Tumor-suppressive activity of retinoic acid receptor-β in cancer

Abstract

Retinoids, a group of structural and functional analogs of vitamin A, are known to regulate a large number of essential biological processes and to suppress carcinogenesis. The effects of retinoids are mainly mediated by nuclear retinoid receptors, which include retinoic acid receptors (RARs) and retinoid X receptors (RXRs). Each receptor has three subtypes (α, β, and γ) and each subtype has different isoforms. Retinoic acid receptor-β (RAR-β) has four isoforms that have different affinities to retinoids and different biological functions. Loss of expression of RAR-β₂ during cancer development is associated with tumorigenesis and retinoid resistance; induction of its expression, on the other hand, can suppress carcinogenesis. Expression of another isoform, RAR-β₄, is increased in various types of cancer. RAR-β₄ transgenic mice develop hyperplasia and neoplasia in various tissues, and induction of RAR-β₄ expression increases the growth of tumor cells that do not express RAR-β₂. Future studies will focus on molecular pathways involving RAR-β₂ and the role of RAR-β₄ in cancer development.

1. Introduction

Retinoids, a group of structural and functional analogs of vitamin A, regulate several essential biological processes [1,2]. Pharmacologically, they have been recognized as important in modulating cell growth, differentiation, and apoptosis and in suppressing carcinogenesis in a broad range of tissue types in vitro, in animal models, and in clinical trials [see reviews in 3,4]. The ability of all-trans retinoic acid (RA) to induce the differentiation of acute promyelocytic leukemia (PML) cells into granulocytes is the basis for its therapeutic activity, and it was its effectiveness in this regard that led the U.S. Food and Drug Administration to approve RA as a therapeutic agent for PML. Promising results of retinoid studies in vitro and in vivo led to retinoid compounds being tested in a battery of clinical prevention and treatment trials being carried out all over the world. In clinical chemoprevention trials, retinoids showed activity in patients with oral leukoplakia [5, 6], cervical dysplasia [7], bronchial metaplasia [8], actinic keratosis [9], and second primary tumors in the aerodigestive tract [10,11]. However, findings from several other studies indicated that retinoids were not effective and could even be harmful [12–15]. Large randomized trials conducted in Europe and the United States showed that moderate doses of natural or synthetic vitamin A compounds were ineffective in reversing premalignancy or in suppressing recurrence of primary tumors [12,13]. Thus, the use of retinoids as chemopreventive agents in human cancers has gone from a zenith in the 1990s to something of a nadir nowadays. Current research efforts focus mainly on the molecular mechanisms underlying the actions of retinoids [16–21] and on use of retinoid X receptor-selective or retinoid receptor-independent retinoids in the prevention of human cancers [22–25].

The effects of retinoids are mediated mainly by two classes of nuclear retinoid receptors: retinoic acid receptors (RARs) and retinoid X receptors (RXRs), both of which are members of the steroid hormone receptor superfamily. Each receptor has three subtypes (α, β, and γ) and each subtype has different isoforms [26,27]. In humans, mRNA for RAR-α is expressed in most tissues; RAR-β expression is prevalent in neural tissues but hardly detectable in skin; and RAR-γ is expressed predominantly in the skin [26,27]. As for the RXRs, RXR-β is found in nearly all tissues; RXR-α is abundantly expressed in the liver, kidney, spleen, and skin; and RXR-γ expression seems to be restricted to mostly muscle and brain [26,27]. Changes in the expression of these receptors have been thought to cause neoplasia and malignant transformation in human cells [see reviews in ref. 3,4]. For example, chromosomal translocations that fuse the PML gene to the RAR-α gene form the chimeric genes PML-RARα and RARα-PML during the malignant transformation of PML [28], and recent evidence suggests that these fused genes are causally linked to the pathogenesis of the disease [29–31]. The RAR- β gene is rearranged in human hepatocellular carcinoma as a result of insertion of a hepatitis B virus sequence [32], although the importance of the rearrangement in causing hepatocellular carcinoma warrants further investigation. However, most alterations in nuclear retinoid receptors result from the reduction or loss of their expression in various premalignant and malignant tissues and cells. For example, loss of RAR-γ expression has been noted in skin cancer or premalignant lesions [33], but loss of RAR-β₂ expression is common in a wide variety of cancers [3,4]. Because much of the work published to date has focused on role of RAR-β₂ in various cancers, the focus of this review is on the molecular role of RAR-β in regulating cell growth and differentiation and in suppressing carcinogenesis.

2. Loss of RAR-β₂ expression as a biomarker in solid tumors

Since the 1990s, loss of the expression of nuclear retinoid receptors, including RAR-β₂, in various cancer cell lines has been detected by northern blotting or reverse transcriptase- polymerase chain reaction techniques [34–36]. Our group was the first to analyze the expression of transcripts of these receptors in formalin-fixed and paraffin-embedded tissues by using in situ hybridization [37]. Afterwards, many studies have demonstrated that among these receptors, loss of RAR-β₂ is the most common and the loss is progressive in premalignant and malignant tissues and cells (Table 1), including those of the head and neck [38,39], breast [40,41], lung [42,43], esophagus [44,45], pancreas [46], cervix [47], and prostate [48,49]. The observations that RAR-β₂ was upregulated in patients after treatment with 13-cis RA and that increased expression of RAR-β₂ correlated with clinical response [39] suggest that RAR-β₂ has an important role in suppressing carcinogenesis. After these findings were published, the methylation status of the RAR-β₂ gene promoter was used as a biomarker for the early detection of malignancy or as an intermediate end-point marker to monitor the efficacy of chemoprevention agents in clinical trials [50–60].

Table 1.

Suppression of Nuclear Retinoid Receptors in Premalignant Lesions and Tumors

Lesion	Lost receptor expression
Oral premalignant lesions	RAR-β2
Bronchial squamous metaplasia	RAR-β2
Head and neck cancer	RAR-β2
Lung cancer	RAR-β2, RAR-γ, RXR-β
Pancreatic cancer	RAR-β2
Esophageal cancer	RAR-β2
Breast cancer	RAR-β2, RAR-α
Prostate cancer	RAR-β2, RXR-β

Methylation of the RAR-β₂ gene promoter, along with methylation of other gene promoters, has been evaluated as a biomarker of breast cancer risk [50,51]. In one study, methylation of the RAR-β₂ gene promoter occurred in 32% of benign breast samples from patients with cancer but in only 9% of similar samples from patients without breast cancer [50]. In another study, random periareolar fine needle aspiration samples from the mammary glands showed methylation of the RAR-β₂ gene promoter in 69% of primary breast cancers tested, and methylation was positively associated with increasing cytologic abnormality in those samples [51]. Hypermethylation of the RAR-β₂ gene promoter is also an important factor in predicting early recurrence of lung cancer and may be useful as a prognostic marker in non-small cell lung cancer (NSCLC) as well, although the clinical implications of these findings need further investigation [52]. Methylation analysis of the RAR-β₂ gene promoter can also aid in the diagnosis of primary lung cancer in bronchial aspirate samples [53] or in serum DNA [54].

In chemoprevention studies, RAR-β₂ has often been used as an intermediate end-point biomarker in clinical trials of various retinoids [39,43,57,58,61]. However, evidence from preclinical studies suggests that RAR-β₂ may also be useful for monitoring the efficacy of other agents, such as inhibitors of the epidermal growth factor receptor (EGFR) and β-cryptoxanthin [59,60].

3. Molecular mechanisms responsible for loss of RAR-β₂ expression

The mechanisms underlying the loss of RAR-β₂ gene expression are not fully understood [reviewed in 3,4]. Early studies showed that loss of RAR-β₂ expression in some forms of cancer (e.g., NSCLC) resulted from chromosome 3p deletion [62]), but in others (e.g., head and neck cancer) neither homozygous deletions nor gene rearrangements have been found [34]. Since those findings were published, other studies have shown that transcriptional deregulation can silence RAR-β₂ expression through decreased levels of co-activators, the presence of co-repressors, or epigenetic mechanisms such as histone deacetylation [63–68]. Expression of RAR-β₂ also depends on the cellular level of retinoids because this receptor is itself an RA–inducible gene. Indeed, RAR-β₂ expression is selectively reduced in several organs during vitamin A–deficient states and is enhanced by RA, as demonstrated in studies of rats [69,70]. In humans, some premalignant tissues are deficient in vitamin A because of reduced uptake of vitamin A from serum or abnormally elevated catabolism of intracellular retinoids [71,72]. We found that the binding of an anti-RA antibody to tissue sections of premalignant lesions was much lower than the binding to normal oral mucosa, suggesting that the lesions had lower levels of retinoids [72]. Lack of RAR-α expression may also contribute to lost RAR-β₂ expression. RAR-α regulates RAR-β₂ transcription by mediating dynamic changes of RAR-β₂ chromatin in the presence and absence of RA. Interfering with RA signaling by silencing RAR-α exacerbates the repression of chromatin for RAR-β₂ and leads to RAR-β₂ transcriptional silencing. RAR-β₂ silencing is also associated with resistance to the growth-inhibitory effect of RA, indicating that RAR-β₂ silencing and RA resistance result from impaired integration of RA signaling at RAR-β₂ chromatin [73].

Nevertheless, many recent studies have shown that the diminished expression of RAR- β₂ in the development of different human cancers results from epigenetic silencing by methylation of cytosine-phospho-guanosine (CpG) islands in the promoter region of the gene [50–56]. Moreover, treatment with 5-aza-2-deoxycytidine, a DNA demethylation agent, was able to induce RAR-β expression in various cancer cell lines [74–77], further supporting the notion that RAR-β₂ is a tumor suppressor gene. Cigarette smoke causes morphologic changes and the loss of RAR-β₂ expression in lung tissues of animals [78], and cigarette smoke and the tobacco carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone have been shown to induce the methylation of the RAR-β₂ gene promoter in murine lung cancer [79]. Our own studies similarly showed that benzo[a]pyrene diol epoxide (BPDE), a carcinogen present in tobacco smoke and environmental pollution, and bile acid, a tumor promoter in the gastrointestinal tract, suppressed RAR-β₂ expression in premalignant and malignant esophageal cells and that BPDE induced the methylation of the RAR-β₂ gene promoter [19,80]. Notably, epigallocatechin gallate, a chemopreventive agent extracted from the tea plant, has been shown to bind to DNA methyltransferases and reactivate RAR-β₂ expression in esophageal cancer cells [81]. Thus, as discussed in section 2, methylation of the RAR-β₂ gene promoter is being evaluated as a biomarker for the early detection or prediction of prognosis for various forms of cancer in humans.

In patients with NSCLC, hypermethylation of the RAR-β gene promoter has different effects on the development of second primary lung cancers (SPLCs) depending on smoking status [82]. In one study, Kim et al [82] showed that current smokers developed more SPLCs when the RAR-β gene promoter was unmethylated than when it was hypermethylated. In contrast, the development of SPLCs in former smokers was higher in patients with hypermethylated than unmethylated RAR-β. This group suggested that in current smokers, the continuous high oxygen tension and free radicals induce apoptosis, which could inhibit the development of SPLC. However, another explanation is that overexpression of RAR-β₄ may be important in the development of SPLCs in active smokers. Three observations prompted this idea: First, expression of both RAR-β₂ and RAR-β₄ is controlled by the same RAR-β promoter (P2; details are given below); second, methylation of the RAR-β gene promoter shuts down the expression of both isoforms; and third, RAR-β₂ expression is lost but RAR-β₄ expression is increased in various forms of cancer, including lung [83]. Collectively, these observations support the idea that hypermethylation of the RAR-β gene promoter may have a protective effect in tumors in which RAR-β₄ is overexpressed. Previous studies have also revealed that exposure to tobacco smoke or BPDE suppresses RAR-β₂ expression through methylation of the RAR-β gene (19,79,80). This may be true for active smokers, but unmethylated RAR-β could trigger RAR-β₄ expression in various cancers [84–86], although the hypothesis that tobacco smoke can increase RAR-β₄ expression requires further testing. In contrast, hypermethylation of the RAR-β gene reduces RAR-β₂ expression, thereby losing any RAR-β₂-protective effects in the suppression of tumorigenesis; therefore nonsmokers or former smokers with hypermethylated RAR-β tend to develop more SPLCs.

In contrast to previous evidence of the importance of RAR-β₂ in suppressing cancer development, our own studies showed that expression of RAR-β in patients with early-stage NSCLC was unexpectedly associated with very poor prognosis [87]. Overexpression of RAR-β correlated with the increased expression of cyclooxygenase-2 (COX-2), an enzyme known to be present at elevated levels in progressive carcinogenesis and a marker of poor prognosis in various malignancies [88], and with increased levels of telomerase [89]. These findings were unexpected, as they contradict the concept of RAR-β₂ functioning as a tumor suppressor gene. In seeking possible mechanisms for this effect, we recently found that reduced expression of RAR-β₂ correlated with increased RAR-β₄ expression in esophageal cancer tissues [84]. This finding may help to clarify this apparent discrepancy between studies, because the in situ hybridization technique we used in the NSCLC studies cannot distinguish among RAR-β isoforms (β₁, β₂, and β₄). Given the finding that the ratio of RAR-β₄ to RAR-β₂ expression in human lung and breast cancer cell lines is higher than that in normal cells [83–86] and that reduced RAR-β₂ expression correlated with increased RAR-β₄ expression [84], we speculate that the in situ hybridization–detected RAR-β mRNA in NSCLC tissues is the RAR-β₄ isoform, which has oncogenic effects on cells. This speculation, however, requires further investigation.

4. RAR-β isoforms

Differences in the actions of P1 and P2, the two known RAR-β promoters (P1 initiates transcription of isoforms 1 and 3, and P2 initiates transcription of isoforms 2 and 4), and alternative splicing [26,27,90,91] give rise to four major isoforms of RAR-β in mice (β₁, β₂, β₃, and β₄) and three in humans (β₁, β₂, and β₄). Additional isoforms (e.g., RAR-β₅ and RAR-β1') have been identified in human cancer cells [92,93]. Briefly, RAR-β₁ is a fetal isoform that may be a master developmental gene in humans; it is also expressed in small-cell lung cancer [94]. In one study involving transgenic mice, RAR-β₁ was found to have unique tumor suppressor activity that could not be entirely compensated by the overexpression of RAR-β₂ and the suppression of RAR-β₄ [95]. RAR-β₂ is the most abundant and the major RA-inducible isoform, and thus the term RAR-β in the literature usually refers to the RAR-β₂ isoform. RAR-β₄ is generated by alternative splicing from the same primary transcripts as those generating RAR-β₂ and is initiated by the CUG codon [90]. The cDNA sequence of RAR-β₄ in regions B to F is identical to that of the same regions of RAR-β₂. The A region of RAR-β₄—only 4 amino acids long—is much shorter than that of RAR-β₂ [90] (Fig. 1). Moreover, this non-AUG codon seems to be relatively inefficient, resulting in alternative use of an internal methionine codon at +448 bp that yields an RAR-β isoform in which all amino acid sequences at the N-terminal of the second finger of the DNA binding domain are truncated. Because the full-length and truncated RAR-β₄ proteins retain the ability to heterodimerize with RXR-α and to interact with transcription cofactors but lack the DNA-binding capacity to regulate gene expression [85,90,96], RAR-β₄ and the truncated RAR-β₄ may act structurally as a dominant-negative form of RAR-β₂ [96]. In addition, RAR-β₄ has an elevated K_d value for RA binding and cannot inhibit AP-1 activity, as RAR-β₂ can [97]. Little is known as yet about the most recently discovered isoforms, RAR-β₅ and RAR-β1'. RAR-β₅ is expressed in human breast cancer cells, and its detection correlates with the RA resistance of estrogen-negative breast cancer cells [92]. RAR-β1' has antitumor activity in lung cancer [93]. In summary, the various RAR-β isoforms in humans have different affinities to RA (at least with regard to RAR-β₂ and RAR-β₄) and different biological functions (e.g., the RAR-β₂ protein is a tumor suppressor whereas RAR-β₄ has oncogenic properties).

Fig. 1.

Comparison of RAR-β₂ and RAR-β₄ cDNA (modified from ref. 90). A. Organization of RAR cDNA is shown at the top and RAR-β isoforms at the bottom. The location of sequences that encode regions B to F and the 3’-untranslated region (UTR) common to all RAR-β isoforms is indicated, as well as those of the 5’-UTR and A region, which is isoform-specific. Numbers correspond to nucleotide positions in the human RAR-β₂ cDNA sequence. The “A” regions of RAR-β₂ and RAR-β₄ are represented as a solid box. The region of RAR-β₂ that is spliced to give RAR-β₄ (between 266–619 bp) is indicated with a dashed line. B. Sequence and schematic representation of donor splice sites for RAR-β₄ and RAR-β₂ (positions 265 and 619, respectively).

The etiology of RAR-β₄ upregulation in cancer cells remains unknown. We speculate that altered expression of microRNAs (miRNA) in cancer tissues may be responsible. miRNA is a class of naturally occurring small noncoding RNA, 18 to 22 nucleotides long, that posttranslationally silences gene expression through binding to complementary target mRNAs, thereby degrading these mRNAs or inhibiting their being translated into protein [see reviews in 98]. This discovery has greatly broadened our understanding of gene regulation mechanisms. Although comprehensive knowledge of the specific mRNA targets of miRNAs is lacking [98,99], bioinformatics analyses may predict their mRNA targets; for example, miR-16, miR-128, and miR-30e can target the RAR-β gene [98]. Further investigation of this exciting discovery is warranted.

5. Role of RAR-β₂ and RAR-β₄ in human carcinogenesis

To date, several convincing lines of evidence have shown correlations between loss of RAR-β₂ expression and increased carcinogenesis in humans, but only a few studies have linked RAR-β₄ expression with cancer development. With respect to RAR-β₂, lung carcinoma cells expressing transfected RAR-β₂ are less tumorigenic in nude mice than the vector-control-transfected cells [100], and transgenic mice expressing antisense RAR-β₂ develop lung cancer [101]. Transfection with an RAR-β₂ expression vector can suppress the growth of various cancer cell lines and restore the sensitivity of those cells to RA treatment [see reviews in 3,4]. Knocking out RAR-β₂ by homologous recombination in F9 mouse teratocarcinoma cells resulted in the loss of RA-associated growth arrest and changes in cell morphology and differentiation [102]. Our own findings have shown that induction of RAR-β₂ expression in esophageal cancer cells suppressed tumor cell growth and colony formation and induced apoptosis [18]. Others have shown that RAR-β₂ also suppressed breast cancer cell metastasis in a mouse xenograft model [103]. In another study, loss of RAR-β₂ expression in patients with neuroblastoma correlated with poor prognosis [104]. Also, esophageal, lung, and breast cancer cell lines that do not express RAR-β₂ are resistant to retinoid treatment [35,44,73]. Finally, in one study of oral dysplastic tissues, half of the cells cultured from those tissues had become immortalized, as indicated by loss of RAR-β₂ and p16 expression, mutations in p53, and induction of hTERT mRNA [105]. Treating these cells with 5-aza-2-deoxycytidine reversed one of the immortal phenotypes and led to re-expression of both RAR-β₂ and p16 [106]. A subsequent study by the same group showed that only the loss of RAR-β and p16 was responsible for the immortalization of oral dysplastic cells [107].

Expression of RAR-β₄, on the other hand, was found to be increased in esophageal cancer tissues and the increase was associated with reduced expression of RAR-β₂ [84]. Pharmacologic concentrations of RA have been shown to increase RAR-β₄ expression in esophageal cancer cell lines (our unpublished data) and in breast cancer cells (86). Several other studies have shown that hyperplasia and neoplasia developed spontaneously in various tissues in RAR-β₄ transgenic mice [83] and that induction of RAR-β₄ expression enhanced the growth of cancer cells that do not express RAR-β₂ (our unpublished data). As noted earlier in this review, RAR-β₄ has an elevated K_d value for RA binding and cannot inhibit AP-1 activity, as does RAR-β₂ [97]. Collectively, these findings suggest that RAR-β₄ either may be an oncogene or may have oncogenic effects. Further investigation of the effects of RAR-β₄ on tumor cell growth, tumorigenesis, and gene expression will clarify whether RAR-β₄ is a dominant-negative form of RAR-β₂ or an oncogene. Such knowledge can be put to use in the development of better strategies to control cancer development.

6. Molecular pathways involved in the mediation of tumor suppression by RAR-β₂

Although the precise molecular mechanisms responsible for RAR-β₂-mediated antitumor activity are not fully understood, we previously found that RAR-β₂ suppressed the expression of EGFR, activating protein-1 (AP-1), and COX-2 and the phosphorylation of extracellular signal- regulated protein kinases 1 and 2 (Erk1/2) [19]. We also found that COX-2 expression was downregulated in RAR-β₂ -transfected esophageal cancer cells and that the restoration of sensitivity to RA was mediated by suppression of COX-2 expression [18]. Further, we unexpectedly found that stable transfection with RAR-β₂ cDNA did not restore the sensitivity of COX-2–negative esophageal cancer cells to RA or inhibit tumor formation in nude mouse xenograft models. These findings indicate that the antitumor effect of RAR-β₂ may require the suppression of COX-2 expression (our unpublished data). To further substantiate this notion, we studied the expression of RAR-β₂ and COX-2 mRNA in tissue specimens and found RAR-β₂ expression to be associated with low levels of COX-2 expression in esophageal cancer tissues. We also found that after a 3-month treatment with 13-cis RA, the induction of RAR-β₂ expression in human oral leukoplakia tissues correlated with a reduction in COX-2 expression and a clinical response (our unpublished data). Collectively, our findings suggest that the antitumor activity of RAR-β₂ occurs through the suppression of COX-2 expression, although further study is needed to determine whether manipulation of COX-2 expression in these cells could antagonize this antitumor activity. We also determined that downregulation of COX-2 expression by RAR-β₂ took place through the inhibition of EGFR, with subsequent decreases in ERK1/2 phosphorylation and AP-1 expression [19]. Most importantly, we showed that induction of COX-2 expression by the carcinogen BPDE depended on both the expression and the inhibition of RAR-β in esophageal cells [19] (Fig. 2). A previous study showed that stable transfection with RAR-β₂ exhibited strong inhibition of AP-1 activity in tumor cells, even in the absence of RA. Moreover, expression of the endogenous AP-1–responsive gene collagenase I was strongly repressed in cancer cells stably transfected with RAR-β₂ (108). Another line of evidence comes from a study showing that a high-fat diet reduced the expression of PPAR-γ and RAR-β₂ mRNA, increased COX-2 and β-catenin levels, and increased the number of aberrant crypt foci in rat colon tissue. However, vitamin A was able to prevent these high-fat-diet–induced alterations of PPAR-γ and RAR-β₂ and the increases in COX-2 and β-catenin [109]—further evidence that that RAR-β₂ can suppress COX-2 expression.

Fig. 2.

A molecular pathway triggered by RAR-β₂ for the suppression of human carcinogenesis. Benzo(a)pyrene diol epoxide (BPDE) exposure is a risk factor in the development of esophageal cancer. BPDE is thought to act by suppressing RAR-β₂ expression and, in turn, upregulating EGFR expression and Erk1/2 phosphorylation, resulting in the induction of AP-1 and COX-2 expression. Further, BPDE can educe the expression of retinoid receptor-induced gene-1 (RRIG1), perhaps through reduction of RAR-β₂. RRIG1 mediates RAR-β₂ effects on the regulation of cancer cell growth and gene expression. RRIG1 protein binds to and inhibits RhoA activity and consequently suppresses Erk1/2 phosphorylation and expression of COX-2 and cyclin D1, resulting in reductions in tumor cell colony formation, invasion, and proliferation.

A cDNA microarray analysis comparison of a parental and an RAR-β₂–transfected lung cancer cell line revealed 27 genes expressed at different levels between the two cell lines [16]. Several of the affected genes code for proteins whose functions augment apoptosis or host immune response. Other investigators identified and characterized genes that are differentially expressed in wild-type cells and RAR-β₂^{−/ −} F9 teratocarcinoma cells by using subtractive hybridization and DNA array analysis [17]. The identified genes, which encode transcription factors, cell surface signal-transduction molecules, and metabolic enzymes, included c-myc, FOG1, GATA6, glutamate dehydrogenase, Foxq1, Hic5, Meis1a, Dab2, midkine, and the PDGF-α receptor [17]. Again, induction of RAR-β₂ expression suppressed the potential for metastasis of breast cancer cells. Another cDNA microarray study showed that RAR-β₂ induced the expression of tumor-cell antigens (CTAG1 and CTAG2), the innate immune response (e.g., RIG-I/DDX58), and a tumor suppressor gene (TYRP1) but reduced the expression of several genes involved in cell adhesion (LSAMP, PCDH11Y, and CD64), nutrient availability (FABP6, SLC38A2, PCSK4, SRPB1), and transcription/AP-1 activity (HOXB7 and JUN) [20].

We identified and cloned a retinoid receptor-induced gene, RRIG1, which is expressed differentially by RAR-β₂-positive and -negative esophageal cancer cells, by using restriction fragment differential display–polymerase chain reaction techniques [21]. RRIG1 is expressed in a broad range of normal tissues but is lost in various types of cancer [21,110]. RRIG1 mediates the effect of RAR-β₂ in the regulation of cancer cell growth and gene expression [21]. The RRIG1 protein is expressed in the cell membrane and binds to and inhibits small GTPase RhoA activity. Whereas induction of RRIG1 expression inhibits RhoA activation and f-actin formation and consequently reduces colony formation, invasion, and proliferation of esophageal cancer cells, antisense RRIG1 increases RhoA activity and f-actin formation and thus induces the colony formation, invasion, and proliferation of these cells. These findings indicate the existence of a novel molecular pathway involving RAR-β₂ regulation of RRIG1 expression and RRIG1–RhoA interactions [21]. Further studies of RRIG1 and the pathways involved may translate into better control of esophageal cancer.

In summary, a large body of evidence has established that RAR-β₂ participates in the regulation of cell growth and the suppression of carcinogenesis, although the molecular mechanisms of action deserve further study. However, the role of RAR-β₄ in cancer development and its molecular signaling transduction pathways has remained virtually unexplored.

7. Future directions

The associations between loss of RAR-β₂ expression and tumorigenesis, and induction of RAR-β₂ expression and the suppression of cancer development, have been well established [3,4]. However, expression of the closely related isoform RAR-β₄ is increased in various types of cancer [83–86], RAR-β₄ transgenic mice developed hyperplasia and neoplasia in various tissues [83], and induction of RAR-β₄ expression increased growth of tumor cells that do not express RAR-β₂ (our unpublished data). Therefore, future studies are needed of the molecular pathways triggered by RAR-β₂ to suppress cancer development, as are other studies of the role of RAR-β₄ in cancer development. Issues remaining to be clarified include whether RAR-β₄-mediated oncogenic effects on normal and cancerous cells are attributable to its modulation of RAR-β₂-mediated signal pathways, or whether RAR-β₄ acts on separate signaling pathways leading to malignant transformation and cancer development. Answers to these questions will provide crucial clues for resolving the occasionally contradictory and controversial findings from studies of retinoids for cancer prevention and treatment.