LETTERdoi:10.1038/nature12037

Visualization of an endogenous retinoic acid gradientacross embryonic developmentSatoshi Shimozono1, Tadahiro Iimura1, Tetsuya Kitaguchi2, Shin-ichi Higashijima3 & Atsushi Miyawaki1,2

In vertebrate development, the body plan is determined by prim-ordial morphogen gradients that suffuse the embryo. Retinoic acid(RA) is an important morphogen involved in patterning theanterior–posterior axis of structures, including the hindbrain1–6

and paraxial mesoderm7,8. RA diffuses over long distances, andits activity is spatially restricted by synthesizing and degradingenzymes9. However, gradients of endogenous morphogens in liveembryos have not been directly observed; indeed, their existence,distribution and requirement for correct patterning remain con-troversial10. Here we report a family of genetically encoded indica-tors for RA that we have termed GEPRAs (genetically encodedprobes for RA). Using the principle of fluorescence resonanceenergy transfer we engineered the ligand-binding domains of RAreceptors to incorporate cyan-emitting and yellow-emitting fluor-escent proteins as fluorescence resonance energy transfer donorand acceptor, respectively, for the reliable detection of ambientfree RA. We created three GEPRAs with different affinities forRA, enabling the quantitative measurement of physiological RAconcentrations. Live imaging of zebrafish embryos at the gastrulaand somitogenesis stages revealed a linear concentration gradientof endogenous RA in a two-tailed source–sink arrangement acrossthe embryo. Modelling of the observed linear RA gradient suggeststhat the rate of RA diffusion exceeds the spatiotemporal dynamicsof embryogenesis, resulting in stability to perturbation. Further-more, we used GEPRAs in combination with genetic and pharma-cological perturbations to resolve competing hypotheses on thestructure of the RA gradient during hindbrain formation and somi-togenesis. Live imaging of endogenous concentration gradientsacross embryonic development will allow the precise assignment ofmolecular mechanisms to developmental dynamics and will acce-lerate the application of approaches based on morphogen gradientsto tissue engineering and regenerative medicine.

RA is a small lipophilic molecule that acts as a ligand for nuclear RAreceptors (RARs). RA is synthesized from retinal by retinaldehydedehydrogenase type 2 (Raldh2) and degraded into polar metabolitesby Cyp26 (Supplementary Fig. 1a). During early vertebrate embryo-genesis, regions of RA synthesis and degradation were mapped bydetecting messenger RNA encoding these metabolizing enzymes. Inzebrafish, raldh2 is expressed in the mid-trunk, whereas cyp26 isexpressed at both the anterior and posterior ends11. We thereforespeculate that a two-tailed gradient of RA could form with the highestconcentration in the mid-trunk and tapering off at each end (Sup-plementary Fig. 2). Gradient formation of other morphogens, all ofwhich are genetically encoded peptides, could be observed by fusingthem to green fluorescent protein (GFP) and introducing them exo-genously into embryos12–16. However, the non-peptidic structure of RAmakes it difficult to image directly17. It therefore remains unknownwhether postulated RA gradients exist, have a linear distribution andare required for normal embryonic development.

To address this gap we developed GEPRAs. The ligand-bindingdomains (LBDs) from mouse RARs were flanked by cyan fluorescent

protein (CFP) and yellow fluorescent protein (YFP), mutants ofAequorea GFP (Fig. 1a). Alterations in the conformation of the LBDin response to RA binding are converted into changes in fluorescenceresonance energy transfer (FRET) from CFP to YFP. Among the con-structs containing the LBD from RAR-b (Supplementary Fig. 3a), theGEPRA-B fusion protein showed the largest response in HeLa cells(Supplementary Fig. 4). In situ calibration for the intracellular RAconcentration ([RA]i) demonstrated an apparent dissociation constant(K9d) of 2 nM (Fig. 1b). Responses to the RA precursors retinal and

1Laboratory for Cell Function Dynamics, Brain Science Institute, RIKEN, 2-1 Hirosawa, Wako-city, Saitama, 351-0198, Japan. 2Life Function and Dynamics, ERATO, JST, 2-1 Hirosawa, Wako-city, Saitama,351-0198, Japan. 3National Institutes of Natural Sciences, Okazaki Institute for Integrative Bioscience, National Institute for Physiological Sciences, Okazaki, Aichi, 444-8787, Japan.

YFP Ratio (CFP/FRET)

a

d

0

0.5

e

Bud

GEPRA-B raldh2/cyp26s

raldh2cyp26s

0

0.5

0

0.45

Transmission

Bud

b

c

GEPRA-G

cyp26sraldh2

O

OH

CFP

YFP

RA

RARLBD

FRET efficiency Ratio

(G

EP

RA

-B)

0.01 1 10 100

Ratio

(GE

PR

A-G

)

0.4

0.5

0.4

0.5

0.3

0.1

Concentration (nM)

Ratio

0.4

0.5raldh2

cyp26s

cyp26s

[RA

]i (nM

)

≥3

2

1

1,5001,0005000

Head-to-tail distance (μm)

0.4

0.3

Ratio

6

2

4

[RA

]i (nM

)

Figure 1 | Development and characterization of GEPRAs. a, Schematicrepresentation of GEPRA. The light and dark yellow denote the unbound andRA-bound states of GEPRA, respectively. b, RA titration curves for GEPRA-B(red) and GEPRA-G (blue). Rmax (RA-depleted form) and Rmin (RA-saturatedform) values for GEPRA-B were slightly higher than those of GEPRA-G. Fittingwith the Michaelis–Menten equation yielded K9d values of 2 and 4 nM forGEPRA-B and GEPRA-G, respectively. Each data point is the mean 6 s.d. fornine experiments. c, Transmission and fluorescence (YFP and FRET ratio)images of a five-somite embryo from the GEPRA-B transgenic zebrafish line.d, e, Ratiometric [RA]i images (left), whole-mount in situ hybridization resultsfor raldh2 and cyp26s (centre), and graphs of the spatial distributions of raldh2and cyp26s (right) in bud-stage embryos expressing GEPRA-B (d) and GEPRA-G (e). Absolute [RA]i values are displayed on the right side of each graph.Scale bars, 200mm.

0 0 M O N T H 2 0 1 3 | V O L 0 0 0 | N A T U R E | 1

Macmillan Publishers Limited. All rights reserved©2013

retinol were almost negligible (K9d? 100 nM; Supplementary Fig. 1b).We speculated that using multiple GEPRAs with different RA affinitieswould allow us to measure [RA]i quantitatively and create two addi-tional GEPRA variants. GEPRA-AA was generated by introducing twoamino-acid substitutions in the LBD of GEPRA-B. Of the constructscontaining the LBD from RAR-c, GEPRA-G produced the best results(Supplementary Fig. 3b). GEPRA-AA and GEPRA-G produced K9d

values of 50 nM (Supplementary Fig. 5a) and 4 nM (Fig. 1b), respectively.We generated transgenic zebrafish lines ubiquitously expressing

GEPRA-B or GEPRA-G. In a five-somite embryo expressing GEPRA-B, the probe was uniformly distributed (Fig. 1c, YFP), but the CFP/FRETratio—that is, [RA]i—was highest in the mid-trunk and lowest inthe head and tail (Fig. 1c, ratio). Next we depleted embryos of RAwith 10mM 4-(diethylamino)benzaldehyde (DEAB), an inhibitor ofRaldh2. DEAB abolished the GEPRA-B-derived high [RA]i signal inthe mid-trunk (Supplementary Fig. 6). This indicates that the probeis RA-specific because DEAB should increase retinal and retinolconcentrations.

We compared signals representing [RA]i with the spatial expressionpatterns of RA-metabolizing enzymes. Immediately after imaging withGEPRA-B (Fig. 1d, left), bud-stage embryos were fixed and subjectedto in situ hybridization with probes specific for raldh2 and cyp26species (cyp26s; a mixture of cyp26a1, cyp26b1 and cyp26c1)5 (Fig. 1d,centre). Comparative spatial profiles indicated that the signals repre-senting [RA]i were high in the raldh2-expressing region and low inthe cyp26s-expressing region (Fig. 1d, right). Two intermediate zonesflanked by the raldh2-expressing and cyp26s-expressing regions wereof particular interest. In the posterior zone, the signal representing[RA]i was graded almost linearly. In the anterior zone, however, theprobe saturated near the raldh2-expressing region. To examine highlevels of [RA]i quantitatively, we imaged GEPRA-G transgenic zebra-fish at the bud stage (Fig. 1e) and found linear [RA]i gradients inboth the anterior and posterior intermediate zones. We also expressedGEPRA-AA transiently by injecting fertilized wild-type eggs withmRNA. GEPRA-AA produced no [RA]i gradient in three-somiteembryos (Supplementary Fig. 5b). On the basis of the K9d values ofthe three GEPRAs, the highest [RA]i situated anteriorly within theraldh2-expressing region was estimated to be 6 nM (Fig. 1e). All ofthe gradients identified by using both GEPRA-B and GEPRA-G(Fig. 1d, e) reached the edges of the cyp26s-expressing regions, indi-cating that the steady-state distribution of RA is linear. This findingsupports the model in which a local source and a local sink togethergenerate a linear gradient in the flanked region based on a simplediffusion model18 within the spatiotemporal parameters of embryo-genesis19. We examined whether a rectangular distribution of [RA]i

could exist stably in the intermediate zones with a computer simu-lation. Assuming that the diffusion coefficient of RA and the width ofthe zone were 10mm2 s21 (ref. 20) and 200mm, respectively, our simu-lation showed that a rectangular distribution created a linear gradientin about 10 min (Supplementary Fig. 7), which is much faster than thetime scale of embryonic growth and supports a source–sink model inthe formation of a linear gradient by simple diffusion.

RA signalling in zebrafish embryos was previously observed usingan exogenous reporter gene system, in which an RA response element(RARE) drives the expression of enhanced YFP (eYFP)21. However,RARE–eYFP transgenic zebrafish did not produce specific fluores-cence signals until very late, at roughly the 18-somite stage (Sup-plementary Fig. 8a), and even at the 20-somite stage eYFP fluorescencewas detected only in old somites. In contrast, the GEPRA-B signalindicating a high [RA]i was distributed from the region near the oticvesicle to the most posterior somite (Supplementary Fig. 8b). At thisstage, raldh2 mRNA accumulated in the eyes and all somites, whereasmRNAs encoding cyp26s were abundant in the head region and tail tip(Supplementary Fig. 8c); these patterns agree with the distributionof GEPRA-B signals representing [RA]i. Thus, RARE–eYFP, whichunlike GEPRAs is an indicator of RA signalling and not [RA]i, is not

sensitive in zebrafish embryos because the eYFP chromophore takes arelatively long time to mature compared with the rapid timescale ofzebrafish embryogenesis.

We verified that the [RA]i gradient in the head region at the budstage (Fig. 1d, e) was contained within the hindbrain field (Sup-plementary Fig. 9). We also found that the signal representing high[RA]i in the mid-trunk was detectable at 75% epiboly and developedinto a clear peak at the tailbud stage (Fig. 2a). Given these observations,we were able to resolve a historical controversy about the putative RAgradient in the hindbrain4–6. The major argument against the pre-sence of this gradient is the observation that embryos depleted ofendogenous RA can be fully rescued with a uniform concentrationof exogenous RA5. To examine whether this approach resulted in arectangular RA distribution, we performed rescue experiments withthe GEPRA-B transgenic line (Fig. 2b). At 4 h after fertilization (hpf),embryos were exposed to 10mM DEAB and various concentrations ofRA. The embryos were assessed for [RA]i at the three-somite to four-somite stages and for morphology at 36 hpf. Treatment with 10mMDEAB (Fig. 2b, second column of images) nearly abolished the signalrepresenting high [RA]i and the imaged embryo developed a kinkedhead, which is a characteristic of RA depletion in zebrafish2. When10 nM RA was applied together with 10mM DEAB (Fig. 2b, fourth

0.5

0.4

1,0000Head-to-tail distance (μm)

Control 10 μM DEAB1 nM RA

10 μM DEAB10 nM RA

10 μM DEAB

Ratio

Transmission

RA

(GEPRA-B)

Morphology

at 36 hpf

a

b

0

0.5

4S 3S 4S 4S

Transmission

RA(GEPRA-B)

75% 90% Bud

0 0.5 1 1.5 2 2.5Time (h)

0

0.5

80% 85%70%

0 1,000Distance (μm)

Ratio

0.3

0.5

0.4

Figure 2 | [RA]i gradient during hindbrain development. a, Time-lapse[RA]i imaging of GEPRA-B in an embryo from 70% epiboly to bud stage. Aseries of transmission images (top), [RA]i images (middle) and ratio profiles(bottom) are shown. The most anterior and most posterior points representedin the ratio profiles are indicated in the transmission images by cyan and greendots, respectively. Times since the start of imaging are shown above the figures.b, Visualization of [RA]i in GEPRA-B-bearing embryos (three-somite (3S) tofour-somite (4S) stage) with and without DEAB to deplete RA, and variousconcentrations of exogenous RA as denoted above the images. In each columnreflecting the various treatment conditions, transmission and [RA]i images areaccompanied by a graph showing the spatial distribution of [RA]i along theanterior–posterior axis of the body. A transmission image of a laterdevelopmental stage (36 hpf) is also presented. A red arrowhead indicates akink in the head region. Scale bars, 200mm.

RESEARCH LETTER

2 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 3

Macmillan Publishers Limited. All rights reserved©2013

column), however, a signal reflecting high [RA]i was detected.Although full recovery was not observed, a substantial [RA]i gradientwas created in the hindbrain field. The gradient was sufficient to allownormal embryo development. Under the same conditions, 1 nM RA(Fig. 2b, third column) resulted in a smaller [RA]i gradient and did notrescue the zebrafish from the effects of RA depletion. The results werereproduced for each treatment condition with multiple embryos (Sup-plementary Fig. 10). These perturbation experiments revealed that thedistribution of RA for hindbrain patterning is reliably governed bylocal degradation of RA, as proposed previously4–6. Such [RA]i gradi-ents depending on RA degradation were also observed during somi-togenesis (Supplementary Fig. 11). In the zebrafish mutant giraffe (gir)containing a mutation in the cyp26a1 gene, there are patterning defectsin various organs22. We injected GEPRA-B mRNA into gir embryos,and did indeed find a global increase in [RA]i in six-somite embryos(Supplementary Fig. 12).

Another essential morphogen that functions during the formationof the embryonic axis includes the fibroblast growth factor (FGF)family7,8,23,24. The interactions between RA and FGF signalling includepositive and negative feedback and feedforward mechanisms. Toexamine how Fgf8 regulates endogenous RA gradients, we performed[RA]i imaging experiments using embryos in which Fgf8 signallingwas suppressed. We injected fgf8 splice-blocking morpholino oligo-nucleotides (MOs)25 into one-cell-stage embryos to phenocopy thewell-characterized zebrafish mutant acerebellar (ace)26,27. As a resultof mutation of the fgf8 gene, homozygous ace embryos lack both acerebellum and organizer in the midbrain–hindbrain boundary whileretaining expression of some rhombomere-marker genes26. BecauseRA signalling is required for the formation of posterior segments ofthe hindbrain2–6, such as rhombomeres 5–7, we assessed whether theRA gradient was retained in the hindbrain field of fgf8 morphantsexpressing GEPRA-B. Time-lapse transmission images confirmedeffective knockdown of fgf8 expression based on a bulge in the area ofthe developing midbrain (arrowheads)26 and shortened tail (Fig. 3a, b).In the MO-treated embryos, the signal representing high [RA]i in themid-trunk was observed until 12.5 hpf (Fig. 3b). Normal posteriori-zation of their hindbrain26 was verified by using in situ hybridizationwith two rhombomere markers (Supplementary Fig. 13). In addition,the expression profiles of raldh2 and cyp26s were unaffected at 10 hpf(Fig. 3c). After 13.5 hpf, however, high [RA]i was attenuated to baselinelevels in fgf8 morphants (Fig. 3b). We found that this decrease in [RA]i

was due to decreased levels of raldh2 expression (Fig. 3d), indicatingthat fgf8 may have a role in the maintenance of raldh2 gene expression.The downregulation of raldh2 expression was previously observedin ace mutants28; however, the GEPRA system allowed us to directlyimage the effect on the RA gradient after the loss of raldh2 expression.

The free diffusion of a lipophilic molecule such as RA may belimited in aqueous environments. A large fraction of RA moleculesare probably transported intracellularly by interaction with cellularRA-binding proteins10,29. On the assumption that most intracellularRAs are bound to carrier proteins, GEPRAs detect the free form of RA.The endogenous RA-buffering system is reminiscent of intracellularCa21 buffering in which loading cells with a large amount of high-affinity Ca21 probes does not markedly affect intracellular Ca21

dynamics30. Thus, the expression of GEPRAs can be increased withoutdisturbing endogenous RA dynamics or the normal embryonic deve-lopment of fish. In support of this, GEPRA transgenic embryos developedwithout apparent morphological aberrations.

Here we have generated fluorescent probes that allowed direct visu-alization of the endogenous RA gradient in live embryos. The existenceof an RA gradient is a matter of historical debate4–6. Our GEPRAsystem clearly demonstrated that a two-tailed linear RA gradient existsduring early embryogenesis, and that Fgf8 is not required for RAgradient formation in the hindbrain but is required for the mainten-ance of raldh2 expression and the RA gradient later in development.The ability of GEPRA to directly reveal endogenous morphogen

gradients such as RA in live embryos will allow a greater under-standing of their roles and mechanisms in patterning the vertebrateembryo and will support gradient-based approaches in medicine andbioengineering.

METHODS SUMMARYThe genes for GEPRAs were constructed from complementary DNAs encodingthe LBD of mouse RAR-b or RAR-c, and subcloned into pCS2 for expression inHeLa cells and for in vitro synthesis of mRNA or into pT2KXIGDin for generationof transgenic zebrafish lines. [RA]i imaging was performed with an inverted con-focal microscope equipped with a 310 objective lens and a 440-nm laser. With theaperture fully open, non-confocal fluorescence images were acquired. The dual-emission ratio imaging was performed under exactly the same conditions for insitu [RA]i calibration (HeLa cells) and in vivo time-lapse two-dimensional [RA]i

imaging (zebrafish embryos). Thus, absolute [RA]i values in the two-dimensionalimage can be obtained from the ratios by using the curves shown in Fig. 1b. RA,retinol, retinal and DEAB were administrated with 0.1% dimethylsulphoxide(DMSO). The morphology of an embryo was observed with the confocal micro-scope simultaneously with [RA]i or with a stereomicroscope afterwards. Thetransgenic zebrafish line (Tg(RARE–gata2:NTD–eYFP)ld1)21 was obtained fromthe Zebrafish International Resource Center (University of Oregon). The eYFPfluorescence was observed with an inverted confocal microscope equipped with a473-nm or 488-nm laser. cDNA encoding raldh2, cyp26a1, cyp26b1, cyp26c1,krox20, hoxd4, otx2 or myoD in pCS2 was used to construct the probe for in situhybridization. All embryos were allowed to grow after live [RA]i imaging, unlessthey were subjected to in situ hybridization. Embryos developed without anymorphological aberrations, indicating that GEPRA expression and light exposureare not morphogenic.

0

0.5

Transmission

RA(GEPRA-B)

10.5 11.5 12.5 13.5 14.5 15.5 16.5

d

a

b

c

raldh2

cyp26s

Control MO fgf8 MO

14 hpf

Control MO fgf8 MO

raldh2

cyp26s

10 hpf 10 hpf

10 hpf 10 hpf

14 hpf

14 hpf 14 hpf

Bud 2S 4S 7S 9S 11S 14S

Transmission

RA(GEPRA-B)

10.5 11.5 12.5 13.5 14.5 15.5 16.5 hpf

0

0.5

0 1,000

0.40.5

Ratio

Head-to-tail distance (mm)

0 1,000Head-to-tail distance (mm)

Ratio

0.30.40.5

Figure 3 | [RA]i gradient is affected by fgf8 expression. a, b, Time-lapse[RA]i imaging of GEPRA-B-bearing embryos from 10.5 to 16.5 hpf afterinjections of control MO (a) or fgf8-specific MO (b). Transmission images (toprow), [RA]i images (middle row) and ratio profiles (bottom row) are presented.The control embryo showed normal somitogenesis; somite stages are labelled inthe transmission images (a). A bulge in the area of the developing midbrain inthe embryo injected with the fgf8-specific MO is indicated by black arrowheads;a signal representing a high [RA]i in the trunk is indicated by white arrows(b). c, d, Comparative in situ hybridizations showing expression of raldh2 andcyp26s in embryos at 10 hpf (c) and 14 hpf (d) after injections with control orfgf8-specific MO. All images are lateral views. Scale bars, 200mm.

LETTER RESEARCH

0 0 M O N T H 2 0 1 3 | V O L 0 0 0 | N A T U R E | 3

Macmillan Publishers Limited. All rights reserved©2013

Full Methods and any associated references are available in the online version ofthe paper.

Received 22 November 2012; accepted 25 February 2013.

Published online 7 April 2013.

1. Stern, C. D. & Foley, A. C. Molecular dissection of Hox gene induction andmaintenance in the hindbrain. Cell 94, 143–145 (1998).

2. Begemann, G., Marx, M., Mebus, K., Meyer, A. & Bastmeyer, M. Beyond the necklessphenotype: influence of reduced retinoic acid signaling on motor neurondevelopment in the zebrafish hindbrain. Dev. Biol. 271, 119–129 (2004).

3. Maves, L. & Kimmel, C. B. Dynamic and sequential patterning of the zebrafishposterior hindbrain by retinoic acid. Dev. Biol. 285, 593–605 (2005).

4. Sirbu, I. O., Gresh, L., Barra, J. & Duester, G. Shifting boundaries of retinoic acidactivity control hindbrain segmental gene expression. Development 132,2611–2622 (2005).

5. Hernandez, R. E., Putzke, A. P., Myers, J. P., Margaretha, L. & Moens, C. B. Cyp26enzymes generate the retinoic acid response pattern necessary for hindbraindevelopment. Development 134, 177–187 (2007).

6. White, R. J., Nie, Q., Lander, A. D. & Schilling, T. F. Complex regulation of cyp26a1creates a robust retinoic acid gradient in the zebrafish embryo. PLoS Biol. 5, e304(2007).

7. Diez del Corral, R. et al. Opposing FGF and retinoid pathways control ventral neuralpattern, neuronal differentiation, and segmentation during body axis extension.Neuron 40, 65–79 (2003).

8. Moreno, T. A. & Kintner, C. Regulation of segmental patterning by retinoic acidsignaling during Xenopus somitogenesis. Dev. Cell 6, 205–218 (2004).

9. Aulehla, A. & Pourquie, O. Signaling gradients during paraxial mesodermdevelopment. Cold Spring Harb. Perspect. Biol. 2, a000869 (2010).

10. White, R. J. & Schilling, T. F. How degrading: Cyp26s in hindbrain development.Dev. Dyn. 237, 2775–2790 (2008).

11. Kudoh, T., Wilson, S. W. & Dawid, I. B. Distinct roles for Fgf, Wnt and retinoic acid inposteriorizing the neural ectoderm. Development 129, 4335–4346 (2002).

12. Teleman, A. A. & Cohen, S. M. Dpp gradient formation in the Drosophila wingimaginal disc. Cell 103, 971–980 (2000).

13. Entchev, E. V., Schwabedissen, A. & Gonalez-Gaitan, M. Gradient formation of theTGF-b homolog Dpp. Cell 103, 981–991 (2000).

14. Gregor, T., Wieschaus, E. F., McGregor, A. P., Bialek, W. & Tank, D. W. Stability andnuclear dynamics of the bicoid morphogen gradient. Cell 130, 141–152 (2007).

15. Yu, S. R. et al. Fgf8 morphogen gradient forms by a source–sink mechanism withfreely diffusing molecules. Nature 461, 533–536 (2009).

16. Muller, P. et al. Differential diffusivity of Nodal and Lefty underlies a reaction-diffusion patterning system. Science 336, 721–724 (2012).

17. Rhinn, M. & Dolle, P. Retinoic acid signalling during development. Development139, 843–858 (2012).

18. Crick, F. Diffusion in embryogenesis. Nature 225, 420–423 (1970).19. Wolpert, L. Positional information and the spatial pattern of cellular differentiation.

J. Theor. Biol. 25, 1–47 (1969).

20. Eichele, G. & Thaller, C. Characterization of concentration gradients of amorphologically active retinoid in the chick limb bud. J. Cell Biol. 105, 1917–1923(1987).

21. Perz-Edwards, A., Hardison, N. L. & Linney, E. Retinoic acid-mediated geneexpression in transgenic reporter zebrafish. Dev. Biol. 229, 89–101 (2001).

22. Emoto, Y., Wada, H., Okamoto, H., Kudo, A. & Imai, Y. Retinoic acid-metabolizingenzyme Cyp26a1 is essential for determining territories of hindbrain and spinalcord in zebrafish. Dev. Biol. 278, 415–427 (2005).

23. Sawada, A. et al. Fgf/MAPK signaling is a crucial positional cue in somite boundaryformation. Development 128, 4873–4880 (2001).

24. Dubrulle, J., McGrew, M. J. & Pourquie, O. FGF signaling controls somite boundaryposition and regulates segmentation clock control of spatiotemporal Hox geneactivation. Cell 106, 219–232 (2001).

25. Draper, B. W., Morcos, P. A. & Kimmel, C. B. Inhibition of zebrafish fgf8 pre-mRNAsplicing with morpholino oligos: a quantifiable method for gene knockdown.Genesis 30, 154–156 (2001).

26. Brand, M. et al. Mutations in zebrafish genes affecting the formation of theboundary between midbrain and hindbrain. Development 123, 179–190 (1996).

27. Reifers, F. et al. Fgf8 is mutated in zebrafish acerebellar (ace) mutants and isrequired for maintenance of midbrain–hindbrain boundary development andsomitogenesis. Development 125, 2381–2395 (1998).

28. Hamade,A.et al. Retinoic acidactivatesmyogenesis in vivo throughFgf8signalling.Dev. Biol. 289, 127–140 (2006).

29. Napoli, J. L. Interactions of retinoid binding proteins and enzymes in retinoidmetabolism. Biochim. Biophys. Acta 1440, 139–162 (1999).

30. Haugland,R. P.& Johnson, I. D. in Fluorescent and LuminescentProbes for BiologicalActivity (ed. Mason, W. T.) 40–50 (Academic, 1999).

Supplementary Information is available in the online version of the paper.

Acknowledgements The authors thank Y. Wada, R. Aoki, M. Sugiyama, F. Picazo andmembers of the Brain Science Institute Research Resource Center for technicalassistance; C. Yokoyama and A. Terashima for critical reading of the manuscript; theFANTOM Consortium for the cDNA clones; and the Zebrafish International ResourceCenter for the transgenic zebrafish. This work was partly supported by grants fromJapan Ministry of Education, Culture, Sports, Science and Technology Grant-in-Aid forScientific Research on Priority Areas ‘Fluorescence Live Imaging’ and ‘Cell Innovation’and the Human Frontier Science Program.

Author Contributions S.S. and A.M. conceived and designed the study. S.S. performedall the experiments, analysed the data and designed the manuscript. T.I. supervised theexperiments on somitogenesis. T.K. and S.H. generated transgenic zebrafish lines. A.M.designed and wrote the manuscript, and supervised the project.

Author Information DNA sequences of GEPRAs are deposited in the DNA Data Bank ofJapan (DDBJ) under accession numbers AB787561–AB787563. Reprints andpermissions information is available at www.nature.com/reprints. The authors declareno competing financial interests. Readers are welcome to comment on the onlineversion of the paper. Correspondence and requests for materials should be addressedto A.M. (matsushi@brain.riken.jp).

RESEARCH LETTER

4 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 3

Macmillan Publishers Limited. All rights reserved©2013

METHODSMaterials. RA, retinol, retinal and DEAB were purchased from Sigma-Aldrich.These compounds were dissolved in dimethylsulphoxide (DMSO) to preparestock solutions. The compounds were administered in a final concentration of0.1% DMSO.Gene construction. cDNA encoding the LBD of RAR-b (FANTOM3 (ref. 31),6820403N24) was amplified using primers containing 59 SphI and 39 SacI sites.The digested product was ligated to a BamHI/SphI fragment encoding CFPand a SacI/XhoI fragment encoding YFP. The ligated product was subcloned intothe BamHI/XhoI sites of the pCS2 vector for mammalian expression. In thisplasmid, genes encoding the donor and/or acceptor were replaced to improvethe RA indicator. Similarly, cDNA encoding RAR-c (FANTOM3 (ref. 31),F730319B19) was used to make a SphI/SacI fragment encoding the LBD.Together with the BamHI/SphI fragment encoding CFP and a SacI/EcoRI frag-ment encoding YFP, cDNA coding for the indicator was subcloned into theBamHI/EcoRI sites of the pCS2 vector. Introduction of two amino-acid substitu-tions (R269A and S280A)32 in the LBD of RAR-b was performed as describedpreviously33.Characterization of the indicators. HeLa cells were grown in DMEM mediumsupplemented with 10% FBS. cDNAs were transfected into cells by usingLipofectamine and Plus reagent (Invitrogen). Two days after transfection, themedium was exchanged with DMEM/F12 without phenol red. Dose–responsecurves were determined with nine transfected cells.Generation of GEPRA transgenic zebrafish lines. GEPRA cDNAs were sub-cloned into the pT2KXIGDin vector, and transgenic lines were created asdescribed previously34. Fish were maintained at 28.0 uC.In vivo RA imaging. A glass bead (Iuchi BZ-1) was placed on a coverslip, and1% agarose (Takara L03) in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.4 mM CaCl2,0.16 mM MgSO4) was poured on the coverslip and allowed to harden. The glassbead was then removed to generate a round chamber. An embryo that had beenanaesthetized with Tricaine at more than 16 hpf was placed in the chamber andcovered with 0.3% agarose in E3 medium. The chamber was submerged in E3medium containing Tricaine. Time-lapse two-dimensional imaging was performedin the xy–t mode with an FV1000 (Olympus) confocal inverted microscope systemequipped with a 310 objective lens (numerical aperture 0.4; UPlanApo) and a 440-nm laser. With the aperture fully open, non-confocal fluorescence images wereacquired. The ratio imaging of embryos was performed under exactly the sameconditions as for calibration experiments with HeLa cells. Thus, absolute [RA]i

values in the image can be obtained from the ratios using the curves shown inFig. 1b. Image processing was performed with ImageJ software (http://rsb.info.nih.gov/ij/). After the background had been subtracted, ratio images were generated andmedian-filtered.Estimation of GEPRA concentration in embryos. Confocal fluorescence imageswere acquired with a GEPRA transgenic embryo and a series of YFP solutions ofvarious concentrations with excitation at 488 nm. The size of the confocal aperturewas about 2 Airy disks. By comparing the intensities of the confocal fluorescenceimages, the GEPRA concentration in the embryo was estimated to be approxi-mately 0.1mM. Considering the concentration35 and affinity for RA (refs 36, 37) of

cellular RA-binding proteins, which were reported to be 10mM and 0.13–2 nM,respectively, only a small fraction of RA molecules should be bound to GEPRA.Transgenic RARE–eYFP zebrafish. Transgenic zebrafish (RARE–gata2:NTD–eYFP)ld1)21 were obtained from the Zebrafish International Resource Center(University of Oregon). Fluorescence was observed with an FV1000 confocalmicroscope (Olympus) with a 488-nm laser or an FV10i confocal microscope(Olympus) with a 473-nm laser.Depletion of endogenous RA and application of exogenous RA. Embryos wereincubated in E3 medium containing 0.1% DMSO (vehicle), 10mM DEAB, 1 nMRA with 10mM DEAB, or 10 nM RA with 10mM DEAB from 4 to 24 hpf. After24 hpf, embryos were bathed in E3 medium. RA imaging was performed at 11 hpf(roughly the three-somite stage). The imaged embryos were kept at 28.0 uC untiltheir morphologies were examined at 36 hpf with a stereomicroscope (MZ16 F;Leica) equipped with a charge-coupled device (CCD) camera (DP50; Olympus).In situ hybridization. cDNA encoding raldh2, cyp26a1, cyp26b1, cyp26c1, krox20,hoxd4, otx2 or myoD was amplified from a cDNA library prepared from 10-hpf or24-hpf embryos and cloned into the pCS2 vector. Probes were labelled with adigoxigenin or fluorescein labelling mix (Roche Diagnostics) and detected withalkaline phosphatase-conjugated antibodies (1:5,000; Roche Diagnostics). Thecolorimetric reaction was performed with BM purple reagent (Roche Dia-gnostics). Images were captured with a stereomicroscope (MZ16 F; Leica)equipped with a CCD camera (DP50; Olympus). Fluorescence and in situ hybridi-zation images were linearly registered with the Image Processing Toolbox inMATLAB software (MathWorks).Knock down of fgf8. Splice-site-targeted morpholino oligonucleotides for the fgf8gene (E2I2 (59-TAGGATGCTCTTACCATGAACGTCG-39) and E3I3 (59-CACATACCTTGCCAATCAGTTTCCC-39)) were used to block its pre-mRNA splic-ing in zebrafish embryos25. One-cell-stage embryos were injected with both E2I2and E3I3 (2.5 ng of each) or a control oligonucleotide (5 ng; 59-CCTCTTACCTCAGTTACAATTTATA-39).

31. Carninci, P. et al. The transcriptional landscape of the mammalian genome.Science 309, 1559–1563 (2005).

32. Zhang, Z. P. et al. Role of Ser289 in RARc and its homologous amino acid residueof RARa and RARb in the binding of retinoic acid. Arch. Biochem. Biophys. 380,339–346 (2000).

33. Sawano, A. & Miyawaki, A. Directed evolution of green fluorescent protein by anew versatile PCR strategy for site-directed and semi-random mutagenesis.Nucleic Acids Res. 15, e78 (2000).

34. Urasaki, A., Morvan, G. & Kawakami, K. Functional dissection of the Tol2transposable element identified the minimal cis-sequence and a highlyrepetitive sequence in the subterminal region essential for transposition. Genesis174, 639–649 (2006).

35. Gustafson, A. L., Donovan, M., Annerwall, E., Dencker, L. & Eriksson, U. Nuclearimport of cellular retinoic acid-binding protein type I in mouse embryonic cells.Mech. Dev. 58, 27–38 (1996).

36. Dong, D., Ruuska, S. E., Levinthal, D. J. & Noy, N. Distinct roles for cellular retinoicacid-binding proteins I and II in regulatingsignalingby retinoicacid. J. Biol. Chem.274, 23695–23698 (1999).

37. Norris, A. W., Cheng, L., Giguere, V., Rosenberger, M. & Li, E. Measurement ofsubnanomolar retinoic acid binding affinities for cellular retinoic acidbinding proteins by fluorometric titration. Biochim. Biophys. Acta 1209, 10–18(1994).

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2013