Retinoids and Rexinoids in Cancer Prevention: From Laboratory to Clinic

Receptor-mediated retinoid signaling

The diversity of the functional effectors of vitamin A signaling is a function of the active metabolites, their receptors and dimerization partners, potential co-regulators and response element polymorphisms. The biological activity of natural retinoids is primarily mediated by selective ligand-activated transcription factors of the nuclear hormone receptor family [20]. The receptors that bind naturally occurring classical retinoids, RARs have three isoforms, α (NR1B1), β (NR1B2) and γ(NR1B3), as well as several splice variants arising from alternative promoters. Another group of retinoid receptors, RXRs, whose ligands were unknown at the time of their discovery, is also comprised of three isoforms. RXR-RAR heterodimers act as ligand-dependent transcriptional regulators by binding to specific DNA-response elements found in the promoter regions of target genes.

Phylogenically and structurally related, RARs and RXRs are both characterized by similar, highly conserved modular domain structures. These contain C-terminal ligand binding domains (LBDs) and central DNA-binding domains (DBDs) that include two zinc-finger modules and two α-helices, as well as variable N-terminal regions. The N-terminal domain that contains naturally disordered protein structures and whose role therefore has initially been elusive, was implicated in a role of providing the flexibility needed for critical posttranslational modification events by kinases and ubiquitin-ligases [21]. Nevertheless, the functions of RARs are not limited to the regulation of cognate target genes. The assembly of co-regulatory protein complexes is directed by the C-terminal ligand-binding domain of RARs. In addition, RARs can act as co-repressors for other gene pathways.

RARs generally dimerize with an RXR as a partner to be activated by a ligand. This is an unequal relationship, where RAR agonists can autonomously activate transcription, while a full response to rexinoids requires binding of an RAR agonist. This is commonly referred to as a ‘non-permissive’ nuclear receptor interaction. In contrast, RXR ligands are able to act on their own through an RXR homodimer, or a variety of partners (PPARs, LXRs, THRs, FXR, etc.). The latter partnering allows RXRs to activate transcription in the absence of a rexinoid. Therefore these are called “permissive” heterodimers (Figure 4). Similar to most type II nuclear receptors, retinoid receptors are retained in the nucleus regardless of their ligand binding status. Bound to regulatory elements of their target genes as RXR/RAR heterodimers, in the absence of ligands, they interact with the co-repressors NCoR, its genetic paralog SMRT, Sin3 or others. These transcriptional repressor complexes, in turn, recruit histone deacetylases that lead to the repression of target gene transcription [22]. Binding of an agonist triggers the repositioning of the H-12 C-terminal helix which results in the dissociation of inhibitory co-regulators and recruitment of cofactor complexes such as CBP/p300, SRCs or CARM1. Co-activators share a consensus α-helical LXXLL motif that fits into the hydrophobic cleft of the receptor LBD that modify local chromatin structure and activate the basal transcription machinery.

While in normal tissues retinoid target genes are readily activated by an agonist, in tumors epigenetic events often interfere with normal retinoid response. Critical elements of the regulatory regions are methylated, which prevents the synthesis of gene specific messenger RNA. A prime example is RARβ, in itself a retinoid-regulated gene, and a key regulator of antineoplastic pathways induced by retinoids. The silencing of RARβ during epithelial carcinogenesis is a prominent reason for retinoid resistance and limited efficacy of retinoid action [23].

The majority of retinoids interact with the ligand binding domains of some or all of the RARs, and some RAR-selective retinoids will bind RXRs, but with a very low affinity. The significance of RXRs is more pronounced in their essential role for the efficient binding to their genomic response elements of the subfamily 1 members of the nuclear hormone receptor superfamily. The latter are generally considered potential RXR partners (Figure 4). When forming dimers with RAR, RXRs are not able to respond to RXR-selective ligands, unless RAR is liganded. This phenomenon is known as RXR subordination. Yet, RXRs are not silent partners in RXR-RAR heterodimers, as simultaneous ligand binding of both receptors results in synergistic transcriptional activation [24]. In contrast, some other receptors (i.e. PPARs, LXRs, FXR) allow dimer activation by an RXR agonist, often replicating the phenotypic response to their own ligands. Rexinoids may modulate gene expression through biochemically detectable dimers of two RXRs; however activation of a target gene as a result of RXR homodimer activation is difficult to show. Rexinoids exert anti-proliferative, proapoptotic and differentiating biological effects on epithelial cells in vitro, and have also been shown to inhibit cancer cell migration, invasiveness and angiogenesis [25]. Naturally occurring ligands of RXRs include 9-cis-RA, docosahexaenoic acid and phytanic acid, and the list extends over a line of synthetic agonists with improved side-effect profiles or heterodimer-selective properties.

Retinoid receptor partners

The transcriptional outcome upon ligand activation is generally specific for the partner receptor and results in the modulation of a transcriptional program commensurate with well-defined biologic responses. For example, vitamin D receptor activation induces genes responsible for maintaining calcium homeostasis and bone formation, LXR agonists activate cholesterol transport mechanisms to protect cells from cholesterol overload, and FXR regulates the enterohepatic circulation of bile acids. Recent in vitro data indicate that retinoid signaling could also be mediated by other nuclear hormone receptors, such as COUP-TF II (NR2F2) [26], RORβ (NR1F2) or TR2 [27], although the relevance of these interactions in vivo remains to be shown.

Molecular mechanisms that modulate retinoid activity. The make-up of a nuclear receptor dimer not only determines which genes it can regulate but also determines which of the many co-regulatory molecules it may bind. The large number of possible complexes formed by dimers and co-regulators lends a level of plasticity almost unparalleled by other therapeutic targets. Furthermore, isotype-selective agonists may dictate differential responses by isotype-restricted co-repressor recruitment [28]. An additional mechanism to modulate retinoid-dependent cell signaling pathways or diminish retinoid signaling has been described in various cancer cells. Altered localization of RXRalpha was identified as a mechanism that provides the means for rapid changes in hepatoma cell response to LPS, as well as a survival strategy for breast cancer cells, which undergo apoptosis upon nucleoplasmic overexpression of RXR [29, 30]. In hepatocellular carcinoma elevated ERK activity phosphorylates RXRalpha on serine 260 and threonine 82, impairing its ubiquitination and inhibiting transcription of its target genes by a dominant-negative mechanism [31]. Phosphorylation events on RAR or co-repressors have also been reported to restore transcriptional competence to RXR and promote differentiation in leukemia cells. This raises the possibility of exploiting mechanisms that modulate heterodimer non-permissivity as a therapeutic option [32].

The magnitude and duration of the hormone response is regulated by the ubiquitin-dependent degradation of RARs and RXRs in the proteasome. In addition, retinoid receptors integrate a series of signaling pathways by means of phosphorylation events which cooperate with retinoids for the control of target gene expression. The highly conserved ligand binding domain (LBD) contains several phosphorylation sites used by the kinases PKA and MSK1, while the N-terminal domain contains a proline-rich motif that serves as substrate for CDKs and MAPKs. The complexity in the transcriptional responses produced by these multiple levels of regulation presents a unique therapeutic opportunity. Although the induced molecular changes overlap, RXR selective agonists differ in their anti-tumor effects from retinoids both in the preclinical and clinical settings.

Extra-nuclear retinoid signaling

In the classical sense, RA is a prototype of hormones that activate nuclear receptor transcription factors that directly regulate gene expression. However, various studies from the past decade demonstrated that several kinase cascades may be activated just minutes after addition of retinol. As these events are transient, the possibility of a transcription-dependent phenomenon can be ruled out. Numerous studies suggested that retinol potentially exerts non-genomic activities through different types of interactions with signaling kinases (Figure 5) [33]. For example, retinol can induce cytokine-dependent activation of JAK2 and subsequently the STAT5 transcription factor, while being exchanged from extracellular RBP to intracellular CRBP-I through Stra6 [34], creating a close link between retinoid homeostasis and immune modulation. Stable association was shown between the regulatory p85 subunit of PI3K and membrane-localized RARalpha in neuroblastoma cells independent of the presence of RA, shifting to an RAR association with the p110 catalytic subunit upon RA liganding [35]. PI3K activation then results in sequential phosphorylation of downstream MAPKs and ribosomal S6 kinases, and is critical for RA-dependent differentiation. Retinol may also induce differentiation in a receptor-independent manner in bronchial epithelial cells by CREB-dependent transcriptional regulation. This mechanism is mediated by protein kinase C through the activation of Erk1/2 and the p90 S6 kinase and is not dependent on retinoid receptors [36]. Extranuclear RARα localized to lipid rafts was shown to trigger the phosphorylation of Rac1, p38 MAPKs and MSK1 through activation of Gαq upon ligand binding (Figure 5) [37]. Activation of MSK1 phosphorylates RARs and ultimately triggers their recruitment to response elements of select target genes, such as Cyp26A1 and RARbeta2 [38]. Nevertheless, RA-induced p38 activation has also been associated with negative regulation of cell differentiation and growth inhibition [39]. In neurons, RA was shown to play a role in neuritogenic differentiation through direct interaction with cytoplasmic c-Src [40]. Cytosolic translocation of RXR has been reported in hepatocytes, where it was implicated in the suppression of the inflammatory response to LPS administration [29]. Another intriguing novel function of RAR was proposed by Chen et al. in a study that localized RARalpha to the dendrites of neurons in association with mRNA granules and revealed its involvement in the regulation of translation [41]. However, many of these new mechanisms were identified in vitro and their relevance to human health and pharmacology is not sufficiently understood yet.

2.1. Retinoids in Dermatology

Commercially retinoids have made the greatest impact in the cosmetics arena, and they are frequently used in creams and ointments to alleviate skin aging. Oral RA is efficacious in keratinizing disorders, but is considered too toxic; therefore, the pharmaceutical industry undertook the synthesis of retinoid derivatives to generate safer alternatives. Minor structural changes, such as forming cis-isomers, resulted in a new class of oral retinoids with strikingly different pharmacological properties. This was a milestone in dermatologic drug discovery. For instance, the isomer 13-cis-retinoic acid demonstrated impressive efficacy in acne and could cure severe acne conglobata [42]. In addition to systemic administration, several retinoids are also available for topical use (Table 1).

The major skin diseases benefiting from retinoids include acne, prosiarsis, disorders of keratinization, as well as precancerous lesions and non-melanoma skin cancer. For the treatment of acne vulgaris retinoids represented a breakthrough in the 1980s. Isotretinoin, the cis isomer of RA, normalizes epidermal differentiation, depresses sebum secretion and reduces inflammation, making long-lasting remission possible. Psoriasis treatment often involves immunosuppressive agents such as cyclosporine, tacrolimus and methotrexate and biologic agents targeting T-cells or tumor necrosis factor-α. Therefore, retinoids play an important role in cases where immunosuppression is not an acceptable side-effect, i.e. in children, patients at high risk of cancer or HIV carriers. Furthermore, the use of retinoids allows a dose reduction of phototherapy with UVB or PUVA, when used in combination. In plaque psoriasis acitretin and etretinate monotherapy may result in substantial improvement in 50% of the cases. In erythrodermic and pustular psoriasis these retinoids represent first-line therapy. Complete clearance requires the combination of retinoids with topical steroids, vitamin D derivatives or phototherapy. Three retinoids are currently approved for systemic therapy, isotretinoin (as Accutane, since 1982) for severe recalcitrant nodular acne, etretinate (Tegison, 1986) for severe recalcitrant psoriasis and acitretin (Soriatane, 1996) for severe psoriasis. Tretinoin tocoferil, the alpha-tocopherol ester of atRA, has been successfully used to treat lichen and macular amyloidosis and skin ulcers [43] (see Table 1).

The prevention of cutaneous malignant diseases was among the first chemopreventive applications of retinoids. Premalignant skin lesions, including actinic keratosis and HPV-induced tumors are frequently treated with retinoids [44]. Patients at high-risk for skin cancer because of genetic disposition (xeroderma pigmentosum), organ transplantation or history of multiple tumors, may be treated with etretinate or acitretin for 6 months [45].

Chronic obstructive pulmonary disease (COPD) and emphysema

Endogeous retinoids play a critical role in the pre- and postnatal development of the lungs. Indeed, the spatial and temporal distribution of retinaldehyde dehydrogenases, enzymes responsible for local RA production define the regions and periods of embryonal lung development [12]. Low retinol levels have been associated with bronchopulmonary dysplasia, a chronic lung condition arising from premature birth. Lung hypoplasia has also been associated with impaired cellular retinol uptake observed in clinical cases of Stra6 mutation [14]. Nevertheless, the importance of retinoid signaling extends to adult life and healthy pulmonary function. An inverse relationship between serum retinol levels and lung function was established in epidemiological studies conducted in smokers with COPD. Therefore, it is an intriguing notion that retinoids might trigger alveolar regeneration in the adult lung after tissue damage has occurred. In spite of the expectations, the Feasibility of Retinoids for the Treatment of Emphysema (FORTE) study using atRA or 13-cis-RA in a 6-month trial with a 3-month crossover period failed to detect significant improvements in CT densitometry score, gas transfer or quality of life [46]. Preclinical models of dexamethasone-impaired alveologenesis and airspace enlargement were used to identify the retinoid receptor subtypes critical for mediating improvement in lung function. 13-cis-RA, retinol, or RXR agonists did not, but 4-oxo RA and RARγ selective agents were able to induce regeneration [47]. Thus, the safety and efficacy of palovarotene (R667), an oral γ-selective retinoid agonist were tested in a randomized, double-blind, placebo-controlled study administering drug for 1 year to 262 patients with emphysema or severe a(1)-antitrypsin deficiency (Table 1) [48]. While none of the differences reached statistical significance, most functional lung parameters changed favorably in the retinoid-treated group versus the placebo group, and pavarotene was generally well tolerated.

Neurologic diseases

There is a growing body of literature that suggests that pharmacological targeting of certain nuclear receptors, particularly RXRs, LXRs and PPARs, is beneficial in neurodegenerative diseases [49, 50]. Genetic studies revealed that markers close to four of six retinoid receptor genes and genes encoding three of the four known retinol-binding proteins had linkage to Alzheimer’s disease (AD). More importantly, Cramer and coworkers showed that oral administration of bexarotene in a mouse model of AD resulted in the clearance of soluble β-amyloid (Abeta) in an apoE-dependent manner, and this was associated with a reversal of cognitive, social, and olfactory deficits [51]. Some subsequent attempts to independently reproduce the original report yielded controversial results, possibly due to the use of different formulations of bexarotene. However, in a similar study PPARgamma and RXRalpha agonists additively enhanced the microglial uptake of Abeta and improved spatial memory [50]. A phase 1b and a phase 2 study are underway to examine the effect of bexarotene on Abeta production and clearance.

In 2013 Parkinson’s disease was added to the list of conditions that could potentially benefit from rexinoids. Screening for small molecules that interact with the orphan nuclear receptor Nurr1 complexes, McFarland et al. showed that bexarotene can activate Nurr1, thereby promoting growth and survival of dopaminergic neurons and preventing motor deficit [52]. While Nurr1 has no known natural agonists, by forming heterodimers with RXR, rexinoids may indirectly modulate Nurr 1 action [53]. Interestingly, these effects were observed when administering markedly lower than the typical clinical doses of bexarotene.

Successful clinical use of retinoids as cancer treatment

Retinoids were shown to promote differentiation and cell death of cancer cells in a number of experimental systems [54, 55]. Some of the most successful therapeutic uses of retinoids are due to their differentiation-inducing effects. The most relevant therapeutic trials and applications of retinoids are summarized in Table 2.

Acute promyelocytic leukemia (APL): The best characterized example of retinoid receptor malfunction in human disease is linked to acute promyelocyte leukemia (APL), in which reciprocal chromosomal t(15,17) translocation between the genes of RARα and the promyelocyte leukemia protein (PML) lead to the alteration of signaling through both proteins. The resulting PML/RXRα fusion protein causes transcriptional repression and is the single causative factor in the etiology of APL. Preclinical investigations revealed that atRA induces terminal differentiation in the leukemic cells by promoting the degradation of the fusion protein through the action of the UBE1L ubiquitin-activating enzyme [56]. Recent studies identified the ubiquitin-specific protease UBP43, an antagonist of UBE1L, as an antineoplastic target whose inhibition destabilizes the t(15;17) PML/RARα fusion protein and induces apoptosis of APL cells [57]. The causative role of a single molecular defect paved the way for differentiation-based therapy and led to successful clinical application of atRA and 13-cis-RA for the treatment of APL (Table 2) [58]. Thus, the addition of atRA to the standard, anthracycline-based therapy increases the odds of long-term survival to 80%, compared to the chemotherapy alone which is curative in only 40% of the cases [58]. The 12 year follow-up of the 379 APL patients enrolled in the First North American Intergroup Study has shown 80% complete responses (CR) in the patients randomized to atRA-treatment for one year after chemotherapy, versus observation [59]. Notably, the differentiating effect of atRA is synergistically enhanced by arsenic trioxide, a compound targeting the PML component of the oncoprotein [60], which achieves remission in the rare cases of relapse after atRA and anthracycline therapy [61]. A recent phase III trial carried out in low-intermediate risk APL patients demonstrated better two-year event-free (97% vs. 86%) and overall survival rates with the combination of atRA with arsenic trioxide than with chemotherapy [62].

The use of retinoids in APL patients fundamentally changed the course of the disease for most patients, and it is bound to transform our approach to cancer treatment in more than one way. First, from a disease with a generally poor prognosis APL was transformed into the most prognostically favorable subtype of acute myeloid leukemia. In addition, the identification of this molecular defect led to the development of one of the first targeted therapies and provided the rationale for personalized medicine. Moreover, the possibility of achieving optimal efficacy without the use of chemotherapeutic drugs marks a novel path in curative therapy that emphasizes improved standard of care.

Kaposi’s sarcoma: 9-cis-RA (alitretinoin) is successfully used in the treatment of refractory cutaneous Kaposi’s sarcoma. Topical treatment with alitretinoin in a randomized, vehicle-controlled phase III trial was associated with a 30% higher response rate with mild to moderate adverse events over 12 weeks (Table 2) [63]. In AIDS-related Kaposi’s sarcoma cases with increased IL-6 levels durable objective response to alitretinoin was reported with improved rates and times to progression [64].

Squamous cell skin cancer: Following positive responses in the initial phase I/II studies with 13-cis-RA administered in combination with interferon-α2a, two out of three phase II/III trials also demonstrated increased progression-free survival in advanced skin, cervical and renal cancers [65]. A phase III randomized trial did not confirm the results, as no prevention of tumor recurrence or second primary tumors were detected, possibly due to a lack of appropriate statistical power [66]. Nonrandomized trials of 13-cis-RA in combination with interferon-α2a and chemotherapy for the treatment of advanced head and neck cancers showed promising results, and are awaiting confirmation by randomized studies [67].

Cutaneous T-cell lymphoma (CTCL): Natural and synthetic analogs of RA have been used for years both as monotherapy and in combinations for the treatment of CTCL [45]. The RXR-selective retinoid bexarotene was found to be active against all stages of the disease. Bexarotene showed a 45–54% response rate in a small phase 2 trial [68] and a 60% response rate in a retrospective study of long-term CTCL patients, earning FDA approval for the treatment of CTCL refractory to at least one previous therapy [68, 69].

Neuroblastoma: More than a decade ago it was suggested that retinoids may be beneficial in the treatment of neuroblastoma [70]. The greatest challenge in the treatment of neuroblastomas is its high recurrence rate. Recently a randomized phase III study established that intermittent dosing of 13-cis-RA significantly improved survival (Table 2) [71]. A phase I trial of fenretinide delivered in an oral powderized lipid complex (LXS) was reported to have anti-tumor activity in patients with refractory neuroblastoma, reaching higher plasma levels yet reduced toxicity as compared to conventional pills [72].

Breast cancer: In the only single agent clinical trial with atRA, this agent failed to achieve the primary response endpoint in patients with metastatic breast cancer [73]. Thus, subsequent studies investigated retinoid efficacy in combination with tamoxifen or interferons [74]. Isotretinoin, alitretinoin and retinyl palmitate were also tested as therapies in late-stage post-menopausal breast cancer patients (Table 2). No potentiation in the effect of tamoxifen or a difference in survival was detected [75], although a clinical response was seen in 55% of the patients treated with retinyl palmitate [76]. Clinical benefit with bexarotene was documented in 20% of patients with hormone-refractory metastatic breast cancers, with lower toxicity than seen with retinoids [77].

Lung cancer: Preclinical findings of anti-proliferative and pro-apoptotic activity in squamous cell cancers inspired a randomized phase II trial in advanced non-small cell lung cancer (NSCLC). The addition of atRA to chemotherapy yielded a benefit of increased response rates and progression-free survival [78]. Furthermore, a phase I/II first-line metastatic study of bexarotene combined with vinorelbine and cisplatin demonstrated a marked extension in median survival [79]. Unfortunately, these results were not confirmed in subsequent phase III trials of bexarotene plus cisplatin and vinorelbine, as progression-free or overall survival of NSCLC patients receiving combined treatment including bexarotene was indistinguishable from the placebo group [80] (Table 2). The Biomarker-integrated Approaches of Targeted Therapy for Lung Cancer Elimination (BATTLE) trial marked a paradigm shift in the design of clinical studies by achieving personalized therapy through adaptive randomization of patients [81]. With one of the four treatment arms incorporating bexarotene in combination with the tyrosine kinase inhibitor erlotinib, the BATTLE trial also revealed further detail on the applicability of bexarotene in lung cancer. Erlotinib plus bexarotene produced a 50% overall 8-week disease control rate (DCR) in pretreated lung cancer patients, and were particularly effective in the groups positive for EGFR mutations and RXRα/cyclinD1 protein expression.

Related clinical results were obtained and reported in a phase II trial conducted in chemotherapy-refractory KRAS mutation-positive NSCLCs [82]. But in addition to showing major clinical responses induced by the combination of bexarotene and erlotinib in advanced NSCLC, these studies defined cyclin D1 as a predictive biomarker, which enables oncologists to make personalized therapy decisions [83]. Cyclin D1, frequently overexpressed in lung cancer, plays a significant role in premalignant lung carcinogenesis, and is therefore a relevant target. Retinoid receptor selective agonists in human bronchial epithelial cells induced cyclin D1 degradation through RARbeta and RXR-dependent ubiquitination [84]. It has been proposed that dual targeting of cyclin D1 will result in increased efficacy, and overcome limitations of retinoid treatment in patients carrying polymorphisms in the cyclin D1 gene that make cyclin D1 resistant to degradation.

Challenges in the Medical Use of Retinoids

Pharmacodynamic considerations: Retinoid bioavailability is a crucial factor for the prediction of retinoid effectiveness and toxicity, particularly in light of the multiple enzymatic steps involved in both the creation and the degradation of active metabolites. Retinoid pharmacokinetics is further complicated by the differential lipid solubility of these agents. For example, etretinate is 50 times more lipophilic than acitretin and is released slowly from adipose tissues, where it is stored [85]. Thus, etretinate levels are detectable in serum for 2–3 years, while acitretin, isotretinoin and its metabolites reach normal levels two weeks after discontinuation of treatment. In vivo, pharmacodynamic mechanisms exert marked pharmacokinetic effects on locally available, intratumoral drug levels and have been shown to be a major determinant of the biological response to treatment [83].

Retinoid toxicity: Systemic retinoid toxicity in adults is similar to hypervitaminosis A, manifested in skin dryness, conjunctivitis and hair loss. The cutaneous and metabolic side-effects of retinoid treatment are dose-dependent and reversible. A minor fraction of APL patients receiving atRA (14–16%) develop retinoic acid syndrome (RAS), encompassing dyspnea, fever, pulmonary hemorrhage and respiratory failure. While these potentially lethal complications can be successfully treated by high-dose steroids, identifying patients who will develop RAS is critical. High initial and increasing white blood cell counts and CD13 expression on leukemic cells indicate predisposition to RAS [86].

Yet, in general, retinoids belong to the group of often well-tolerated pharmacologic agents. Many patients have been treated with retinoids, particularly etretinate or isotretinoin, continuously for as many as 15 years without developing signs of chronic toxicity [87]. However, because retinoids play an important role in the differentiation of embryonic tissue and organ formation, exposure of the fetus to therapeutic doses of retinoids is associated with teratogenic side effects. Thus, systemic use of these agents in patients of child-bearing age requires contraception for the duration of sustained retinoid levels in the body.

In contrast to retinoids, RXR-selective ligands have reduced teratogenicity and cutaneous toxicity. The most frequent adverse effects associated with rexinoid use are hyperlipidemia including triglycerides and cholesterol, the suppression of thyroid function and osteoporosis. Hyperlipidemia is treated with dose adjustment and lipid lowering drugs [88]. Individuals requiring life-long treatment should be subject to close monitoring. In comparison, long-term treatment with rexinoids is associated with a lower level of adverse effects than retinoids.

Retinoid resistance: Malignant transformation of epithelial cells may profoundly change retinol metabolism and contribute to the development of retinoid resistance [15]. Cellular retinoid uptake is regulated by LRAT, the enzyme that esterifies retinol to retinyl esters, and the RBP receptor Stra6. LRAT expression is suppressed in many human tissues undergoing malignant transformation, including breast, prostate, skin, kidney or bladder, resulting in low intracellular retinyl ester levels [89]. Stra6 mutations are associated with developmental errors [90]. Breast and prostate cancers also express aberrantly low levels of ALDH1a2, the enzyme forming RA from retinaldehyde [91]. ALDH1a2 has therefore been proposed as a tumor suppressor [92]. Mechanisms of retinoid clearance may also be affected in cancer cells, mainly through the over-expression of CYP26A1, the enzyme that oxidizes RA to more polar metabolites [93].

Strategies to minimize retinoid toxicity

Beneath the wealth of pharmacological benefits, toxicity represents the main limitation to the chronic use of retinoids. Therefore, it is imperative to work out viable strategies to minimize toxicity without compromising therapeutic effectiveness. One successful approach includes administering high, intermittent doses of a retinoid, which may be more effective than continuous low-dose therapy, as exemplified by 13-cis-RA to improve event-free survival of neuroblastoma patients [150]. Muindi et al. determined in pharmacokinetic measurements of the FORTE study that intermittent therapy with high-dose atRA would result in the greatest atRA exposure [151]. Other efforts to minimize retinoid-related adverse effects encompass modifications to the basic retinoid structure. For example, the alpha-tocopheryl ester of atRA, tretinoin tocoferil, is 150 times less toxic than the parent compound and lacks teratogenicity [152]. A synthetic analogue of 9-cis-retinoic acid, 9cUAB30, primarily binds RXR, yet lacks the hypertriglyceridemic effect of most rexinoids [153]. The first pharmacokinetic studies aim to determine the optimal dose and side-effects of this drug [154]. However, new drug development based on structure-function relationship is expensive and very time-consuming.

Alternatively, the use of combinations harbors great potential with respect to both increasing efficacy, as well as limiting toxicity. Combinations synergistic for the cancer preventive effect or its surrogate are expected to retain efficacy even at low doses of the agents, where drug-specific side-effects are reduced. Synergistic interactions may result from affecting the same or related pathways, or may involve unrelated pathways.

A unique mechanism of cooperation (or synergy) was shown between vitamin K2 and both acyclic retinoids or 9-cis-RA, offering a perspective for a potential chemopreventive combination. It was proposed that vitamin K2 prevents the phosphorylation and accumulation of inactive RXRα by inhibiting Ras and extracellular signal-regulated kinase activation, allowing retinoids to exert their cancer preventive effect in hepatic cells [155]. ACRs themselves may be able to restore impaired receptor function through inhibition of RXR phosphorylation. Another approach is to examine potential enhancing interactions among the multitude of partnering nuclear hormone receptors. This way anti-proliferative and anti-tumor cooperation was demonstrated between RXRs and PPARs in premalignant mammary epithelial cells, as well as in vivo [156, 157].

Targeting the same pathways by two agents may accentuate an overlapping spectrum of adverse effects. Therefore, current preclinical studies take an unbiased approach to identify synergistic combinations in vitro in phenotypic compound screens of premalignant precursor cells or more recently using computational methods. Such combinations might consist of a component that exhibits proven cancer preventive characteristics and a second agent selected from a pool of compounds based on its ability to enhance the effect of the primary chemopreventive agent. Candidate combinations are subsequently tested in mechanistic studies and animal models, with a goal of being translated into clinical settings. The inclusion of agents that are already FDA-approved and clinically tested, albeit with a different scope of indications, will require “drug repurposing”, but might be the fastest, and most efficient ways to arrive at clinically testable, low-toxicity alternatives for cancer prevention [158]. One novel prospect for a breast cancer preventive combination includes the rexinoid bexarotene and a non-selective beta-adrenergic inhibitor, carvedilol. Identified in high-throughput compound screens of premalignant breast epithelial cells for combinations synergistically anti-proliferative over a wide dose-range, low doses of these two agents are being tested in transgenic mouse models of breast cancer to establish their preventive capacity (I.U. unpublished observation). Beyond the immediate clinical applicability of the two FDA-approved agents, this and similar investigations will likely reveal potentially exploitable novel interactions of adrenergic signaling pathways and rexinoid-dependent anti-cancer activity.

Summary - The future of retinoids in the prevention of cancer

The medical use of retinoids extends over more than half a century and marks some of the successes in pharmacology. Topical treatments are valuable in conditions associated with differentiation defects and the application of Tazarotene was decisive for the curative outcome of APL therapy. However, apart from leukemias, skin cancers and neuroblastoma, among other examples, currently retinoids play a more limited role in cancer treatment and prevention than would be expected based on their favorable pharmacologic properties.

Retinoids and rexinoids are ligands recognized by specific receptors which directly interact with the genome and regulate concerted programs of gene expression. These programs elicit coordinated physiologic changes that can antagonize malignant transformation, promote differentiation, cause apoptosis, insulin sensitivity, oxidative metabolism, and immune modulation, or suppress inflammation. The large number of possible complexes of nuclear receptor dimers and co-regulators formed in response to ligand binding confers tissue-selective modulation of cell function that collectively exert antineoplastic effects.

Although both retinoids and rexinoids possess a spectrum of pleiotropic effects, it is unlikely that retinoid monotherapy will be a preventive strategy for most cancers. However, as multifunctional agents, retinoids are especially well-suited to combination therapies [159]. From a long line of studies it appears that to best exploit the potential of retinoid pharmacology optimal combinations with other agents will have to be identified and the issues of resistance and toxicity addressed systematically in preclinical models and relevant clinical trials. There is a need to identify validated biomarkers in order to predict retinoid response in the target tissue as clinical data demonstrate that retinoid action is often dependent on specific cellular contexts and epigenetic modifications. Such biomarkers may be derived from preclinical studies or from patient material from window-of-opportunity trials. To take advantage of the full potential of retinoids such biomarkers will have to be used to better define clinically-responsive populations.

Useful knowledge can be gained from retrospective studies asking pharmacogenomic questions, by linking specific genetic loci or single nucleotide polymorphims to therapeutic efficacy [133]. Ample preclinical and ex vivo data are available about the downstream transcriptional responses of target organs that should allow the monitoring of the target-specific activity after retinoid treatment. These changes in the expression of retinoid target genes would also be important for bioavailability assessment.

From the prevention aspect of clinical use, rexinoids seem to outweigh retinoid agonists in their activities. While exhibiting modest activity against existing tumors (with the exception of CTCL), rexinoids are far more effective preventive agents than are retinoids in preclinical models. Furthermore, long-term treatment with rexinoids is associated with significantly fewer adverse effects than retinoids. Therefore, rexinoids should be clinically tested for their ability to suppress tumor formation in organ sites where preclinical and biomarker studies have demonstrated efficacy. Future hypothesis-driven clinical trials are needed to take full advantage in the clinic of the promise of retinoid and rexinoid pharmacology.