The Targeted Disruption of Both Alleles of RARβ2 in F9 Cells Results in the Loss of Retinoic Acid-associated Growth Arrest*

F9 teratocarcinoma cell lines, carrying one or two disrupted alleles of the RARβ2 gene, were generated by homologous recombination to study the role of RARβ2 in mediating the effects of retinoids on cell growth and differentiation. Retinoic acid (RA) does not induce growth arrest of the RARβ2−/− cells, whereas the F9 WT and RARβ2+/− heterozygote lines undergo RA-induced growth arrest. The RARβ2+/− lines also exhibit a faster cell cycle transit time in the absence of RA. The RARβ2−/− stem cells exhibit an altered morphology when compared with the F9 WT parent line, and after RA treatment, the RARβ2−/− cells do not exhibit a fully differentiated cell morphology. As compared with F9 WT cells, the RARβ−/− cells exhibited a markedly lower induction of several early RA-responsive genes and no induction of laminin B1, a late response gene. The induction of RA metabolism in the F9 RARβ2−/− cells following differentiation was not impaired. The research presented here, and prior research suggest that RARβ is required for RA-induced growth arrest in a variety of cell types and that RARβ also functions in mediating late responses to RA. These findings are significant in view of the reduced expression of RARβ transcripts in a number of different types of human carcinomas.

ety of cells (4, 7, 8, 10 -12). RAR␤ exhibits a restricted pattern of expression during development as well as in the mature organism (4,6,13,14). This pattern of expression is different from those of the other RARs and suggests that RAR␤ performs specific functions distinct from those of RAR␣ and RAR␥.
F9 teratocarcinoma cells (15) express all known RA receptors, but RAR␤ mRNA is only present in high amounts after RA addition (10), consistent with the fact that RAR␤ 2 is the predominant RAR␤ isoform expressed in F9 cells (4,6). F9 cells have recently been used in our laboratories to inactivate the RAR␣ and RAR␥ genes by homologous recombination (16,17). Both F9 RAR␣ and RAR␥ null cell lines exhibit marked modulation of a variety of genes when compared with the F9 wildtype cells. This gene knockout approach has allowed the identification of a series of genes that are direct or indirect targets of RAR␣ or RAR␥, such as the "homeobox" genes of the Hoxb and Hoxa clusters, and the genes encoding the extracellular matrix proteins laminin and collagen IV(␣1) (16,17).
RAR␤ mRNA expression has been reported to be greatly reduced in breast cancer cells (18,19), oral and epidermal squamous cell carcinoma (cell lines and tissues) (20 -24), and lung carcinoma lines and tissues (25)(26)(27)(28). In breast cancer cell lines, RA is not able to induce RAR␤ mRNA expression, even though these cells can transcriptionally activate an exogenous RAR␤ RARE (retinoic acid response element) (19). When the expression of the RAR␤ gene is restored in breast cancer cell lines via an exogenous RAR␤ cDNA expression vector, the cells acquire sensitivity to RA-mediated apoptosis and growth arrest (29 -31). Transfection of a human epidermoid lung cancer cell line with RAR␤ causes decreased tumorigenicity (32). In addition, overexpression of RAR␤ 2 in HeLa cells induces growth inhibition (33), while expression of RAR␤ antisense mRNA decreases RA sensitivity in responsive cell lines (31) and causes an increased frequency of carcinomas in transgenic mice (34). The loss of RAR␤ may be an essential step in neoplastic progression, since there is evidence of a progressive decrease in RAR␤ mRNA expression during breast carcinogenesis (35) and greatly reduced RAR␤ mRNA expression in morphologically normal tissue adjacent to breast carcinomas (36). Therefore, characterizing the role of RAR␤ in the control of RA-mediated differentiation and growth arrest may lead to a greater understanding of the mechanisms underlying the development and progression of cancer.
We have generated RAR␤ 2 knockout F9 cells by homologous recombination in order to study the role of RAR␤ 2 in mediating the effects of retinoids on cell growth and differentiation. RA is incapable of inducing growth arrest in these RAR␤ 2 Ϫ/Ϫ cells, indicating that in F9 cells RAR␤ 2 is required for the growth inhibitory actions of RA. This finding is significant in light of the reduced expression of RAR␤ transcripts in a number of carcinomas, and it suggests that RAR␤ could regulate RAinduced growth arrest in a variety of cell types.

Generation of Targeted Disruptions in the RAR␤ Gene-
The RAR␤ 2 isoform was inactivated with a disruption vector described previously (37). The neomycin-resistance gene was inserted into the exon 4 of the RAR␤ gene, which encodes the 5Ј-untranslated and the A2 region sequences of the RAR␤ 2 isoform. The insertion of the neomycin resistance gene in this region disrupts only the RAR␤ 2 isoform of the RAR␤ gene. F9 WT cells were electroporated and selected in G418. Resistant colonies were isolated and screened by Southern analysis for the expected change in size of the 6.5-kb wild-type (WT) KpnI genomic fragment (the mutated allele is 4.3 kb), using a probe derived from DNA sequence located 5Ј of the sequence con-tained in the targeting construct. Two heterozygous lines were generated, F9 WT-␤2-291 and F9 WT-␤2-270, out of a total of 800 colonies screened (Fig. 1A). To target the second allele, the heterozygous lines were selected in high concentrations of G418 for 21 days (16,17,38). One clone was isolated, out of 300 surviving colonies (F9 WT-␤2-270-110), as a result of a second recombination event in which the second WT allele was disrupted ( Fig. 1A). This clone was then subcloned, and eight subclone lines were generated that showed similar behavior (data not shown).
In order to show that the RAR␤ 2 Ϫ/Ϫ line lacked detectable RAR␤ 2 mRNA, RT-PCR analysis was performed (Fig. 1B). Whereas bands of the appropriate size for RAR␤ 2 were detected in the F9 WT and RAR␤ 2 ϩ/Ϫ 270 lines, no bands of that size were detected in the RAR␤ 2 Ϫ/Ϫ line. Thus, the RAR␤ 2 Ϫ/Ϫ cells do not express the RAR␤ 2 transcripts which are expressed in the F9 WT cells. The expression of retinoid receptors was evaluated by Northern and Western analysis (Fig. 1C). As expected, there is no detectable induction of RAR␤ transcripts upon treatment of the F9 RAR␤ 2 Ϫ/Ϫ cells with 1 M RA. In contrast, in the heterozygous lines and the F9 WT line, there was a large (Ϸ20-fold) induction of the RAR␤ transcripts after RA treatment. The expression of RAR␣ and RXR␤ mRNA was similar in all the lines. The expression of RAR␥ mRNA increased slightly (approximately) 2-fold after RA treatment of the F9 RAR␤ 2 Ϫ/Ϫ cells, in contrast to the decrease in RAR␥ mRNA observed in RA-treated F9 WT and RAR␤ 2 ϩ/Ϫ cells. The levels of RXR␣ protein (Fig. 1C) were lower in the RAR␤ 2 ϩ/Ϫ and RAR␤ 2 Ϫ/Ϫ than in the wild type cells. In summary, the RAR␤ 2 Ϫ/Ϫ line has a lower level of RXR␣ protein and (after RA treatment) a higher amount of RAR␥ mRNA. The other retinoid receptors are present at levels similar to those seen in F9 WT cells (Fig. 1C).
The levels of the RAR␤ protein in the F9 WT, F9 RAR␤ 2 ϩ/Ϫ, and F9 RAR␤ 2 Ϫ/Ϫ cell lines were evaluated by Western analysis (Fig. 1D). After 48 h of RA treatment, the RAR␤ protein is induced in the F9 WT cells and is induced to a lesser extent in the F9 RAR␤ 2 ϩ/Ϫ cells. No RAR␤ protein was detected in the F9 RAR␤ 2 Ϫ/Ϫ cells (Fig. 1D). Analysis of the Growth of the F9 RAR␤ 2 Ϫ/Ϫ Cells-The responsiveness of the F9 RAR␤ 2 Ϫ/Ϫ line to RA-induced growth arrest was evaluated (Fig. 2). Unlike the F9 WT line and the heterozygote lines, the F9 RAR␤ 2 Ϫ/Ϫ line did not growth arrest in response to RA. Other retinoids and retinoid agonists were tested, including a RXR pan-agonist (BMS 188,649), and these were also not able to induce growth arrest in the F9 RAR␤ 2 Ϫ/Ϫ cell line (Fig. 3), although they were effective to varying degrees in arresting the growth of the parent F9 WT cell line. This suggests that RAR␤ 2 is required for all of these compounds to exert their growth inhibitory effects in F9 cells.
Interestingly, both of the independently generated F9 RAR␤ 2 ϩ/Ϫ heterozygote lines grow faster that the WT F9 cells in the absence of RA (Fig. 2, note the different y axis scales). The mean doubling time for the parent WT cell line was 22 h, whereas the F9 RAR␤ 2 ϩ/Ϫ 270 heterozygote line had a mean doubling time of 17 h, indicating that reduced levels of RAR␤ 2 protein are associated with a decrease in the cell cycle transit time.
Morphological Characterization of the F9 RAR␤ 2 Ϫ/Ϫ Cells-The morphology of the RAR␤ 2 Ϫ/Ϫ cells is markedly different from that of the parent F9 WT cells (Fig. 4). In the absence of RA, there are noticeable differences between the F9 WT cells and the F9 RAR␤ 2 Ϫ/Ϫ cells; the latter clump together in cell aggregates, which do not exhibit the irregular borders characteristic of F9 WT cells. Following RA treatment, F9 WT cells form long cellular processes that are absent in the RAR␤ 2 Ϫ/Ϫ cells, which look like undifferentiated cells (Fig. 4). F9 WT cells exhibit even more dramatic morphological alterations after the addition of cAMP and theophylline (47), unlike the RAR␤ 2 Ϫ/Ϫ cells. It should be noted that it is possible to observe a few morphological changes occurring in the F9 RAR␤ 2 Ϫ/Ϫ cells after treatment with RA; the RA-treated F9 RAR␤ 2 Ϫ/Ϫ cells become more flattened and exhibit more vacuoles. The F9 RAR␤ 2 ϩ/Ϫ cells exhibit a similar morphology to F9 WT cells in both the absence and presence of RA (data not shown), indicating that both alleles of RAR␤ 2 must be inactivated in order to observe alterations in the morphological phenotype.
Effects of RA on Gene Expression in F9 RAR␤ 2 ϩ/Ϫ and F9 RAR␤ 2 Ϫ/Ϫ Lines-The F9 WT, F9 RAR␤ 2 ϩ/Ϫ, and F9 RAR␤ 2 Ϫ/Ϫ cell lines were cultured in the presence of 1 M all-trans-RA for the indicated times, and the expression of several genes which are transcriptionally activated by RA was evaluated (Fig. 5, A and B). In the heterozygous F9 RAR␤ 2 ϩ/Ϫ lines, the induction of laminin B1, CRABP II, and P450RAI (CYP26) by RA was similar to or slightly lower than that in the F9 WT cells. The expression of the REX-1 gene, which in F9 WT cells was reduced by ϳ2-fold after 48 h of RA treatment, was reduced similarly in the RAR␤ 2 ϩ/Ϫ cells (2.4-fold). The greatest difference between F9 WT and F9 RAR␤ 2 ϩ/Ϫ cells was that the Hoxa-1 gene was always induced to a greater extent at earlier time points in the heterozygote lines than in the F9 WT cells. In the experiment shown, there was an 18-fold induction of Hoxa-1 mRNA in the heterozygote lines cultured in the presence of RA for 48 h versus a 4.7-fold induction in F9 WT cells cultured under the same conditions (Fig. 5, A and B). mRNA were comparable in the F9 WT and the RAR␤ 2 ϩ/Ϫ lines (7.3-and 6.5-fold induction).
In the F9 RAR␤ 2 Ϫ/Ϫ line, the expression of early response genes such as p450RAI and CRABP II was RA-responsive, although the magnitude of the induction of these mRNAs was much lower in the F9 RAR␤ 2 Ϫ/Ϫ cells than in the F9 WT cells.
In contrast, the induction of the Hoxa-1 message after 48 h of RA treatment was only 2-fold lower in the RAR␤ 2 Ϫ/Ϫ cells than in the F9 WT cells. The reduction in the levels of REX-1 message after 24 h of RA treatment was comparable in the RAR␤ 2 Ϫ/Ϫ cells and the F9 WT cells (Fig. 5, A and B). Interestingly, the reduced REX-1 mRNA levels were not maintained in the RAR␤ 2 Ϫ/Ϫ cells, and by 96 h REX-1 levels were the same as those in the untreated cells (see Fig. 5B for normalized data). Laminin B1, a late response gene, was not induced in the RAR␤ 2 Ϫ/Ϫ cells (Fig. 5, A and B). In summary, the F9 RAR␤ 2 Ϫ/Ϫ cells exhibited a reduction in the magnitude of the induction of a number of RA-responsive genes following RA treatment.
Metabolism of RA in the F9 RAR␤ 2 Ϫ/Ϫ Cells-Since we observed that the induction of p450RAI message after RA exposure was lower in the RAR␤ 2 Ϫ/Ϫ cells than in the F9 WT cells (Fig. 5, A and B), we compared the amounts of RA metabolized by the F9 WT and F9 RAR␤ 2 Ϫ/Ϫ lines. The two cell lines were treated for 72 h with vehicle or with 1 M RA and were then cultured in the presence of [ 3 H]all-trans-RA for 1 h or for 3 h, after which both the intracellular and the extracellular levels of the polar metabolites of [ 3 H]all-trans-RA were quantitated by reverse-phase HPLC (Fig. 6). Like the F9 WT cells, the F9 RAR␤ 2 Ϫ/Ϫ cells were able to metabolize RA and the levels of RA metabolites produced were dramatically increased after RA-induced differentiation. Since the levels of metabolites produced from [ 3 H]RA were similar in the F9 WT and the RAR␤ 2 Ϫ/Ϫ cell lines, the lower induction of the p450RAI transcripts after RA in the RAR␤ 2 Ϫ/Ϫ cells (Fig. 5) was not associated with an impairment of the ability to metabolize [ 3 H]RA under the conditions of our assays. DISCUSSION We have previously reported the generation of F9 teratocarcinoma cell lines carrying targeted disruptions of the RAR␣ and RAR␥ genes (16,17). We now continue the characterization of the roles of the RARs in mediating RA-induced F9 cell differentiation by generating cell lines carrying one or two disrupted alleles of the RAR␤ 2 gene. The RAR␤ 2 disruption completely abrogates the growth arrest that is one of the hallmarks of RA action in a variety of cell types. The F9 RAR␤ 2 Ϫ/Ϫ cells did not growth arrest in response to RA or a variety of other retinoids (Figs. 2 and 3). Thus, the data presented in this report, together with prior data from various F9 RAR and RXR knockout lines, strongly argue for a critical role for RAR␤ 2 in mediating growth arrest in response to RA in F9 cells (Figs. 2  and 3). Whereas in the absence of RAR␤ 2 no growth arrest in response to RA is observed in the F9 RAR␤ 2 Ϫ/Ϫ cells (Fig. 2), the RAR␣Ϫ/Ϫ and RAR␥Ϫ/Ϫ lines growth arrest like WT cells, and the RXR␣Ϫ/Ϫ and RXR␣Ϫ/Ϫ RAR␣Ϫ/Ϫ lines partially growth arrest after RA addition (16, 17, 48 -50). Only the RXR␣Ϫ/Ϫ RAR␥Ϫ/Ϫ line shows no growth arrest in response to RA. Although in this line RAR␤ 2 mRNA induction is decreased by 2-3-fold (49), this relatively small impairment of RAR␤ 2 expression probably does not contribute to the observed lack of growth arrest of the RXR␣Ϫ/Ϫ RAR␥Ϫ/Ϫ cells (50). Since the RAR␤ 2 Ϫ/Ϫ line has a lower level of RXR␣ protein than F9 WT cells (Fig. 1C), it is possible that in these RAR␤ 2 Ϫ/Ϫ cells the RXR␣:RAR␤ heterodimer is involved in mediating the RAinduced growth arrest.
The mechanism by which RA induces growth arrest is not known, although several reports have linked RAR␤ to growth regulation and apoptosis (see Introduction). Interestingly, the two independent F9 RAR␤ 2 ϩ/Ϫ heterozygote lines, which arrested their growth normally in response to RA, showed a faster cell cycle transit time in the absence of RA (Fig. 2). At present we do not have an explanation for this observation. The finding that two independently derived RAR␤ϩ/Ϫ lines grow faster than the F9 WT cells in the absence of RA (Fig. 2) is particularly intriguing in light of the greatly reduced expression of RAR␤ in a variety of human cancers and preneoplastic cells (see Introduction). Our data support the hypothesis that reduced levels of RAR␤ confer a selective growth advantage to preneoplastic cells and, therefore, that the loss of one allele of RAR␤ may be a key step in neoplastic progression.
The RAR␤ 2 disruption alters the morphology of these cells as compared with the F9 WT parent line (Fig. 4). In the absence of RA, this difference in morphology may result from the presence of low levels of RAR␤ 2 transcripts in the F9 WT cells and in the RAR␤ 2 ϩ/Ϫ heterozygote cells, but not in the RAR␤ 2 Ϫ/Ϫ line. After RA treatment, the F9 RAR␤ 2 Ϫ/Ϫ cells do not exhibit a typically differentiated morphology. A partial or complete lack of morphological differentiation following RA treatment has been previously observed in the F9 RAR␥Ϫ/Ϫ lines (16), the F9 RXR␣Ϫ/Ϫ lines (48), and in the double mutant lines F9 RXR␣Ϫ/ϪRAR␥Ϫ/Ϫ and F9 RXR␣Ϫ/ϪRAR␣Ϫ/Ϫ (50).
There was a markedly lower induction of several RA-responsive genes in the F9 RAR␤ 2 Ϫ/Ϫ cells when compared with F9 WT cells (Fig. 5). In contrast, in the F9 RAR␣Ϫ/Ϫ and RAR␥Ϫ/Ϫ lines, the inactivation of RAR␣ and RAR␥, respectively, caused a severely impaired induction of particular RAresponsive genes, whereas the induction of other RA-responsive genes was like that in F9 WT cells (16,17). The modulation of gene expression in response to RA in the RAR␤ 2 Ϫ/Ϫ cells was actually strongest at early time points when compared with the F9 WT cells. This RA induction of gene expression diminished at later times, and laminin B1, a late response gene, was not induced in the F9 RAR␤ 2 Ϫ/Ϫ cells. Our data suggests that in F9 WT cells the initial response to RA is mediated via the RAR␣ and RAR␥ receptors, but that the large increase in RAR␤ 2 receptors after RA treatment (which occurs after Ϸ16 -24 h) is required for the maximal expression of RA-responsive genes at later times after RA addition. Alternatively, the lack of growth arrest in response to RA observed in the RAR␤ 2 Ϫ/Ϫ cells may contribute to a reduction in the magnitude of the differentiation response in these cells. That the initial RA response with respect to transcriptional activation of RA-responsive genes such as Hoxa1 is present in the F9 RAR␤ 2 Ϫ/Ϫ cells is consistent with prior data indicating that RXR␣ and either RAR␣ or RAR␥ are the essential receptor pairs for these early responses during the RA-induced conver-sion of F9 stem cells to primitive endoderm cells (49). Collectively, the research presented here as well as prior research (48 -52) argue against a role for RAR␤ in the initial events in the RA-induced differentiation process.
The RA-inducible P450RAI has been implicated in the metabolism of RA in a variety of cell types (53,54). Previous data suggested that the RAR␥⅐RXR␣ heterodimer is responsible for the induction of p450RAI in F9 cells (17,54). Although in RAR␤ 2 Ϫ/Ϫ cells the levels of p450RAI transcripts attained after RA treatment were much lower than in the F9 WT cells (Fig. 5), there was no corresponding impairment in the ability of the RAR␤ 2 Ϫ/Ϫ cells to metabolize [ 3 H]RA (Fig. 6). Therefore, it is likely that the lower level of p450RAI mRNA present in RAR␤ 2 Ϫ/Ϫ cells is sufficient to metabolize 50 nM [ 3 H]RA at the same rate as that seen in F9 WT cells. In summary, our data indicate that RAR␤ 2 is not required for the induction of RA metabolism in F9 cells.
Since most of the genes evaluated (with the exception of laminin B1) were induced in the F9 RAR␤ 2 Ϫ/Ϫ cells following RA treatment, albeit at reduced levels, it is possible that in F9 cells RAR␤ does not activate the transcription of the early response genes tested, but instead is required to achieve the maximum induction of these RA-responsive genes. This result, together with the high degree of functional redundancy among the different RARs in the context of mouse knockouts (55) may explain why mice carrying two disrupted alleles of RAR␤ 2 were phenotypically normal (37) and why mice in which all isoforms of RAR␤ were disrupted (56, 57) presented a relatively mild phenotype when compared with the phenotypes of other RAR knockouts (58,59). This may also account for the observations that RAR␤Ϫ/Ϫ mice, as well as RAR␤Ϫ/Ϫ mice knocked out for the p53 gene or carrying a MMTV-ras transgene, did not exhibit any increase in tumor formation over a period of at least 14 months. 2