Retinoid-Dependent Gene ExpressionRegulates Early Morphological Events inthe Development of the Murine Retina

DEBORAH. L. STULL* AND KENNETH. C. WIKLER

Section of Neurobiology, Yale School of Medicine, New Haven, Connecticut 06520

ABSTRACTEndogenous retinoids have been implicated in the axial patterning of the embryonic verte-

brate retina; however, no studies have directly examined how asymmetric retinoid-dependentgene expression regulates early morphological events in the development of the retina. Here weused a line of indicator mice that possess a retinoid-dependent transgene to examine therelationship between retinoic acid (RA)-dependent gene expression and events occurring duringearly eye morphogenesis, such as the closure of the optic disc. We found that retinoid-regulatedgene expression shifts along the dorsal/ventral axis of the embryonic retina; at embryonic day (E)E11.5 transgene expression is restricted to the neuroepithelium in dorsal retina, and by E14.5only immature cells located in ventral retina and the dorsal retinal margins demonstratetransgene activation. By manipulating RA levels, we were not only able to systemically alterRA-dependent gene expression along the dorsal/ventral axis, but also to affect retinal morphol-ogy. In particular, reducing RA availability resulted in the abnormal closure of the optic fissure.These results indicate that asymmetric levels of RA regulate early RA-dependent gene expressionin the eye and demonstrate that the normal pattern of retinoid-dependent gene transcriptionalong the dorsal/ventral axis is critical for the proper development of the vertebrate retina. J.Comp. Neurol. 417:289–298, 2000. © 2000 Wiley-Liss, Inc.

Indexing terms: retinoic acid; retinogenesis; transgenic

Retinoic acid (RA), a derivative of vitamin A, has beenshown to regulate axial patterning during the develop-ment of numerous systems in species ranging from chickto human (reviewed in Means and Gudas, 1995). For ex-ample, studies examining the effects of a vitamin A defi-cient diet on vertebrate embryogenesis found that elimi-nating RA in utero results in abnormalities in the axialdevelopment of the eye (Warkany and Schraffenberger,1946; Wilson et al., 1953; Abramovici et al., 1978). Inaddition, ocular and retinal malformations are also ob-served after RA levels are perturbed during retinogenesisin the zebrafish; for example, depleting RA levels resultsin the elimination of the ventral retina (Marsh-Armstronget al., 1994). These studies indicate that proper levels ofRA are important for normal retinal development. How-ever, they do not address the mechanisms that underlieRA-regulation of morphological events during the devel-opment of the vertebrate retina.

One possible mechanism for RA-regulation has beensuggested by studies examining rates of RA synthesisalong the dorsal/ventral axis of the developing retina.These studies provide evidence that RA availability iscompartmentalized along this axis during early retinal

development; enzymes that synthesize RA from retinalde-hyde are asymmetrically expressed along the dorsal/ventral axis and synthesize RA at different rates through-out development (McCaffery et al., 1991, 1992, 1993).These findings predict that RA regulates early retinalmorphogenesis by being differentially available to dorsaland ventral retina. An alternative hypothesis, suggestedby work done in zebrafish (Hyatt et al., 1992; Marsh-Armstrong et al., 1994), proposes that RA specifically reg-ulates the development of ventral retina, with the devel-opment of dorsal retina representing a default pathway.Neither of these possibilities, however, has been fully ex-amined as a potential mechanism underlying retinal de-velopment in the mouse. For example, these studies do notexamine whether there is functional significance, such asasymmetry in RA-mediated gene transcription, to the pro-

Grant sponsor: NIH; Grant number: EY09917.* Correspondence to: Dr. Deborah Stull, Department of Molecular and

Cellular Biology, Harvard University, 16 Divinity Avenue, Cambridge, MA02138. E-mail: stull@fas.harvard.edu

Received 17 June 1999; Revised 4 October 1999; Accepted 5 October 1999

THE JOURNAL OF COMPARATIVE NEUROLOGY 417:289–298 (2000)

© 2000 WILEY-LISS, INC.

posed asymmetry in RA availability. In addition, the po-tential connection between the proposed asymmetry in RAlevels and the development of one compartment over an-other along the dorsal/ventral axis of the developing mu-rine retina remains unexplained.

Here we have examined the pattern of retinoid-dependent gene expression in the neuroepithelium of theembryonic murine retina during the formation of the opticcup using indicator mice that possess a retinoid-dependent transgene (Balkan et al., 1992). In addition, wehave examined the effects of perturbing RA availability onboth the early morphological events involved with thedevelopment of the dorsal/ventral axis of the murine ret-ina and the regional distribution of RA-dependent geneexpression. If RA availability is critical for the develop-ment of ventral retina, then we would expect RA-inducedchanges in RA-dependent gene expression to be correlatedwith changes in the morphological and antigenic develop-ment of ventral retina specifically. However, if RA avail-ability regulates the formation of the dorsal/ventral axisspecifically, then disrupting this gradient will disrupt themorphological and antigenic compartmentalization of ret-inal neuroepithelium into dorsal and ventral regions.

Our analysis of retinae from the indicator mice revealeda striking restriction of retinoid-dependent transgene ex-pression in the retinal neuroepithelium; during the forma-tion of the optic cup transgene expression is restricted todorsal retina, but shifts to ventral retina by the onset ofcellular differentiation. Additionally, perturbing RA levelsduring development resulted not only in changes to theregional expression of the retinoid-dependent transgene,but also in retinal malformations, such as retardedventral/temporal growth during the invagination of theoptic cup and the appearance of ectopic pieces of ventralretina. These results indicate that the asymmetric synthe-sis of RA along the dorsal/ventral axis is associated withasymmetric RA-mediated gene expression. In addition,our results indicate that the regional restriction of RA-dependent gene expression during early retinal develop-ment is involved in regulating the closure of the opticfissure, thus participating in the early establishment ofthe dorsal/ventral axis.

MATERIALS AND METHODS

Animals

CD-1 (Charles River Laboratories, Wilmington, MA) fe-male mice were mated overnight with male mice homozy-gous for an RA responsive element (RARE)-thymidine ki-nase promoter (tk)-lacZ (lacZ) transgene (RARE-tk-lacZ,Balkan et al., 1992). The presence of a vaginal plug in themorning was considered embryonic day (E) E0.5. To ex-amine the developmental pattern of retinoid-dependentgene expression RARE-tk-lacZ 1/- animals were sacrificedat E10, E11.5, E13.5, E14.5, E15.5, postnatal day (P) P0,or as adults. Five embryos or animals were examined foreach age.

To examine the effects of perturbing RA availability onearly retinal morphology CD-1 female mice were crossedwith RARE-tk-lacZ 1/1 males and gavage fed either all-trans RA (80 mg/kg in sunflower oil; Sigma Chemical Co.,St. Louis, MO) or citral (3,7-dimethyl-2, 6-octadien-1-al;3.0 g/kg in sunflower oil; Sigma Chemical Co., St. Louis,MO), which is a competitive inhibitor of aldehyde dehy-drogenases (Connor and Smit, 1987; Connor, 1988; Schuh

et al., 1993) at different developmental time points. Ani-mals were sacrificed at either E11.5, at the onset of cellu-lar differentiation, or E14.5, at the peak of cone genesis.Retinoic acid was administered fourteen hours prior tosacrifice (Colbert et al., 1993, 1995; Anchan et al., 1997)and citral was administered 4 days prior to sacrifice (An-chan et al., 1997); citral was administered at E7.5 forsacrifice at E11.5 and at E10.5 for sacrifice at E14.5.

Additional animals were treated at E7.5 with citral andsacrificed at either E9.5, E10.5, or E14.5 to examine theeffects of decreased RA availability on the formation of theoptic fissure (E9.5, E10.5) or on its closure (E14.5). Controlanimals received sunflower oil alone. Animals were fed atconsistent times to control for daily fluctuations in metab-olism. Five indicator embryos for each treatment group(citral at E9.5, E10.5, E11.5, E14.5; RA at E11.5, E14.5)were compared to control animals at those ages.

Immunohistochemistry

Mice were euthanized with an overdose of sodium pen-tobarbital (65 mg/kg, IP), as approved by Yale Universityto conform to NIH guidelines. To examine the adult dis-tribution of RA-dependent gene expression, eyes weremarked at the superior rectus to indicate orientation, enu-cleated, and placed in chilled phosphate buffered saline(PBS). Embryos and eyes were fixed in 4% paraformalde-hyde, cryoprotected through a graded sucrose series, andfrozen. To examine transgene expression serial coronalsections (12 mm) were immunoreacted for b-galactosidase(1:500; polyclonal; 5 Prime3 3 Prime, Inc., Boulder, CO).To analyze the axial development of the retina adjacentsections were labeled with polyclonal antibodies againstPax2 (1:200; G. Dressler, University of Michigan MedicalCenter), an early marker of embryonic ventral retina(Nornes et al., 1990), or aldehyde dehydrogenase class-1isoform (AHD-2; 1:3,500; J. Hilton, Johns Hopkins Uni-versity), a marker of embryonic dorsal retina (McCafferyet al., 1992). All sections were counterstained with bisBen-zimide (10 mM; Sigma Chemical Co., St. Louis, MO), anuclear stain. b-galactosidase labeling was detected byimmunofluorescence after incubation in a goat anti-rabbitantibody conjugated to cyanine (cy3; 1:200; Jackson Im-munoResearch Laboratories, Inc., West Grove, PA) andcoverslipped with Vectashield (Vector Laboritories, Bur-lingame, CA). AHD-2 and Pax2 labeling were detected byfirst incubating tissue in biotinylated secondary antibody(1:200; Vector Laboritories, Burlingame, CA) and then inan avidin-biotin-peroxidase complex (Vector Laboratories,Burlingame, CA). The horseradish peroxidase (HRP) wasvisualized with a 0.005% 3,3-diaminobenzidine HCl and a0.003% hydrogen peroxide (H2O2) solution. Images werephotographed using conventional photography and Kodak3200 black and white film. Composite images were thencomplied by scanning individual film negatives into AdobePhotoShop. Scanned images were minimally brightened tocompensate for scanning.

Morphometric analysis

Every second bisBenzimide labeled section through theretinae of control (n55), citral-treated (n55), and RA-treated (n55) embryonic eyes was measured using a Zeissmicroscope and Sony video camera coupled to a Macintoshcomputer equipped with an image grabbing board (Nu-vista) (Wikler et al., 1990). Retinal and optic disc lengthswere measured along both the dorsal/ventral and nasal/temporal axes and retinae were reconstructed to show the

290 D.L. STULL AND K.C. WIKLER

size, shape, and dorsal/ventral position of the optic stalk,as well as retinal size and shape. To determine if RA andcitral treatments affected the size and shape of the retinaand optic disc, the difference of means (F-test) and anal-ysis of variance tests were performed on these measure-ments. A P-value of less than 0.05 was considered to bestatistically significant (Tables 1 and 2).

Optic cup reconstructions

Every third bisBenzimide stained section through theretinae of control (n55), citral-treated (n55), and RA-treated (n55) eyes at E10.5 (control and citral-treated),E11.5 (control, citral-treated, and RA-treated), and E14.5(control and citral-treated) was acquired on the computersystem described above. These images were outlined andthen rendered into three-dimensional structures using im-aging software from Imaging Research, Inc. This software,MCID, which runs on a PC using the Windows NT oper-ating system, allows the alignment of consecutive sectionsin order to reconstruct a three-dimensional structure. Theschematic retinal sections depicted in Figures 5 and 6represent a subset of the outlines made from these images,while the schematic whole retinae represent the renderedthree-dimensional images.

RESULTS

Normal RA-dependent gene expressionshifts from dorsal to ventral retina

Previous studies have suggested that asymmetric levelsof RA are critical for the proper development of the dorsal/ventral axis of the vertebrate retina (McCaffery et al.,1992, 1993; Hyatt et al., 1992, 1996; Marsh-Armstrong etal., 1994). The asymmetric expression of the enzymes thatsynthesize RA from retinaldehyde also suggest that theremight be a corresponding gradient of RA-dependent geneexpression. To determine the temporal and spatial patternof retinoid-activated gene expression in the embryonicmurine retina we analyzed retinae from transgenic indi-cator mice that possess a retinoid-dependent (RARE-tk-lacZ) construct. Transgene expression in these mice, asrepresented by b-galactosidase immunoreactivity, indi-cates the presence of activated retinoic acid receptors (en-dogenous retinoids bound to receptors), thus revealing thedistribution of retinal cells responding to endogenous RA(Balkan et al., 1992).

As early as E9.5, when the optic vesicles are evident astwo lateral outgrowths of the forebrain (Pei and Rhodin,1970), transgene expression is present throughout thedorsal aspect of the optic vesicle (data not show). By E11.5(Fig. 1A,B), after the optic vesicle has invaginated to formthe optic cup, b-galactosidase immunoreactivity is sharply

restricted to cells located dorsal to the optic disc (Fig. 1B).By E13.5 (Fig. 1C,D), however, transgene expression isstrongly evident in ventral retina (Fig. 1D), with fewercells in dorsal retina expressing the transgene. This ex-pression pattern continues one day later, at E14.5, thoughtransgene expression is further reduced in dorsal retina(Fig. 1E), leaving an unresponsive stripe of neuroepithe-lium reminiscent of the expression pattern of the RA-degrading oxidase, CYP26 (McCaffery et al., 1999). AtE15.5 (Fig. 1F), and postnatal day (P) P0 (Fig. 1G), trans-gene expression is almost exclusively restricted to ventralretina, though it is significantly weaker at E15.5 (Fig. 1F).These results demonstrate that there is a pronounceddorsal to ventral shift in the pattern of retinoid-inducedtransgene activation beginning with the formation of theoptic cup and proceeding through embryonic retinogen-esis.

In the adult retina, however, b-galactosidase positivecells are confined mainly to the lower third of the innernuclear layer (INL; Fig. 1H) in all regions along thedorsal/ventral axis of the retina, suggesting that they areamacrine cells. Additionally, a small number of eitherdisplaced amacrine cells or ganglion cells are labeled inthe ganglion cell layer (GCL; Fig. 1H).

These results indicate that the retinal neuroepitheliumresponds differentially to endogenous retinoids along thedorsal/ventral axis during distinct stages of retinal devel-opment. In addition, these results suggest that RA-regulated genes are also asymmetrically expressed alongthe dorsal/ventral axis, as predicted by the asymmetricexpression of the RA-synthesizing enzymes. Although thepattern of transgene expression does not correlate withthe activity pattern of the synthesizing enzymes, theremight be alternative sources of RA, as found in the adultretina (Edwards et al., 1992), which should be consideredin interpreting these results. However, our results do pro-vide indirect evidence that RA participates in mediatingthe molecular events underlying early patterning eventsof the developing retina.

RA manipulation alters retinal morphology,but does not affect the expression of two

markers of the embryonic retina

Our results indicate that RA-dependent gene expressionis compartmentalized along the dorsal/ventral axis duringthe development of the retina. Since it has been suggestedthat RA levels along this axis are critical for the propergrowth and development of ventral retina (Marsh-Armstrong et al. 1994), we examined the effects of per-turbing RA availability on the formation of the dorsal/ventral axis. In addition we examined morphological

TABLE 2. Perturbing RA does not significantly affect the size and shapeof the E14.5 retina1

Characteristics Control Citral RA

N/T length (mm) 528.8 (36.4) 525.0 (41.3) 466.7 (35.0)*D/V height (mm) 1,793.1 (133.3) 1,769.9 (76.9) 1,798.5 (108.1)Retina dorsal to optic

disc (%)50.5 (0.93) 55.3 (0.82)** 55.0 (1.1)**

Optic disc width (mm) 68.8 (11.3) 71.8 (11.7) 70.0 (16.7)Optic disc height (mm) 117.3 (17.8) 119.5 (29.2) 126.3 (41.6)

1Measurements of the size and shape of the E14.5 indicator retinae after treatmentwith oil (control), citral (citral), or exogenous RA (RA). Significance was calculatedusing a difference of means test and P , 0.05 was considered to be statisticallysignificant. *, P , 0.05. **, P , 0.01. Mean values are given with standard deviationsindicated in parentheses.

TABLE 1. Perturbing RA alters the size and shape of the E11.5 retina1

Characteristics Control Citral RA

N/T length (mm) 405.0 (13.8) 335.7 (16.2)** 302.0 (52.9)**D/V height (mm) 1,040.2 (62.2) 1,138.0 (134.2) 848.0 (59.4)**Retina dorsal to optic

disc (%)51.3 (0.82) 58.9 (8.1)* 66.0 (2.8)**

Optic disc width (mm) 50.0 (8.9) 80.0 (17.3)* 95.0 (25.5)**Optic disc height (mm) 41.3 (10.1) 102.3 (24.5)** 115.8 (30.2)**

1Measurements of the size and shape of the E11.5 indicator retinae after treatmentwith oil (control), citral (citral), or exogenous RA (RA). Significance was calculatedusing a difference of means test and P , 0.05 was considered to be statisticallysignificant. *, P , 0.05. **, P , 0.01. Mean values are given with standard deviationsindicated in parentheses.

291RA-DEPENDENT GENE EXPRESSION IN THE RETINA

events as well as the expression patterns of two region-specific genes that are expressed during embryogenesis ineither dorsal retina, aldehyde dehydrogenase class-1 iso-form (AHD-2, McCaffery et al. 1992), or ventral retina,Pax2 (Nornes et al., 1990). These morphometric and anti-genic markers have proven to be sensitive assays of RA-induced changes in the development of the dorsal/ventralretinal axis (Hyatt et al., 1996).

We found that perturbing RA levels during the forma-tion of the optic cup (E7.5, citral; E11, RA) dramaticallyaltered retinal morphology (see Table 1, Fig. 3). For ex-ample, E11.5 embryos exposed to exogenous RA in uterohad retinae that were significantly smaller along both thenasal/temporal (25%; P50.0004) and dorsal/ventral (18%;P50.0001) axes compared to control retinae. Optic discsize was also affected; optic discs in RA-treated retinae

Fig. 1. Retinal morphology and retinoid-dependent transgene ex-pression during retinogenesis in the mouse. Sagittal sections throughE11.5 (A,B), E13.5 (C,D), E14.5 (E), E15.5 (F), P0 (G), and adult (H)eyes were immunoreacted for b-galactosidase (B,D–H) and counter-stained with bisBenzimide (A,C). At E11.5 (A,B) transgene expressionis restricted to the neuroepithelium dorsal to the optic disc (B). ByE13.5 (C,D) transgene expression has begun to shift to ventral regionsof the retinal neuroepithelium (D), while expression in dorsal regions

is reduced. At E14.5 (E) transgene expression is restricted to ventralneuroepithelium and the dorsal retinal margins. At E15.5 (F) and P0(G) transgene expression continues to be restricted to ventral regionsof the retina. In the adult retina (H) b-galactosidase positive cells aredistributed across the retina in the INL and GCL (white arrows, H).INL, inner nuclear layer; GCL, ganglion cell layer. Scale bars 5 50mm in A–D; 200 mm in E,F; 100 mm in G,H.

292 D.L. STULL AND K.C. WIKLER

were both wider (nasal/temporal axis; P50.0001) andlonger (dorsal/ventral axis; P50.0001). Finally, while op-tic discs in control retinae were positioned close to themidpoint of the dorsal/ventral axis (with 49% of the retinalneuroepithelium ventral to the optic disc), RA-treatmentmarkedly reduced the length of the retina below the opticdisc (34% ventral to the optic disc; P50.0001).

We observed that early citral treatment also shortenedthe retina along the nasal/temporal axis by 17%(P50.0001). Early citral treatment affected optic disc sizeas well; optic discs were significantly larger along both thenasal/temporal (P50.0027) and dorsal/ventral (P50.0001)axes. These results indicate that our citral treatmentcauses alterations in a few parameters of retinal develop-ment in a similar fashion as our RA treatment did, per-haps a result of a negative feedback pathway initiated bythe inhibition of RA synthesis by citral. However, in oneimportant parameter, dorsal/ventral height, citral and RAappear to have opposing effects (Table 1).

Perturbing RA levels later in development (E10.5, cit-ral; E14, RA) did not significantly alter most aspects ofretinal size and shape (Table 2), including retinal andoptic disc length along the dorsal/ventral axis. ExogenousRA did, however, significantly shorten the retina along thenasal/temporal axis by 12% (P50.0075). In addition, treat-ment with RA or citral later in development also resultedin a more ventrally positioned optic disc (Table 2).

These results, therefore, indicate that perturbing earlyRA availability has profound effects on the overall sizeand shape of the retina, as well as on the ratio of retinalneuroepithelium dorsal and ventral to the optic disc. How-ever, using region-specific markers of embryonic dorsal

(AHD-2, McCaffery et al., 1992) and ventral (Pax2, Norneset al., 1990) retina, we did not find that these morpholog-ical changes were correlated with the ectopic expression ofthese markers (Fig. 2). Therefore, based on the treatmentsused here, RA availability appears to regulate the mor-phological parcellation of the retinal neuroepitheliumalong the dorsal/ventral axis without altering the anti-genic identities of the dorsal and ventral regions of theneuroepithelium located above and below the optic disc.

Intriguingly, early citral treatment consistently (5/5embryos) resulted in the presence of an additional regionof ventral retina positioned near the optic disc (Fig. 3B).These “secondary” retinae were never observed in eithercontrol (Fig. 3A) or RA-treated (Fig. 3C) animals, or inE14.5 citral-treated retinae (Fig.4E). These additionalpieces of ventral retina were separated from the primaryretina by a layer of cells arising from the retinal pigmentepithelium (RPE) and were present in the optic disc regionof the retina when examined in serial sections through theeye. These secondary retinae were immunoreactive forPax2 (data not shown). These results suggest that manip-ulating RA-availability during early retinogenesis influ-ences mechanisms underlying differential growth of thevertebrate dorsal/ventral axis.

Reconstructions reveal that perturbing RAlevels affects the closure of the optic fissure

The appearance of ectopic pieces of ventral retinal neu-roepithelium in the optic disc regions of retinae treatedwith citral early in retinogenesis (Fig. 3B) suggested thatRA levels might be important for the normal developmentand closure of the optic fissure. In addition, the observed

Fig. 2. Expression patterns of AHD-2 (A–C) and Pax2 (D–F) inE11.5 retinae. AHD-2 expression is restricted to regions of the retinalneuroepithelium dorsal to the optic disc in control (A), citral-treated(B), and RA-treated (C) retinae. Pax2, normally expressed in ventral

region of the retinal neuroepithelium (D), continues to be expressed inventral retina in citral-treated (E), and RA-treated (F) retinae. Scalebar 5 100 mm.

293RA-DEPENDENT GENE EXPRESSION IN THE RETINA

asymmetry in RA-dependent gene expression suggests arelationship between retinoid-dependent gene transcrip-tion and the appearance of these ectopic pieces of ventralretina. To investigate this possible relationship we exam-ined the distribution of RA-dependent gene expressionafter perturbing RA availability by treating animals witheither RA or citral. Because these ectopic pieces of ventralretina are found only in animals exposed to citral duringearly retinogenesis, we sacrificed citral-treated animals atadditional time points to examine the development ofthese “secondary retinae” both prior to and after E11.5.We found that as early as E9.5 retinoid-dependent trans-gene expression is apparent in the presumptive neural

retina in both control and citral-treated animals (data notshown). By E10.5 retinoid-dependent transgene expres-sion is restricted to the dorsal region of the optic cup incontrol animals (Fig. 5), as is observed at E11.5 (Figs. 1B,4A). In contrast, citral treatment induces transgene ex-pression across the entire E10.5 retina, including the ven-tral region of neuroepithelium where the optic fissureforms (Fig. 5).

At E11.5 ectopic transgene expression in citral-treatedretinae persists in ventral retina, though its expression indorsal retina is diminished (Figs. 4B, 6). At this age,RA-treated retinae demonstrate transgene expressionthroughout the retina (Fig. 4C), indicating the loss of the

Fig. 4. Expression of a retinoid-dependent transgene in E11.5(A–C) and E14.5 (D–F) indicator retinae exposed to oil (A,D), citral(B,E), or RA (C,F). Note that in the control E11.5 retinab-galactosidase positive cells are restricted to dorsal retina (A). Citraltreatment induces transgene expression in ventral retina, while re-ducing expression in dorsal retina (B). Treatment with RA induces

b-galactosidase expression throughout the retina (C). Citral (E) andRA (F) treatments do not alter the regional distribution of retinoidresponsive cells in the E14.5 retina; instead citral decreases trans-gene expression (E), while RA increases expression (F). RA, retinoicacid. Scale bars 5 50 mm in A–C; 100 mm in D–F.

Fig. 3. Retinal morphology of control (A), citral- (B), and RA-(C)treated E11.5 retinae visualized with a nuclear stain, bisBenzimide.Citral and RA treatments alter the size and shape of the retina andoptic disc (B,C) but not the overall gross morphology of the retinal

neuroepithelium. Both treatments result in an increase in the lengthof retinal neuroepithelium located dorsal to the optic disc. Citraltreatment also results in additional retinal regions located ventral tothe primary retina (white arrow, B). Scale bar 5 100 mm.

294 D.L. STULL AND K.C. WIKLER

asymmetry in RA-dependent gene expression seen in con-trol retinae (Figs. 1B, 4A). Finally, by E14.5, when mor-phological alterations have been shown to be less severe(Table 2), retinoid-dependent transgene expression is

largely limited to ventral retina in control (Fig. 4D), citral-treated (Fig. 4E), and RA-treated animals (Fig. 4F),though to greater and lesser extents. These results indi-cate that administering citral during the development of

Fig. 5. Schematic and three-dimensional representations of con-trol and citral-treated retinae. Every third bisBenzimide stained sec-tion through control and citral-treated E10.5 retinae was acquiredand outlined as described in Methods. Representative schematic sec-tion are shown here. Red labeling depicts transgene expression. All

sections were rendered into three-dimensional structures, shown tothe left of the schematic outlines using imaging software. Whitearrows indicate optic fissures. Note that the growth of the ventral/temporal retina is delayed in citral-treated animals. D, dorsal; T,temporal; V, ventral; N, nasal.

Fig. 6. Schematic and three-dimensional representations of control,citral-, and RA-treated retinae. Every third bisBenzimide stainedsection through control, citral-, and RA-treated E11.5 retinae wasacquired and outlined as described in Methods and in the legend for

Figure 5. White arrows indicate optic fissures. White arrowhead incitral-treated retina indicates ectopic piece of ventral retina. Notethat the optic disc is larger in citral- and RA-treated retinae. RA,retinoic acid; D, dorsal; T, temporal; V, ventral; N, nasal.

295RA-DEPENDENT GENE EXPRESSION IN THE RETINA

the retinal optic fissure induces retinoid-dependent geneexpression in the ventral neuroepithelium, providing in-direct evidence for the involvement of RA in the formationand/or closure of this structure.

To investigate the possible relationship between RA-dependent gene expression and the development of theoptic fissure we examined the morphology of citral-treatedretinae. Although the neuroepithelium of citral-treatedretinae appears morphologically similar to that of thecontrol at E9.5 (data not shown), by E10.5 the growth ofthe temporal/ventral retina appears to be delayed incitral-treated retinae (Fig. 5). Interestingly, this region ofthe retina corresponds to the location of the ectopic piecesof ventral retina observed in citral-treated animals (Fig.3B), which are only located in the region of the opticfissure (Fig. 6). By E14.5 the ectopic pieces of ventralretina are no longer present and the optic fissure hasfused, resulting in a wider than normal optic disc (data notshown). These results indicate that though the optic fis-sure is eventually able to fuse in citral-treated animals,the development of the optic disc region is not normal,perhaps due to the citral-induced disruption of relativegrowth of dorsal and ventral retina.

DISCUSSION

Asymmetric retinoid-dependent geneexpression during retinogenesis

Because studies have implicated RA in the axial pat-terning of numerous systems such as the limb (Tabin,1991) and inner ear (Kelley et al., 1993), the present studyfocused on the role of RA in development of the dorsal/ventral axis of the murine retina. The experiments de-scribed here were designed to address two interrelatedhypotheses; the first hypothesis proposes that RA-dependent gene expression is compartmentalized alongthe dorsal/ventral axis of the embryonic murine retina, ashas been suggested for RA availability (McCaffery et al.,1992, 1992, 1993) and the second is that RA availability iscritical for the development of the dorsal/ventral axis. Inparticular, these experiments focused on whether asym-metry in RA-dependent gene expression is required for theproper formation and development of the vertebrate ven-tral retina.

To determine the spatial and temporal expression ofretinoid-dependent gene transcription in the embryonicretina we analyzed a line of indicator mice that possess aretinoid-dependent transgene. This transgene has beensuccessfully used to map the location of retinoid-dependent neuroblasts in the developing olfactory systemand spinal cord (Balkan et al. 1992; Colbert et al., 1993,1995; Anchan et al., 1997) in two separate lines of trans-genic mice, indicating that the expression pattern of thistransgene is reproducible and not dependent on insertionsite (Balkan et al., 1992). In addition, this RARE-tk-lacZtransgene is 100 times more sensitive to all-trans RA thanto other retinoid isomers (Balkan et al., 1992), indicatingthat transgene expression in these mice represents RAREactivation by all-trans RA-bound receptors.

In the present study, we found that early in retinogen-esis (E10.5) retinoid-dependent gene expression is re-stricted to immature retinal cells located in the region ofthe retina that is dorsal to the optic disc. This regionaldistribution parallels the restriction of AHD-2, the dorsalRA-synthesizing enzyme, and cellular RA binding protein

(CRABP) I to dorsal retina during early retinogenesis(McCaffery et al., 1992, 1993). This co-localization ofAHD-2 and CRABP I to dorsal retina could result inhigher levels of RA in this region, since it has been sug-gested that CRABP I increases intracellular RA levels(Means and Gudas, 1997); however, no studies have di-rectly measured in vivo levels of retinoids in the earlyembryonic murine retina (see Mey et al., 1997 for retinoidlevels in chick).

We observed that retinoid-activated transgene expres-sion shifts from dorsal retina at E11.5 to ventral retinaand the dorsal retinal margins by E14.5. This change inthe regional location of activated cells suggests a parallelshift in all-trans RA levels along the dorsal/ventral axis.Consistent with this hypothesis McCaffery and Dragerhave determined that at this later age ventral retina ismore effective than dorsal retina at inducing retinoid-dependent transgene activation when co-cultured with areporter cell line (McCaffery et al., 1992). This result,together with the data presented here, supports the hy-pothesis that RA levels are elevated in dorsal retina dur-ing the initial formation of the optic disc and in ventralretina later in development, perhaps due to the differen-tial activity of the retinoid-synthesizing enzymes (McCaf-fery et al., 1992, 1993). This asymmetric retinoid-dependent gene expression, as well as the probability ofasymmetric RA levels, during the initiation of retinal de-velopment suggests that RA availability is an importantaspect of the formation of the dorsal/ventral axis of thedeveloping murine retina.

RA-induced alterations in retinalmorphogenesis

It has been suggested that RA acts specifically to regu-late the growth and development of ventral retina (Hyattand Dowling, 1997). This hypothesis arose from studiesthat demonstrated that perturbing RA levels affects mul-tiple aspects of dorsal/ventral axis development of theretina in zebrafish (Hyatt et al., 1992, 1996; Marsh-Armstrong et al., 1994). For example, augmenting RAlevels induces the duplication of ventral retina (Hyatt etal., 1992), while depleting retinoid levels with citral re-duces the size of ventral retina, creating a “half-retina”(Marsh-Armstrong et al., 1994). Exogenous RA applica-tion not only induces the ectopic appearance of ventralcharacteristics, such as pax2[b] expression and a second-ary optic fissure, but also reduces the activity of the dorsalmarker, aldehyde dehydrogenase (Hyatt et al., 1996).These results support the hypothesis that in zebrafish RAis necessary for the normal development of ventral retina;perturbations in RA availability appear to alter the devel-opment of ventral retina and the expression of ventralretinal characteristics specifically.

The results of the present study, however, do not sup-port the hypothesis that perturbing RA availability ven-tralizes the embryonic murine retina, thus suggestingthat in the mammalian retina RA levels do not specificallyregulate the development of ventral retina. First, wefound that RA increased the length of retinal neuroepithe-lium dorsal, not ventral, to the optic disc, without eitherinducing ectopic expression of Pax2 or reducing the ex-pression of AHD-2, which was expressed throughout theregion dorsal to the optic disc.

Second, citral treatment in the mouse did not result inthe complete elimination of ventral retina as was observedin zebrafish (Marsh-Armstrong et al., 1994). In fact, the

296 D.L. STULL AND K.C. WIKLER

length of the region of retinal epithelium located ventralto the optic disc in citral-treated E11.5 animals was com-parable to that measured in controls (Table 1). Interest-ingly, longer retinae were observed in embryonic chickretinae after citral treatment (Abramovici et al., 1978),though the regional proportions of retinal neuroepithe-lium dorsal and ventral to the optic disc were not mea-sured. Thus, these results indicate that perturbing RAavailability in the developing mammalian and avian ret-ina does not result in the specific changes in ventral reti-nal identity observed in zebrafish. Instead, in the mam-malian retina, RA appears to regulate the grossmorphology of the optic cup and fissure along the dorsal/ventral axis.

These differences in RA-mediated effects in the mouseand zebrafish may result from species differences in themechanisms of retinoid-dependent gene transcription andthe regulation of axial development. This hypothesis issupported by evidence that retinal development in thezebrafish differs from that in other vertebrates, such asthe mouse, in one important aspect; in the zebrafish theaxial development of the eye involves a series of rotations(Schmitt and Dowling, 1994). The developmental dorsal/ventral axis, therefore, is perpendicular to the dorsal/ventral axis in the adult zebrafish retina. This phenome-non suggests that the mechanisms for determining axialpositioning in the developing zebrafish may differ fromthose in other vertebrates to account for these rotations.

Alternatively, manipulating RA levels in the zebrafishmay affect multiple events involved in the development ofthe retina that are temporally separated in the mouse. Forexample, the formation of the optic cup in zebrafish occursin 12 hours (Schmitt and Dowling, 1994), while this eventtakes 2.5 days in the mouse (Pei and Rhodin, 1970). Ourobservation of a delay in the growth of ventral/temporalretina indirectly supports this explanation. Our results,therefore, suggest that RA regulation of early morpholog-ical events in the mammalian retina differs from thatobserved in the zebrafish; in the murine retina, RA avail-ability appears to regulate the formation of the dorsal/ventral axis of the retina without altering ventral retinalidentity specifically. Further examination with additionaltreatments would elucidate these mechanisms.

RA availability appears to regulate theclosure of the optic fissure

The appearance of ectopic pieces of ventral retina follow-ing early treatment with citral offers the opportunity toexamine the role of RA in the early development of themurine retina. There are three possible hypotheses thatmight explain the appearance of these “secondary” retinae.For example, these additional pieces might be regions ofdorsal retina that have been shifted ventrally, which wouldoccur if the invagination of the optic cup was abnormal.Alternatively, these pieces might represent additionalgrowth of the ventral neuro-epithelium, which would sug-gest that the absence of RA regulates region-specific prolif-eration within the retinal neuroepithelium. Finally, thesepieces might be regions of the ventral neuroepithelium thathave become segregated from the primary retinal sheet,perhaps during the closure of the optic fissure.

Our results support this last hypothesis. The first hy-pothesis is not supported by our finding that these ectopicpieces of ventral retina express Pax2, a marker of embry-onic ventral retina, suggesting that these pieces are ven-tral retinal neuroepithelium in origin. The second possi-

bility also seems unlikely since the citral-treated retinaeare not only not significantly longer along the dorsal/ventral axis (see Table 1), but the mean length of neuro-epithelium ventral to the optic disc in the citral-treatedretina (466 mm) is actually shorter than that measured inthe control (509 mm).

These results indicate that though the ectopic piecesappear to be ventral neuroepithelium in origin, they arenot the result of additional growth; instead, these piecesappear to represent regions of the ventral epithelium thathave been segregated from the rest of the retina duringdevelopment. The abnormal fusion of the optic fissuremight result in this separation. This hypothesis is indi-rectly supported by the delay in the growth of temporal/ventral neuroepithelium, observed in the E10.5 citral-treated retina, which may cause the two margins of theoptic fissure to overlap rather than abut. The restriction ofthese pieces to the optic disc region of the retina supportthis hypothesis, as does our observation of retinal pigmentepithelial cells separating the ectopic pieces. The ectopicpieces, therefore, appear to represent regions of the nasal/ventral retina that underlie the temporal/ventral siderather than abutting it.

Treatment with citral not only resulted in ectopic piecesof ventral retina, but also in the ectopic expression of theretinoid-dependent transgene. Citral-treatment inducedectopic retinoid-dependent transgene expression in theventral retina at an age, E10.5, when the growth of thisregion is delayed. Citral-treatment also suppresses RA-dependent gene expression in dorsal retina. These resultssuggest that this “flipping” of the normal expression pat-tern inhibits the growth of ventral retina, thus resultingin the abnormal closure of the optic fissure discussedabove. These results suggest that RA-dependent genetranscription might be important in dorsal retina first toinitiate growth of the nasal and temporal regions and inventral retina later to ensure that these sides meet to formthe optic fissure.

Our results, therefore, indicate that RA-dependent geneexpression is asymmetric along the dorsal/ventral axisthroughout the embryonic development of the murine ret-ina. In addition, perturbations in this expression patternresult in alterations in retinal morphology, includingchanges in optic fissure closure and in the partitioning ofthe retinal neuroepithelium above and below the opticdisc. The consequences of these morphological changes onmore mature aspects of retinal development, such as thespecification of different cell types, however, are unknown,and are the focus of the work ongoing in this laboratory.

ACKNOWLEDGMENTS

We thank Dr. A.-S. LaMantia and Dr. E. Linney fortheir gift of the RARE-tk-lacZ indicator mice, Dr. G.Dressler for his gift of the Pax2 serum, Dr. J. Hilton for hisgift of the AHD-2 antibody, and Jue Bao for expert tech-nical assistance.

LITERATURE CITED

Abramovici A, Kam J, Liban E, Barishak RY. 1978. Incipient histopatho-logical lesions in citral-induced microphthalmos in chick embryos. DevNeurosci 1:177–185.

Anchan RM, Drake DP, Haines CF, Gerwe EA, LaMantia A-S. 1997.Disruption of local retinoid-mediated gene expression accompanies ab-normal development in the mammalian olfactory pathway. J CompNeurol 379:171–184.

297RA-DEPENDENT GENE EXPRESSION IN THE RETINA

Balkan W, Colbert M, Bock C, Linney E. 1992. Transgenic indicator micefor studying activated retinoic acid receptors during development. ProcNatl Acad Sci USA 89:3347–3351.

Colbert MC, Linney E, LaMantia A-S. 1993. Local sources of retinoic acidcoincide with retinoid-mediated transgene activity during embryonicdevelopment. Proc Natl Acad Sci USA 90:6572–6576.

Colbert MC, Rubin WW, Linney E, LaMantia A-S. 1995. Retinoid signalingand the generation of regional and cellular diversity in the embryonicmouse spinal cord. Dev Dyn 204:1–12.

Connor M. 1988. Oxidation of retinol to retinoic acid as a requirement forbiological activity in mouse epidermis. Cancer Res 48:7038–7040.

Connor M, Smit MH. 1987. Terminal-group oxidation of retinol by mouseepidermis: inhibition in vitro and in vivo. Biochem J 244:489–492.

Edwards ARB, Adler AJ, Dev S, Claycomb RC. 1992. Synthesis of retinoic acidfrom retinol by cultured rabbit Muller cells. Exp Eye Res 54:481–90.

Hyatt GA, Schmitt EA, Marsh-Armstrong NR, Dowling JE. 1992. Retinoicacid induced duplication of the zebrafish retina. Proc Natl Acad SciUSA 89:8293–8297.

Hyatt GA, Schmitt EA, Marsh-Armstrong N, McCaffery P, Drager UC,Dowling JE. 1996. Retinoic acid establishes ventral characteristics.Development 122:195–204.

Hyatt GA, Dowling JE. 1997. Retinoic acid: a key molecule for eye andphotoreceptor d development. Invest. Ophthamol Vis Sci 38:1471–1475.

Kelley MW, Xu X-M, Wagner MA, Warchol ME, Corwin JT. 1993. Thedeveloping organ of Corti contains retinoic acid and forms supernumer-ary hair cells in response to exogenous retinoic acid in culture. Devel-opment 119:1041–1053.

Marsh-Armstrong, N, McCaffery P, Gilbert W, Dowling JE, Drager UC.1994. Retinoic acid is necessary for development of the ventral retina inzebrafish. Proc Natl Acad Sci USA 91:7286–7290.

McCaffery P, Lee M-O, Wagner MA, Sladek NE, Drager UC. 1992. Asym-metrical retinoic acid synthesis in the dorsoventral axis of the retina.Development 115:371–382.

McCaffery P, Posch KC, Napoli JL, Gudas L, Drager UC. 1993. Changingpatterns of the retinoic acid system in the developing retina. Dev Biol158:390–399.

McCaffery P, Wagner PE, O’Neill J, Petrovich M, Drager UC. 1999.Dorsal and ventral retinal territories defined by retinoic acid syn-thesis, break-down, and nuclear receptor expression. Mech Dev 82:119 –130.

Means AL, Gudas LJ. 1995. The roles of retinoids in vertebrate develop-ment. Annu Rev Biochem 64:201–233.

Means AL, Gudas LJ. 1997. The CRABP I gene contains two separable,redundant regulatory regions active in neural tissues in transgenicmouse embryos. Dev Dyn 209:59–69.

Mey JP, McCaffery P, Drager UC. 1997. Retinoic acid synthesis in thedeveloping chick retina. J Neurosci 17:7441–7449.

Nornes HO, Dressler GR, Knappik EW, Deutsch U, Gruss P. 1990. Spa-tially and temporally restricted expression of Pax2 during murineneurogenesis. Development 109:797–809.

Pei YF, Rhodin JAG. 1970. The prenatal development of the mouse eye.Anat Rec 168:105–126.

Schmitt, EA, Dowling JE. 1994. Early eye morphogenesis in the zebrafish,Brachydanio rerio. J Comp Neurol 344:532–542.

Schuh TJ, Hall BL, Kraft JC, Privalsky ML, Kimelman D. 1993. verbAand citral reduce teratogenic effects of all trans retinoic acid andretinol respectively, in xenopus development. Development 119:785–798.

Tabin CJ. 1991. Retinoids, homeoboxes, and growth factors: toward molec-ular models for limb development. Cell 66:199–217.

Warkany J, Schraffenberger S. 1946. Congenital malformations induce inrats by maternal vitamin A deficiency. I. defects of the eye. ArchOphthamol 35:150–169.

Wikler KC, Williams RW, Rakic P. 1990. Photoreceptor mosaic: numberand distribution of rods and cones in the rhesus monkey retina. J CompNeurol 297:499–508.

Wilson, JG, Roth CB, Warkany J. 1953. An analysis of the syndrome ofmalformations induced by maternal vitamin A deficiency: effects ofrestoration of vitamin A at various times during gestation. Am J Anat92:189–217.

298 D.L. STULL AND K.C. WIKLER