Correspondence to: Gregg Duester, Development and Aging Program, Sanford-Burnham Medical Research Institute, 10901 North Torrey Pines Road, La Jolla, California 9203. E-mail: email@example.com
Retinoic acid (RA) is an active metabolite of vitamin A that functions as an important signaling molecule during neural development. Although the role of endogenous RA during hindbrain patterning is firmly established (Gavalas and Krumlauf, 2000), studies on RA function during forebrain development have led to conflicting results (Schneider et al., 2001; Halilagic et al., 2003; Ribes et al., 2006; Molotkova et al., 2007; Siegenthaler et al., 2009; Chatzi et al., 2011). RA functions as a ligand for nuclear RA receptors that bind gene regulatory elements as heterodimers with retinoid X receptors (Germain et al., 2002; Mic et al., 2003). During early embryogenesis, RA is generated in a two-step process by metabolism of retinol to retinaldehyde primarily by retinol dehydrogenase-10 encoded by Rdh10 (Sandell et al., 2007) followed by metabolism of retinaldehyde to RA by retinaldehyde dehydrogenases encoded by Raldh1, Raldh2, and Raldh3 (Aldh1a1, Aldh1a2, and Aldh1a3) (Duester, 2008). In mouse embryos, Rdh10, Raldh1, Raldh2, and Raldh3 expression in optic and olfactory tissues generate diffusible RA that may reach the forebrain although none of these enzymes are expressed in the forebrain before embryonic day (E) 12.5. Beginning at E12.5, Raldh3 expression is detected in the forebrain lateral ganglionic eminence (LGE; a region within the basal ganglia) and expression of both Rdh10 and Raldh2 is detected at E12.5–E14.5 in the meninges that surrounds the forebrain cortex (Li et al., 2000; Smith et al., 2001; Mic et al., 2002; Cammas et al., 2007; Siegenthaler et al., 2009).
Early forebrain patterning is controlled by several signals including Fgf8 expressed anteriorly in the anterior neural ridge, Shh expressed ventrally in the prechordal plate, and Wnt genes expressed dorsally (Gunhaga et al., 2000, 2003). RA was also reported to play a role in forebrain patterning by activating Fgf8 and Shh expression (Schneider et al., 2001; Halilagic et al., 2003; Ribes et al., 2006) or Gli3 expression (Gongal et al., 2011). However, studies on Raldh2−/−;Raldh3−/− double mutant mice suggested that RA is not required for early forebrain patterning, and instead revealed that RA generated by Raldh3 in the LGE after E12.5 is required during forebrain neuronal differentiation to activate dopamine receptor-2 (DRD2) expression in the nucleus accumbens (Molotkova et al., 2007) and to stimulate γ-aminobutyric acid (GABA)ergic differentiation in the basal ganglia (Chatzi et al., 2011). Investigation of an ethylnitrosourea-induced Rdh10 mutant with reduced RA synthesis in the meninges was reported to exhibit reduced radial expansion of forebrain cortical neurons at E14.5 (Siegenthaler et al., 2009). However, RA-rescued Raldh2−/− embryos that lack meningeal RA synthesis do not exhibit a defect in forebrain cortical expansion (Chatzi et al., 2011).
Here, a strain of Rdh10−/− null mice was generated, and the role of RA in early brain patterning and forebrain cortical expansion was further examined. We found that Rdh10 is required for RA synthesis and RA activity in the early head during the stages when patterning of the hindbrain and forebrain occurs. Also, Rdh10 is required for RA synthesis later in the meninges, although Rdh10 is not required for RA synthesis in the forebrain basal ganglia. Rdh10−/− embryos exhibit defective hindbrain patterning but no defect in forebrain expression of Fgf8, Shh, or Gli3 was detected, suggesting that RA activity in the early anterior head (located mostly in the optic vesicles that are out-pocketings of the forebrain) is unnecessary for early forebrain patterning. Our findings also show that meningeal RA synthesis later in development is not required for radial expansion of the embryonic forebrain cortex.
Retinaldehyde Rescue of Mouse Rdh10−/− Embryos
Previously reported Rdh10−/− null mouse strains and an ethylnitrosourea-induced Rdh10trex mutant strain completely lacking enzyme activity were found to exhibit varying degrees of growth deficiency or lengths of embryonic survival from E10.5–E14.5 that likely reflect differing strain backgrounds (Sandell et al., 2007; Cunningham et al., 2011; Rhinn et al., 2011; Sandell et al., 2012). Another ethylnitrosourea-induced mutant strain Rdh10m366Asp retains approximately 40% enzyme activity and survives until E17.5 or in some cases to birth (Ashique et al., 2012). Analysis of survival for our Rdh10−/− mouse strain revealed that embryonic lethality or resorption is often observed between E10.5–E14.5 similar to previous Rdh10−/− and Rdh10trex strains, and we did obtain a Rdh10−/− embryo at E14.5 that appears essentially identical to E14.5 Rdh10trex embryos, exhibiting prominent defects in the shape of the forebrain, eye, and frontonasal region, stunting of forelimbs (although normal hindlimbs), and anemia in the liver and throughout the embryo (Fig. 1A,B).
As Raldh2−/− embryos can be rescued by maternal dietary RA treatment (0.1–0.25 mg/g food) to generate embryos that survive to E10.5–E14.5 (Niederreither et al., 2001; Zhao et al., 2009), we reasoned that retinaldehyde treatment may rescue our Rdh10−/− strain. We found that dietary supplementation of pregnant mice from conception onward with retinaldehyde at 0.1 mg/g food allowed postnatal survival of Rdh10−/− mice to approximately postnatal day 7 (P7); from four litters treated this way we obtained 23 newborn pups, 7 of which were Rdh10−/− that survived until P2 (n = 2), P3 (n = 4), or P7 (n =1). We found that a higher dose of retinaldehyde (0.25 mg/g food) did not improve survival of Rdh10 knockouts beyond P7 and in fact decreased survival as no offspring survived birth. While these dietary studies were underway, a report appeared showing that another Rdh10−/− strain can be efficiently rescued by maternal dietary retinaldehyde treatment (0.4 mg/g food) limited to E7–E11, generating mice that survive to adulthood (Rhinn et al., 2011). Thus, it appears that retinaldehyde treatment (particularly 0.25 mg/g food or higher) is toxic to embryos if provided beyond E11.5.
Examination of E14.5 Rdh10−/− embryos treated with retinaldehyde (0.25 mg/g food) from E7–E9 demonstrated that several early onset defects had been rescued including the observation of relatively normal morphology of the forebrain, eye, frontonasal process, and forelimb, plus normal distribution of blood in the liver and throughout the body (Fig. 1C; n = 3). This method thus provides conditionally rescued Rdh10−/− embryos for analysis of late embryonic Rdh10 function that do not suffer gross morphological and craniofacial defects due to the early role of Rdh10 in RA synthesis up to E9.5.
Detection of RA Activity in Forebrain Tissues
The retinaldehyde-rescued E14.5 Rdh10−/− embryos described above were used to determine whether Rdh10 is required for RA synthesis in forebrain tissues. As these mutants were treated only until E9 and had 5 days to clear the administered retinaldehyde before analysis, we reasoned that this protocol would allow us to analyze Rdh10 function in tissues of older embryos that express Rdh10 such as the meninges.
We examined RA activity in brain tissue from E9.5–E14.5 using a tissue explant RA bioassay in which tissues cultured as explants on top of an F9-RARE-lacZ RA reporter cell line are assayed for detection of diffusible RA by examining lacZ expression; these cells are sensitive to 1 nM RA added to the culture medium (Wagner et al., 1992; Luo et al., 2004; Chatzi et al., 2011). For wild-type embryos at E9.5 we found that RA activity was easily detectable in eye explants but not adjacent forebrain tissue (Fig. 2A,B; n = 2). For E10.5 wild-type embryos, RA activity was detectable in eye but not forebrain cortex or lateral ganglionic eminence (LGE) which can be dissected separately by this stage (Fig. 2C–E; n = 3). For E14.5 wild-type embryos, RA activity was detected in meninges consistent with meningeal expression of Rdh10 and Raldh2 at this stage, although no RA was detected in cortex, while RA activity was detected in LGE consistent with Raldh3 expression at this stage (Fig. 2F–H; n = 5). For E14.5 Rdh10−/− embryos (rescued with retinaldehyde to E9), meninges RA activity was lost, cortex was still negative, and LGE remained positive; using this RA-reporter assay we cannot state that all RA is lost from meninges, only that there is a considerable loss as the assay is sensitive to 1 nM RA (Fig. 2I–K; n = 3). A lack of detectable RA activity in meninges of conditionally rescued E14.5 Rdh10−/− embryos indicates that this model can be used to examine Rdh10 function in forebrain meninges, particularly in comparison to previously published forebrain studies using the Rdh10m366Asp ENU mutant that retains approximately 40% of its RA-generating activity (Siegenthaler et al., 2009; Ashique et al., 2012).
Forebrain Neuronal Differentiation in Rdh10−/− Embryos
Previous studies with Rdh10m366Asp mutants that exhibit reduced meningeal RA synthesis (Ashique et al., 2012) suggested that RA secreted by the meninges is required for radial expansion of the forebrain cortical postmitotic neuronal layer (Siegenthaler et al., 2009). However, unrescued Rdh10 mutants exhibit massive head deformities affecting forebrain shape that may negate the conclusion that RA acts directly in the cortex to control forebrain cortical expansion. Here, we analyzed forebrain differentiation in conditionally rescued Rdh10−/− embryos that display relatively normal head shape and lack detection of RA activity in the meninges (Figs. 1C, 2I), effectively cutting off the previously proposed source of RA for the cortex (Siegenthaler et al., 2009). We analyzed expression of the neuronal marker Tuj1 and the proliferative marker Ki67 in coronal brain sections of both wild-type and conditionally rescued Rdh10−/− embryos at E14.5. Double immunostaining for Tuj1 and Ki67 revealed no alterations in radial expansion of the postmitotic Tuj1-expressing cortical layer, nor the Ki67 proliferative zone, in the Rdh10−/− cortex when compared with wild-type (Fig. 3A). Additionally, cortical layer thicknesses for the ventricular zone, subventricular zone, intermediate zone, and cortical plate in E14.5 conditionally rescued Rdh10−/− mutants and wild-type were not significantly different (Fig. 3B). These findings indicate that RA synthesized by Rdh10 in the meninges is not required for cortical expansion in the forebrain. Instead, our findings are consistent with an earlier function for Rdh10 in RA synthesis for the cranial neural crest to establish normal head shape as a prerequisite for normal cortical expansion, but a direct role for RA signaling in the cortex is not supported. As the meninges is derived from the neural crest, its ability to synthesize RA by E12.5 may relate to further regulation of craniofacial development by RA signaling.
Previous studies demonstrated a loss of GABAergic differentiation in the basal ganglia of Raldh3−/− embryos that lack RA activity in the LGE (Chatzi et al., 2011). Here, we examined conditionally rescued E14.5 Rdh10−/− embryos for GABA immunoreactivity in the LGE and found essentially no difference with the wild-type control (Fig. 3C; n = 2). Thus, Rdh10 is unnecessary for GABAergic differentiation in the LGE, consistent with our RA explant bioassay showing that RA activity in the LGE is not affected by loss of RDH10 function (Fig. 2K). This observation indicates that a different retinol-metabolizing enzyme is needed to supply the retinaldehyde substrate for RALDH3 in the LGE; a potential candidate may be RDHE2 (SDR16C5) that has recently been found to play a role in neural retinol metabolism (Belyaeva et al., 2012).
Effect of Rdh10−/− Genotype on Early Head RA Activity and Brain Patterning
To investigate the early role of head RA activity in brain patterning, we examined unrescued Rdh10−/− embryos from E7.75–E8.75. Previous studies with mouse embryos carrying the RARE-lacZ RA-reporter transgene have shown that RA activity in the head is first observed in the posterior hindbrain at E7.5 followed by the optic vesicles at E8.5–E9.5, which are an outpocketing of the forebrain, but most of the forebrain does not express RARE-lacZ during these stages (Rossant et al., 1991). These observations are consistent with our RA explant bioassay showing that E9.5 eye tissue contains detectable RA while adjacent forebrain tissue does not (Fig. 2A,B). In E8.5 Rdh10−/− embryos carrying RARE-lacZ, we found that all RARE-lacZ expression was lost in the anterior head (eye/forebrain) and most but not all expression was lost in the hindbrain (Fig. 4A). As some previous studies have questioned the sensitivity of RARE-lacZ for detection of endogenous RA, we sought to validate RARE-lacZ sensitivity by culturing E8.5 Rdh10−/− embryos in the absence or presence of RA. We found that addition of as little as 0.25 nM RA was able to induce RARE-lacZ in the eye/forebrain region of Rdh10−/− embryos, and addition of 1.0 nM RA resulted in more extensive induction of RARE-lacZ in the head of both wild-type and Rdh10−/− embryos; control untreated wild-type and Rdh10−/− embryos exhibited RARE-lacZ expression indistinguishable from uncultured embryos (Fig. 4A; n = 2 for control; n = 2 for 0.25 nM RA; n = 3 for 1.0 nM RA). As 0.25 nM RA was able to stimulate a response in the Rdh10−/− head, these observations demonstrate that the eye/forebrain of Rdh10−/− embryos contains less than 0.25 nM RA, much less than wild-type anterior head, which has been reported to contain approximately 30 nM RA (Horton and Maden, 1995). Our ability to detect induction of RARE-lacZ in the eye/forebrain region with 0.25 nM RA demonstrates extremely high sensitivity for RARE-lacZ as this concentration is near the dissociation constant (kD) of RA with its three receptors (0.2–0.7 nM) (Allenby et al., 1993). These observations demonstrate conclusively that Rdh10−/− embryos lack physiological RA activity in the early forebrain.
Several studies have suggested that RA signaling is required to activate expression of Fgf8 and Shh in chick/mouse forebrain (Schneider et al., 2001; Halilagic et al., 2003; Ribes et al., 2006) or Gli3 (a downstream target of Shh signaling) in zebrafish forebrain (Gongal et al., 2011). We examined E8.75 Rdh10−/− embryos by whole-mount in situ hybridization and found that brain expression of Fgf8 in the anterior neural ridge (n = 3), Shh in the ventral forebrain (n = 3), and Gli3 in the dorsal forebrain (n = 3) was indistinguishable from that observed in wild-type embryos (Fig. 4B). Because we found above that Rdh10−/− embryos do not possess residual RA activity (Fig. 4A), these observations demonstrate that RA is not required to activate expression of these important forebrain patterning genes.
In contrast, we found that Rdh10−/− embryos do exhibit defects in hindbrain patterning. During hindbrain development, Hoxb1 is broadly induced by RA signaling at E7.75 in the posterior hindbrain, then expression becomes restricted by RA to rhombomere 4 (Marshall et al., 1994; Studer et al., 1994; Niederreither et al., 2000; Sirbu et al., 2005). We found that Rdh10−/− embryos lack the anterior extension of Hoxb1 into the posterior hindbrain normally observed at E7.75, and then at E8.5 Hoxb1 expression that is normally expressed tightly in rhombomere 4 was expanded along the anteroposterior axis (Fig. 5A–D; n = 7). EphA2 is another gene that is normally expressed only in rhombomere 4 along the hindbrain (Becker et al., 1994), but E8.5 Rdh10−/− embryos exhibited ectopic expanded expression of EphA2 along the anteroposterior axis of the hindbrain (Fig. 5E,F; n = 4). Expression of Krox20 is normally expressed in two tight domains (rhombomeres 3 and 5) by E8.5 (Schneider-Maunoury et al., 1993); however, Rdh10−/− embryos display ectopic expansion of Krox20 posterior to rhombomere 5 (Fig. 5G,H: n = 2). Similar hindbrain patterning defects were reported for another Rdh10−/− strain that loses most but not all RA in the hindbrain (Rhinn et al., 2011). Also, the hindbrain defects we observe are similar to the intermediate effects of RA removal with an RAR antagonist in chick hindbrain which retained some RA activity (Dupé and Lumsden, 2001). Thus, while loss of RA activity in the forebrain/eye of Rdh10−/− embryos does not affect early forebrain patterning, clear effects on hindbrain patterning are observed even though all RA activity has not been eliminated from the hindbrain.
The studies reported here demonstrate that RDH10 plays a major role in generating retinaldehyde for RA synthesis early in the head and later in the meninges. Disagreement exists as to whether RA signaling is required for patterning or differentiation of the forebrain. With regard to forebrain patterning, inconsistent results were obtained between early studies using drugs to inhibit RA signaling that suggested an RA requirement (Schneider et al., 2001) or later genetic loss-of-function studies eliminating RA synthesis that demonstrated no RA requirement (Molotkova et al., 2007). In general, genetic loss-of-function studies are more reliable to determine biological function compared with drug treatments that may create side-effects. However, inconsistencies have arisen between genetic studies using Raldh2−/− embryos to examine potential RA function during early forebrain patterning based upon whether analysis was performed at E9.5, which suggested that RA is required (Ribes et al., 2006), or E8.75, which suggested that RA is not required (Molotkova et al., 2007). Because Raldh2−/− embryos suffer significant body and heart growth defects by E8.5 and fail to grow beyond E8.75, they are unlikely to provide reliable data beyond this point. Here, we show that Rdh10−/− embryos, which do not suffer severe growth defects at E8.75, can be used to conclude that RA is not required for activation of Fgf8, Shh, and Gli3 during early forebrain patterning; it is clear that RA signaling has been drastically reduced in the head of Rdh10−/− embryos because hindbrain rhombomere patterning has been disturbed. Furthermore, in past studies it has not been clear how much RA activity is missing in the forebrain when using either pharmacological or genetic methods. Here, we have determined that RARE-lacZ is very sensitive in the head at E8.75, thus demonstrating that Rdh10−/− embryos lack significant RA activity in the eye/forebrain region (i.e., < 0.25 nM RA). We suggest that the early RA activity domain in the optic vesicle/anterior forebrain at E8.75 is designed for control of eye and cranial neural crest development as previously described (Matt et al., 2005; Molotkov et al., 2006) but not early forebrain patterning. Although it is possible that species-specific differences may exist in how forebrain patterning genes are regulated as suggested by studies in chick (Halilagic et al., 2003) and zebrafish (Gongal et al., 2011), it appears more likely that the role of early anterior RA in those species is also to control eye and cranial neural crest development as is the case for mouse.
In addition, the studies reported here demonstrate that conditionally rescued Rdh10−/− embryos are highly superior to unrescued Rdh10 mutants to explore potential RA function at later stages in forebrain differentiation. Previous studies on an Rdh10m366Asp ENU mutant that retains approximately 40% of its RA-generating activity (Siegenthaler et al., 2009; Ashique et al., 2012) and a Foxc1 mutant with reduced meningeal Rdh10 expression that exhibits a 20% reduction in meningeal RA levels (Siegenthaler et al., 2009) concluded that RA generated in the meninges by RDH10 is required for forebrain cortical expansion at E14.5. However, those Foxc1 and Rdh10 mutants also exhibit severe cranial skeletal deformations that were not taken into account as pointed out previously (Chatzi et al., 2011). A recent study following up on these conclusions further asserted that CoupTFI interacts with RA signaling to control forebrain cortical expansion (Harrison-Uy et al., 2013). Here, conditionally rescued E14.5 Rdh10−/− embryos that lack cranial defects and exhibit a loss of detectable RA activity in the meninges were shown to not suffer the previously reported defects in forebrain cortical expansion. Thus, we conclude that neither meningeal RA signaling nor an interaction between CoupTFI and meningeal RA signaling are necessary for radial expansion of the forebrain cortex during late embryonic development. Instead, our Rdh10−/− conditional rescue studies show that RA activity in the head at E8.5–E9.5 is required for early craniofacial development which is a prerequisite for proper formation of all head structures including the forebrain cortex.
Our findings demonstrate that RA activity observed at E12.5–E14.5 in the meninges (derived from the cranial neural crest) is not required during forebrain cortical neuron generation. The function of meningeal RA synthesis is unknown, but we suggest that it may be a source of RA needed to regulate the late stages of craniofacial development based upon previous studies showing an RA receptor requirement for normal function of postmigratory cranial neural crest cells (Dupe and Pellerin, 2009). Therefore, our findings suggest that a direct role for RA signaling in embryonic forebrain development may be limited to the previously reported roles in basal ganglia and interneuron differentiation (Krezel et al., 1998; Molotkova et al., 2007; Liao et al., 2008; Urban et al., 2010; Chatzi et al., 2011; Crandall et al., 2011). However, further studies on postnatal forebrain development may determine whether RA plays additional roles in the forebrain such as the one proposed in the hippocampus for memory formation (Chiang et al., 1998; Misner et al., 2001; Jacobs et al., 2006; McCaffery et al., 2006).
Generation of Rdh10 Null Mice and Genotyping
The Rdh10−/− mouse line used here was generated from mutant embryonic stem (ES) cells purchased from the Knockout Mouse Project (KOMP, University of California, Davis). KOMP Rdh10 mutant ES cells carrying a lacZ gene-trap insert in intron 1 flanked by FRT sites, plus a floxed exon 2 (EPD0149_1_D08), were injected into C57BL/6 blastocysts to generate mice. Mice carrying the Rdh10 gene trap mutation were mated to mice with germline expression of Flp recombinase on a C57BL6 background obtained from the Jackson Laboratory (stock #009086 B6.129S4-Gt(ROSA)26Sortm1(FLP1)Dym/RainJ). Conditional Rdh10-floxed mice lacking the gene trap following FLP recombination were crossed with mice expressing a germline Cre recombinase on a BALB/c background obtained from the Jackson Laboratory (stock #003465 BALB/c-Tg(CMV-cre)1Cgn/J) to obtain germline null Rdh10−/− mice. Further matings were performed with FVB/N wild-type mice to generate Rdh10−/− mice lacking the FLP and Cre transgenes, and Rdh10−/− mice carrying the RARE-lacZ transgene were generated (Rossant et al., 1991).
For timed matings, noon on the day of vaginal plug detection was considered embryonic day 0.5 (E0.5). Embryos derived from timed matings were genotyped by polymerase chain reaction (PCR) analysis of yolk sac DNA; Rdh10−/− mutant primers: 5′-AAGGCGCATAACGATACCAC-3′ and 5′-CGAGCTCAGACCATAACTTCG-3′ (164-bp PCR product); wild-type primers: 5′-GGCACCCCAATCTACATCTTATGTG-3′ and 5′-AAGCAACAATCTAAACTCATTCAAAGCC-3′ (472-bp PCR product). All mouse studies conformed to the regulatory standards adopted by the Animal Research Committee at the Sanford-Burnham Medical Research Institute.
Retinaldehyde Rescue of Rdh10−/− Phenotype
Rdh10−/− embryos were treated by maternal dietary retinaldehyde supplementation as follows. Briefly, all-trans-retinaldehyde (Sigma Chemical Co.) was dissolved in ethanol then diluted in corn oil and mixed with powdered standard mouse chow to provide a final concentration of 0.1 mg/g or 0.25 mg/g. For long-term studies, pregnant mice or mice with litters were provided this diet continuously. To generate conditionally rescued Rdh10−/− embryos, pregnant mice were provided the 0.25 mg/g retinaldehyde diet fresh twice a day beginning the morning of E7 and extending until the morning of E9 when mice were returned to normal mouse chow until E14.5 when embryos were dissected.
Whole-Mount in Situ Hybridization
Whole-mount in situ hybridization was performed using digoxigenin-labeled antisense RNA probes as previously described (Wilkinson, 1992). Wild-type and mutant embryos were stained for the same length of time.
E14.5 heads were fixed overnight at 4°C in 4% paraformaldehyde and coronal paraffin sections (7 μm) were processed for immunofluorescence as described (Shen et al., 2008; Chatzi et al., 2011). The primary antibodies included rabbit anti-GABA 1:500 (Millipore), rabbit anti-Tuj1 1:200 (Abcam), and rabbit anti-Ki67 1:200 (Abcam). The thickness of the cortical layers was measured using ImageJ analysis. Data are presented as mean ± SEM; for pair-wise analysis analysis of variance test was used.
Retinoic Acid Detection
To detect RA in whole embryos, we used mice carrying the RARE-lacZ RA-reporter transgene which places lacZ (encoding β-galactosidase) under the control of a RA response element (RARE) (Rossant et al., 1991); wild-type and mutant embryos were stained for 18 hr. For detection of RA in embryonic forebrain tissues we performed an RA tissue explant bioassay using Sil-15 F9-RARE-lacZ RA-reporter cells (Wagner et al., 1992). Briefly, dissected tissues of approximately equal size from wild-type and mutant embryos were incubated overnight on top of the reporter cells to allow diffusion of RA in the tissue to the reporter cells, followed by detection of β-galactosidase activity as described previously (Luo et al., 2004).
We thank the following for plasmids used to produce mouse in situ hybridization probes: C.C. Hui for Gli3, R. Krumlauf for Hoxb1, G. Martin for Fgf8, A. McMahon for Shh, D. Wilkinson for EphaA2 and Krox20. We also thank M. Wagner for the Sil-15 F9-RARE-lacZ RA-reporter cell line and J. Rossant for the RARE-lacZ transgenic mouse. G.D. is funded by the National Institutes of Health.