An Analysis of Retinoic Acid-induced Gene Expression and Metabolism in AB1 Embryonic Stem Cells*

Murine embryonic stem cells such as the AB1 cell line undergo differentiation in the presence of retinoic acid (RA) into an extraembryonic epithelial cell type. This results in the activation of genes such as Hoxa-1, Hoxb-1, laminin, collagen IV( (cid:97) 1), tissue plasminogen activator, RAR (cid:98) , and CRABPII. The CRABPI gene is regulated in an unusual fashion; CRABPI message and protein levels are induced at low concentrations of RA, but induction is diminished at higher concentrations. AB1 cells take up RA rapidly from the medium, and the addition of low, exogenous concentrations of RA to the culture medium results in very high intracellular RA concentrations. For example, AB1 stem cells cultured in 5 n M [ 3 H]RA have an internal [ 3 H]RA concentration of 1–2 (cid:109) M within the first hour. AB1 cells also metabolize [ 3 H]RA to more polar RA derivatives. The half-life of RA in AB1 cells not previ- ously exposed to RA is about 2–2.5 h versus 40–45 min in cells cultured for 2–3 days in 1 (cid:109) M exogenous RA. Thus, the enzyme(s) which metabolize RA are induced or acti- vated by RA. Furthermore, the local concentration of RA required to elicit some biological responses may be higher than previously thought. containing the labeled retinoids was separated from the aqueous phase by centrifugation. The upper organic phase was collected and dried in a SpeedVac (Savant). Samples were stored under argon for no longer than 1 week at (cid:50) 70 °C. High pressure liquid chromatography (HPLC) analysis was performed using a Waters Millenium system (Millipore Corp.) to separate the various retinoids. Samples were ap- plied to an analytical 5- (cid:109) m reversed-phase C 18 column (Vydac, Hespe-ria, CA) at a flow rate of 1.5 ml/min and the retinoids were separated using a gradient from 0% acetonitrile (15 m M ammonium acetate, pH 6.5) to 100% acetonitrile (75). Non-labeled retinoid standards were run concurrently and were monitored at a wavelength of 340 nm while a Packard A-500 radiochromatography detector (Packard Instruments, Downers Grove, IL) was employed to monitor the labeled retinoids.

Retinoic acid (RA), 1 a derivative of vitamin A, plays a critical role in many cellular processes, including cellular differentiation (1, 2) and embryogenesis (3,4). RA can also regulate the growth and differentiation of a large variety of pre-malignant and neoplastic cell types, both in vivo and in culture (2,(5)(6)(7)(8)(9). RA and several of its analogs have proven to be efficacious as agents in the treatment of dermatologic disorders (10) and as chemotherapeutic agents in the treatment of a number of human cancers (11)(12)(13)(14)(15). However, the usefulness of retinoids such as RA in the treatment of human disease is limited by their potent teratogenicity (16 -18) and rapid metabolism in many cell types (12,19).
The mechanisms by which RA exerts its actions are not fully understood, but several classes of proteins involved in mediating the cellular response to RA have been identified. These cellular mediators include at least two families of nuclear ret-inoid receptors (RARs and RXRs) and two cellular retinoic acid-binding proteins (CRABPI and CRABPII). The nuclear receptors are members of the steroid hormone receptor superfamily and act as high affinity ligand-dependent transcription factors (20,21). In the presence of various RA isomers, RARs and RXRs form heterodimers (21)(22)(23)(24)(25)(26)(27) that activate differentiation-specific gene transcription by binding to retinoic acid response elements in the promoter regions of RA-responsive target genes (28 -33). The CRABPs are cytoplasmic, low molecular weight proteins which are members of the fatty acidbinding protein superfamily (34). The CRABPI and CRABPII proteins are highly related to one another and bind to alltrans-RA with high, but varying affinities (34 -38). The functions of the CRABP proteins are not well understood, but they are thought to play an important role in regulating the metabolism of RA (39 -41), and in regulating the intracellular levels of RA in various tissues of the developing embryo. This hypothesis is supported by the fact that CRABPI and CRABPII messages are expressed in specific and distinct temporal and spatial patterns in many tissues of the developing embryo, including those tissues that are sensitive to the teratogenic effects of RA (42)(43)(44)(45)(46)(47)(48).
In response to RA, many teratocarcinoma cell lines (such as F9, P19, and NT2) (49 -52) and embryonic stem cell lines (such as D3 and CCE) (53)(54)(55) differentiate into cell types resembling those found in early stages of the developing embryo. The AB1 murine embryonic stem cell line (56) is a pluripotent stem cell line that has been used extensively by many groups to investigate the effects of homozygous deletions of a variety of developmentally-regulated (activin, inhibin, follistatin, Hox b4) (57)(58)(59)(60)(61) and growth-related genes (p53, c-myc) (59,62) in tissue culture and in mice. Although AB1 cells have been widely used to study the effects of gene knockouts, these cells have not been well characterized previously for their response to RA. It is known, however, that AB1 cells undergo differentiation after treatment with RA. In this paper, we characterize the RA response of the AB1 embryonic stem cell line with regard to RA-inducible gene expression and RA metabolism.

MATERIALS AND METHODS
Cell Culture and Reagents-AB1 murine embryonic stem cells (56) were kindly provided by Dr. Allan Bradley (Baylor College of Medicine, Houston, TX). AB1 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum, 2 mM glutamine, 1 ϫ antibiotic/antimycotic (consisting of 100 units/ml penicillin G sodium, 100 g/ml streptomycin sulfate, 0.25 g/ml amphotericin B), 1 mM sodium pyruvate, 100 M minimal essential medium non-essential amino acids, 1 M ␤-mercaptoethanol, and 10 3 units/ml leukemia inhibitory factor (LIF) at 37°C in 5% CO 2 . Embryonic stem cells are routinely cultured in the presence of feeder cells to prevent spontaneous differentiation, however, addition of LIF to the culture medium inhibits spontaneous differentiation (63) and allows AB1 cells to be grown without feeder cells. Addition of RA overcomes the inhibitory effects of LIF and induces differentiation (64). All cell culture reagents were obtained from Life Technologies, Inc. (Grand Island, NY). Retinoic acid (Sigma) was dissolved in 100% ethanol and stored at 4°C.
Retinoic acid was added to cells 24 h after plating; the final ethanol concentration did not exceed 1% of the total volume. [ 3 H]All-transretinoic acid and [␣-32 P]dCTP were obtained from DuPont NEN (Boston, MA).
Probes-An 800-base pair EcoRI fragment containing the entire CRABPI cDNA was excised from the pMT-CRABPI plasmid (40) for use as a probe for Northern blots. Similarly, an 876-base pair EcoRI/HindIII fragment containing the entire CRABPII cDNA was excised from the pMT-CRABPII plasmid. 2 In addition, cDNA fragments corresponding to the murine Hox b1, laminin B1, collagen IV (␣1), Tam 2.5 (tissue plasminogen activator), J6 serpin, Rex1, and actin genes were excised from the following plasmids for use as probes. The pHox b1 (also known as Hox 2.9) plasmid was obtained from Dr. Joseph Grippo (Hoffman LaRoche, Nutley, NJ) and contains a 435-base pair partial cDNA specific for the murine Hox b1 gene within the EcoRI/HindIII site of pGEM72f(ϩ). The plamB1 5.5 plasmid contains a partial cDNA specific for the murine laminin B1 gene within the EcoRI site of pUC9 (65). The pcol 381 plasmid contains a partial cDNA specific for the murine collagen IV (␣1) gene within the EcoRI site of pUC9 (65). The Tam 2.5a plasmid was kindly provided by Dr. Sidney Strickland (SUNY Stony Brook, Stony Brook, NY) and contains the full-length murine tissue plasminogen activator cDNA cloned into the EcoRI site of pKSϩ/Ϫ (66,67). The pJ6-3 contains a partial cDNA specific for the murine serpin gene within the PstI site of pUC9 (68). The pRex1-2-5R plasmid contains a 1.7-kilobase pair fragment of the murine Rex-1 gene cloned into the EcoRI site of pUC9 (69). The pUC Act1 plasmid contains a partial cDNA specific for the murine ␤ actin gene within the PstI site of pUC9 (70).
RNA Isolation and Northern Blot Analysis-Total cellular RNA was isolated by the guanidine isothiocyanate method of Chirgwin et al. (71). Cells were plated in medium containing 10 3 units/ml LIF and treated with varying concentrations of RA 24 h after plating. For the half-life experiment, AB1 cells were cultured in the presence or absence of RA for 72 h, followed by addition of 2 g/ml actinomycin D-mannitol (Sigma) for varying time periods. The mRNA was size fractionated on 1.2% agarose, 2.2 M formaldehyde gels, transferred to nitrocellulose, and hybridized to random primer-labeled probe (72). Blots were washed once in 2 ϫ SSC, 0.1% SDS for 20 min at room temperature, once in 2 ϫ SSC, 0.1% SDS at 60°C, and once in 0.2 ϫ SSC, 0.1% SDS at 60°C. Autoradiographs were quantitated by scanning densitometry using an LKB densitometer (Pharmacia LKB, Piscataway, NJ). All of the Northern experiments described in this paper were performed at least twice from two independent RNA preparations.
Preparation of Cytoplasmic Protein Extracts and Polyacrylamide Gel Electrophoresis-Cytoplasmic protein extracts were prepared as described (40). Briefly, cells were resuspended in 10 mM Tris, pH 7.4, and 7 mM ␤-mercaptoethanol and solubilized by Dounce homogenization. Supernatants were obtained by ultracentrifugation at 100,000 ϫ g for 1 h at 4°C. Protein concentration was determined by Bio-Rad protein assay (Bio-Rad) as per the manufacturer's instructions. Bovine serum albumin (Sigma) was used as a standard. 100 g of protein extract was incubated with 50 nM [ 3 H]all-trans-retinoic acid (DuPont NEN) and 2 mM dithiothreitol for 16 h at 4°C in the presence or absence of a 500-fold excess of unlabeled retinoic acid. The samples were then analyzed by the nondenaturing polyacrylamide gel electrophoresis (PAGE) technique described in Siegenthaler and Saurat (73) and Siegenthaler et al. (74), except that samples were resolved on a 12% polyacrylamide gel and 0.32 M Tris, pH 8.4, was used in the separating gel. The sample containing lanes of the gel were sliced into 2-mm bands. Each gel slice was incubated overnight at 50°C in SOLVABLE tissue solubilizer (DuPont NEN) and tritiated counts were quantitated by liquid scintillation counting in ATOMLIGHT scintillant (DuPont NEN).
Extraction of Retinoids and HPLC Chromatography-Cells were plated in 60-mm tissue cultures dishes at a density of 2 ϫ 10 5 cells/dish in medium containing 10 3 units/ml LIF. 24 h after plating, varying concentrations of RA were added to the cells for 72 h. On the day of labeling, medium containing 5 or 50 nM [ 3 H]all-trans-RA (50 Ci/mmol) was added to cells for various lengths of time as indicated. A separate control consisting of radiolabeled RA in medium without cells was included concurrently during the incubation period. Cells and onequarter of the labeled media were collected. Cells were scraped into phosphate-buffered saline and retinoids were extracted immediately with acetonitrile/butanol (50:50, v/v) and saturated potassium phosphate as described previously by Boylan and Gudas (41). The organic phase containing the labeled retinoids was separated from the aqueous phase by centrifugation. The upper organic phase was collected and dried in a SpeedVac (Savant). Samples were stored under argon for no longer than 1 week at Ϫ70°C. High pressure liquid chromatography (HPLC) analysis was performed using a Waters Millenium system (Millipore Corp.) to separate the various retinoids. Samples were applied to an analytical 5-m reversed-phase C 18 column (Vydac, Hesperia, CA) at a flow rate of 1.5 ml/min and the retinoids were separated using a gradient from 0% acetonitrile (15 mM ammonium acetate, pH 6.5) to 100% acetonitrile (75). Non-labeled retinoid standards were run concurrently and were monitored at a wavelength of 340 nm while a Packard A-500 radiochromatography detector (Packard Instruments, Downers Grove, IL) was employed to monitor the labeled retinoids.

AB1 Cells Express Retinoic Acid-inducible
Genes-AB1 cells were cultured in the presence of LIF, either without (AB1 stem cell control), or with 1 M RA for various lengths of time. Upon treatment with RA, AB1 cells differentiate into cells resembling extraembryonic endoderm. AB1 cells also behave similarly to other embryonic stem cell lines and teratocarcinoma cell lines with respect to the expression of many RA-responsive genes (Fig. 1). Some early-response genes such as the homeobox-containing genes, Hox b1 ( Fig. 1) (76,77) and Hox a1 (data not shown) (78), are induced by RA, while the zinc-finger containing Rex1 gene (55, 69) is down-regulated by RA (Fig. 1). In addition, late-response genes such as collagen IV (␣1) (65), tissue plasminogen activator (Tam 2.5) (66, 67), laminin B1 (65), and J6 serpin (68,79) are induced after RA treatment (Fig. 1). The RAR␤ gene is also induced in response to RA (data not shown). All of these genes are expressed to similar degrees and with similar kinetics as compared to what has been observed previously in F9 teratocarcinoma cells (6, 65, 67-69, 78, 80). These results demonstrate that AB1 cells respond to RA by differentiating along an epithelial pathway to parietal endoderm (2) and by inducing or down-regulating a large number of genes.
Expression of CRABPI and CRABPII Genes-The genes encoding the cellular retinoic acid-binding proteins, CRABPI and CRABPII, are also regulated by RA (34). While CRABPII mRNA expression is induced after RA treatment of both F9 and 2 A. C. Chen, C. Stoner, and L. J. Gudas, submitted for publication. P19 cells (36), CRABPI message is reduced by RA in F9 cells (35), but induced by RA in P19 cells (50). The mechanism for the differential regulation of the CRABPI gene in F9 and P19 cells is not known. While less is known about the regulation of the CRABPI gene (35,81,82), the transcriptional regulation of the CRABPII gene has been reported to occur via high affinity RA response elements (32,33). AB1 cells were cultured in LIF in the presence of a range of RA doses for varying lengths of time. The CRABPI message is expressed in AB1 stem cells, but is induced to much higher levels upon treatment with RA ( Fig. 2). Induction of CRABPI gene expression is similar to the induction of "late" response genes since CRABPI message levels are still increasing 72 h after RA addition. Most intriguing, though, is that the CRABPI message is induced at low exogenous RA concentrations, but is almost extinguished by higher RA concentrations. CRABPI mRNA is induced to the greatest extent after AB1 cells have been exposed to 100 pM to 1 nM RA (Fig. 2). In contrast, CRABPII mRNA is expressed at increasingly higher levels as exogenous RA concentrations are raised, with maximal expression detected upon addition of 1 M RA (Fig. 2).
We have also examined the expression of other RA-responsive genes such as Hox a1, Hox b1, collagen IV (␣1), laminin B1, and tissue plasminogen activator. Thus far, all of the genes that have been examined appear to behave like the CRABPII gene; the genes are most highly induced at the highest RA concentrations (data not shown). These data suggest that the CRABPI gene is regulated by RA in a different manner from the CRABPII gene as well as from other RA-responsive genes.
Higher Levels of Functional CRABPI Protein Are Produced at Lower Exogenous RA Concentrations-We next wanted to de-termine if the increased level of CRABPI message expressed in cells treated with lower doses of RA resulted in increased production of functional CRABPI protein. AB1 cells were cultured in LIF and either 100 nM or 1 M RA for 72 h. Cytoplasmic protein extracts were prepared from cells, incubated with [ 3 H]RA, and resolved by nondenaturing polyacrylamide gel electrophoresis (PAGE) analysis (73,74). The amount of functional CRABPI and CRABPII protein expressed in a particular cell line can be visualized by virtue of its ability to bind to [ 3 H]RA. In addition, this procedure allows us to separate and directly compare the amount of CRABPI and CRABPII protein produced in a given cell line. The migration profiles demonstrate that significant amounts of functional CRABPI and CRABPII protein are produced in AB1 cells after treatment with 1 M RA for 72 h (Fig. 3A). CRABPI protein levels are about 1.4-fold higher than CRABPII protein levels. After treatment with 100 nM RA (Fig. 3B), AB1 cells express an increased level of CRABPI protein and a decreased level of CRABPII protein when compared to the levels of CRABPI and CRABPII proteins expressed in cells treated with 1 M RA (compare Fig.  3, A and B). This results in a 5.0-fold higher level of CRABPI than CRABPII at 100 nM RA. These data indicate that as the concentration of exogenously added RA is increased, CRABPII message and protein levels also increase while CRABPI message and protein levels decrease.
Increased Induction of CRABPI Expression at Low RA Concentrations Is Not a Result of Message Stabilization-To investigate whether the increased RA responsiveness of the CRABPI gene at low RA concentrations was a result of increased stability of CRABPI message, we determined the half-life of the CRABPI message in cells treated with high versus low exoge-

FIG. 2. Expression of CRABPI and CRABPII genes in AB1 cells.
Total cellular RNA was prepared from AB1 cells grown in LIF in the presence or absence of the indicated concentrations of RA for various lengths of time. Northern blots were hybridized to random primer-labeled cDNA probes for CRABPI, CRABPII, and actin. Actin message levels are not affected by RA treatment in AB1 cells. Therefore, the Northern analyses are shown together with the CRABPI and CRABPII gene expression quantitated and normalized to actin. This experiment was performed twice; one representative experiment is shown. nous RA concentrations. AB1 cells were cultured in LIF and in the presence or absence of 1 M or 1 nM RA for 72 h, then followed by treatment with actinomycin D-mannitol for varying lengths of time. As shown previously, CRABPI mRNA is induced after addition of either 1 nM or 1 M RA, however, the highest message levels are found at 1 nM RA (Fig. 4, fourth vs. eighth lanes ). The half-life of the CRABPI message is the same for AB1 cells grown in 1 M or 1 nM RA (Fig. 4). In each case, the half-life of CRABPI mRNA is approximately 9 h. CRABPII message is induced to high levels only after treatment with 1 M RA and the half-life of CRABPII message is longer than 9 h, since CRABPII message levels have not started to decline 9 h after treatment with actinomycin Dmannitol. These results indicate that the increased level of CRABPI mRNA expressed at low RA concentrations does not result from increased stabilization of CRABPI message. Rather, these data suggest that CRABPI mRNA induction by RA is regulated at the transcriptional level.
Kinetics of Retinoic Acid Metabolism in AB1 Cells-In the next series of experiments, we characterized AB1 cells for their ability to metabolize retinoic acid. Studies in our laboratory and by others have shown that many cell types, including other teratocarcinoma cells and embryonic stem cells, have the ability to metabolize RA to various polar derivatives (41,(83)(84)(85). 2 In F9 teratocarcinoma cells, RA is metabolized relatively slowly in undifferentiated cells (85); however, the rate of RA metabolism is dramatically increased when F9 cells are cultured in the presence of 1 M RA. Thus, RA can induce its own metabolism in F9 cells. Work in our laboratory has suggested that enzymes involved in the rapid metabolism of RA are induced by 8 h after RA treatment of F9 cells. 3 Most of the studies investigating RA metabolism in RAdifferentiated cells have been performed after culturing cells in the presence of high exogenous RA concentrations (i.e. 1 M RA) (85). Fewer studies have been done to analyze RA metabolism in cells cultured in the presence of varying concentrations of RA, and to identify possible qualitative differences in the types of polar RA metabolites generated.
AB1 cells were cultured in LIF in the absence or presence of 1 M or 1 nM RA for 72 h, followed by exposure to 5 or 50 nM Analysis  (Fig. 5B, closed squares). We also found that AB1 cells exhibited similar kinetics of  (Fig. 5C). Thus, there is a striking increase in the rate of [ 3 H]RA metabolism when AB1 cells are initially cultured in 1 M RA as compared to lower concentrations of RA (Fig. 5C).
In another series of experiments, a much higher concentration of [ (Table II) (Table II).
Formation of Polar Metabolites in AB1 Cells-As previously mentioned, AB1 cells metabolize [ 3 H]RA into various polar derivatives, which are then found in both the cells and in the media (Fig. 5A). We measured the levels of 3 H-polar metabolites in cells (Fig. 6, top panel) and in the media (Fig. 6, bottom panel) generated by AB1 cells initially cultured in the presence or absence of 1 M or 1 nM exogenous RA for 72 h, and then labeled for 1 or 2 h with 50 nM [ 3 H]RA. The results are plotted as the levels of 3 H-polar metabolites (expressed as counts per minute) versus each of the culture conditions. When we examine the levels of 3 H-polar metabolites present in the cell, we find that AB1 cells initially cultured in the presence of 1 M RA for 72 h produce higher levels of 3 H-polar metabolites in 1 h than AB1 stem cells or cells previously cultured in 1 nM RA (Fig. 6, top panel, open bars). The increased production of 3 H-polar metabolites in cells cultured in 1 M RA presumably results from the increased rate of [ 3 H]RA metabolism (Fig. 5, B and C). After 2 h of labeling with [ 3 H]RA, the levels of 3 H-polar metabolites in untreated stem cells and in cells initially cultured in 1 nM RA continue to increase, while 3 H-polar metabolite levels in cells initially cultured in 1 M RA have already begun to decline as these 3 H-polar metabolites are further metabolized to compounds which are not extracted in organic solvents (Fig. 6, top panel, compare open bars to hatched bars).   H]RA in the 2 ml of starting medium is equal to 5 or 50 nM RA. The volume of AB1 cells has been measured and 1 ϫ 10 6 cells have a volume of about 1 l. Thus, the volume of one AB1 cell is about 1 pl. Cell volumes have been reported from 0.5 pl for lymphoid cells to Ͼ10 pl for a large fibroma cell (86,87). Therefore, to calculate the intracellular RA concentration, we used the following formula:  6. Formation of polar metabolites in AB1 cells. Cells were cultured as described in Table II, then labeled for 1 or 2 h with 50 nM [ 3 H]RA. Cells and media were harvested, retinoids were extracted, and reverse-phase HPLC analysis was performed. The amounts of 3 H-polar metabolites produced by cells grown under the different conditions were quantitated by taking the sum (in counts per min) of all the [ 3 H]retinoids eluting in the polar region (i.e. between minutes 6.8 and 12.3) of the HPLC tracing. These data are plotted as tritiated counts per min of cell-associated polar metabolites or 3 H-polar metabolites in the medium versus the culture condition over the prior 72 h. This experiment was performed three times with similar results; one representative experiment is shown. The data presented in this figure were obtained from the same experiment as the data presented in Fig. 5B, left panel. medium (Fig. 6, bottom panel, open bars, 10 Ϫ9 M RA) than AB1 stem cells (Fig. 6, top and bottom panels, open bars, stem). Since the kinetics of [ 3 H]RA metabolism by AB1 stem cells versus AB1 cells cultured 72 h in 1 nM RA are similar ( Fig. 5B and C), we conclude that the levels of 3 H-polar metabolites are higher within AB1 cells initially cultured in 1 nM RA because a greater proportion of the polar metabolites remains in these cells than in AB1 stem cells. Under these same culture conditions (72 h of culture in the presence of 1 nM RA), the CRABPI mRNA level is very high (Fig. 2). Therefore, one potential explanation for the high intracellular levels of 3 H-polar metabolites under these conditions is that the polar metabolites are bound by CRABPI protein in the cells. This explanation is consistent with data from Fiorella and Napoli (88). There researchers previously demonstrated that 4-hydroxy-RA and 4-oxo-RA, when bound to CRABPI, were metabolized very slowly in testis microsomal extracts; in contrast, unbound 4-hydroxy-RA and 4-oxo-RA had elimination t1 ⁄2 values of 40 and 9 min, respectively. DISCUSSION RA-responsive Gene Expression-AB1 murine embryonic stem cells differentiate into extraembryonic, parietal endoderm-like cells in response to RA treatment (Fig. 1). As such, these cells behave similarly to RA-responsive teratocarcinoma and other embryonic stem cell lines characterized previously. All of the differentiation-specific genes examined in this paper have been characterized previously in other RAresponsive cell culture models as being RA-responsive genes (2, 6, 65, 67-69, 78, 80). These include both early response genes, such as homeobox genes (Hox a1 and Hox b1) and zinc-finger containing transcription factors (Rex1), as well as late response genes, such as CRABPII, RAR␤, laminin B1, collagen type IV (␣1), and tissue plasminogen activator.
AB1 cells respond to exogenous RA in a dramatically different fashion depending on whether they were previously cultured in the presence of high or low exogenous concentrations of RA. This difference is marked in particular by the expression of the CRABPI gene (Fig. 2). When cultured for 72 h in the presence of high exogenous RA concentrations (i.e. 1 M RA), AB1 cells express all of the differentiation-specific, RA-responsive target genes discussed above, including the CRABPI and CRABPII genes. In contrast, when cultured in the presence of low exogenous RA concentrations (i.e. 1 nM RA), AB1 cells express the CRABPI gene at a much higher level (Fig. 2). This effect is also seen at the protein level. Higher amounts of functional CRABPI protein are observed in cells treated with 100 nM RA when compared to those treated with 1 M RA (Fig.  3). This regulation of the CRABPI gene presumably occurs at the transcriptional level and not through message stabilization since the half-life of the CRABPI message is similar in AB1 cells cultured in either high or low RA concentrations (Fig. 4). That the CRABPI gene is transcriptionally activated to a greater extent by low RA concentrations than by high exogenous RA concentrations is of interest because concentrations of RA in the range of 100 pM and 1 nM are more physiologically relevant and more analogous to the concentrations of RA reported in various tissues of the developing embryo (4). Additionally, our data suggests that the CRABPI and CRABPII genes would be expressed in an opposing fashion as the external RA concentration increased from low to high concentration; the CRABPI mRNA level would decrease while the CRABPII mRNA level would increase.
Our results regarding the higher levels of expression of the CRABPI gene in AB1 cells cultured in the presence of low versus high doses of exogenous RA (Fig. 2) can be interpreted in several ways. One possibility is that the CRABPI promoter may contain both positive and negative regulatory elements which are involved in the differential response of the CRABPI gene in AB1 cells treated with low versus high doses of RA. For example, a positively acting transcription factor could be induced at both low and high doses of RA. This factor would then bind to the CRABPI promoter and activate CRABPI gene expression at low doses of RA. However, if a negatively acting factor were induced in AB1 cells only at high exogenous RA concentrations, this negative factor could then bind to a regulatory element in the CRABPI promoter to repress CRABPI expression. Since the regulatory elements of the CRABPI gene have not been well characterized, this hypothesis cannot be tested at present. Alternatively, the pluripotential AB1 cells may differentiate into different cell types as a result of low versus high dose RA treatment. A third potential explanation for the CRABPI gene expression data presented here is that specific metabolites of RA such as 4-oxo-RA influence the expression of the CRABPI gene; differences in the production of these polar metabolites under different culture conditions (Fig. 6) could result in different levels of CRABPI mRNA. The relationships among RA metabolism, CRABPI gene expression, and the ability of cells to differentiate deserve further analysis in these AB1 embryonic stem cells and in other cell types in the future.
Metabolism of Retinoic Acid-In addition to the differences in gene expression described above, we also observed differences in RA metabolism between AB1 cells cultured in the presence of high or low concentrations of RA. After culture in the presence of high exogenous RA concentrations, AB1 cells exhibited a dramatically increased rate of [ 3 H]RA metabolism to 3 H-polar derivatives ( Fig. 5; Table I). In contrast, AB1 cells cultured in 1 nM RA did not show an increase in the rate of [ 3 H]RA metabolism; they metabolized [ 3 H]RA at about the same rate as stem cells (Fig. 5; Table I). From the experiments reported here it is clear that in AB1 cells, RA induces its own metabolism to polar metabolites via a high capacity, low affinity enzyme system. These polar metabolites are then further metabolized via an inducible enzyme system. The uptake of RA appears to occur by a rapid, high capacity mechanism such that under certain conditions, e.g. in AB1 stem cells, the enzyme(s) which metabolize RA are rate-limiting and large quantities of RA accumulate in the cells. Some information is available concerning the enzymes which metabolize RA, and it is known that a number of cytochrome P450 enzymes can metabolize RA (89 -96). Reynolds et al. (97) have shown that metabolism of RA occurs in human keratinocytes, but the enzyme(s) involved have not been thoroughly studied. Various purified isozymes of microsomal cytochrome P450 have been shown to be capable of metabolizing RA and other retinoids (98). The identities of the enzyme(s) involved in RA metabolism in these embryonic cells are unknown.
One of the most striking results of this study is that the actual intracellular [ 3 H]RA concentrations are extremely high when AB1 cells are cultured in the presence of 1 nM to 1 M exogenous RA in the medium (Table II). How much of this cell-associated RA is associated with intracellular membranes versus bound to proteins such as CRABPI, CRABPII, and the RARs is not known. The K d of CRABPI for RA has previously been determined to be about 7 nM (38,39,99). Thus, the intracellular RA concentration would be far above the K d of CRABPI even when these AB1 cells are cultured in the presence of nanomolar concentrations of exogenous RA (Table II). It would be expected that virtually all of the CRABPI protein in the cell would be bound to RA, and that most of the RA in the cell would be present as free RA or RA bound to membranes and/or other proteins.
We have made similar observations concerning the high intracellular RA levels in other cell types cultured in the presence of exogenous RA, 4 so this result is not unique to AB1 cells. These results have important implications for studies in which cells or tissues are cultured in the presence of high exogenous concentrations of RA. The local concentrations of RA required to elicit some biological responses may be higher than previously thought, based on the exogenous concentrations of RA used to elicit biological effects. Furthermore, such high intracellular concentrations of RA may lead to the activation of extraneous receptors not normally activated by endogenous retinoids. The fact that RA is so highly teratogenic may also be explained in part by this ability of cells to accumulate RA from the surrounding medium to extremely high intracellular levels.