The previously reported mouse Rdh10^trex point mutation disrupts RDH10 function and causes numerous embryonic defects culminating in embryo death between E.10.5–14.5 [23]. In Rdh10^trex/trex mutant embryos RA synthesis is reduced but not completely eliminated, as residual RA signaling is detected in the trunk neural tube. This raised the possibility that Rdh10^trex/trex may be a severe hypomorph rather than a complete null. For comparison with other mutants in the present study, we examined Rdh10^trex/trex embryos at E9.5 and E10.5 (Fig. 1A,B,D,E). Although the Rdh10^trex/trex phenotype is somewhat variable at E9.5, all embryos have numerous discernible abnormalities. First and second pharyngeal arches form in most Rdh10^trex/trex mutant embryos, but third and more posterior pharyngeal arches are absent or fused to the second arch (Fig. 1B,E). Otic vesicles are usually small, duplicated and displaced posteriorly (Fig. 1B). Mutant embryos exhibit cardiac abnormalities, including heart edema and looping defects (Fig. 1B). At E10.5 a reduction in size of the forelimb bud and absence of posterior pharyngeal arches are clearly evident (Fig. 1E). This constellation of defects observed in the Rdh10^trex/trex embryos is consistent with retinoid deficiency but is less severe than that observed in embryos in which RALDH2 function has been eliminated.

As part of the same chemical mutagenesis screen that yielded the Rdh10^trex mutation, we also generated a mutant originally named “grimace”[29] that exhibited interesting abnormalities in formation of the pharyngeal arches (Fig. 1C). Mapping, cloning and sequence analysis were used to identify the grimace mutation, which was determined to be a single base change within the Aldh1a2 locus. We therefore henceforth refer to this mutation as Aldh1a2^gri. Sequence analysis of genomic DNA reveals that the Aldh1a2^gri mutation alters the 5′ splice consensus sequence at the boundary of exon 4 and intron 4–5, changing the second intronic nucleotide from a conserved T to a C (Fig. 1G). The altered splice consensus sequence disrupts splicing of Aldh1a2 mRNA and leads to a reduction in Aldh1a2 transcript level. Quantitative RT-PCR reveals that Aldh1a2 transcript in homozygous Aldh1a2^gri/gri mutants is reduced to less than 1% of that of wild type embryos (Fig. 1H). We therefore interpret the new Aldh1a2^gri mutation is either a null allele of Aldh1a2 or an extreme hypomorph. Consistent with the loss of Aldh1a2 transcript, we observe the phenotype of homozygous Aldh1a2^gri/gri mutants to be very similar to that of established Aldh1a2 knockout alleles, each exemplifying the abnormalities associated with severe retinoid deficiency (Fig. 1C,F). Mutant Aldh1a2^gri/gri embryos die by E10.5 exhibiting a number of striking defects including cessation of growth at ∼E8.5–E9.0, defects in axial turning, dilated unlooped hearts, small somites and tiny otic vesicles. Whereas established knock alleles of Aldh1a2 exhibit no axial turning, many Aldh1a2^gri/gri mutants display a slightly milder phenotype with a small degree of embryo turning. The slightly milder turning defect may represent a difference of strain background on the axial turning phenotype or may indicate that the Aldh1a2^gri allele is not an absolute null.

Comparison of the Rdh10^trex versus the Aldh1a2^gri phenotypes illustrates the less extreme defects elicited by Rdh10^trex mutation relative to that of Aldh1a2^gri (compare Fig. 1 E, F). At E9.5 Aldh1a2^gri/gri mutants have extreme defects in extension and turning of the body axis, with some embryos completely unturned and others with an abnormal half-turned “J” shaped body axis when viewed from the front. In contrast, nearly all E9.5 Rdh10^trex/trexembryos are completely turned. Aldh1a2^gri/gri mutant embryos lack all pharyngeal arches posterior to the first arch, whereas many Rdh10^trex/trex embryos form second arches and, in some cases, rudimentary third arches. The milder phenotype of the Rdh10^trex/trex embryos relative to the severe phenotype of Aldh1a2^−/− mutants, together with the presence of residual RA activity detected using RARElacZ reporter mice [23], implies the presence of residual RDH function within Rdh10^trex/trex embryos. The residual RA in these embryos could result from a small amount of RDH activity remaining for the mutant RDH10 protein, or to the function of unidentified RDH or ADH enzymes.

In order to directly evaluate the contribution of RDH10 in embryonic RA synthesis we sought to completely eliminate RDH10 function from developing mouse embryos. To that end, we generated mice null for RDH10 activity utilizing ES cells produced by the Chori-Sanger-UC Davis (CSD) arm of the international Knockout Mouse Project (KOMP) consortium. The CSD KOMP knockout construct is designed to create a series of alleles that can be generated by sequential deletion through exposure to Cre or FLP recombinase (Fig. 2A). The initial targeted knock-in construct inserts a splice acceptor-lacZ gene trap cassette between exon 1 and exon 2 of the Rdh10 gene, an allele we call Rdh10^βgeo. The gene trap insertion is expected to produce a null allele by disrupting the endogenous Rdh10 transcript and producing instead a reporter lacZ fusion transcript. Exposure of the gene trap allele to FLP-recombinase can then be used to excise the disruptive splice acceptor cassette (flanked by FRT sites). FLP-mediated excision of the gene trap cassette creates a functionally wild type/conditional knockout allele of Rdh10 with loxP sites flanking exon 2, an allele we call Rdh10^flox. Cre-mediated recombination can then be used to excise the loxP-flanked exon 2, thereby producing a null allele we call Rdh10^delta. Exon 2 was chosen for deletion because its absence is predicted to create a frameshift mutation thereby truncating any resulting protein product and targeting the transcript for nonsense mediated decay. Prior to distribution, correct targeting of the gene trap construct to the Rdh10 genomic locus in ES cells was assessed by 5′ and 3′ PCR junction fragment assays at the KOMP production and distribution centers. The quality control tests included long range PCR assays pairing generic targeting cassette primers with Rdh10 locus-specific genomic primers homologous to the sequences upstream and downstream of the target site.

In order to generate mice bearing an Rdh10 knockout allele, we obtained from the KOMP consortium ES cell clone Rdh10_D08, which is heterozygous for the Rdh10 gene trap allele. These ES cells were injected into C57BL6/J blastocysts to produce chimeric mice that were identified by PCR amplification of the βgeo fusion gene from tail DNA samples. Male mice chimeric for the Rdh10 gene trap allele were then mated to identify individual animals transmitting the targeted allele. Mice carrying the Rdh10^βgeo gene trap allele were crossed with mice carrying ubiquitously expressed FLP recombinase [30] to generate a conditional Rdh10^flox allele. Exposure of the gene trap allele to FLP recombinase resulted in excision of the gene trap/βgeo portion of the targeting cassette and leaving a pair of loxP sites flanking exon 2 to create the conditional allele. Rdh10^flox mice were then crossed with mice bearing a female germline expressed Cre recombinase [31]. The resulting exposure of the Rdh10^flox allele to Cre recombinase caused excision of the loxP-flanked exon 2 thereby generating Rdh10^delta allele.

Embryos heterozygous for the Rdh10^βgeo gene trap alleles were assessed for expression of the reporter βgeo transcript by staining for β-galactosidase activity. In these embryos β-galactosidase activity was very weak but the Rdh10^βgeo gene trap transcript could be detected in a pattern similar to that of the endogenous Rdh10 transcript (Fig. 2 B–C). Both the endogenous Rdh10 mRNA, detected by RNA in situ hybridization and the Rdh10^βgeo gene trap reporter are present in the posterior pharyngeal region and the lateral plate mesoderm between the fore- and hind-limb buds. Expression of Rdh10 mRNA and the Rdh10^βgeo gene trap reporter are also present in the isthmus of the midbrain and in the dorsal somites. These sites correspond to the previously reported expression domains for Rdh10 [23], [32].

Animals heterozygous for the Rdh10^βgeo allele are viable and fertile, with no overt abnormalities. Homozygous Rdh10^βgeo/βgeo embryos, however, exhibit all the hallmarks of extreme retinoid deficiency, with a phenotype remarkably similar to embryos lacking Aldh1a2. At E9.5, Rdh10^βgeo/βgeo embryos display defects in embryo growth and axial turning (Fig. 3A–B). Posterior pharyngeal arches are absent. Otic vesicles are present, but reduced in size. Somites are small and extension of the body axis is impaired. Hearts are unlooped and edemic. The parallels between the Rdh10^βgeo/βgeo phenotype and phenotypes of Aldh1a2^−/− mutants indicate that RDH10 deficiency is as deleterious to embryonic growth and patterning as is ALDH1A2 deficiency.

Mice carrying the Rdh10^βgeo allele were crossed with Rdh10^trex to generate compound heterozygote Rdh10^βgeo/trex embryos. The compound heterozygous embryos displayed a range of phenotypes (Fig. 3 C–D). The most severely affected resembled Rdh10^βgeo/βgeo embryos and the mildest were similar to Rdh10^trex/trex embryos. Failure of the Rdh10^βgeo allele to complement the Rdh10^trex mutation provides genetic confirmation that the Rdh10 conditional construct is correctly targeted to the Rdh10 locus.

Similar to the Rdh10^βgeo mice, animals heterozygous for the Rdh10^delta mutation are also viable and fertile with no obvious defects. Rdh10^delta/delta embryos can be recovered at the expected Mendelian frequency at E9.5 and E10.5 but viable homozygous embryos are not recovered at later stages (Rdh10^delta/delta genotype: E9.5 n = 12/47, E10.5 n = 4/10, E11.75–E12.5 n = 0/32). In the homozygous state, the Rdh10^delta allele produces all the classic hallmark phenotypes of severe retinoid deficiency (Fig. 4 A–D). Rdh10^delta/delta mutants have defects in growth and axial turning. Axial turning defects are most apparent at E9.5, when many Rdh10^delta/delta embryos appear to be caught in an abnormal leftward “J” shape or, occasionally, a rightward “L” body shape when viewed from the front (Fig. 4 B). Some embryos have a milder phenotype in which the tailbud is positioned in front of the embryo, indicating that axial turning is complete. The growth defect in Rdh10^delta/delta mutants is most obvious at E10.5. For wild type embryos, overall embryo size increases substantially between E9.5 and E10.5 (Fig. 4 A, C). Rdh10^delta/delta embryos, in contrast, appear to cease growth and axis extension around E9.25. Rdh10^delta/delta embryos have abnormalities in many tissues and structures affected by retinoid deficiency including the heart, otic vesicle, pharyngeal arches and forelimb buds. Cardiac defects are observed in all Rdh10^delta/delta embryos. Hearts fail to undergo normal looping and chamber formation, remaining, instead, simple tubes oriented vertically and enlarged by edema. Otic vesicles are present, but are tiny and occasionally duplicated. Pharyngeal arch defects are also apparent in Rdh10^delta/delta embryos. First arches are small and malformed, while arches posterior to the first are absent altogether in most embryos. Forelimb buds are reduced in size or absent.

The Rdh10^delta construct was designed to cause a null mutation via deletion of exon 2, generating a stop codon that should disrupt the open reading frame of the Rdh10 mRNA and target the transcript for nonsense mediated decay. In order to verify that the Rdh10^delta transcript encodes a premature stop codon, we characterized Rdh10 cDNA from wild type and Rdh10^delta/delta mutant embryos. The wild type Rdh10 mRNA encodes an open reading frame that spans the 6 spliced exons of the transcript (Fig. 4 E). Amplification of Rdh10 cDNA from Rdh10^delta/delta embryos produced a distinct product 235 bp shorter than the wild type cDNA. Direct sequencing of the mutant Rdh10 cDNA revealed that exon 2 is absent and exon 1 is spliced directly to exon 3, producing a frame shift and generating a stop codon within the open reading frame in the third codon of exon 3 (Fig. 4 E). Protein produced from the Rdh10^delta transcript would be encoded by exon 1 only. Quantitative RT- PCR was performed to assess transcript levels of the Rdh10^delta transcript. Transcript from Rdh10^delta/delta embryos was detected at ∼50% the level produced by wild type embryos, suggesting that the Rdh10^delta transcript is targeted for degradation by nonsense mediated decay (Fig. 4 F).

RDH10 and RALDH2 play sequential roles in synthesis of RA. In order to assess the extent to which the new Rdh10^delta and Aldh1a2^gri mutations disrupt RA synthesis, we utilized the RARE-lacZ reporter transgene, which allows the level and distribution of RA signaling to be visualized by staining for β-galactosidase activity [33]. Loss of RALDH activity in Aldh1a2^gri/gri mutant embryos causes a severe reduction in RA signaling (Fig. 5 A–B), as has been observed for previously reported knockout alleles of Aldh1a2 [8], [27]. Whereas wild type littermate embryos have strong RA signaling throughout the trunk and within the forebrain, homozygous Aldh1a2^gri/gri mutant embryos are almost completely devoid of RA signaling. A small residual amount of RA signaling can be detected in mutant embryos in the forebrain, consistent with the RA signaling in previously reported alleles [8], [27]. In the mutants we examine here, there is also a very small amount of residual RA signaling detected within the trunk, specifically within the neural tube region at the level of the first few somites. These data demonstrate that the Aldh1a2^gri mutation is either a null allele or extreme hypomorph. The residual RA signaling observed within the forebrain of mutant embryos corresponds to the signal in previously reported knockout alleles of Raldh2 and likely results from activity mediated by RALDH1 and RALDH3. The small amount of RA signaling detected in the trunk of Aldh1a2^gri/gri mutant embryos, which has also been observed in one of the established Aldh1a2 knockout alleles [34], may result from a tiny fraction of wild type activity remaining in the Aldh1a2^gri allele or from variation in the null phenotype on different strain backgrounds.

Analogous to the RA attenuation resulting from disruption of Aldh1a2, loss of RDH10 function in the Rdh10^delta/delta mutants also elicits a dramatic reduction in RA signaling (Fig. 5 C–J). At E8.5 RA signaling in mutant embryos is essentially absent (Fig. 5 C–F). At E9.5 residual RA signaling can be detected in Rdh10^delta/deltaembryos within the trunk neural tube region, the extent and pattern being slightly broader than is seen for the Aldh1a2^gri/gri mutants (Fig. 5 G–J). Residual RA signaling within the forebrain, which is observed in Aldh1a2^gri/gri mutants, is completely abolished in Rdh10^delta/delta embryos. For the Rdh10^delta/delta mutant embryos, as with the Rdh10^trex/trex mutant embryos, the residual RA signaling in the trunk region is localized to the neural tube. Rdh10^delta/delta embryos also retain a small amount of residual signaling in the heart and tissue dorsal to the heart. Rdh10^delta/delta embryos have variable residual RA signaling that is somewhat reflective of their variability in phenotype. These data indicate that, at the critical E8.0–E8.5 stage of development, RDH10 is essential for nearly all RDH activity within the early developing embryo. At E9.5 and later stages, the residual RA detected indicates that other enzymes can contribute a small amount of RDH activity to metabolism of Vitamin A into RA and the small amount of RA produced results in signaling concentrated within the trunk neural tube.

The severe phenotype of the Rdh10^βgeo/βgeo and Rdh10^delta/delta mutants and dramatic reduction in RA synthesis, indicate that RDH10 is responsible for most RDH activity within the early embryo and raises the possibility that the RDH reaction could function as a control point in converting Vitamin A to RA. We therefore sought to determine if embryo RA levels feed back to regulate transcription of the Rdh10 gene. To that end we evaluated Rdh10 expression in Aldh1a2^gri/gri mutant embryos in which RA production is severely limited. Rdh10 expression levels were assessed by quantitative RT-PCR of E9.5 Aldh1a2^gri/gri mutant embryos and their wild type and heterozygous littermates (Fig. 6 A). The RA deficient Aldh1a2^gri/gri embryos have significantly elevated Rdh10 expression levels relative to wild type embryos, with expression levels in homozygous mutant embryos ranging between 1.3-fold and 1.8-fold higher than wild type and heterozygous littermates. The elevated Rdh10 transcript in Aldh1a2^gri/gri mutant embryos was expressed in the normal spatial distribution within the embryo at E8.5 (Fig. 6B). These data showing elevation of Rdh10 expression in RA-deficient Aldh1a2^gri/gri mutant embryos clearly demonstrate that Rdh10 transcription is regulated by feedback from the metabolic product RA in mice. Interestingly, Rdh10 contains a potential RAR-RXR (RARE) sequence within exon 5, however to date it is not known whether this is a functional motif nor if it acts in a negative or positive fashion in response to retinoid signaling as has been observed for other RAREs. Nonetheless, our results provide compelling evidence that the RDH reaction is an important control point in regulating metabolism of Vitamin A into RA in mammals and sets the stage for future examination of the complexities of feedback and feedforward enzymatic regulation during this process.

Vertebrate embryonic development depends upon precise spatial and temporal control of RA signaling. Two sequential enzymatic steps are required for metabolism of retinol into RA, the RDH first reaction and the RALDH second reaction (Fig. 6C). Biochemically, either step could function as a control point for regulating synthesis of RA. However, because the RDH first reaction was initially attributed to ubiquitous or redundant members of the ADH family and was thought to occur in an unregulated fashion, the initial step of Vitamin A metabolism was interpreted as being unimportant for the control of RA synthesis. Over time this led to acceptance of the view that that control of Vitamin A metabolism and RA synthesis was driven predominantly by the RALDH second step.

The view that the RDH first reaction occurs in a ubiquitous and unregulated fashion was challenged by the identification of Rdh10^trex, a mutation that dramatically impairs RA synthesis via disruption of RDH10, a spatially and temporally regulated membrane-bound SDR. However, because the RDH reaction had previously been thought to be carried out by redundant or ubiquitous ADH family members, and because the Rdh10^trex mutant embryos exhibited a milder phenotype and survived longer than embryos lacking Aldha2, the RDH first reaction continued to be thought of as relatively unimportant for the spatiotemporal control of RA synthesis. The residual production of RA and the less severe phenotype of Rdh10^trex embryos compared to Aldh1a2 mutant embryos indicated that some RDH activity remained intact in the Rdh10^trex embryos, mediated possibly by remaining activity of the hypomorphic Rdh10^trex point mutant enzyme, or by other RDH enzymes, or by ADH enzymes. Given that genetic or environmental perturbations causing excessive or insufficient levels of RA cause embryonic malformations, it is important to resolve the question of which enzymes catalyze the RDH reaction during embryogenesis and, further, to understand whether these enzymes function redundantly and ubiquitously or if their regulated expression may serve as a control point for production of RA.

To gain insight into the enzymes and regulation of RA synthesis, we have generated a knockout mutation of Rdh10 and demonstrate that complete loss of the RDH10 enzyme results in embryonic death ∼E10.5. Rdh10^delta/delta embryos have severe defects in growth, axial turning, and cardiac and pharyngeal development. The severe embryonic phenotypes of Rdh10^delta/delta mutants unequivocally demonstrate that redundant or ubiquitous enzymes, including the members of the ADH family, do not contribute significantly to physiologically relevant production of RA during embryogenesis. Moreover, the severity of the phenotypes resulting from loss of an individual enzyme, RDH10, provides clear proof that the first RDH reaction of Vitamin A metabolism is as critical a control point in the synthesis of RA as the second RALDH reaction.

Although the phenotype of Rdh10^delta/delta mutant embryos is very severe, it is not identical to the extreme phenotype of loss of Aldha2. Loss of RALDH2 function results in embryos that apparently cease to grow at by E8.5, whereas loss of Rdh10 results in embryos that apparently cease to grow at around E9.0. The slightly milder phenotype of the Rdh10^delta/delta embryos is likely due to the small amount of residual RA produced at E9.5 within the trunk region of affected embryos. The identity of the enzymes responsible for the residual trunk RA at E9.5 is unknown. It is possible the small amount of residual RA production in Rdh10^delta/deltamutant embryos represents the contribution of other, as yet unidentified, SDR enzymes closely related to RDH10 or, alternatively, ADH enzymes. Recently, we demonstrated that a compound knockout of multiple Adh genes does not impair the residual RA synthesis in Rdh10^trex embryos nor exacerbate the phenotype [35], a result which argues against a role for ADHs. Whichever enzymes are ultimately found to be responsible for the small amount of residual RA, it is clear that they are insufficient to produce physiologically relevant levels of RA and are unable to compensate for loss of Rdh10.

Concurrent with the present study, a distinct knockout allele of Rdh10, a deletion of Rdh10 exon 4, was reported in the literature [36]. The phenotypes of the Rdh10 exon 4 deletion allele (Rdh10^−/−)are largely consistent with the Rdh10^delta allele we describe here. In both cases, the majority of embryos survive until ∼E10.5 and exhibit defects in growth and extension of the body axis, absence of posterior pharyngeal arches, compacted somites, and small or duplicated otic vesicles. RA signaling is also notably absent from the forebrain region in both knockout models at this stage. The primary difference is that we do not observe embryo survival beyond E10.5 in our Rdh10^delta mutants, whereas a small fraction (∼10%) of exon 4 deletion mutants survive until E12.5. This difference may be related to variation in strain background or the relative importance of the individual exons for Rdh10 activity that were deleted. We also examined RA signaling in Rdh10^delta/delta embryos in a broader range of embryos and observe that between the 1–6 somite stages (Fig. 5C–F), RA signaling is not detected in Rdh10^delta/delta embryos compared to the robust detection of reporter activity in control embryos. Taken together, these data indicate the important role of Rdh10 in RA signaling during the early critical E8.0–8.5 window of development.

In addition to showing that RDH10 is the predominant enzyme responsible for embryonic conversion of retinol to retinal, the data presented here reveal that Rdh10 gene expression is modulated by presence or absence of RA, being upregulated by RA deficiency in Aldha2^gri/gri mutants. The upregulation of Rdh10 expression when RA synthesis is impaired demonstrates that Rdh10 is a point for feedback control in Vitamin A metabolism and RA synthesis in mammals (Fig. 6C). A similar mode of regulation has been observed in Xenopus and zebrafish, indicating that the mechanistic feedback control of Rdh10 to modulate the RA synthesis pathway is evolutionarily conserved [24], [25]. The evolutionary conservation of feedback regulation of the Rdh10 gene by RA lends further support to the concept that RDH10-mediated oxidation of retinol to retinal plays an important role in regulating the key process of embryonic Vitamin A metabolism and RA synthesis.

The conversion of retinol to retinal is a reversible reaction. Enzymes of the SDR family can mediate both the forward and the reverse directions. Transcriptional regulation of SDR enzymes to mediate RA homeostasis is observed in other contexts. In zebrafish embryos, excess RA induces expression of dhrs3a, an SDR that catalyzes the reduction of retinal into retinol [25]. In adult mouse liver, chronic exposure to ethanol causes complex changes in RA metabolism and is associated with transcriptional changes in RA metabolic genes including induction of the SDR Dhrs9 [37]. Together these data suggest that reversible interconversion of retinol to retinal by SDRs serve as a key control step in RA homeostasis. Consistent with this, biochemical analysis of Vitamin A metabolism in embryos has indicated that the first enzymatic reaction, the oxidation of retinol to retinal, is the rate-limiting step in the synthesis of retinoic acid [38].

The finding that the RDH reaction, mediated primarily by RDH10, is a major control point in RA synthesis has implications for understanding the tissue specificity and sub-cellular localization of the sequential steps of RA production. With respect to tissue distribution of Vitamin A metabolism and RA synthesis, it is notable that the expression pattern of Rdh10 within an embryo is very similar to, but not identical to that of Aldh1a2 [23], [32]. In chick, the expression pattern of Rdh10 has also been shown to overlap with that of the gene encoding the retinol transporter Stra6 [26]. Together, these data suggest that retinol is imported into cells, oxidized into retinal and subsequently into the product RA all largely within the same population of cells. We have previously shown that maternal administration of intermediate retinal, which exposes all cells of the embryo to the retinal intermediate, can rescue RDH10 deficient embryos to birth [39] and recently this has now been extended postnataly [36]. Although viable, these rescued embryos still exhibit some abnormalities including limb and behavioral defects. The defects remaining in retinal rescued animals indicate that while embryo viability is not dependent on spatially restricted retinal production, proper tissue patterning and development are.

With respect to sub-cellular localization of Vitamin A metabolism, the medium chain reductase ADH enzymes are known to function within the cytosol, while the SDR RDHs function in a membrane bound capacity. RDH10 is associated with the membrane-bound compartment [40] and its membrane association is necessary for its function [41]. Recently, we demonstrated that the RDH reaction occurs not in the cytosol where ADHs reside, but in a membrane-bound cellular compartment such as that occupied by RDH10 [39]. Within the cytosol, RBP1 binds to retinol and inhibits oxidation. The membrane localization of RDH10 is functionally important, presumably because it allows RDH10 to oxidize RBP1-free pools of retinol available in the membrane compartment (Fig. 6C). These findings suggest that ADHs do not contribute significantly to retinol oxidation because they cannot oxidize the RBP1-bound retinol in the cytosol and are unable to access RBP1-free retinol from membranes (Fig. 6C). They explain why the cytosolic ADHs, although widely expressed, cannot compensate for the loss of RDH10 during embryogenesis.

The lack of function of ADHs in embryonic retinol oxidation is consistent with studies of individual and compound Adh knockout mice, which suggest that ADHs function primarily in governing post-natal retinoid homeostasis (reviewed in [35]). In adult mice, ADH1 provides considerable protection against vitamin A toxicity [42], [43] whereas ADH4 promotes survival during vitamin A deficiency [44], a difference which illustrates distinct physiological functions for these two enzymes in retinol metabolism. In contrast, ADH3 appears to function redundantly in both the prevention of retinol toxicity and Vitamin A deficiency, a dual role consistent with its ubiquitous expression. Collectively, these biochemical and compound mutant analyses strongly argue against a significant role for ADHs in Vitamin A metabolism and RA synthesis during embryonic development.

Given that Rdh10 is the predominant enzyme oxidizing retinol to retinal and the mounting evidence that ADHs make little if any contribution, we conclude that the conversion of retinol to retinal within an embryo must occur predominantly, if not completely, within a membrane-bound compartment. Collectively these data demonstrate the critical roles played by Rdh10 in mediating Vitamin A metabolism and synthesis of RA and as a nodal point in governing the feedback regulation of RA production. Considering the relatively discrete sites of Rdh10 expression, the allelic series of Rdh10 alleles including conditional alleles described here, will facilitate important strategies to characterize the consequences of impaired Vitamin A metabolism and RA synthesis in a precisely controlled spatiotemporal manner.

The Rdh10 mouse strains used for this research project were generated from ES cells obtained from the trans-NIH Knockout Mouse Project (KOMP) Repository, a NCRR-NIH supported strain repository. ES cells were created by CSD consortium from funds provided by the KOMP. To inquire about KOMP products go to www.komp.org or email service@komp.org. The Rdh10_D08 (full clone name, EPD0149_1_D08) ES cell clone was produced by the CSD consortium (Childrens Hospital of Oakland Research Institute, The Sanger Institute, and UC Davis). Quality control for correct targeting was assessed by the KOMP production and distribution centers using PCR based assays that spanned the 5′ and 3′ junction fragments. The parental ES cell line was JM8.N4, which is a C57BL/6N background. For the mice described in this report, we obtained Rdh10_D08 ES cells from KOMP and injected them into C57BL/6 blastocysts. The Rdh10 gene trap line was thereby established and subsequently maintained on a C57BL6/J background. Some animals, such as those crossed with transgenic reporter lines, were mixed with other backgrounds. The KOMP repository independently generated mice from the same clone of ES cells. The strain name for the KOMP generated mice is C57BL/6N-Rdh10^{tm1a(KOMP)Wtsi}.

Aldh1a2^gri mutant mice were originally described as grimace mutants obtained as part of a chemical mutagenesis screen for defects in early craniofacial development [29]. Mice carrying the FLP recombinase (FLPeR) in the Rosa26 locus on a C57BL6 background [30] available from the Jackson Laboratory as stock #009086 B6.129S4-Gt(ROSA)26Sor^tm1(FLP1)Dym/RainJ. Mice carrying a Cre recombinase transgene expressed in oocytes driven by ZP3 promoter sequences [45] are available from the Jackson Laboratory as stock #003651 C57BL/6-TgN(Zp3-Cre)93Knw.

All genotyping was accomplished by real-time PCR assays performed by the commercial genotyping service Transnetyx (Transnetyx.com). For the wild type Rdh10 allele assay, a probe denoted “Rdh10-1 WT” was used to detect a fragment spanning the region of the unmodified genomic locus corresponding to the insertion site of the knock in construct. For the Rdh10^βgeo allele, an assay was used to detect the presence of lacZ and neomycin, corresponding to the βgeo cassette. For the Rdh10^delta allele, a probe denoted “Rdh10-1EX” was used, which detects an amplified region spanning the junction fragment covering the single recombined FRT site and the 3′ loxP site. To test for homozygosity of the null allele, samples were screened for absence of Rdh10 exon 2 using an assay probe designated “Rdh10-2 WT”. The following sequences indicate the relevant amplified regions, excluding primers:

Rdh10 wild type allele

GCTCATTCATTGCTTGGCTTTGAATGAGTTTAGATTGTTGCTTACAGTTTGGATTTGTGAGCCTTAAAGTGAATTTC

Rdh10^delta allele

AGTTCCTATTCCGAAGTTCCTATTCTCTAGAAAGTATAGGAACTTCGTCGAGATAACTTCGTATAGCATACATTATACGAAGTTATGGTCTGAGCTCGCC

To assay for Aldh1a2 expression in Aldh1a2^gri mutant embryos, qRT-PCR TaqMan assay was used with the following probe and primers:

Aldh1a2Fwd-TGA TAA AAT TCA CGG ATT GAC CAT

Aldh1a2Rev-GGG CTC GTG TCT TGT GAA AGT AA

Aldh1a2TaqMan Probe - CCTGTAGATGGAGACTAT

To assay for Rdh10 expression in wild type and Aldh1a2^gri mutant embryos, qRT-PCR SYBR assay was used with the following primers:

Rdh10-F: TGG TGT GGT TTC TGG ACA TCA CCT

Rdh10-R: CTC TAG CAT CGT TGG AAG AAA GGC

To assay for altered Rdh10 mRNA in Rdh10^delta mutant embryos, qRT-PCR SYBR assay was used with the following primers:

Rdh_q1F: CTGGGACTGTTCAGCACTG

Rdh_q1R: TCTACAAGGTAAGGGCAAACC

Whole mount in situ hybridization for Rdh10 was performed as in [23]. Embryo morphology was visualized by staining whole embryos with DAPI to label all cell nuclei and imaging fluorescent signal. Embryos were fixed in 4% formalin overnight at 4°C then stained with 2 ug/ml DAPI in PBS and imaged by confocal microscopy using an LSM5 Pascal confocal microscope. For each embryo, a Z-stack of confocal slices was collapsed to form a “pseudo-SEM” image. RA signaling was visualized by staining embryos carrying the RARE-lacZ transgene for β-galactosidase activity. A subset of the brightfield embryo images (Fig. 2D, Fig. 5 G–J, 6B) were processed with Helicon Focus image processing software to compile and render the focused regions of multiple focal planes as a single in-focus image. A minimum of 5 embryos were used for each parameter analysed unless otherwise stated.

Transnetyx automated qPCR genotyping: www.Transnetyx.com

Knockout Mouse Project: www.KOMP.org

Rdh10 conditional knockout vector: http://www.sanger.ac.uk/htgt/report/project_gene_report?project_id=34040