Synthetic Retinoids Beyond Cancer Therapy

Abstract

While the uses of retinoids for cancer treatment continue to evolve, this review focuses on other therapeutic areas in which retinoids [retinol (vitamin A), all-trans retinoic acid (RA), and synthetic retinoic acid receptor (RAR)α-, β-, and γ-selective agonists] are being used and on promising new research that suggests additional uses for retinoids for the treatment of disorders of the kidneys, skeletal muscles, heart, pancreas, liver, nervous system, skin, and other organs. The most mature area, in terms of US Food and Drug Administration–approved, RAR-selective agonists, is for treatment of various skin diseases. Synthetic retinoid agonists have major advantages over endogenous RAR agonists such as RA. Because they act through a specific RAR, side effects may be minimized, and synthetic retinoids often have better pharmaceutical properties than does RA. Based on our increasing knowledge of the multiple roles of retinoids in development, epigenetic regulation, and tissue repair, other exciting therapeutic areas are emerging.

INTRODUCTION

Since the discovery in 1991 that all-trans retinoic acid (RA), an endogenous metabolite of vitamin A (VA, ROL, retinol), was incredibly effective for the treatment of patients with a particular type of leukemia, acute promyelocytic leukemia (1), RA and various retinoids (defined as vitamin A, its metabolites, and related synthetic derivatives with vitamin A–like effects) have been studied and used for the prevention and treatment of several types of cancer (2–4). While the uses of retinoids for cancer treatment continue to evolve (5–7), here the focus is on other therapeutic areas where retinoids are currently being used and on promising new research that suggests more uses for retinoids in the future. Possible treatments of various disorders and injuries of the kidneys, liver, pancreas, heart, muscle, nervous system, and skin are discussed. Because of this focus on the molecular actions of retinoids and space limitations, the immune system is not covered, though retinoids have effects on immune cells in many of the tissues discussed below; likewise, not all relevant publications can be cited. The focus of this review is on newer areas of research, developing themes, and emerging therapeutic opportunities in terms of mechanisms of action of retinoids in both the prevention and the treatment of disease.

RETINOIDS: A BRIEF OVERVIEW OF METABOLISM AND MOLECULAR MECHANISMS OF ACTION

VA is a micronutrient that is required for life itself, and we obtain all of our required VA from our food (8, 9). Within the body, VA is stored primarily in the ester form, as retinyl palmitate, in stellate cells, fat-storing cells in the liver perisinusoidal space (10). The enzyme lecithin:retinol acyltransferase (LRAT), which converts VA (retinol) to retinyl palmitate by transferring palmitate from the sn-2 position of lecithin (Figure 1), is a critical protein for the uptake of VA from the blood (11). Though the focus here is on the actions of VA that are similar to those of hormones in the body, one form of VA, 11-cis-retinaldehyde, is converted to all-trans retinaldehyde in the retina by photons, which allows us to see (12).

Figure 1.

Uptake, metabolism, and transcriptional actions of vitamin A (i.e., retinol or ROL) and RA. Retinol is taken up by cells via Stra6 or other yet-to-be-identified membrane proteins, and then ROL can be esterified by LRAT to retinyl esters, considered to be a storage form of the vitamin. Retinol can also be metabolized to RAL via many different RDHs and then to RA via ALDH1a2. RA then can be further oxidized by Cyp26a1 to various hydroxylated metabolites, some of which are excreted from the cell, or transported to the nucleus. CRABP2 plays a role in this movement of RA to the nucleus, but CRABP2 is not required for RA to regulate transcription in the nucleus. The RARs (RARα, β, and γ), which are members of the nuclear receptor family, heterodimerize with RXRs and, with RA bound as an agonist, activate the transcription of target genes. Note that the membrane protein Stra6 is expressed in only some tissues. Abbreviations: ALDH1a2 (RALDH2), aldehyde dehydrogenase 1a2; CRABP2, cellular retinoic acid–binding protein 2; CRBP1 (RBP1), cellular retinol-binding protein; CYP26a1, cytochrome P450 enzyme; LRAT, lecithin:retinol acyltransferase; RA, all-trans retinoic acid; RAL, retinaldehyde; RAR, retinoic acid receptor; RARE, retinoic acid response element on the DNA; RDH, retinol dehydrogenase; RE, retinyl ester; RetSDR (Dhrs3), dehydrogenase/reductase SDR member 3; ROL, vitamin A, retinol; RXR, retinoid X receptor.

The absorption and metabolism of VA to RA are complex and differ by cell type (13–15) (Figure 1). The liberation of VA from storage forms and the production of biologically active VA metabolites are key steps in the biological actions of retinoids, and these steps, which occur at the cell or tissue level, are not well understood in terms of their regulation (Figure 1).

Once liberated from storage, VA is metabolized, and RA is the most biologically active form of VA. RA is the primary, biologically active, endogenous agonist for all three retinoic acid receptors (RARs) (RARα, β, and γ), and VA’s actions are most often mediated by these receptors, which act as transcription factors (16). These RARs are bound in a complex with the retinoid X receptors (RXRs) (RXRα, β, and γ) (Figure 1), forming heterodimers with the RXRs on DNA at enhancers/retinoic acid response elements (Figure 1) to activate transcription of their primary response genes in the presence of agonist (17). The RXRs and RARs are members of the steroid/thyroid/vitamin D/peroxisome proliferator-activated receptor (PPAR) family of nuclear receptors, which function as ligand-dependent transcription factors (18). RARβ and γ can also act as transcriptional repressors in the absence of ligand (19). RARα, β, and γ are expressed at various levels in all cell types in the body and regulate different, but sometimes overlapping, target genes in numerous tissues at different times during development (20, 21). RA induces the differentiation of multiple types of stem cells (both pluripotent stems cells and more committed precursors) by activating the transcription of sets of key target genes (22). In addition, there are compelling data that RARα can act in the cytoplasm in a nontranscriptional manner in neurons to regulate synaptic strength and other aspects of learning and memory (23, 24). In neuronal dendrites, RA treatment causes rapid accumulation of RARα in RNA granules, resulting in increased local translation of the AMPA-receptor subunit GluR1 (25, 26). This type of cytoplasmic local translation of selective proteins in neurons has not been clearly shown for RARβ and γ.

RETINOID AGONISTS MAY HAVE ADVANTAGES AS POTENTIAL THERAPEUTICS

Whereas RA binds RARα, β, and γ with high affinity and has a very short biological half-life, synthetic analogs can exhibit a selective affinity for one type of RAR. These synthetic agonists can exhibit pharmacodynamic and pharmacokinetic properties that diverge from those of RA, making them potentially useful drugs (27). In this review, we discuss some of these synthetic agonists of the RARs with a focus on how these synthetic retinoids are currently used or have the potential to be used to treat disease. Of course, we also briefly review the cell type–specific roles of each of the RARs in normal physiology, since the functions of these receptors in specific cell types impact the use of RAR agonists for disease treatments. There are compelling reasons to develop additional receptor-selective synthetic retinoids for the treatment of many diseases, and literature supporting this conclusion is summarized in the following seven sections.

RETINOIDS IN KIDNEY DISEASE

Generally, it has been thought that the adult human kidney contains an average of one million nephrons, but much more variability in this number has recently been discovered; up to a 13-fold variation in the nephron number can occur in normal individuals (28). The nephron number is lower in kidneys of hypertensive patients compared to normotensive, age-matched controls (29, 30). Such variations in nephron number could explain differences in susceptibility to systemic hypertension and chronic kidney disease (CKD) progression (31). During pregnancy, VA must be acquired by the embryo from the maternal circulation via the placenta (32). Low nephron number is often associated with gestational exposure to poor nutrient status in the mother, including VA deficiency (33–37). In rodents, even mild VA deficiency (50% decrease) in the mother during gestation results in a ~20% decrease in nephron number in the pups (34, 38). It is intriguing that a common human ALDH1a2 (aldehyde dehydrogenase 1a2) gene variant [rs7169289(G)] associated with higher RA levels in umbilical cord blood is also associated with increased kidney volume (∼22% increase) in newborns, adjusted for body surface area (39). Collectively, these data suggest that maternal malnutrition, particularly with respect to VA deficiency, can result in a permanent reduction in nephron number in her offspring. This reduction in nephron number could make an individual more prone to CKD as an adult.

Currently, CKD is a health problem of epidemic proportions, affecting approximately 10% of the world’s population. A major factor in the progression of CKD is fibrosis; CKD is generally marked first by more acute tubular injury and subsequently by interstitial fibrosis (40). Kidney proximal tubule injury leads to activation of fibroblasts in the kidney to become α-smooth muscle actin (α-SMA)⁺-activated myofibroblasts. The signals for this injury-associated fibroblast activation are not completely understood; however, researchers have noted an association between myofibroblast distribution and nephron damage after injury. Myofibroblasts surround tubules only after injury in a rat acute kidney injury model (41). Strikingly, these myofibroblasts greatly increase the level of the enzyme ALDH1a2 (Figure 1), the rate-limiting enzyme in the production of RA from VA, shortly after injury to the proximal tubules. Prior to injury, the proximal tubules themselves express ALDH1a2, but the proximal tubule epithelial cells lose this ALDH1a2 expression upon injury (42). These results suggest that additional endogenous RA is needed during the repair of tubule injury and that when tubules are injured, RA is synthesized from VA in the myofibroblasts rather than in the tubules (Figure 2). Thus, RA possibly acts in a paracrine manner (Figure 2), and over the short-term, activation of fibroblasts may be beneficial for repair of an injury to the tubules rather than harmful. However, when the activation of fibroblasts continues over a more extended period of time, such as in CKDs, the fibrosis becomes more extensive and deleterious.

Figure 2.

Model of activation of aldehyde dehydrogenase 1a2 (ALDH1a2) after injury to epithelial cells, and potentially to muscle and neurons, to facilitate repair. Epithelial cells (yellow), after injury, produce less retinoic acid (RA) (red), and activated myofibroblasts transiently produce RA to stimulate repair, cell division, and then differentiation of the injured cells in an autocrine or paracrine process. It is possible that this endogenous repair process could be reproduced by providing exogenous RA or synthetic retinoids. ALDH1a2 is the enzyme that metabolizes retinaldehyde to RA. Ki-67 is a nuclear protein associated with cell proliferation, and its expression is transiently elevated after injury.

All three RARs are expressed in multiple cell types in the kidney. Endogenous RA is required for proper development of the kidney during embryogenesis (43, 44). RA at pharmacological doses reduces fibrosis in a rat model of diabetic nephropathy by downregulating transforming growth factor beta (TGFβ) (45) and by lowering the inflammatory response mediated by Toll-like receptor 4 (TLR4) in this same rodent model (46). Conversely, we have reported that when RARα is selectively deleted in some of the proximal tubule cells in mice, massive fibrosis results. Our data indicate that RARα is necessary for proper proximal tubule function and that the chronic injury to the tubules from the loss of RARα leads to fibrosis (47).

Research in a murine model of human immunodeficiency virus (HIV)-associated renal failure showed that exogenously added RA, the endogenous agonist for all three RARs (Figure 1), improves podocyte function (48–50). Moreover, RA also improves podocyte function in several other glomerular pathologies, including focal segmental glomerulosclerosis and puromycin aminoglycoside–induced nephrosis (51–53). RA at pharmacological doses reduces renal injury in some additional rodent disease models, but RAR-selective agonists were not tested (54–57). RA improved glomerulonephritis via the activation of RARα signaling in podocytes (58). Overexpression of retinol dehydrogenase 9 (DHRS9), an enzyme that converts retinol to retinaldehyde, in podocytes attenuates kidney injury and restores podocyte differentiation in an HIV kidney disease model (59). All of these studies, collectively, suggest that VA has an important role in kidney physiology and that RA, used pharmacologically, has beneficial effects on both proximal tubules and podocytes in the context of CKD.

In most of the research discussed above, RA, a pan-RAR agonist with poor pharmacokinetic properties, was used. Interestingly, various RAR-selective synthetic retinoids were able to reduce renal damage and mesangial cell proliferation in a glomerulonephritis rat model (60). Additionally, RARα synthetic agonists reduced proteinuria and improved kidney injury in HIV-1 transgenic mice, Tg26, a model of human nephropathy associated with HIV (49, 61). In this murine model, differentiation is restored and podocyte cell proliferation is reduced by RARα agonists (48). AM80 (tamibarotene), a RARα and β agonist, ameliorated fibrosis in a unilateral ureteral obstruction model of fibrosis (62). We recently showed that a RARβ2-selective agonist could ameliorate pathology in a mouse model of diabetic nephropathy (63). Few clinical trials of retinoids in kidney disease have been carried out, which suggests an opportunity for testing RARα- or RARβ-selective synthetic retinoids. One small trial with 13-cis RA of patients with glomerulosclerosis was performed (NCT00098020), but results are not yet available.

HEPATIC STEATOSIS, LIVER FIBROSIS, AND CIRRHOSIS

Nonalcoholic fatty liver disease (NAFLD) is a general term that describes liver diseases of varying severity, including hepatic steatosis and steatohepatitis; these progressive conditions can lead to liver fibrosis, cirrhosis, and ultimately, hepatocellular carcinoma. A general feature of NAFLD is excess fat (steatosis) in hepatocytes (64). Shockingly, the number of NAFLD cases in the United States is expected to reach ~100 million by 2030 (64). While NAFLD is often associated with obesity, a large number of individuals with NAFLD in Asia have a normal body mass index. NAFLD is also associated with many other pathological conditions, including type 2 diabetes, myocardial remodeling, kidney disease, and heart failure (65). A related disease, alcohol-associated liver disease (ALD), is the most common type of chronic liver disease, and ALD patients show steatosis, oxidative stress, and inflammation in their livers. Most people who consume over 40 grams of alcohol per day develop ALD (66). In NAFLD, ALD, and nonalcoholic steatohepatitis (NASH), hepatic stellate cells, which are fat-storing cells in the perisinusoidal space, acquire a proliferative, myofibroblast-like phenotype, concomitant with loss of VA (retinol) from the lipid droplets and increased extracellular matrix production (67). Hepatic stellate cells, in the absence of liver injury, contain 80–90% of the VA that is stored in the liver (67), even though these stellate cells only account for approximately 8% of the total number of cells in the liver. Furthermore, about 90% of the VA in the entire body is stored in the liver (68). Thus, when these hepatic stellate cells become activated and lose their VA stores, there could be major consequences not only in the liver but in many other tissues and organs in the body.

It should be noted that while we are focused here on obesity-associated NAFLD and NASH causing hepatic depletion of VA, infections, such as with measles virus, can cause a major depletion of VA in patients (69), as can some drugs. Thus, delineating the mechanisms by which hepatic stellate cells store and release VA has relevance beyond that in NAFLD, ALD, and NASH.

In a definitive publication linking low retinoid signaling with NAFLD and hepatocellular carcinoma, researchers demonstrated that a dominant negative RAR transgene, which inhibited almost all signaling by all three RARs (α, β, and γ) specifically in hepatocytes, resulted in microvesicular steatosis in the transgenic mice at 4 months of age (70). At 12 months of age, these transgenic mice exhibited hepatocellular carcinomas and adenomas of the liver. Because the transgene was not 100% effective in blocking the functions of the RARs, addition of pharmacological doses of RA was able to act via the RARs to reverse the steatosis and greatly limit the hepatocellular carcinoma development in these transgenic mice (70). This research clearly shows that appropriate retinoid signaling in hepatocytes is required to prevent steatosis, fibrosis, and ultimately, hepatocellular carcinoma.

Two different groups, studying both murine and human stellate cell activation in cell culture and in cultured liver slices, respectively (71, 72), demonstrated that early in the time course of hepatic stellate cell activation there is a large but transient increase in RA concentration, concomitant with a transient increase in ALDH1a2 levels (Figure 2). Of course, RA is generated from VA, via retinaldehyde, by ALDH1a2. These data suggest that hepatic stellate cells behave in some respects like the activated myofibroblasts in the kidney, discussed in the prior section, after proximal tubule injury (42) in that the stellate cells may transiently provide endogenous RA to the hepatocytes injured by lipid toxicity associated with obesity. Consistent with this idea is a recent report showing that RA at pharmacological doses reversed both steatosis and fibrosis in a NASH mouse model. The fibrosis reversal by exogenously added RA involves induction of MerTK cleavage in macrophages via the AKT-P38-Adam17 signaling pathway, which then reduces the production of TGFβ and limits stellate cell activation. This suppression of fibrosis by RA in this NASH model apparently does not occur via actions of the RARs and is dependent on MerTK cleavage in macrophages, but notably, the steatosis reversal induced by pharmacological doses of RA is independent of MerTK cleavage (73). These data also fit with the model (Figure 2) that when stellate cells become activated and release VA, endogenous RA is generated in an effort to control the extent of fibrogenesis in the early stages of hepatic injury. However, persistent liver injury leads to the longer-term depletion of VA because most of the stellate cells have differentiated into a new, myofibroblast-like cell type and no longer store lipid and VA.

Can synthetic retinoids prevent or reduce these pathological processes, and which RARs are involved? Some experimental data address these questions. RARβ, in the presence of RA, plays a role in enhancing hepatic fatty acid oxidation, ketogenesis, and energy expenditure in the mice through transcriptional activation of fibroblast growth factor receptor 2 (FGF2) (74). Our laboratory reported that a selective, synthetic RARβ2 agonist, given orally to mice on a high-fat diet and to genetically obese mice (db/db, ob/ob mice), reduced hepatic steatosis (75). In contrast, a synthetic RARα agonist, AM80, did not improve hepatic steatosis in a similar dietary obesity model (76), indicating that the different RARs can play unique roles. We also found that a synthetic RARβ2 agonist greatly limited activation of hepatic stellate cells to myofibroblasts in these mouse models (77).

Fatty acids arrive in the liver from the diet, from lipolysis of triglyceride in adipose tissue, and via the synthesis of fatty acids from glucose and fructose by de novo lipogenesis. This last pathway is primarily responsible for the greater hepatic lipid content in NAFLD patients (78). Importantly, almost all fructose (sucrose is metabolized to fructose and glucose in the intestine) is taken up from the portal blood by the liver, where it is used in de novo lipogenesis. A key enzyme in this lipogenesis pathway is ketohexokinase (KHK), and inhibitors of the enzyme are being investigated to treat NAFLD (79). If fatty acids can be diverted from the liver into adipose tissue or muscle, this can reduce NAFLD, as shown by studies of agonists of PPARγ. However, PPARγ agonists such as pioglitazone increase peripheral fat and result in weight gain (80), limiting their therapeutic usefulness. We have shown that a synthetic RARβ2 agonist can limit the increase in KHK, thus being of potential value in the treatment of NAFLD.

DIABETES AND THE HEALTH OF PANCREATIC β CELLS

Our group and the Blaner group have made complementary and related discoveries about the actions of VA on pancreatic islets. We found that LRAT null and wild-type adult mice, made VA deficient by removing VA from their diets for 10 weeks, become hyperglycemic and exhibit massive islet β cell apoptosis (81). (The LRAT null mice cannot store VA as retinyl esters in their livers, so they become VA deficient more quickly than wild-type mice.) Brun et al. (82) showed, through the use of a dominant negative RAR transgene like the one used above in the liver (70), that RA and RAR signaling in the adult pancreas are required for maintaining β cell function. These results collectively indicate that VA is required for the β cells in the islets of adult mice to function properly.

Are these findings in mice indicating that VA is required for proper pancreatic β cell function relevant to humans? Could retinoid signaling be aberrant or lacking in the human population? While most people in countries like the United States obtain enough VA in their diets, a large proportion of the population in the United States suffers from overnutrition. Our research and that of others have shown that obesity itself can lower VA levels in many organs, including the liver, pancreas, lungs, and kidneys, even when the mice are kept on a high-fat diet with adequate VA (83, 84). This organ VA deficiency is not reflected in the serum VA level and thus would not be detected by measuring VA in the sera of obese individuals. Thus, it is possible that obese individuals develop type 2 diabetes and lose pancreatic β cell mass at least in part because of VA deficiency in the pancreas from overnutrition and obesity. Finally, RARβ2 agonists can normalize blood glucose and reduce fasting insulin levels in obese mice with hyperglycemia, suggesting that, in addition to its effects on pancreatic β cells, RARβ acts, either directly or indirectly, to decrease peripheral insulin resistance (75). Thus, this type of agonist may be a useful treatment to prevent the development or slow the progression of type 2 diabetes (75).

RETINOIDS AND HEART DISEASE

There is much evidence that endogenous retinoids have actions in the heart, though no synthetic retinoids are approved by the US Food and Drug Administration (FDA) for use to treat any type of heart disease. The RARs and RA binding proteins are expressed in the heart during development and in the adult (85–90). In addition, from large-scale single-cell and single-nucleus transcriptome analysis of adult human hearts, Litviňuková et al. (91) found that transcripts of ALDH1a2 (RALDH2), the enzyme that carries out the rate-limiting step in RA synthesis from retinaldehyde, are present at very high levels in atrial cardiomyocytes, strongly suggesting that local production of RA occurs in these cardiomyocytes. Interestingly, a single-nucleotide polymorphism in ALDH1a2 (Figure 1), rs261316 allele (T), which may decrease its function, is highly associated with increased odds of having uncontrolled blood pressure on combination therapy with a thiazide diuretic and a beta-blocker (92). ALDH1a2 is also expressed in cardiac fibroblasts and is a gene target of the transcription factors Tbx20 and Gata4 in these cells. Furthermore, after injury, endogenous RA signaling is active, for example, in the postischemic heart (50, 93). Notably, the levels of RAR target gene transcripts transiently increase in the peri-infarct area in cardiac fibroblasts (93).

RA given exogenously also limits fibrosis and cardiomyocyte apoptosis in obese mice (94). VA (retinol) likely plays a role in remodeling of the heart in part by inhibiting the renin-angiotensin system (95, 96), but other mechanisms are also possible. Exogenous RA displays protective effects against cardiac arrhythmias (97). In adult rats, VA deficiency causes left ventricular dilatation that leads to a major decrease in cardiac function and remodeling (98, 99). Supplementation with RA at pharmacological doses prevents left ventricular dilatation and preserves ventricular function post myocardial infarction in rats (100). In zebrafish, which can regenerate their hearts, overexpression of Cyp26a1, which metabolizes RA (see Figure 1), or expression of a dominant-negative RAR causes both reduced RA signaling and a decrease in cardiomyocyte proliferation during the regeneration process (101). Additionally, within 3 h after ventricular injury, ALDH1a2 (RALDH2) is activated in the endocardium, and at one day after injury, ALDH1a2 is localized to the endocardial cells at the infarct site where regenerative cardiomyocyte proliferation takes place (101, 102). These data show that increased RA at the injury site is required for cardiogenesis and that the endocardium is one source of RA after tissue injury in zebrafish. The RARβ2-selective agonist AC261066 may be a valuable addition to the drugs used in the prevention and treatment of myocardial infarction and cardiac remodeling, based on data in mouse models (103).

MUSCLE REPAIR AND BLOCKAGE OF HETEROTOPIC OSSIFICATION BY SYNTHETIC RETINOIC ACID RECEPTOR γ AGONISTS

Abnormal, ectopic bone formation can take place in soft tissues, such as muscle or connective tissue, after various injuries from trauma or surgery. Essentially, muscle and connective tissue such as tendons become replaced by bone, with the bone forming outside the skeleton. Approximately 65% of severely injured soldiers develop this pathology, which can limit motion, cause pain, and result in problems with mobility (104, 105). Additionally, an activating mutation in the bone morphogenetic protein I (BMPI) receptor ACVR1/ALK2^(R206H) causes a congenital condition called fibrodysplasia ossificans progressiva, which is characterized by massive heterotopic ossification and is often fatal. Since attenuation of retinoid signaling is necessary for chondrogenic differentiation (106), RARγ agonists, and in particular the RARγ-selective agonist palovarotene (107), were tested and shown to be effective in reducing chondrogenesis, thereby limiting heterotopic ossification in several mouse models of limb injury (108, 109). These results have led to several clinical trials with an RARγ-selective agonist that are underway for treatment of fibrodysplasia ossificans progressiva patients (e.g., NCT02279095, NCT03312634, NCT02190747).

Synthetic selective RARγ agonists, including palovarotene (R667), CD1530, and CD437, also greatly improved the repair of injured skeletal muscle in a murine model (110). Injuries in the tibialis anterior muscles of adult mice were repaired by regenerating muscle cells, fibrous scar tissue, and adipose tissue. Strikingly, when injured mice were then treated with RARγ agonists [versus corn oil (vehicle)], fibrous or adipose tissue covered a smaller area in the RARγ agonist–treated mice, indicating better muscle repair. Moreover, RARγ-deficient mice showed a great reduction in muscle repair. Similar to the elevation in ALDH1a2 seen in the kidney myofibroblasts after proximal tubule injury discussed above (42), there is a very large increase in retinoid signaling immediately after this skeletal muscle injury. ALDH1a2, CRABP1, and the RAR transcripts dramatically increase in skeletal muscle between 8 h and 4 days after injury, and then these transcript levels decline back to their preinjury levels (110).

What is the function of retinoid signaling in the skeletal muscle? Though the answer is not totally clear, Shirasawa et al. (111) provided some insight by focusing on the stem cells in skeletal muscle. Skeletal muscle contains two types of stem cells: satellite cells that are myoblast-specific stem cells, and fibroadipogenic progenitors (FAPs) that are capable of differentiating into osteoblast, chondrocyte, fibroblast, and adipocyte cells. These FAPs play roles in heterotopic ossification, fibrosis, and fatty infiltration (112). After a rotator cuff tear in the subscapularis muscle in a mouse model, daily oral administration of the synthetic RARγ agonist CD1530 significantly decreased fatty infiltration in the muscle, in part by decreasing the levels of PPARγ and CCAAT/enhancer-binding protein α (C/EBPα) transcripts (111). Administration of adapalene, a selective RARγ and RARβ agonist, after the rotator cuff tear injury gave a similar result: less fatty infiltration in the muscle. Thus, these selective, synthetic retinoid agonists may improve muscle repair by influencing the survival and differentiation pathways of two types of stem cells, satellite cells (113) and FAPs (111). Notably, Akt/mechanistic target of rapamycin (mTOR) signaling, the transcription factors sterol regulatory element binding protein 1 (SREBP1) and PPARγ, and C/EBPα and fatty acid synthase (FASN) proteins were increased in a similar rotator cuff tear injury model in rats. Another drug, rapamycin, an inhibitor of mTOR, also decreased fatty infiltration at the injury site (114). The potential link between mTOR signaling and retinoid signaling would be worth investigating in this model. The use of RARγ-selective agonists could improve muscle repair in athletes, soldiers, and older adults, especially since people can become more prone to muscle injury as they age.

LEARNING AND MEMORY, NEURONAL INJURY, AND SYNTHETIC RETINOID AGONISTS

A number of years ago, researchers demonstrated that the three RARs are expressed in the adult central nervous system (115). RARβ is expressed in discrete regions of the brain and is essential for normal learning and memory in mice (116). As rodents age, reduced endogenous retinoid signaling appears to disrupt hippocampal circuits necessary for short-term/working memory and long-term declarative memory (117), and RA at pharmacological doses can alleviate some of the aging-related memory deficits in mice (118). Recent work has begun to unravel the mechanism by which RA signaling influences synaptic function and related behavior (24). Hummel et al. (119) reported that adult male mice subjected to experimental traumatic brain injury and then given RA for the first 3 days after injury exhibited reduced brain lesion size, less reactive astrogliosis, and reduced axonal injury at day 7 postinjury relative to mice not given RA. Furthermore, in a model of spinal cord injury, a RARβ-selective agonist, CD2019, promoted axonal outgrowth and enhanced functional recovery (120). This functional recovery required NG2⁺ cells near the neuronal cells to produce RA by activating ALDH1a2 in response to the RARβ agonist (121). Synthetic RARβ agonists also showed efficacy in preclinical neuropathic pain models (27, 122). From the publications summarized here, it is clear that synthetic RARβ agonists should continue to be tested for efficacy in models of various types of neuronal injuries and in models of neurodegenerative disorders (123).

SYNTHETIC RETINOIDS IN DERMATOLOGY

RA (all-trans RA, tretinoin) and 13-cis RA (isotretinoin) are considered to be first-generation retinoids for acne treatment and treatment of other skin disorders such as psoriasis, etretinate and acitretin are second-generation retinoids, and tazarotene and adapalene are third-generation retinoids (124) (Figure 3). These retinoids regulate epidermal cell differentiation, reduce inflammatory processes, and may act as antioxidants in the skin (124, 125). Adapalene is a selective agonist for RARβ and γ, as is tazarotenic acid, the active form of tazarotene (AGN190168) (126). A topical formulation of adapalene was approved by the FDA in 2016 as the first over-the-counter, nonprescription retinoid for treatment of acne.

Figure 3.

Structures of retinoids for treatment of various skin diseases.

For mild acne, topical treatments such as tretinoin or adapalene are recommended. For severe acne, oral isotretinoin is quite useful. The risk of teratogenicity from use of topical retinoids has been suggested (127), and although available data are limited, these retinoids should not be used during pregnancy (128). A number of new RARγ-selective, fourth-generation synthetic retinoids are being developed for the treatment of acne and other skin diseases (129). Trifarotene is a potent, RARγ-selective drug that is relatively stable in the skin but is rapidly metabolized in liver microsomes, suggesting an excellent safety profile with respect to potential liver toxicity. Trifarotene exhibits strong anti-inflammatory properties and comedolytic (inhibition of acne lesion formation) activity (130) (Figure 3). The FDA approved trifarotene cream for severe acne treatment in 2019, the first new retinoid to receive FDA approval in over 20 years. Since no clinical trials on acne patients comparing trifarotene cream with tretinoin lotion have been performed to date, more evidence is required to determine whether trifarotene is more efficacious than tretinoin. Among the many transcripts altered by trifarotene, reduced levels of fibroblast growth factor receptor 2 (FGFR2) may be of special interest (130) because some gain-of-function mutations in the FGFR2 gene cause Apert syndrome, a syndrome associated with severe acne (131). These data suggest that a reduction in FGFR2 mRNA could be a key mechanism by which trifarotene reduces acne in patients (130), especially since the FGF/FGFR pathway also plays a role in regulating differentiation in the sebaceous gland.

While it is known that these retinoids for acne treatment act via binding to the RARs, how retinoids act at a molecular or cellular level to treat acne successfully is not well understood. Recently, Oulès et al. (132) showed that the transcription factor GATA binding protein 6 (GATA6) in the upper pilosebaceous unit (which includes the hair shaft, hair follicle, and sebaceous gland) of skin regulates keratinocyte proliferation, limits epidermal differentiation in the unit, and limits lipid production during sebaceous differentiation. Moreover, GATA6 levels are lower in human acne skin, and overexpression of GATA6 phenocopies the effects of RA on sebocytes. That RA increases GATA6 levels and reduces lipid accumulation while 5 alpha-dihydro-testosterone reduces GATA6 levels and increases lipids are striking results, likely indicating that RA is an effective treatment for acne via its ability to increase GATA6 levels in the pilosebaceous unit.

Synthetic retinoids are used in the treatment of psoriasis, an inflammatory disorder that involves the immune system and is present in approximately 2% of the world’s population. There are several types of psoriasis, including plaque, pustular, guttate, and erythrodermic. Rapid proliferation and abnormal differentiation of keratinocytes occurs in psoriasis. Recently, treatment of psoriasis has been transformed by the use of drugs that influence immune functions, including tumor necrosis factor α (TNF-α) inhibitors, IL-12/IL-23 inhibitors, and IL-17 inhibitors (133). However, retinoids are still occasionally used in the treatment of psoriasis, even though the mechanism(s) of action of synthetic retinoids, such as tazarotene or acitretin, in this disease is unclear (134). Many transcripts regulated by retinoids in keratinocytes are involved in transcriptional regulation (135), suggesting that retinoids act by modulating the aberrant differentiation in psoriasis, but tazarotene is also reported to reduce inflammation in this disease (134). Although retinoids as monotherapy for psoriasis are not first-line treatments, acitretin in combination with phototherapy can lower the frequency and dosage of UVB (280–320 nm) radiation needed (136) and also allow for a lower dose of acitretin. Acitretin also enhances the efficacy of psoralen + UVA (320–400 nm) radiation (PUVA) treatment for psoriasis (137). Long-term PUVA use is clearly associated with an increased risk of developing squamous cell carcinoma of the skin (138). Since acitretin has suppressive effects on squamous cell carcinoma formation and it allows for reduced cumulative PUVA exposure, these modalities are frequently used in combination (139).

In 2000, the RXR agonist bexarotene (LGD-1069; Targretin) was approved for treatment of cutaneous T cell lymphomas. Cutaneous lymphomas are defined as extranodal non-Hodgkin’s lymphomas with skin infiltration. Mycosis fungoides is the most common type of cutaneous T cell lymphoma. In a trial of orally administered bexarotene, patients exhibited positive clinical responses, but adverse events included hypothyroidism, hypercholesterolemia, and hypertriglyceridemia in the majority of individuals (140, 141).

In the area of cosmeceuticals (a cosmetic claimed to have medicinal properties), the use of retinoids for the treatment of age-related skin wrinkling and hyperpigmentation is expanding. Topical RA was the first retinoid used to improve photoaged skin (142). Tazarotene has also been shown to have about the same degree of efficacy as RA (143). Others, such as adapalene, have also been shown to reduce facial wrinkling of photoaged skin and improve skin hydration (144). Alitretinoin (9-cis-RA) is approved by the FDA as a topical drug for cutaneous Kaposi’s sarcoma, and alitretinoin, which binds all RARs, also shows antiproliferative effects and some positive effects on photoaged skin (145). For treatment of photoaged skin, VA (retinol) itself is comparable to RA, but retinol is less potent than RA. Retinol and RA, applied topically, increase the thickness of the epidermis. Many of the beneficial effects of retinol and RA are also effected through their actions on the dermis, where they greatly increase collagen synthesis (146, 147). Currently, a cream (Tri-Luma) that contains three active drugs, fluocinolone acetonide (a corticosteroid), hydroquinone (a melanin-synthesis inhibitor), and tretinoin (RA), is available by prescription and is approved by the FDA to treat hyperpigmentation of the skin. While hyperpigmentation treatments are generally just given until a reduction in pigmentation occurs, the use of retinoids for skin wrinkling associated with aging requires chronic treatment. Notably, there is little information in the literature about the long-term effects of retinoids on the stem cell population in the skin epidermis and on the cells of the dermis.

CHALLENGES OF DEVELOPING SYNTHETIC RETINOIDS INTO DRUGS

The opportunities summarized above provide only a glimpse of the therapeutic potential of synthetic retinoids, but there are important challenges that must be recognized. To use RAR-selective agonists more effectively in the future, some of the same problems that affect the use of RA as a drug must be overcome. These include, first, the poor solubility of retinoids in aqueous solutions and the fact that RA increases its own catabolism when administered intravenously, greatly limiting its therapeutic efficacy (148). While not all of the synthetic RAR-selective agonists will display the same pharmacokinetic profiles as RA, poor aqueous solubility is a common feature of retinoids.

Second, RA is also photosensitive, which makes its topical use for skin disorders difficult without careful formulation. Recent developments in terms of delivery of RA and RAR-selective retinoids include delivery in nanoparticles (149–151) and delivery in liposomes (152, 153). Interestingly, the properties of VA itself have been used to enhance specific delivery of cationic liposomes containing short hairpin RNAs to stellate cells in the liver (154). An excellent review on the formulation and delivery of retinoids was recently published (155).

Third, teratogenicity can pose a major problem. While many of the synthetic, RAR-selective agonists studied in rodent models have not been tested for teratogenicity and are not FDA approved, RA is teratogenic when used as a drug, and it is likely that all members of the retinoid signaling family have teratogenic properties if given systemically to pregnant women. This teratogenicity will certainly hinder or limit the clinical testing and the therapeutic uses of the synthetic, RAR-selective agonists for the treatment of the disorders discussed above, even though many of the disorders reviewed here primarily affect older individuals. Since RA is an endogenous agonist for all three RARs, RA itself is not an ideal drug, as pharmacological doses of RA are associated with side effects that are possibly related to RA binding all three different RARs rather than a specific RAR. Given the broad distribution of RARs in cells throughout the body, side effects must be monitored carefully. Nevertheless, retinoid agonists offer a fertile field both for scientific research and for the development of therapeutics.

CONCLUSIONS

To generalize from the findings reported in this review, we summarize the following seven key points. First, synthetic retinoids have great potential for the treatment of many diseases (Figure 4). Because they can be designed to act through a specific RAR, synthetic retinoids may have fewer side effects. Likewise, they may have improved pharmacokinetic and pharmacodynamic properties. Second, challenges to the development of synthetic retinoids include the fact that many types of cells respond to retinoids, making side effects a significant problem. Third, in terms of disease treatments, more positive results are likely if receptor-selective retinoids are employed, given both the pharmacokinetic characteristics and the side effects associated with RA. Fourth, the most mature area of retinoid use, in terms of FDA-approved, RAR-selective agonists, is for the treatment of skin disorders. Fifth, localized or cell type–specific delivery of receptor-selective agonists would be quite beneficial in terms of limiting systemic side effects. Sixth, ALDH1a2, the enzyme that converts retinaldehyde to RA, is often transiently activated in surrounding cells when proximal tubule cells in the kidney, hepatocytes in the liver, and cardiomyocytes (in zebrafish) are injured. These data indicate that increased RA is temporarily needed to promote cell proliferation and repair after injury and that the RA comes from cells surrounding the injured cells (Figure 2). The molecular signals for the transcriptional activation of ALDH1a2 in these activated myofibroblasts in the kidney and liver following injury are not well delineated, and this should be a focus of future research. Such research could lead to methods to produce active retinoids in a cell type–specific manner. Finally, exactly how RA promotes repair and proliferation after injury is not understood, but given the actions of RA in promoting stem cell differentiation by suppressing canonical Wnt signaling (7, 22, 156, 157) and mediating major epigenetic changes (158–160), it is likely that such mechanisms are involved. Therefore, a better understanding of VA actions in the repair and regeneration of cells may also guide further therapeutic development.

Figure 4.

Summary of some of the major results of retinoid therapies in various cell types and organs in experimental animal models and/or in humans. (Note that the anti-inflammatory effects of retinoids on the immune system and the antiproliferative effects on cancer cells are not covered in this review; the reader is referred to references within this review in which these topics are covered). Abbreviations: RA, all-trans retinoic acid, TGFβ, transforming growth factor beta.