Characterization of Retinoic Acid Receptor-deficient Keratinocytes*

Retinoids are essential for normal epidermal growth and differentiation and show potential for the prevention or treatment of various epithelial neoplasms. The retinoic acid receptors (RAR a , - b , and - g ) are transducers of the retinoid signal. The epidermis expresses RAR g and RAR a , both of which are potential mediators of the effects of retinoids in the epidermis. To further investigate the role(s) of these receptors, we derived transformed keratinocyte lines from wild-type, RAR a , RAR g , and RAR ag null mice and investigated their response to retinoids, including growth inhibition, markers of growth and differentiation, and AP-1 activity. Our results indicate that RAR g is the principle receptor contributing to all- trans -retinoic acid (RA)-medi-ated growth arrest in this system. This effect partially correlated with inhibition of AP-1 activity. In the absence of RARs, the synthetic retinoid N -(4-hydroxyphenyl)-retin-amide inhibited growth; this was not observed with RA, 9- cis RA, or the synthetic retinoid (E)-4-[2-(5, 5, 8, 8 tetra-methyl-5,6,7,8-tetrahydro-2-naphthalenyl)-1-propenyl] benzoic acid. Finally, both RAR a and derived from the mean ( 6 S.D.) of four replicate wells. Transient Transfection and AP-1 Activity Assay— Transfections were performed using Lipofect ACE reagent (Life Technologies, Inc.). Briefly, cells were plated in 6-well cluster plates at 4 3 10 4 cells/well. Trans- fections consisted of 0.5 m g of AP-1 reporter or appropriate control (41), either alone or with expression vectors encoding c-Fos, c-Jun, CBP, p300, or RARs. Total DNA (5 m g; normalized with KS 1 ) was mixed with 10 m l of lipid and added to 100 m l of serum-free S-minimal essential medium. The lipid/DNA mixture was then added to the cells in 1 ml of complete medium and incubated at 37 °C overnight. Medium was changed daily, and luciferase activity was assessed 48 h post-transfec-tion. Results were corrected for protein concentration and are expressed as the mean ( 6 S.D.) from three independent transfections. All experiments were repeated at least three times with comparable results. from each cell protein using DC protein assay kit (Bio-Rad). Electrophoretic mobility shift assays ; of interest were detected by incubation with the desired antibodies and detection with an ECL kit (Amersham Pharmacia Biotech) as per the manufacturer’s instructions. Antibodies were purchased from Santa Cruz Biotechnology. Northern Blot Analysis— Fifteen micrograms of total RNA, isolated by Trizol reagent (Life Technologies, Inc.), were size fractionated on a 1% agarose-formaldehyde gel in MOPS buffer and transferred to a MAGNA nylon membrane (MSI). Fragments were isolated by restric-tion digestion of cDNAs followed by purification by Geneclean and used to generate probes by labeling with [ a - 32 P]CTP by random priming with an oligo labeling kit (Amersham Pharmacia Biotech). Membranes were hybridized according to the manufacturer’s directions.

Vitamin A derivatives (retinoids) play central roles in embryonic development and maintenance of various tissues in the adult (1)(2)(3). Retinoids also exhibit potent antitumorigenic properties in diverse model systems and show potential for the treatment of a number of human malignancies, including diverse epithelial cancers or pre-cancerous lesions (4 -9).
The retinoid signal is transduced by two families of nuclear receptors, the retinoic acid (RA) 1 receptors (RAR␣, -␤, and -␥ and their isoforms) and the retinoid x receptors (RXR␣, -␤, and -␥) (10 -13). RARs function as ligand-inducible transcription regulators by binding, together with an RXR partner, to specific cis-acting response elements (RAREs). RARs can be activated by both RA and its stereoisomer, 9-cis RA, whereas RXRs are activated only by 9-cis RA (14). RXRs are also essential heterodimeric partners for a number of other nuclear receptor signaling pathways, including thyroid hormone, vitamin D, and certain orphan receptors (15,16). Although 9-cis RA is not obligatory for transcriptional regulation via these pathways, some results suggest that RXR-specific ligands can elicit transcriptional activation in certain settings (e.g. Refs. [17][18][19][20]. RARs, like several other nuclear receptors, can function in a ligand-dependent manner to inhibit AP-1 activity, and it has been suggested that the affect of retinoids on the growth of transformed cells may occur through this trans-repression mechanism (21)(22)(23). This inhibition is believed to be due, at least in part, to competition for limiting amounts of transcriptional co-factors, such as CBP and/or its homologue p300, common to both pathways (24,25). Other mechanisms, such as inhibition of the expression of AP-1 family members or c-Jun N-terminal kinase (JNK), may also contribute to this cross-talk (26 -29).
Gene targeting of the various RARs has revealed essential and diverse roles for these receptors (2,30,31). However, because of perinatal or embryonic lethality inherent to many of these RAR null backgrounds, there is a void in our knowledge of RAR function in a number of contexts, such as tumorigenesis.
Exogenous retinoids can attenuate the effects of tumor promoters in the two stage skin carcinogenesis protocol (9,32). Among the retinoid receptors, normal epidermis expresses RAR␥ and RAR␣␥ as well as RXR␣ and RXR␤, with RAR␥ and RXR␣ as the predominant heterodimer (33,34). This pattern of expression prompted us to investigate the roles of RAR␣ and RAR␥ in mediating the antitumorigenic effects of retinoids in epithelial keratinocytes. To this end, we established RAR␣, RAR␥, and RAR␣␥ null keratinocyte lines by transformation with a dominant-negative p53 expression vector and compared the properties of these various lines. Our results demonstrate that RAR␣ and RAR␥ affect different aspects of retinoid response in these transformed cells, with RAR␥ being the primary mediator of RA-induced growth inhibition. However, other synthetic ligands affected proliferation independent of the RARs. RAR-dependent, but not -independent, growth inhibitory effects generally correlated with the attenuation of AP-1 transcriptional activity. Finally, the effects of RAR␣ and RAR␥ on expression of certain keratinocyte markers suggests that each RAR may perform a subset of specific functions, which cannot be entirely fulfilled by other RARs in this cell type.

EXPERIMENTAL PROCEDURES
Primary Keratinocyte Culture and Immortilization-The RAR null mice used in these studies have been described previously (35,36). RAR␣, RAR␥, and RAR␣␥ mutants were generated from the appropriate matings, whereas wild-type offspring were obtained from RAR␥ ϩ/Ϫ intercrosses. Fetuses were procured by caesarean section at 18.5 days post coitus, and genotype was determined by polymerase chain reaction as described (37). Primary keratinocyte cultures were established from * This work was supported by the National Cancer Institute of Canada with funds from the Canadian Cancer Society (to D. L.) and by personnel support from the Medical Research Council of Canada (to P. G.), the Cancer Research Society, Inc. (to C. F. C.), and the Fonds de la Recherches en Santé de Québec (to D. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
A single 10-cm plate of cells (ϳ2 ϫ 10 6 ) of each genotype was harvested, and cells were resuspended in 800 l of medium. The cells were then electroporated (250 mV, 960 microfarads in a 0.4-cm gap cuvette) with 25 g of a linearized expression vector harboring a mutated p53 from the Friend erythroleukemia cell line CB7 (39). Cells were plated and routinely subcultured until past crisis. All experiments were performed using cultures between passage 16 and 26.
Growth Assays-Transformed keratinocytes were seeded into 96well plates at a cell density of 500 cells/well and were treated the following day with vehicle (Me 2 SO) or the appropriate retinoid (RA, 9-cis RA, 4-HPR, or TTNPB). Medium was replenished every second day. Growth was assessed either in response to varying concentrations of retinoid at eight days post-plating or over time in response to 10 Ϫ6 M ligand. DNA content was assessed as a measure of cell growth using crystal violet staining as described previously (40). Relative dye binding was assessed by OD at 590 nm using a microplate reader. Results were expressed either as A 590 values or as growth relative to untreated controls and were derived from the mean (Ϯ S.D.) of four replicate wells.
Transient Transfection and AP-1 Activity Assay-Transfections were performed using Lipofect ACE reagent (Life Technologies, Inc.). Briefly, cells were plated in 6-well cluster plates at 4 ϫ 10 4 cells/well. Transfections consisted of 0.5 g of AP-1 reporter or appropriate control (41), either alone or with expression vectors encoding c-Fos, c-Jun, CBP, p300, or RARs. Total DNA (5 g; normalized with KSϩ) was mixed with 10 l of lipid and added to 100 l of serum-free S-minimal essential medium. The lipid/DNA mixture was then added to the cells in 1 ml of complete medium and incubated at 37°C overnight. Medium was changed daily, and luciferase activity was assessed 48 h post-transfection. Results were corrected for protein concentration and are expressed as the mean (Ϯ S.D.) from three independent transfections. All experiments were repeated at least three times with comparable results.
Electrophoretic Mobility Shift Assays and Western Blot Analysis-Cells were cultured in 10-cm plates in the presence of RA (10 Ϫ6 M) or vehicle for 48 h prior to harvest. Nuclear proteins were isolated from each cell line, and protein concentration was determined using the DC protein assay kit (Bio-Rad). Electrophoretic mobility shift assays were performed essentially as before (42). Briefly, binding reactions containing ϳ2 ng of probe (50,000 cpm) and 5 g of nuclear protein were separated by electrophoresis through a 6% polyacrylamide gel containing 0.25 ϫ Tris borate and EDTA. Specificity of binding was assessed by competition with a 10-fold excess of unlabeled RARE (5Ј-GGGTAGGG-TTCACCGAAAGTTCACTCGCA) or AP-1 (5Ј-GATCCGATGAGTCAG-CCA) double-stranded oligonucleotides.
For Western blot analysis, 40 g of nuclear protein from the various cell lines were size fractionated on a 10% SDS-polyacrylamide gel electrophoresis and electroblotted to Immobilon-P polyvinylidene difluoride membrane as recommended by the supplier (Millipore). Proteins FIG. 1. Electrophoretic mobility shift assay of an RARE sequence by nuclear extracts from transformed keratinocyte lines. Nuclear extracts (5 g) were incubated with a labeled doublestranded RARE oligonucleotide probe (50,000 cpm) and bound and free probe separated by polyacrylamide gel electrophoresis. Binding was competed with 10-fold excess RARE probe (C) but not by a nonspecific probe comprised of an AP-1 recognition sequence (NSC). Note the presence of a nonspecific complex (N/S) migrating slightly faster than the RAR-containing complex (RAR). WT, wild type. of interest were detected by incubation with the desired antibodies and detection with an ECL kit (Amersham Pharmacia Biotech) as per the manufacturer's instructions. Antibodies were purchased from Santa Cruz Biotechnology.
Northern Blot Analysis-Fifteen micrograms of total RNA, isolated by Trizol reagent (Life Technologies, Inc.), were size fractionated on a 1% agarose-formaldehyde gel in MOPS buffer and transferred to a MAGNA nylon membrane (MSI). Fragments were isolated by restriction digestion of cDNAs followed by purification by Geneclean and used to generate probes by labeling with [␣-32 P]CTP by random priming with an oligo labeling kit (Amersham Pharmacia Biotech). Membranes were hybridized according to the manufacturer's directions.

Generation of RAR Null Cell
Lines-Primary cultures of wild-type and RAR null keratinocytes showed no major differences in morphology, growth, or immortalization with dominant-negative p53. All lines grew well for at least 40 passages, suggesting that they were immortalized. None of the lines formed colonies in soft agar or were tumorigenic in nude mice. 2 Electrophoretic mobility shift assay revealed that, relative to wild-type extracts, disruption of RAR␣, and to a greater extent RAR␥, decreased specific binding to an RARE, and association was completely abolished in extracts from RAR␣␥ double null cultures (Fig. 1). Northern blot analysis confirmed the disruption of RAR␣ and/or RAR␥ message in the appropriate cell line (data not shown). RAR␤ transcripts were undetectable in all lines by Northern blot or polymerase chain reaction approaches, consistent with previous studies indicating that this receptor type is not expressed in epidermal keratinocytes (33,43). These data suggest that there is no compensatory upregulation of the remaining receptors in response to disruption of a given RAR.
Contribution of Specific RARs to Retinoid-mediated Growth Inhibition-All transformed cell lines exhibited similar morphology and growth characteristics in the absence of retinoid treatment (Fig. 2). However, wild-type and RAR␣ Ϫ/Ϫ cultures were growth inhibited by 10 Ϫ6 M RA (Fig. 2). In marked contrast, RAR␥ Ϫ/Ϫ cells were highly resistant and RAR␣␥ Ϫ/Ϫ cultures were completely resistant to these effects. The growth arrest observed in wild-type and RAR␣ null cultures was likely because of the inhibition of proliferation as opposed to apoptosis, as judged by thymidine incorporation and programmed cell death assays (data not shown).
Dose-response experiments were performed to determine the relative sensitivity of the various cell lines to growth arrest by RA, 9-cis RA, or the synthetic retinoids TTNPB or 4-HPR. As shown in Fig. 3, wild-type keratinocytes exhibited a significant reduction in proliferation at 10 Ϫ9 M RA, with the maximal affect at 10 Ϫ7 -10 Ϫ6 M RA. RAR␣ Ϫ/Ϫ keratinocytes exhibited a similar profile, although their response to RA was slightly more pronounced than wild-type cultures. Consistent with timecourse analysis, RAR␥ Ϫ/Ϫ keratinocytes were only marginally inhibited by the highest dose of RA examined (10 Ϫ6 M), and RAR␣␥ Ϫ/Ϫ keratinocytes were not significantly affected by RA at any dose tested. 9-cis RA is a ligand for both RARs and RXRs, and RXR agonists have been shown to induce effects on growth or differentiation in several model systems. Proliferation of both wildtype and RAR␣ Ϫ/Ϫ cultures was inhibited by 9-cis RA, although higher concentrations were required compared with RA (Fig.  3). Interestingly, 9-cis RA had no significant outcome on the growth of either RAR␥ Ϫ/Ϫ or RAR␣␥ Ϫ/Ϫ cultures. This finding suggests that RXR activation does not lead to growth arrest in this model system, at least in the absence of RARs. Whether RXR-specific signaling has other biological consequences remains to be investigated.
The RAR agonist TTNPB was a very potent inhibitor of growth in wild-type or RAR␣ Ϫ/Ϫ cultures with an effect evident at 10 Ϫ11 -10 Ϫ10 M (Fig. 3). However, TTNPB affected RAR␥ Ϫ/Ϫ and RAR␣␥ Ϫ/Ϫ cultures only at the highest dose tested (10 Ϫ6 M). Whether this is indicative of effects on other pathways or is because of nonspecific cytotoxicity is unknown.
The synthetic retinoid 4-HPR has been shown to be a potent inducer of growth arrest and/or apoptosis in several model systems (44 -46). This compound was the least efficient of all those tested in inhibiting proliferation of wild-type and RAR␣ Ϫ/Ϫ cultures (Fig. 3). However, in marked contrast to the other retinoids, 4-HPR affected the growth of RAR␥ and RAR␣␥ cultures at high doses, consistent with receptor-dependent and -independent mechanisms of action for this compound (47)(48)(49)(50).
RAR Regulation of AP-1 Transcriptional Activity-RARs can repress AP-1 transcriptional activity, and this mechanism of action has been proposed to underlie at least some of the antitumorigenic effects of retinoids (8,9,43,51). In transient transfection assays, we found that RA (10 Ϫ6 M) inhibited AP-1 activity 8 -10-fold in wild-type and RAR␣ Ϫ/Ϫ cultures (Fig. 4A). AP-1 activity in RAR␥ Ϫ/Ϫ cultures was more modestly affected, typically exhibiting 10 -30% reduction, whereas activity in RAR␣␥ Ϫ/Ϫ cultures was not affected. The latter line was capable of response following re-introduction of either RAR␣ or RAR␥ by transient transfection (Fig. 4B). Thus, attenuation of AP-1 activity requires the presence of at least one functional RAR, although there does not appear to be discrimination between receptor types for this outcome.
Dose-response studies revealed a close parallel between AP-1 activity and growth inhibition mediated by all four compounds in wild-type cultures (Fig. 5). However, growth arrest induced by 4-HPR in RAR␥ and RAR␣␥ mutant lines never correlated with a reduction of AP-1 activity (data not shown). This finding underscores a unique and unknown mechanism of action for this retinoid in affecting proliferation.
We next determined the effect of RA on the expression of AP-1 members in wild-type and RAR null lines. Both the basal mRNA levels and RA response of several of the AP-1 members varied across the different RAR null lines. In untreated cells, c-fos expression was comparable across all four lines, although it was slightly reduced in RAR␣␥ cells (Fig. 6). RA strongly inhibited c-fos in both wild-type and RAR␣ Ϫ/Ϫ lines but had no effect in RAR␥ or RAR␣␥ cultures. This pattern was also observed at the protein level (Fig. 7). In contrast, treatment affected c-jun expression only in RAR␣ null cultures, although basal mRNA levels varied across the lines. However, c-Jun protein did not reflect its cognate mRNA levels and was reduced by RA treatment in wildtype, RAR␣ Ϫ/Ϫ , and RAR␥ Ϫ/Ϫ cultures. Phosphorylated c-Jun (P-Jun) levels paralleled those of c-Jun, suggesting that variations in phosphorylation were because of alterations in total c-Jun levels, rather than affects on JNK activity.
Fra-1 expression was barely detectable in wild-type, RAR␣ Ϫ/Ϫ , and RAR␣␥ Ϫ/Ϫ lines but was elevated in RAR␥ Ϫ/Ϫ cultures; Western blot analysis was inconclusive, as the signal was too weak to be distinguished (data not shown). junB and junD expression did not vary significantly with the exception that junB levels were slightly reduced in the RAR␣␥ Ϫ/Ϫ line ( Fig. 6 and data not shown). A number of mechanisms have been suggested to underlie retinoid repression of AP-1 activity. These include inhibition of JNK activity, which is unlikely given the observation that P-Jun levels appear to change as a function of c-Jun levels. Alternatively, the observed down-regulation of c-Fos and/or c-Jun proteins might play a role, especially if either of them are limiting. A third mechanism involves competition for limiting ancillary factors common to both RAR and AP-1 transcriptional complexes, such as p300/CBP (24). We addressed the latter two possibilities by assessing the ability of exogenous CBP, p300, c-Fos, or c-Jun to negate the effects of RA treatment on AP-1 activity.
CBP or p300 transfection in wild-type cells resulted in a dose-dependent increase in AP-1 activity in the absence of RA (Fig. 8A). Interestingly, p300 appeared to be more potent in affecting AP-1 activity, suggesting that it may be preferred over CBP in this context. Despite this increase in activity, expressing the data as fold inhibition indicated that both factors resulted in only a modest reversal of inhibition (Fig. 8B).
Overexpression of either c-Fos or c-Jun also resulted in an increase in basal AP-1 activity, again with only a marginal reduction in fold repression mediated by RA (Fig. 8). Although this rescue effect was more pronounced when both c-Jun and c-Fos were co-transfected, repression was not completely abolished (Fig. 8B). These observations suggest that several mechanisms, including titration of limiting co-factors and inhibition of expression of AP-1 family members, act in concert in an RAR-dependent manner to attenuate AP-1 activity in these transformants.
Effect of RAR Ablation on Gene Expression-Northern blot analysis was performed to study the effect of receptor disruption on the expression levels of several genes implicated in keratinocyte growth and differentiation. The major integrin isoforms found in epidermis are integrin ␣ 2 , ␣ 3 , ␣ 6 , ␤ 1 , and ␤ 4 . These are expressed in basal keratinocytes, and a decrease in integrin expression is generally correlated with differentiation and loss of proliferative potential (52,53). Northern blot analysis revealed that, with the exception of the ␣ 3 isoform, all integrins were down-regulated by RA treatment in wild-type, RAR␣ Ϫ/Ϫ , and RAR␥ Ϫ/Ϫ cultures in a manner that correlated with the effects of treatment on proliferation (Fig. 9). Moreover, although integrin expression was not affected by RA treatment in RAR␣␥ cultures, basal expression of integrins ␣ 2 , ␣ 3 , and ␤ 4 was substantially reduced. The seemingly contradictory observation that both RA excess and RAR loss can reduce expression of several integrins is perhaps indicative of an altered differentiation state in the double mutant line. Interestingly, integrin ␤ 1 expression decreased in untreated RAR␣ Ϫ/Ϫ cells and was up-regulated in untreated RAR␥ mutant cultures.
Keratin expression patterns reflect the differentiation states of the various epithelial strata (52,53). K5/K14 are expressed in basal epidermal cells, whereas K1/K10 are associated with early differentiation steps and predominate in suprabasal cells. K6 and K19 are not expressed in normal epidermal keratinocytes but are often observed in situations of aberrant proliferation, such as psoriasis, wound healing, and propagation in tissue culture. With the exception of RAR␣␥ null cultures, RA treatment repressed expression of K10 (Fig. 9). This observation may be related to the fact that RA excess can inhibit keratinocyte differentiation (54,55). However, RAR disruption did not result in up-regulation of this differentiation marker, suggesting that K10 is not normally regulated by the RARs but responds to pharmacological levels of RA. RA suppressed K6 expression in wild-type, RAR␣ Ϫ/Ϫ , and (to a lesser extent) RAR␥ Ϫ/Ϫ cultures, consistent with the effect of treatment on both AP-1 activity and proliferation. K19 expression was induced in wild-type and RAR␣ Ϫ/Ϫ cultures. Interestingly, this gene was also up-regulated in RAR␥ Ϫ/Ϫ cells in the absence of treatment. These data indicate that, as for the integrins, the roles of the various RARs on keratin expression vary depending on both the receptor and the gene of interest. Moreover, most of the integrin and keratin markers were affected in both RAR␣ and RAR␥ null cultures. This demonstrates that both receptor types transduce effects on expression of many responsive genes. DISCUSSION We present, for the first time, the effects of RAR disruption on the characteristics and RA response of transformed epidermal keratinocytes. These data indicate that each RAR type plays both specific as well as overlapping roles in events related to keratinocyte growth and gene expression.
RARs Are Not Necessary for Survival and Growth of Transformed Keratinocytes in Culture-Previous work using a dominant-negative RAR␣ under the control of a basal-keratinocytespecific promoter suggested that RA signaling is essential for normal keratinocyte differentiation (56). We found that wildtype and RAR␣␥ null keratinocytes are comparable in regards to growth and morphology, although expression of some markers, such as integrin ␣ 2 , did differ. It is unlikely that RAR␤ plays any compensatory role in these cells, as we have never observed expression of this receptor in these cultures irrespective of RAR status. Moreover, the RAR␣␥ line was completely resistant to excess RA with respect to all outcomes examined. Thus, these cells are likely completely devoid of functional RARs. The difference between the relatively mild effects observed in the present study, compared with a dominant-negative RAR (56), suggests that transgene expression affects other pathways, perhaps by sequestration of RXRs. Alternatively, we cannot exclude an unrecognized compensation mechanism in the RAR null animals and derivative cells that may mask certain roles for these receptors in skin.
RAR␥ Is the Principle Mediator of Growth Arrest in Transformed Keratinocytes-Analysis of the effects of the various RARs on growth inhibition suggests that the major player in transducing this effect is RAR␥ with only a negligible contribution by RAR␣. This may simply be because of the prevalence of the former receptor type in keratinocytes (57) rather than indicative of receptor-specific function. Nevertheless, irrespective of the basis for this finding, these data suggest that targeting RAR␥ is a logical strategy to affect disorders of keratinocyte proliferation. In the absence of the RARs, the RXR ligand 9-cis RA had no effect on proliferation, suggesting that liganded RXRs do not impact on keratinocyte growth, at least in the absence of RARs. This is in contrast to previous reports, which suggests that growth inhibitory effects can be mediated by RXR-selective agonists in several other transformed cell types (17, 19, 58 -61). This suggests either that RXR ligands inhibit growth in a cell type-specific manner or that these synthetic agonists have effects that cannot be mimicked by 9-cis RA. It will be of interest to determine if such ligands exert an effect in the RAR␣␥ null line.
Although 9-cis RA inhibited the growth of wild-type and RAR␣ null lines, it was consistently less potent than RA or TTNPB. Because both RA and 9-cis RA have comparable affinities for the RARs (62), this observation suggests either that 9-cis is more labile than RA or is titered away from RAR signaling. In contrast to 9-cis RA, the synthetic RAR agonist TTNPB was a strong inhibitor of growth. As displacement assays suggest that TTNPB binds to the RARs with a lower affinity than RA, the greater relative potency of this analog likely lies in its enhanced stability and/or weaker association with cellular RA-binding proteins relative to RA (63). Moreover, although full manifestation of growth inhibition by TTNPB required the RARs, a slight inhibition at the highest dose used was seen in RAR␥ and RAR␣␥ null cultures, suggesting either nonspecific cytotoxicity or effects via nonreceptormediated pathways.
Evidence for RAR-independent Mechanisms of Growth Inhibition by 4-HPR-4-HPR can inhibit growth and induce apoptosis in a number of model systems (44,64,65). Although this compound can act directly via the RARs, some of its effects may also be mediated by receptor-independent mechanisms (47,66). Indeed, we found that 4-HPR inhibited proliferation in all lines assessed but that this occurred at lower concentrations in RARpositive cultures. This is consistent with 4-HPR acting through both RAR-dependent and -independent mechanisms. As 4-HPR is well tolerated at high doses (65), it should mediate effects even in those epithelial tumors that lack retinoid responsiveness to "pure" RAR agonists.
RARs and Inhibition of AP-1 Activity-RA is a potent inhibitor of AP-1, and this mechanism of action has been proposed to underlie some of the antitumorigenic effects of retinoids (2, 67-70, 72, 73). We found that both RAR␣ and RAR␥ can repress AP-1 activity in p53-transformed keratinocytes in culture. However, attenuation of AP-1 activity in RAR␥ null cultures (ϳ30%) did not agree well with growth inhibition in this cell line, which was minimal. A lack of correlation was especially notable in the case of 4-HPR, which never repressed AP-1 reporter activity in RAR␥ or RAR␣␥ null cultures at concentrations that inhibited their growth. This observation supports the existence of retinoid-mediated mechanisms of growth inhibition unrelated to AP-1 activity, at least in this model.
Multiple Pathways for RA in Inhibition of AP-1 Activity-Several mechanism have been proposed to underlie the crosstalk between AP-1 and RA signaling pathways. These include titration of common transcriptional co-regulators, such as CBP/ p300 (24), effects on expression of AP-1 family members (27,74,75), or inhibition of JNK activity (28). We found evidence that supports the two former possibilities in our model.
RA inhibited c-Fos and c-Jun (and P-Jun) protein levels in a manner that partially correlated with the observed attenuation of AP-1 activity. Although c-Jun levels paralleled the effect of treatment on AP-1 across the various cell lines, c-Fos was not affected by RA in RAR␥ or RAR␣␥ null cultures. These data underscore a specific role for RAR␥ in inhibiting c-Fos expression, whereas either RAR␣ or RAR␥ affected c-Jun. In this regard, it is interesting to note that c-Jun was also elevated in untreated RAR␣ and RAR␣␥ lines, indicating that RAR␣, presumably in its unliganded state, represses c-Jun expression. Consistent with this possibility, prior studies also suggest that certain RARs may be antagonistic to one another in F9 cells (71,76). FIG. 9. Northern blot analysis of keratinocyte gene expression. Wild-type (WT) and RAR null keratinocytes were treated with carrier (Ϫ) or 10 Ϫ6 M RA (ϩ) for 48 h. 15 g of total RNA from each cell line were used to prepare Northern blots, which were probed with cDNAs encoding various keratinocyte markers, as denoted to the right of the figure. Ribosomal RNA (18S) was used as loading control. The results are typical of at least two experiments.
Transfection of either CBP or p300 induced basal AP-1 expression in wild-type keratinocytes and modestly attenuated the effects of RA on AP-1 activity. The persistent inhibition of AP-1 by RA, even following transfection of higher levels of CBP/p300 expression vectors, suggests that the RARs remain in excess. Alternatively, as c-Fos and/or c-Jun also appear to contribute to this relationship, the combination of several distinct events probably underlies retinoid antagonism of AP-1.
RAR␣ and RAR␥ Play Both Overlapping and Specific Roles in Keratinocytes-The phenotype of RAR mutant mice clearly indicates that these receptors are highly but not completely redundant. This is supported by work using RAR null F9 embryocarcinoma cells (e.g. Refs. 71 and 76). Our present findings suggest that specificity also exists regarding RAR function in keratinocytes. Although certain aspects of this selectivity may be explained by the relative levels of expression, with RAR␥ being the predominant receptor type, this is not always the case. For example, repression of integrin ␣ 6 and integrin ␤ 4 was more affected by the loss of RAR␣ than the loss of RAR␥.
In addition to specific functions, it is also interesting to note that RAR␣ may actually attenuate the effects of RAR␥ in certain instances. For example, growth inhibition mediated by RA, 9-cis RA, or TTNPB was greater in RAR␣ null cells relative to wild-type cultures, suggesting that RAR␥ is a more efficient inducer of growth arrest in the absence of RAR␣. Other complex interactions were also observed. RA induced keratin 19 expression in wild-type cells, and this induction was greatly reduced in RAR␣ null lines, indicating that RAR␣ positively regulates this gene. However, in RAR␥ null cultures, K19 basal expression was induced, and RA induction was lost. One mechanistic explanation for this is that RAR␥ normally mediates repression of this gene in the absence of RA and that RAR␣ induces its expression in the presence of ligand. Consistent with this, K19 expression in RAR␣␥ cells returns to basal values, and regulation is lost, supporting a model for opposing function between RAR␣ and RAR␥ in regulating this gene; such antagonism has been suggested previously (71). Although analysis of the regulatory sequences governing this response is necessary to understand the nature of these observations, these examples offer strong evidence that RAR␣ and RAR␥ are not completely functionally equivalent in this model system.