997) and Pa2 (Nornes et al., 1990), all of which are absent from adjacprospective RPE. Other homeodomain proteins are expresthroughout the OV and then become restricted to either [Six-3 (Oliver et al., 1995), Pax-6 (Grindley, et al., 1995)] to the RPE [Otx-1 (Bonicello et al., 1993) and Otx-(Bovolenta et al., 1997)] by the time optic cup forms. Otxshows a clear down-regulation in the presumptive NR eventhe early OV stage. In one transdifferentiation assexogenous FGF2 in the media induces expression of Chxfrom RPE tissue (Zhao et al., 1997).

We propose a model in which the entire OV is competenform NR (Fig. 6). As the distal OV contacts the SE, FGF fistabilizes NR cell fates in that region of the OV. Proximal zonof the OV, without reinforcement, slowly lose the potential realize a NR fate and eventually become “fixed” as RPE. Tmodel can be tested since the downstream consequenceFGF on gene transcription in the eye can be analyzConsistent with this model is our demonstration that, wherethe OV sees FGF, a NR cell fate is chosen. As the NR has

buthisthege

heinWehetsvo,e(s) NRingingPEbyns, et

rlyatednupes3;

isneghNRbyrsne,

91),x-

entsed

NRor2-2 at

ay,-10

t torstestohiss ofed.ver no

significance to the organism unless it forms in proper relatioto the rest of the eye, FGF signaling, by refining the NRcompetent field, assures that the NR only differentiates at distal tip of the OV.

The authors thank Henri Furenaux, Michael Marusich and StevMcLoon for their generous sharing of antibodies; D. A. Fischman, DHerzlinger, J. Sparrow, A. Alvarez-Buylla and J. Kuhlman for thecomments and discussions on this study and Ms L. Miroff, L.Ong aL.Cohen-Gould for their technical assistance. This work wasupported in part by grants from the NIH, and the MatheFoundation. T. M. is an Irma T. Hirschl Scholar. This work is in partiafulfillment of the doctoral degree of J. H. in the Graduate School Medical Sciences of Cornell University.

REFERENCES

Altshuler, D., LoTurco, J. J., Rush, J. and Cepko, C.(1993) Taurinepromotes the differentiation of a vertebrate retinal cell type in vitroDevelopment 119, 1317-1328.

Bonicello, E., Gulisano, M. and Broccoli, V.(1993). Emx and Otx homeoboxgenes in the developing mouse brain. J. Neurobiol. 24, 1356-1366.

Bovolenta, P., Mallamaci, A., Briata, P., Corte, G. and Boncinelli, E.(1997). Implication of OTX2 in pigment epithelium determination andneural retina differentiation. J. Neuroscience 17, 4243-4252.

Coulombre, J. L. and Coulombre, A. J.(1965). Regeneration of neuronalretina from the pigmented epithelium in the chick embryo. Dev. Biol. 12,79-92.

DeIongh, R. and McAvoy, J. W.(1993). Spatio-temporal distribution of acidicand basic FGF indicates a role for FGF in rat lens morphogenesis. Dev. Dyn.198, 190-202.

Detwiler, S. R. and VanDyke, R. H. (1954). Further experimentalobservations on retinal inductions. J. Exp. Zool. 126, 135-155.

Dorris, F. (1938). Differentiation of the chick eye in vitro. J. Exp. Zool. 78,385-415.

Dragomirov, N. I. (1937). The influence of the neighboring ectoderm on thorganization of the eye rudiment. Dokl. Akad. nauk 15, 61-64.

Furough, R., Zhan, X., MacPhee, M., Friedman, S., Engleka, K. A.,Sayers, T., Wiltrout, R. H. and Maciag, T. (1993). Differentialtransforming abilities of non-secreted and secreted forms of humFibroblast Growth Factor-1. J. Biol. Chem. 268, 2960-2968.

Grindley, J. C., Davidson, D. R. and Hill, R. E.(1995). The role of Pax-6in eye and nasal development. Development 121, 1433-1442.

Guillemot, F. and Cepko, C. L. (1992). Retinal fate and gagnlion celldifferentiation are potentiated by acidic FGF in an in vitro assay of earretinal development. Development 114, 743-754.

Hicks, D. and Courtois, Y.(1988). Acidic fibroblast growth factor stimulatesopsin levels in retinal photoreceptor cells in vitro. FEBS Lett. 234, 475-479.

Hicks, D. and Courtois, Y. (1992). Fibroblast growth factor stimulatesphotorecptor differentiation in vitro. J. of Neuroscience 12, 2022-2033.

Itoh, N., Mima, T. and Mikawa, T. (1996). Loss of fibroblast growth factorreceptors is necessary for terminal differentiation of embryonic limmuscles. Development 122, 291-300.

Jacobson, A. G. (1966). Inductive processes in embryonic developmenScience 152, 25-34.

Jeanny, J.-C., Fayein, N., Moenner, M., Chevalier, B., Barritault, D. andCourtois, Y. (1987). Specific fixation of bovine brain and retinal acidic andbasic fibroblast growth factors to mouse embryonic eye basememembranes. Exp. Cell Res. 171, 63-75.

Kostyk, S. K., Wheeler, E. L. and Wagner, J. A.(1996). Unusual expressionof the HU paraneoplastic antigen in the visual system. Neuro Report 7, 1549-1552.

Liu, I. S. C., Chen, J., Ploder, L., Vidgen, D., van der Kooy, D., Kalnis, V.I. and McInnes, R. R.(1994). Developmental expression of a novel murinehomeobox gene (Chx10): Evidence for roles in determination of thneuroretina and inner nuclear layer. Neuron 13, 377-393.

Lopashov, G. V.(1963). Developmental Mechanisms of Vertebrate Rudimen.New York: Pergamon Press.

Lopashov, G. V. and Stroeva, O. G.(1961). Morphogenesis of the vertebrateeye. Adv. Morphol. 1, 331-377.

877FGF patterns the optic vesicle

m

n

in

n

n

ype

al

l

Marusich, M. F., Furneaux, H. M., Henion, P. D. and Weston, J. A.(1993).Hu neuronal proteins are expressed in proliferating neurogenic cellsJ.Neurobiology 25 143-155.

Mathers, P. H., Grinberg, A., Mahon, K. A. and Jamrich, M. (1997). TheRx homeobox gene is essential for vertebrate eye development. Nature 387,603-607.

McAvoy, J. W. and Chamberlain, C. G.(1990). Growth factors in the eye.Prog. Growth Factor Res. 2, 29-43.

McFarlane, S., Cornel, E., Amaya, E. and Holt, C. E.(1996). Inhibition ofFGF receptor activity in retinal. Neuron 17, 245-254.

McLoon, S. C. and Barnes, R. B.(1989). Early differentiation of retinalganglion cells: An axonal protein expressed by premigratory and migratretinal ganglion cells. J. Neuroscience 9, 1424-1432.

Mikami, Y. (1939). Reciprocal transformations of the parts in the developieye-vesicle, with special reference to the inductive influence of leectoderm on the retinal differentiation. Zool. Mag. (Japan) 51, 253-256.

Mikawa, T. (1995). Retroviral targeting of FGF and FGFR in cardiomyocytand coronary vasculogensis: Discontinuous formation of coronary vessAnn. NY Acad. Sci. 752, 506-516.

Mima, T., Ueno, H., Fischman, D. A., Williams, L. T. and Mikawa, T.(1995). Fibroblast growth factor receptor is required for in vivo cardiproliferation at early embryonic stages of heart development. Proc. Natn.Acad. Sci., USA 92, 467-471.

Monaghan, A. P., Davidson, D. R., Sime, C., Graham, E., Baldock, R.,Bhattacharya, S. S., and Hill, R. E.(1991). The Msh-like homeobox genesdefine domains inthe developing vertebrate eye. Development 112, 1053-1061.

Nornes, H. O., Dressler, G. R., Knapik, E. W., Deutsch, U. and Gruss, P.(1990). Spatially and temporally restricted expression of Pax2 during murneurogenesis. Development 109, 797-809.

Oliver, G., Mailhos, A., Wehr, R., Copeland, N. G., Jankins, N. A. andGruss, P. (1995). Six3, a murine homologues ofthe sine oculis gendemarcates the most anterior border of the developing neural plate anexpressed during eye development. Development 121, 4045-4055.

Orts-Llorca, F. and Genis-Galvez, J. M.(1960). Experimental production ofretinal septs in the chick embryo. Differentiation of pigment epithelium inneural retina. Acta Anat. 42, 3170.

.

ing

ngns-

esels.

ac

ine

e,d is

to

Park, C. M. and Hollenberg, M. J.(1989). Basic FGF growth factor inducesretinal regeneration in vivo. Dev. Biol. 134, 201-205.

Park, C. M. and Hollenberg, M. J. (1991). Induction of retinal regenerationin vivo by growth factors. Dev. Biol. 148, 322-333.

Pittack, C., B., Grunwald, G. B. and Reh, T. A.(1997). Fibroblast growthfactors are necessary for neural retina but not pigmented epitheliudifferentiation in chick embryos. Development 124, 805-816.

Reh, T. A. (1992). Cellular interactions determine neuronal phenotypes irodent retinal cultures. J. Neurobiology 23, 1067-1083.

Rydel, R. E. and Greene, L. A.(1987). Acidic and basic fibroblast growthfactors promote stable neurite outgrowth and neuronal differentiation cultures of PC12 cells. J. Neuroscience 7, 3639-3653.

Sparrow, J. R., Hicks, D. and Barnstable, C. J.(1990). Cell commitmentand differentiation in explants of embryonic rat neural retina. Comparisowith the developmental potential of dissociated retina. Dev. Brain Res. 51,69-84.

Stroeva, O. G.(1960). Experimental analysis of the eye morphogenesis imammals. J. Embryol. Exp. Morph. 8, 349-368.

Tcheng, M., Fuhrman, G., Hartman, M-P., Courtois, Y. and Jeanny, J.-C.(1994). Spatial and temporal expression patterns of FGF receptor genes t1 and type 2 in the developing chick retina. Exp. Eye Res. 58, 352-358.

Togari, A., Dickens, G., Kazuya, H. and Guroff, G.(1985). The effect offibroblast growth factor on PC12 cells. J. Neurosci. 5, 307-316.

Wanaka, A., Milbrand, J. and Johnson Jr., E. M.(1991). Expression of FGFreceptor gene in rat development. Development 111, 455-468.

Watanabe, T. and Raff, M. C.(1992). Diffusible rod-promoting signal in thedeveloping rat retina. Development 114, 899-906.

Yang, Y. and Niswander, L. (1995). Interaction between the signalingmolecules WNT7a and SHH during vertebrate limb development: Dorssignals regulate anteroposterior patterning. Cell 80, 939-947.

Zhao, S., Thornquist, S. C. and Barnstable, C. J.(1995). In vitrotransdifferentiation of embryonic rat retinal pigment epithelium to neuraretina. Brain Res. 677, 300-310.

Zhao, S., Rizzolo, L. J. and Barnstable, C. J.(1997). Differentiation andtransdifferentiation of the retinal pigment epithelium. Inter. Rev. Cytology171, 225-266.