Drs. Ulven and Glover contributed equally to the execution of this work.
Quantitative axial profiles of retinoic acid in the embryonic mouse spinal cord: 9-Cis retinoic acid only detected after all-trans-retinoic acid levels are super-elevated experimentally
Article first published online: 2 OCT 2001
Copyright © 2001 Wiley-Liss, Inc.
Volume 222, Issue 3, pages 341–353, November 2001
How to Cite
Ulven, S. M., Gundersen, T. E., Sakhi, A. K., Glover, J. C. and Blomhoff, R. (2001), Quantitative axial profiles of retinoic acid in the embryonic mouse spinal cord: 9-Cis retinoic acid only detected after all-trans-retinoic acid levels are super-elevated experimentally. Dev. Dyn., 222: 341–353. doi: 10.1002/dvdy.1184
- Issue published online: 2 NOV 2001
- Article first published online: 2 OCT 2001
- Manuscript Accepted: 27 JUN 2001
- Manuscript Received: 17 JAN 2001
- Norwegian Medical Research Council
- Throne Holst Fund
- Nansen Fund
- vitamin A;
- 9-cis-retinoic acid;
- all-trans-retinoic acid;
- spinal cord;
Studies using bioassays in normal mice and gene activation in transgenic reporter mice have demonstrated peaks of retinoic acid receptor (RAR) signaling in the brachial and lumbar regions of the spinal cord. Recently, Solomin et al. (Solomin et al.  Nature 395:398–402) detected a retinoid X receptor (RXR) signal in the same region of the developing spinal cord at a slightly later stage than the RAR signal. This finding raises the question of which retinoid ligands underlie RAR and RXR signaling in this part of the embryo. Quantitative measurements of regional differences in retinoid profiles have not been reported previously due to limitation in the sensitivity and specificity of available retinoid detection methods. Here, by using a recently developed ultrasensitive HPLC technique (Sakhi et al.  J. Chromatogr. A 828:451–460), we address this question in an attempt to identify definitively the endogenous retinoids present in different regions of the spinal cord at the stages when regional differences in RAR and RXR signaling have been reported. We find a bimodal distribution of all-trans retinoic acid (at-RA), the ligand for RARs, and relate this to the expression of several retinoid-synthesizing enzymes. However, we do not detect 9-cis-retinoic acid (9-cis-RA), the putative RXR ligand, in any region of the spinal cord unless retinoid levels are massively increased experimentally by gavage feeding pregnant mice with teratogenic doses of at-RA. This study provides for the first time quantitative profiles of endogenous retinoids along the axis of the developing spinal cord, thereby establishing a foundation for more definitive studies of retinoid function in the future. It sets definite limits on how much 9-cis-RA potentially is present and demonstrates that at-RA predominates over 9-cis-RA by at least 30- to 180-fold in different spinal cord regions. © 2001 Wiley-Liss, Inc.
The function of vitamin A during embryonic development is mediated through the action of two classes of specific retinoid receptors, the retinoic acid receptor (RAR) and the retinoid X receptor (RXR). Both classes are ligand-dependent transcription factors belonging to the nuclear hormone receptor superfamily (Mangelsdorf and Evans, 1995; Mangelsdorf et al., 1995; Chambon, 1996). Little is known about the endogenous retinoid ligands activating these receptors in vivo. A large number of in vitro experiments have shown, however, that members of the RAR family are activated by several naturally occurring retinoids, including all-trans-retinoic acid (at-RA), 9-cis-retinoic acid (9-cis-RA), all-trans-4-oxo-retinoic acid (at-4-oxo-RA), all-trans-4-oxo-retinal (at-4-oxo-RAL), all-trans-4-oxo-retinol (at-4-oxo-ROH), and all-trans-3,4-didehydro-retinoic acid (at-dd-RA). Members of the RXR family are more selective, being efficiently activated in vitro by 9-cis-RA, 9-cis-3,4-didehydro-retinoic acid (9-cis-dd-RA) and 9-cis-4-oxo-retinoic acid (9-cis-4-oxo-RA) (Mangelsdorf and Evans, 1995; Mangelsdorf et al., 1995; Chambon, 1996). These ligands are usually synthesized in vivo by complex metabolic systems, involving numerous enzymes and binding proteins (see Ulven et al., 2000 and references therein).
The central nervous system (CNS) is a major site of retinoid action, as both vitamin A deficiency and excess cause abnormal neural patterning and development (Durston et al., 1989; Altaba and Jessell, 1991; Maden et al., 1997; Niederreither et al., 2000). Moreover, null mutations of specific combinations of nuclear retinoid receptors produce the same spectrum of abnormalities (Lohnes et al., 1994; Dupe et al., 1999). In addition to these teratogenic effects, retinoids have been implicated as pivotal regulators of the normal determination and differentiation of neurons. For example, experimental manipulation of retinoid synthesis modifies gene expression patterns and the differentiation of specific neuron classes in the developing spinal cord (Forehand et al., 1998; Sockanathan and Jessell, 1998; Pierani et al., 1999). At least some of these effects seem to derive from the neighboring paraxial mesoderm, either directly by retinoids synthesized in the mesoderm or potentially through the action of other factors whose synthesis in the mesoderm is regulated by retinoids (Forehand et al., 1998; Gould et al., 1998; Sockanathan and Jessell, 1998).
Several studies have addressed the axial localization of retinoid signaling in the developing spinal cord. Gene activation in transgenic reporter mice (Reynolds et al., 1991; Colbert et al., 1993; Solomin et al., 1998; Mata et al., 1999) and bioassay systems (Colbert et al., 1993; McCaffery and Drager, 1994) have demonstrated peaks of retinoid signaling in the brachial and lumbar regions of the spinal cord. In situ hybridization in mouse embryos and immunohistochemistry in chicken embryos has demonstrated strong expression of the retinoic acid synthesizing enzyme RALDH2 in the somites, the meninges, the roof plate of the neural tube, and particularly in motoneurons in the brachial and lumbar regions, the last-mentioned, thus, paralleling the bimodal axial profile of retinoid signaling (Niederreither et al., 1997; Sockanathan and Jessell, 1998; Berggren et al., 1999).
Recently, Solomin et al. (1998) developed an elegant transgenic reporter mouse model in which the DNA-binding domain of GAL4 is fused to the ligand-binding domain of either RAR or RXR. They demonstrated activation of the RAR construct in the brachial and lumbar part of the spinal cord starting at 10.5 days post coitum (dpc). They also observed activation of the RXR construct in the same regions slightly later, starting at 11.0–12.0 dpc.
Transgenic reporter mouse models demonstrate unequivocally whether a specific type of receptor has been activated, but they provide only qualitative information about the signaling involved. They do not show which ligands have been bound or in which concentrations. Nor do they provide a reliable indication of temporal dynamics, because the time course of the reporter gene signal is not only due to ligand-receptor and receptor-response element interactions, but also to reporter gene transcript and protein lifetimes, which could last hours beyond termination of ligand-receptor binding. Thus, to advance our understanding of how retinoid signaling regulates spinal cord development, direct, quantitative measurements of endogenous retinoid ligands and receptors are essential.
By using high-pressure liquid chromatography (HPLC), Horton and Maden (1995) detected substantial amounts of at-RA in homogenates from whole spinal cords of 10.5 dpc mouse embryos, but did not detect other candidate ligands such as 9-cis-RA, 3,4-didehydro-retinoids, or 4-oxo-retinoids. Presumably, the endogenous at-RA is responsible for the various reports of RAR activation. Because 9-cis-RA is considered the principal ligand for RXRs, the absence of detectable 9-cis-RA raises the question as to which ligands drive the RXR activation observed by Solomin et al. (1998) starting at 11 dpc. One possibility is that 9-cis-RA appears at later stages than assayed by Horton and Maden (1995), another is that 9-cis-RA is present but below the limit of detection, and a third is that RXR activation is driven by other ligands (Eager et al., 1992; Harmon et al., 1995; Kitareewan et al., 1996) or by ligand-independent mechanisms such as phosphorylation (Lefebvre et al., 1995; Rochette-Egly et al., 1995).
As we have recently developed a substantially more sensitive HPLC method (Sakhi et al., 1998), we now re-address this question in an attempt to identify definitively the endogenous retinoid ligands present at the stages when Solomin et al. (1998) report RAR and RXR receptor activation. Moreover, by microdissection of over 100 embryonic spinal cords, we extend the findings of Horton and Maden (1995) by measuring different regions of the spinal cord separately to obtain for the first time quantitative axial profiles. We find a bimodal distribution of at-RA, which we relate to the expression of RALDH2 (assayed by immunohistochemistry) and of several other retinoid-synthesizing enzymes (assayed by reverse transcription-polymerase chain reaction [RT-PCR]). However, we do not detect 9-cis-RA in any region of the spinal cord unless at-RA levels are massively increased experimentally by gavage feeding pregnant mice with teratogenic doses of at-RA. Therefore, these results set a definite ceiling on the maximum amount of 9-cis-RA that could be present (if indeed 9-cis-RA is normally present at all) in the different spinal cord regions.
Axial Profiles of Endogenous Retinoids
Studies using bioassays and transgenic reporter mice have revealed axial profiles of retinoid synthesis and signaling along the developing spinal cord with peaks in the brachial and lumbar regions (Reynolds et al., 1991; Colbert et al., 1993; McCaffery and Drager, 1994; Solomin et al., 1998; Mata et al., 1999). To identify which retinoids are actually present, and to quantitate their axial differences, we used a combination of microdissection and sensitive HPLC techniques to measure endogenous retinoids in the brachial, thoracic, and lumbar regions of 11.0–12.0-dpc spinal cord. The tissue isolation and preparation, chromatographic separation, and electrochemical detection methods we used have been fully validated and show excellent precision and reproducibility (Sakhi et al., 1998). The mole limits of detection are 70 fmoles for at-ROH, 27 fmoles for at-RA, and 67 fmoles for 9-cis-RA.
In initial experiments, we dissected 55 spinal cords into three regions (Fig. 1), made tissue homogenates, and injected homogenate volumes equivalent to each region from 40 spinal cords into the HPLC system (homogenate volumes equivalent to 4.4 spinal cords were used for protein measurements and homogenate volumes equivalent to 4.4 spinal cords were used for at-ROH quantification, see below). We detected prominent peaks corresponding to at-ROH, at-RA, and an unidentified compound (Fig. 2A). The identification of at-ROH and at-RA was confirmed by using HPLC-mass spectrometry (MS) and full-spectrum ultraviolet (UV) analysis on a separate homogenate of the lumbar regions from 10 spinal cords dissected on a separate occasion (Fig. 3). By comparison with retention times of known standards, we could exclude all of the compounds listed in Table 1 as possible candidates for the unidentified prominent peak, which moreover did not exhibit the typical UV-spectral characteristics of a retinoid. We also detected minor peaks. One of these corresponded either to at-RAL, or 13-cis-ROH, or both; we could not discriminate these by full-spectrum UV analysis because the sample was too small. Several other minor peaks could not be identified, but could be excluded from the list in Table 1 on the basis of retention time (Fig. 2A). The same results were obtained in a separate assay of 15 spinal cords dissected on a separate occasion.
|13-cis-retinoic acidc||13-cis-RA||After gavage feeding|
|9-cis-retinoic acidc||9-cis-RA||After gavage feeding|
|All-trans-4-oxo-retinoic acidc||At-4-oxo-RA||After gavage feeding|
Because retinoid receptors can also be activated by compounds other than retinoids (Eager et al., 1992; Harmon et al., 1995; Kitareewan et al., 1996), and we obtained a prominent peak that is not a retinoid, we assayed other potential ligands, focusing on relevant fatty acids. Arachidonic acid (ARA), docosahexaenoic acid (DHA), and eicosapentaenoic acid (EPA) standards were all detectable with our ECD system, but none of these corresponded to any of the peaks in the tissue chromatograms. In general, nonconjugated fatty acids in physiological plasma concentrations give weak or negative responses in the described ECD system.
Both at-ROH and at-RA, as well as the unidentified prominent peak, exhibited bimodal axial profiles with the highest level in brachial, the lowest level in thoracic, and an intermediate level in the lumbar region (Table 2). Amounts of at-ROH substantially exceeded those of at-RA, being approximately 3.5, 6, and 2 times higher in the brachial, thoracic, and lumbar regions, respectively, when normalized to protein (Fig. 4A).
To estimate the tissue concentrations of the extant at-RA, we divided the measured amounts by the approximate volumes of the dissected brachial, thoracic, and lumbar spinal cord regions. This method provides a minimum concentration that assumes an even distribution of the at-RA throughout the tissue and probably underestimates the effective concentrations. The volumetric concentration estimates were 74 nM (brachial), 36 nM (thoracic), and 200 nM (lumbar).
To compare the retinoid profiles in the spinal cord with those in the body wall, we analyzed the retinoid contents of the brachial, thoracic, and lumbar regions of the remainder of the embryos after evisceration and removal of the spinal cord (Figs. 1, 2B). This analysis was done three times, each on homogenates from 15 embryos, with samples representing an additional two or three embryos taken for protein and separate ROH analyses. The profiles in the body wall (Fig. 4B) differed from those in the spinal cord in two ways. First, neither at-ROH nor at-RA exhibited bimodal axial profiles in the body wall. Indeed, at-RA levels were higher in thoracic than in brachial and lumbar regions, despite the inclusion of the extremities in the brachial and lumbar regions. Second, at-ROH levels exceeded at-RA levels by much more in the body wall than in the spinal cord (being approximately 19, 11, and 10 times higher in the brachial, thoracic, and lumbar regions, after normalization to protein).
Nondetection of 9-cis-RA
We did not detect 9-cis-RA in any region of the spinal cord or body wall (Fig. 2). However, nondetection does not necessarily mean absence. Given the detection limit of 9-cis-RA in our electrochemical detection system, if 9-cis-RA were present in any part of the spinal cord, it would have to be in amounts less then 67 fmoles. This amount would give ratios of at-RA to 9-cis-RA greater than 179, 30, and 108 in brachial, thoracic, and lumbar regions, respectively. As will be described in a later section, our assay system is capable of detecting 9-cis-RA in embryonic tissue, so the lack of detection in the normal spinal cord at this stage of development represents a definite upper limit equivalent to the detection limit.
To test whether the lack of detection of 9-cis-RA was due to nonenzymatic isomerization during dissection, we carried a 9-cis-RA standard solution through the entire procedure. To test whether lack of detection was due to degradation of 9-cis-RA during postdissection procedures, we added known amounts of 9-cis-RA to parallel tissue homogenates (cells in homogenates were lysed by freeze-thawing to release intracellular contents), both before and after protein extraction. In no case did we observe degradation or isomerization of the 9-cis-RA.
mRNA Expression of Retinoid Metabolic Enzymes
By using RT-PCR we analyzed the expression of genes coding for enzymes known to oxidize retinol into retinoic acid. In each experiment, we assessed β-actin expression as an internal control, and we used either mRNA isolated from adult liver, or from whole embryos (9.5 or 10.5 dpc), tissues known to express the various enzymes, as positive controls for each primer pair. One sample in each experiment was run without reverse transcriptase to ensure that there was no amplification of genomic DNA.
We analyzed two classes of enzymes. One includes the alcohol dehydrogenases (ADH) and the short chain dehydrogenase/reductases (SDR, includes cis-retinol/androgen dehydrogenases (CRAD) and retinol dehydrogenases (RDH or RoDH)), which catalyze the conversion of retinols to retinals. Of these, ADH4 and CRAD2 are able to convert both at-ROH and 9-cis-ROH, whereas ADH1, RoDH1, and RoDH3 are specific for all-trans-ROH versus 9-cis-ROH and CRAD1 and RDH5 have the opposite specificity. The other enzyme class includes the aldehyde and retinaldehyde dehydrogenases (ALDH, RALDH), which catalyze the conversion of retinals to retinoic acids. Of these, both ALDH1 and RALDH2 oxidize at-RAL and 9-cis-RAL into the respective retinoic acids.
We did not detect CRAD2, RoDH1, or RoDH3 transcripts in whole 11.5-dpc embryos; therefore, we did not assay for these in isolated spinal cord regions. Transcripts coding for ADH1, ADH4, RDH5, CRAD1, ALDH1, and RALDH2 were detected in each region of the spinal cord (Fig. 5). CRAD1 and RALDH2 were the only transcripts that in all measurements showed a consistently bimodal pattern, with a lower amount of product in the thoracic relative to the brachial and lumbar regions. This finding was true even after normalization to protein (obtained from the same dissection), which raises the relative thoracic amount (Fig. 5). The other transcripts gave more variable patterns that were not convincingly bimodal.
Axial Profiles of RALDH2 Immunoreactivity
Antibodies to RALDH2, but not to CRAD1, are available. Previous studies have demonstrated the presence of RALDH2 immunoreactivity in the spinal meninges, in the roof plate, and in motoneurons of the lateral motor column in the brachial and lumbar regions (Niederreither et al., 1997; Sockanathan and Jessell, 1998; Berggren et al., 1999). We assessed the axial distribution of RALDH2 immunoreactivity in 11.0- to 12.5-dpc mouse embryos to correlate its spatial distribution with the retinoid and RALDH2 transcript data described above.
We confirmed the presence of RALDH2 immunoreactivity in brachial and lumbar motoneurons (Fig. 6). We also found lightly stained cells within the cervical region of the spinal cord. We saw no RALDH2-positive cells in the thoracic region of the spinal cord. In contrast to the discontinuous distribution of RALDH2 immunoreactivity within the spinal cord, immunoreactivity in the surrounding mesenchyme, in particular that which was condensing into the spinal meninges, was continuous along the entire extent of the spinal cord. Nevertheless, the meningeal staining was modulated rostrocaudally, with highest intensity in the cervical and brachial regions, lowest intensity in the thoracic region, and an intermediate intensity in the lumbar region.
Changes in Retinoid Levels After Teratogenic Doses of at-RA
Solomin et al. (1998) showed that RXR activation-mediated lacZ expression, which is normally restricted to the brachial and lumbar regions, has spread throughout the spinal cord 12 hr after the administration of a teratogenic dose of at-RA to pregnant mice at 11.0 dpc. To assess how such treatments affect the levels of the various retinoids in the spinal cord, we similarly administered teratogenic doses of at-RA following several paradigms. Three and 6 hr after administration of 100 mg/kg at-RA to 11.5-dpc pregnant mice (Satre and Kochhar, 1989), there was a massive increase (approximately 200 times) of at-RA levels in a homogenate of five whole spinal cords. Several additional peaks appeared, including at-4-oxo-RA and 13-cis-RA. Moreover, a minor peak corresponding to 9-cis-RA appeared (Fig. 7). We confirmed the identity of this peak as 9-cis-RA by using coelution of an added 9-cis-RA standard.
The amount of 9-cis-RA per spinal cord was 0.16 pmoles at 3 hr and 0.15 pmoles at 6 hr after administration. Amounts in the rest of the embryo were 0.6 pmoles after 3 hr falling to 0.3 pmoles after 6 hr. In comparison, the amount of endogenous at-RA per spinal cord in a homogenate of five whole spinal cords from untreated control embryos was 0.55 pmoles (this is nearly the same as for the measurements shown in Table 2, where addition of brachial, thoracic, and lumbar amounts equals 0.58 pmoles). Thus, despite super-elevation of at-RA levels, 9-cis-RA levels do not reach the normal at-RA level.
Might the 9-cis-RA observed after gavage feeding represent an artefact? We checked the at-RA we administered to the pregnant mice assiduously for purity and detected no contamination by 9-cis-RA. To check whether the 9-cis-RA seen in embryonic tissues after gavage feeding arose by isomerization of at-RA during our procedures, we performed control experiments: (1) at-RA standards in buffer were incubated in Eppendorf tubes under the same dissection, freezing, and storage procedures; (2) the same amount of at-RA as we measured in the embryonic tissues after teratogenic doses was added to homogenates of native embryonic tissue, either before or after protein extraction. Retinoids were assayed in these test cases either immediately or after storage at −70°C. In no case did we observe any isomerization of exogenous at-RA to 9-cis-RA, indicating that our procedures did not artefactually isomerize endogenous at-RA to 9-cis-RA.
Surprisingly, when we administered teratogenic doses of at-RA by using the paradigm of Solomin et al. (1998), namely analysis 12 hr after administration of 20 mg/kg at-RA to 11.0-dpc pregnant mice, we found no increase in at-RA levels in homogenates of brachial, thoracic, or lumbar regions from 20 spinal cords. Nor was 9-cis-RA detected. The absence of an increase is almost certainly because retinoid elevations generated by this treatment have already subsided within 12 hr, based on time courses of retinoid elevation published previously (Creech Kraft et al., 1987, 1989; Satre and Kochhar, 1989; Kochhar et al., 1995; Ward and Morriss-Kay, 1995).
We report here the first quantitated axial profiles of endogenous retinoids in the developing spinal cord, obtained with an advanced HPLC system with separation capability and sensitivity sufficient to provide definitive identification of individual retinoids. The main findings are (1) Amounts of at-RA normalized to protein are approximately 5 and 4 times higher in brachial and lumbar regions than in the thoracic region, and amounts of the precursor at-ROH are approximately 3 and 1.5 times higher in brachial and lumbar regions than in the thoracic region. These bimodal profiles support previous reports of retinoid synthetic activity (McCaffery and Drager, 1994; Berggren et al., 1999) and retinoid-driven reporter gene assays (Reynolds et al., 1991; Colbert et al., 1993; Solomin et al., 1998; Mata et al., 1999). (2) 9-cis-RA is not detected in the untreated spinal cord. If endogenous 9-cis-RA is present at all, it cannot exceed 0.6, 3.4, and 0.9% of the amount of endogenous at-RA in brachial, thoracic, and lumbar regions respectively. (3) 9-cis-RA is detected, however, after super-elevation of at-RA levels. When at-RA levels are increased approximately 200-fold by gavage feeding of pregnant dams, the amount of 9-cis-RA in the entire spinal cord reaches approximately 160 fmoles, or approximately 3.5 times lower than the amount of endogenous at-RA normally present. (4) Transcripts coding for several retinoid synthesizing enzymes are present in the spinal cord, and transcripts for two of these, namely CRAD1 and RALDH2, exhibit bimodal profiles. (5) RALDH2 immunoreactivity exhibits a bimodal profile similar to that seen in the chicken embryo, as it is expressed in the lateral motor column (Sockanathan and Jessell, 1998; Berggren et al., 1999). In addition, there is strong expression in the mesenchyme surrounding the spinal cord, particularly that which is coalescing into the spinal meninges. Meningeal expression is continuous along the entire length of the spinal cord but appears to be modulated with a profile similar to that seen for RALDH2 transcripts and for endogenous retinoids. Finally, RALDH2 immunoreactivity is present in the roof plate with similar intensity along the entire spinal cord.
Bimodal Axial Profiles
In previous studies, bioassays and reporter mice demonstrated the presence of retinoids in the developing neuraxis including the spinal cord, with higher concentration at brachial and lumbar levels relative to thoracic levels (Reynolds et al., 1991; Rossant et al., 1991; Colbert et al., 1993; Maden et al., 1998; Solomin et al., 1998; Mata et al., 1999). This bimodal axial distribution was correlated with retinoid synthesis (McCaffery and Drager, 1994) and with RALDH2-positive motoneurons, which are restricted to brachial and lumbar levels (Niederreither et al., 1997; Sockanathan and Jessell, 1998; Berggren et al., 1999). However, none of these studies used techniques that allowed definitive identification of the retinoids in question. We have shown here that both at-RA and at-ROH exhibit a clear bimodal axial distribution along the spinal cord (see Table 1). An extensive list of other retinoids were assayed (see Table 1), including all that have been implicated as ligands for RARs and RXRs. Of these, only at-RAL and/or 13-cis-ROH were present in detectable amounts. Neither of these functions as ligands. We failed to detect several other compounds that have been implicated as ligands. We wish to emphasise that our HPLC system is substantially more sensitive than any others used for embryonic retinoid measurements previously. Therefore, it seems that at-RA is the principal, if not the only, retinoid that can generate the bimodal retinoid signals seen in previous studies. We cannot exclude that a transient binding of ligand to RXRs at earlier stages of development may have produced the signal seen at 11.0–12.0 dpc.
Our RT-PCR assays of enzyme transcripts shed light on which synthetic pathways may be involved in the synthesis of RA in the spinal cord. The synthesis of RAs is a two-step reaction, involving several enzymes that can oxidize retinols to retinaldehydes, and several enzymes that can further oxidize retinaldehydes to retinoic acids. Transcripts coding for several enzymes participating in each of these steps were present in the spinal cord. CRAD1 and RALDH2 mRNA were the only transcripts that exhibited a bimodal profile.
Antibodies to RALDH2, but not CRAD1, are available. Therefore, we performed an immunohistochemical analysis of RALDH2 expression to correlate the spatial distribution of this enzyme with that of retinoids and enzyme gene transcripts. The immunostaining confirms the tissue distribution of RALDH2 demonstrated previously in both mouse and chicken embryos. RALDH2 is expressed in the roof plate, in motoneurons of the lateral motor column (restricted to brachial and lumbar regions), and in the surrounding meninges and adjacent mesenchyme, as well as other sites further removed from the spinal cord (Zhao et al., 1996; Niederreither et al., 1997; Sockanathan and Jessell, 1998; Berggren et al., 1999; Haselbeck et al., 1999). These earlier studies did not assess axial differences in RALDH2 expression in tissues outside the neural tube. Our findings supplement these by showing that the meningeal expression is also modulated along the length of the cord, in a profile similar to that of endogenous at-RA. Thus, the bimodal profile of retinoids can be related to intramedullary as well as extramedullary sites of synthesis. However, the detection of retinoid signaling in reporter mice clearly has a threshold that is not reached by the lower levels of retinoids present in the thoracic region.
Several manipulations in chicken embryos have shown that retinoid signaling plays a pivotal role in patterning the developing spinal cord. Sockanathan and Jessell (1998) and Pierani et al. (1999) have demonstrated retinoic acid-dependent regulation of ventral progenitor proliferation and ventral neuron differentiation. In particular, retinoic acid is required for the differentiation of motoneurons of the lateral subdivision of the lateral motor column, which are presumably exposed to retinoic acid as they migrate past RALDH2-positive motoneurons in the medial subdivision of the lateral motor column (Sockanathan and Jessell, 1998). Thus, the restriction of RALDH2 expression to brachial and lumbar levels within the ventral part of the spinal cord can explain why the lateral subdivision of the lateral motor column only differentiates at these levels. Pharmacologic experiments using receptor-specific antagonists suggest that this effect of retinoic acid requires the activation of both RARs and of RXRs. Retinoic acid is also required for the differentiation of specific classes of ventral interneurons (Pierani et al., 1999), but because these interneuron classes are also found in the thoracic region, it would seem that extramedullary sources of retinoic acid, such as the meninges, are involved in their differentiation.
In another study, Forehand et al. (1998) applied at-RA or RA synthesis-blocking drugs to individual somites at thoracic levels and observed changes in the differentiation of sympathetic preganglionic neurons. Of interest here is the combined intrasegmental and intersegmental pattern of differentiation of these neurons, a feature that led Forehand et al. (1998) to propose the combined action of segmentally iterated signals (presumably from the somites) and a longitudinally graded signal in setting up the pattern. It will be interesting in this regard to assess the contribution of meninges-derived and motoneuron-derived retinoic acid to the longitudinally graded component.
Lack of Detection of 9-cis-RA
Solomin et al. (1998) recently observed a bimodal activation pattern of a Gal4-RXR fusion protein in the spinal cord of 11.0- to 12.0-dpc mouse embryos. This finding demonstrates the activation of RXRs and suggests the presence of an endogenous RXR ligand in the brachial and lumbar regions of the cord. Because 9-cis-RA generally is considered the principal ligand for RXRs, it is important to determine whether 9-cis-RA is indeed present in the spinal cord at these stages.
We did not detect 9-cis-RA under normal conditions, whereas at-RA was readily detectable. The limit of detection for 9-cis-RA in our HPLC assay system gives a predominance of at-RA over 9-cis-RA of at least 180-fold in the brachial region, 30-fold in the thoracic region, and 110-fold in the lumbar region. These ratios could in fact be much higher. Although we cannot eliminate the possibility that 9-cis-RA is present, it is clearly far less abundant than at-RA.
Our estimates of volumetric concentrations of at-RA ranged from 36 to 200 nM. Such estimates must be interpreted with caution. First, retinoids may be sequestered into subpopulations of cells within a tissue, giving potentially higher concentrations. For example, if at-RA were localized to motoneurons only, our volumetric concentration estimates would be several 10-fold too low. On the other hand, our measurements are of extracted retinoids. Binding of retinoids to proteins in vivo, for example cellular retinoic acid binding protein, could substantially lower their free concentrations. Therefore, it is entirely speculative to extrapolate from the limit of 9-cis-RA detectability, which effective concentrations 9-cis-RA might obtain in the spinal cord. What we do know is that the signal observed by Solomin et al. (1998) at 11.5 dpc was throughout the dorsoventral extent of the spinal cord, i.e., it was not limited to a minor subpopulation of spinal cord cells. Therefore, it seems likely that 9-cis-RA, if present, cannot be concentrated to much more than a few nM, and possibly substantially less. Whether such concentrations are sufficient to generate the observed signal remains to be seen.
Solomin et al. (1998) also observed an activation of the Gal4-RXR fusion protein throughout the longitudinal axis of the embryonic spinal cord after gavage feeding of pregnant dams with at-RA. Their interpretation was that the exogenous at-RA was isomerized to 9-cis-RA and, thereby, activated RXRs throughout the spinal cord. We assayed the retinoid contents of spinal cords treated by using a slightly different paradigm (Satre and Kochhar, 1989) and found a massive increase in at-RA content and the appearance of a small 9-cis-RA peak. The amount of 9-cis-RA in this nonphysiological situation was approximately 30% of the amount of at-RA present normally. The concurrent super-elevation of at-RA, on the other hand, generated a ratio of over 600-fold between at-RA and 9-cis-RA. Thus, both under normal and experimental conditions, 9-cis-RA is far less abundant than at-RA.
The appearance of 9-cis-RA after gavage feeding with at-RA suggests a conversion of at-RA to 9-cis-RA. Our control experiments seem to rule out the possibility of photoisomerization or chemical isomerization during handling and preparation of the embryonic tissue. Moreover, no enzymes are known to catalyze this conversion. Therefore, a likely possibility is that isomerization to 9-cis-RA occurs within the pregnant dam, with the relatively poor transfer of cis-isomers across the placenta (Creech et al., 1989; Kochhar et al., 1995), limiting the amount accumulating in the embryo.
It is important to note that oxidation of 9-cis-ROH to 9-cis-RA in the spinal cord is formally possible given the presence of CRAD1, ADH4, RDH5, ALDH1, and RALDH2 transcripts and RALDH2 protein. Nevertheless, the extant enzymatic machinery is clearly generating a more substantial and/or stable pool of at-RA than 9-cis-RA.
How Are RXRs Activated in the Embryonic Spinal Cord?
It is clear from several studies that RXRs play important roles in embryonic development. Selective activation of RXRs with RXR-specific ligands affect the development of embryonic tissues in vivo and in vitro (Lu et al., 1997; Sockanathan and Jessell, 1998). Gene knockouts have shown that RXRα is necessary for normal development of the mouse, whereas the other RXRs may be dispensable (Krezel et al., 1996). Selective genetic manipulation of RXRα, which prevents its functional interaction with coactivators and, therefore, its interaction with the general transcriptional machinery, produces specific developmental defects (Mascrez et al., 1998). RXRα is expressed in the spinal cord of both chicken embryos and mouse embryos (Dolle et al., 1994; Hoover and Glover, 1998). Thus, the demonstration by Solomin et al. (1998) that RXR-mediated signaling occurs in the spinal cord has important implications for our understanding of retinoid-dependent differentiation in this part of the central nervous system.
But does 9-cis-RA activate RXRs in the spinal cord? RXRs are much more efficiently activated by 9-cis-RA than at-RA in vitro (Heyman et al., 1992; Allenby et al., 1993). Nevertheless, reporter assays using chimeric receptor constructs show that all-trans isomers at submicromolar concentrations can bind to RXR homodimers to activate response elements up to 10% or more of the maximum activation produced by 9-cis-RA (see Fig. 5 in Allenby et al., 1993). Our estimates of volumetric concentrations for at-RA lie in this range. Thus, in the absence of 9-cis-RA, or in a situation where 9-cis-RA is far less concentrated than at-RA, at-RA could produce at least as strong an activation of RXRs as 9-cis-RA.
Alternatively, other molecules might function as ligands for RXRs. Phytanic acid and eicosanoids, for example, have been shown to activate RXRs (Eager et al., 1992; Kitareewan et al., 1996), and other ligands may exist as well. It is important to realize that our chromatograms include only those molecular species for which the extraction, separation, and detection procedures are designed, and do not rule out the presence of nonretinoid ligands. Phytanic acid cannot be assayed by our system. The fatty acids ARA, DHA, and EPA can, but we did not detect them in the spinal cord.
All-trans-retinoic acid, 9-cis-retinoic acid, 13-cis-retinoic acid, all-trans-retinol, 13-cis-retinol, all-trans-retinal, 13-cis-retinal, 9-cis-retinal, arachidonic acid, cis-5,8,11,14,17-eicosapentaenoic acid, and cis-4,7,10,13,16,19-docosahexaenoic acid were obtained from Sigma Aldrich (St. Louis, MO). The retinoids 11,13-di-cis-retinol, 9-cis-retinol, all-trans-3,4-didehydro-retinal, 9-cis-3,4-didehydro-retinal, 9-cis-3,4-didehydro-retinoic acid, 9-cis-3,4-didehydro-retinol, all-trans-3,4-didehydro-retinol, 4- oxo-all-trans-retinoic acid and 4-oxo-13-cis-retinoic acid were gifts from F. Hoffmann-La Roche (Basel, Switzerland). All-trans-3,4-didehydro-retinoic acid was a generous gift from A. Vahlquist (Uppsala, Sweden).
Animal protocols were in accordance with the official governmental guidelines on the care and use of laboratory animals in Norway. Embryos from natural matings of 6- to 8-week-old F1 hybrids (C57BL/6J×CBA/J) were collected at 11.5 days of gestation. The morning of the day of vaginal plug appearance was defined as 0.5 dpc. For prenatal retinoid treatment, pregnant mice were gavage-fed with all-trans-RA suspended in 300 μl of corn oil at a concentration of 20 or 100 mg/kg body weight.
Dissection of Embryos
All of the following procedures, except those involving tissues destined for RT-PCR analysis, were conducted under red lighting, yellow lighting, or both, with wavelengths above 595 nm to prevent photoisomerization of endogenous retinoids, which occurs within a few minutes in room light (unpublished results). Pregnant mice were killed by cervical dislocation after which the embryos with placentae were removed by means of an abdominal incision. Before dissection, the embryos in their amnions were maintained in ice-cold saline (0.9% NaCl). The embryos were staged according to morphology of visceral arches and limb buds (Kaufman, 1992) and then decapitated. Stages ranged from 11.0 to 12.0 (predominantly 11.5 to 12.0) for HPLC measurements of endogenous retinoids. Stage 12.5 embryos were used for control experiments to test for retinoid isomerization, and stage 12.0 to 12.5 embryos were used for RT-PCR experiments.
The spinal cord was dissected in cold saline into three portions, the first of which (B) encompassed the brachial region, the second of which (T) was restricted to the thoracic region, and the third of which (L) encompassed the lumbar region (Fig. 1). All dissections were performed by the same person (J.C.G.) experienced in microdissection techniques. Dissected spinal regions were collected, gassed with argon, and frozen in liquid nitrogen or dry ice, one by one. The time from the start of dissection to freezing typically ranged from 12 to 20 min and never exceeded 35 min. After all tissues were collected and frozen they were stored at −70°C until assays were performed. We wish to emphasize that these dissections involved a substantial amount of labor. Tissue for RT-PCR was dissected in cold phosphate buffered saline under RNAse-free conditions, and then stored at −70°C.
The dissected spinal cord segments were completely free of surface ectoderm, notochord, and paraxial mesoderm (somites), but there was some contamination by the developing meninges. All regions were contaminated by slight amounts of ventral meninges, which was impossible to visualize under the red lighting. Laterally and dorsally disposed meninges could be clearly split away from the cord and was consistently removed from the brachial region, but the lateral portion was difficult to remove from thoracic and lumbar regions because of higher adhesivity to the spinal cord. We estimated visually that the thoracic and lumbar regions were each contaminated by no more than 10% of the total meningeal covering. By contrast, the meninges were removed completely from all spinal cords destined for RT-PCR experiments, as these were dissected under normal light.
Embryonic tissue was homogenized and assayed for protein levels with the Micro BCA protein assay reagent kit (Pierce). Bovine serum albumin was used as internal standard.
Solid Phase Extraction - HPLC
To preserve the configuration of the extremely labile retinoids, on-line solid phase extraction, microcolumns, and column switching were used in combination with diode array detection (DAD), coulometric electrochemical detection (ECD), or MS detection.
Retinoid analysis was performed according to a recently published method (Sakhi et al., 1998). Tissue was thawed and homogenized with a motorized pellet grinder in a clear Eppendorf tube. The homogenate of 330 μl was adjusted to 340 μl with buffer. Then, 510 μl of cold acetonitrile was added. After thorough mixing and centrifugation at 5,300 × g for 10 min (5°C), 750 μl of the clear supernatant was transferred to an amber glass vial. An aliquot of 250 μl of water was added with subsequent mixing, resulting in a final acetonitrile concentration of 45%. The entire procedure was performed under red light, and the samples were kept on ice under argon atmosphere whenever possible. An aliquot of 1,000 μl was then injected into the HPLC system. The samples were submitted to on-line solid phase extraction followed by automated transfer of the extract to the analytical column by column switching. The detection system consisted of three electrochemical cells. The guard cell was set to +750 mV and was used to oxidize any trace organic compounds in the separating mobile phase ensuring a very low background. The screening cell lowered the amount of oxidizeable components in the injected sample and was set to +450 mV. The analytical cell was set to +750 mV relative to the palladium reference electrode and provided the signal by oxidizing the double bond in the polyene chain of the retinoid.
Atmospheric Pressure Electrospray Ionization-Mass Spectrometry (AP-ESI-MS) and Diode Array Detection (DAD)
Tissue was homogenized in 30 μl of 0.9% NaCl in an Eppendorf vial by using a motorized pellet grinder (Kontes) and 60 μl of ice-cold acetonitrile was added. After thorough mixing the vial was snap frozen in liquid nitrogen before re-homogenization, mixing, and centrifugation at 10,000 × g for 1 min at ambient temperature. An aliquot of 80 μl was injected on a 2.1 × 250-mm suplex pKb-100 column. Separation was obtained with a gradient starting with 95% solution A (69.5% AcN-30% water + 0.5% HCOOH), changing linearly over 40 min to 40% solution B (69.5% AcN-30% MTBE-0.5% HCOOH). The flow was 0.4 ml/min, and the temperature was 40°C. Full spectral analysis in the range 250–600 nm was performed with a HP 1100 DAD equipped with a 2-μl flow cell set to an optical resolution of 1.2 nm, a bandwidth of 30 nm, and a 2-sec filter. The liquid chromatograph was interfaced by AP-ESI to the single quadrupole mass spectrometer operated in positive mode (Hewlett Packard). Single ion monitoring (SIM) was performed for selected retinoids. For RA, the [M+1]+ ion of protonated RA with a mass to charge ratio (m/z) of 301.2 was selected. For ROH, [M-17]+ resulting from dehydration of the alcohol was selected. Tuned to give maximum sensitivity for these fragments, the settings of the mass spectrometer was as follows: fragmentor, 70 V; nebulizer pressure, 30 psi; drying gas, 10 L/min; drying gas temperature, 350°C; capillary voltage, 4,000 V; gain 5.
mRNA Isolation and RT-PCR Analysis
Dissected tissues were frozen in liquid nitrogen immediately and then stored at −70°C until use. mRNA was isolated by using the MicroFast mRNA isolation kit (Invitrogen). cDNA synthesis and PCR (40 cycles) was performed as described by Ulven et al. (1998). The specific primer pairs are listed in a previous report (Ulven et al., 2000).
Each RT-PCR analysis was also carried out without reverse transcriptase as a negative control. β-Actin was used in all experiments as a control. Each experiment was repeated at least three times with similar results.
Stage 11.5 to 12.5 embryos were decapitated, fixed in 4% paraformaldehyde in phosphate buffer (pH 7.4) for several hours at 4°C, rinsed with PBS, and cryoprotected in 20% sucrose in PBS. They were frozen in OTC compound (ChemiTeknik, Oslo) and sectioned transversely at 14 μm. Immunohistochemistry for RALDH2 was performed according to Berggren et al. (1999), by using fluorescein- or rhodamine-conjugated secondary antibodies (Amersham, 1:200).
Tissue Volume Estimates
The volumes of brachial, thoracic, and lumbar regions dissected for measurement of endogenous retinoids were estimated by measuring cross-sectional areas in digitized sections obtained for immunohistochemistry and multiplying by the lengths of the respective regions.
We thank Dr. Peter McCaffery for the gift of RALDH2 antiserum.
- 1993. Retinoic acid receptors and retinoid X receptors: interactions with endogenous retinoic acids. Proc Natl Acad Sci USA 90: 30–34. , , , , , , , , , .
- 1991. Retinoic acid modifies the pattern of cell differentiation in the central nervous system of neurula stage Xenopus embryos. Development 112: 945–958. , .
- 1999. Differential distribution of retinoic acid synthesis in the chicken embryo as determined by immunolocalization of the retinoic acid synthetic enzyme, RALDH-2. Dev Biol 210: 288–304. , , , .
- 1996. A decade of molecular biology of retinoic acid receptors. FASEB J 10: 940–954. .
- 1993. Local sources of retinoic acid coincide with retinoid-mediated transgene activity during embryonic development. Proc Natl Acad Sci USA 90: 6572–6576. , , .
- 1987. Low teratogenicity of 13-cis-retinoic acid (Isotretinoin) in the mouse corresponds to low embryo concentrations during organogenesis: comparison to the all-trans isomer. Toxicol Appl Pharmacol 87: 474–482. , , , .
- 1989. Teratogenicity and placental transfer of all-trans-, 13-cis-, 4-oxo-all- trans-, and 4-oxo-13-cis-retinoic acid after administration of a low oral dose during organogenesis in mice. Toxicol Appl Pharmacol 100: 162–176. , , , , .
- 1994. Developmental expression of murine retinoid X receptor (RXR) genes. Mech Dev 45: 91–104. , , , .
- 1999. Key roles of retinoic acid receptors alpha and beta in the patterning of the caudal hindbrain, pharyngeal arches and otocyst in the mouse Development 126: 5051–5059. , , , , .
- 1989. Retinoic acid causes an anteroposterior transformation in the developing central nervous system. Nature 340: 140–144. , , , , , , .
- 1992. Structural and functional evidence for activation of a chick retinoid X receptor by eicosanoids. Proc R Soc Lond B Biol Sci 250: 63–69. , , , .
- 1998. Segment-specific pattern of sympathetic preganglionic projections in the chicken embryo spinal cord is altered by retinoids. Proc Natl Acad Sci USA 95: 10878–10883. , , , , .
- 1998. Initiation of rhombomeric Hoxb4 expression requires induction by somites and a retinoid pathway. Neuron 21: 39–51. , , .
- 1995. Activation of mammalian retinoid X receptors by the insect growth regulator methoprene. Proc Natl Acad Sci USA 92: 6157–6160. , , , .
- 1999. Distinct functions for Aldh1 and Raldh2 in the control of ligand production for embryonic retinoid signaling pathways. Dev Genet 25: 353–364. , , .
- 1992. 9-cis retinoic acid is a high affinity ligand for the retinoid X receptor. Cell 68: 397–406. , , , , , , .
- 1998. Regional pattern of retinoid X receptor-alpha gene expression in the central nervous system of the chicken embryo and its up-regulation by exposure to 9-cis retinoic acid. J Comp Neurol 398: 575–586. , .
- 1995. Endogenous distribution of retinoids during normal development and teratogenesis in the mouse embryo. Dev Dyn 202, 312–323. , .
- 1992. The atlas of mouse development. Academic Press, London. .
- 1996. Phytol metabolites are circulating dietary factors that activate the nuclear receptor RXR. Mol Biol Cell 7: 1153–1166. , , , , , , , , , , .
- 1995. Placental transfer and developmental effects of 9-cis retinoic acid in mice. Teratology 51: 257–265. , , , .
- 1996. RXR gamma null mice are apparently normal and compound RXR alpha +/- /RXR beta -/-/RXR gamma -/- mutant mice are viable. Proc Natl Acad Sci USA 93: 9010–9014. , , , , , .
- 1995. Protein phosphatases 1 and 2A regulate the transcriptional and DNA binding activities of retinoic acid receptors. J Biol Chem 270: 10806–10816. , , , , .
- 1994. Function of the retinoic acid receptors (RARs) during development. I. Craniofacial and skeletal abnormalities in RAR double mutants. Development 120: 2723– 2748. , , , , , , , .
- 1997. Ligand-bound RXR can mediate retinoid signal transduction during embryogenesis. Development 124: 195–203. , , .
- 1997. Positional apoptosis during vertebrate CNS development in the absence of endogenous retinoids. Development 124: 2799–2805. , , , , .
- 1998. The distribution of endogenous retinoic acid in the chick embryo: implications for developmental mechanisms. Development 125: 4133–4144. , , , .
- 1995. The RXR heterodimers and orphan receptors. Cell 83: 841–850. , .
- 1995. The nuclear receptor superfamily: the second decade. Cell 83: 835–839. , , , , , , , , , .
- 1998. The RXRalpha ligand-dependent activation function 2 (AF-2) is important for mouse development. Development 125: 4691–4707. , , , , , .
- 1999. Feedback-inducible nuclear-receptor-driven reporter gene expression in transgenic mice. Proc Natl Acad Sci USA 96: 13270–13275. , , .
- 1994. Hot spots of retinoic acid synthesis in the developing spinal cord. Proc Natl Acad Sci USA 91: 7194–7197. , .
- 1997. Restricted expression and retinoic acid-induced downregulation of the retinaldehyde dehydrogenase type 2 (RALDH-2) gene during mouse development. Mech Dev 62: 67–78. , , , , .
- 2000. Retinoic acid synthesis and hindbrain patterning in the mouse embryo. Development 127: 75–85. , , , , .
- 1999. A sonic hedgehog-independent, retinoid-activated pathway of neurogenesis in the ventral spinal cord. Cell 97: 903–915. , , , .
- 1991. Activity of the beta-retinoic acid receptor promoter in transgenic mice. Mech Dev 36: 15–29. , , .
- 1995. Phosphorylation of the retinoic acid receptor-alpha by protein kinase A. Mol Endocrinol 9: 860–871. , , , , , , .
- 1991. Expression of a retinoic acid response element-hsplacZ transgene defines specific domains of transcriptional activity during mouse embryogenesis. Genes Dev 5: 1333–1344. , , , , .
- 1998. Quantitative determination of endogenous retinoids in mouse embryos by high-performance liquid chromatography with on-line solid-phase extraction, column switching and electrochemical detection. J Chromatogr A 828: 451–460. , , , , .
- 1989. Elevations in the endogenous levels of the putative morphogen retinoic acid in embryonic mouse limb-buds associated with limb dysmorphogenesis. Dev Biol 133: 529–536. , .
- 1998. Motor neuron-derived retinoid signaling specifies the subtype identity of spinal motor neurons. Cell 94: 503–514. , .
- 1998. Retinoid-X receptor signalling in the developing spinal cord. Nature 395: 398–402. , , , , , , , , .
- 1998. Expression of retinoic acid receptor and retinoid X receptor subtypes in rat liver cells: implications for retinoid signalling in parenchymal, endothelial, Kupffer and stellate cells. Eur J Cell Biol 77: 111–116. , , , , , .
- 2000. Identification of endogenous retinoids, enzymes, binding proteins, and receptors during early postimplantation development in mouse: important role of retinal dehydrogenase type 2 in synthesis of all-trans-retinoic acid. Dev Biol 220: 379–391. , , , , , , , , .
- 1995. Distribution of all-trans-, 13-cis- and 9-cis-retinoic acid to whole rat embryos and maternal serum following oral administration of a teratogenic dose of all-trans-retinoic acid. Pharmacol Toxicol 76: 196–201. , .
- 1996. Molecular identification of a major retinoic-acid-synthesizing enzyme, a retinaldehyde-specific dehydrogenase. Eur J Biochem 240: 15–22. , , , , , , .