Retinoic acid, a regeneration-inducing molecule



Retinoic acid (RA) is the biologically active metabolite of vitamin A. It is a low molecular weight, lipophilic molecule that acts on the nucleus to induce gene transcription. In amphibians and mammals, it induces the regeneration of several tissues and organs and these examples are reviewed here. RA induces the “super-regeneration” of organs that can already regenerate such as the urodele amphibian limb by respecifying positional information in the limb. In organs that cannot normally regenerate such as the adult mammalian lung, RA induces the complete regeneration of alveoli that have been destroyed by various noxious treatments. In the mammalian central nervous system (CNS), which is another tissue that cannot regenerate, RA does not induce neurite outgrowth as it does in the embryonic CNS, because one of the retinoic acid receptors, RARβ2, is not up-regulated. When RARβ2 is transfected into the adult spinal cord in vitro, then neurite outgrowth is stimulated. In all these cases, RA is required for the development of the organ, in the first place suggesting that the same gene pathways are likely to be used for both development and regeneration. This suggestion, therefore, might serve as a strategy for identifying potential tissue or organ targets that have the capacity to be stimulated to regenerate. Developmental Dynamics 226:237–244, 2003.© 2003 Wiley-Liss, Inc.


It is our intention in this review to highlight the data which show that, in several systems, retinoic acid (RA) induces regeneration not only of individual tissue types but also of entire organs. This ability is not limited to the animals that are classically known to regenerate, that is the urodele amphibians, but it also extends to mammals. In many branches of medical research, the induction of regeneration is the ultimate goal because “self-repair” circumvents the massive problems of transplant rejection. One can think of many areas for which such regenerative discoveries would be of immense benefit—cardiac muscle regeneration; nerve regeneration in the damaged spinal cord; nerve cell regeneration in the many forms of neurodegenerative disease such as Alzheimer's, Parkinson's, motor neuron disease, or after stroke; islet cell regeneration in diabetes, myelin regeneration in muscular dystrophy. The list is endless. Pharmaceutical companies are continually searching for new small molecules that are biologically active in these medical conditions. Perhaps it may transpire that the biologically active small molecule retinoic acid or a closely related synthetic derivative will be used in the treatment of some of these conditions in the future.


RA is an endogenous molecule in the embryonic and adult vertebrate, which is derived from vitamin A. It is of low molecular weight (300 Da) and is lipophilic. In the adult, vitamin A is obtained from the diet in the form of retinyl esters present in animal meat or β-carotene present in plants. Cells of the embryo or adult that require RA obtain it from the blood system where it circulates as retinol bound to retinol-binding protein. Inside the cell, the sequestered retinol is enzymatically converted first to retinal by the retinol or alcohol dehydrogenases (ADHs) and then to RA by the retinal dehydrogenases (RALDHs; Duester, 2000). RA is further metabolized by a cytochrome P450 enzyme called Cyp26 to supposedly inactive products such as 4-oxo-RA, 4-OH-RA, 18-OH-RA, and 5,18-epoxy-RA (White et al., 1996; Fujii et al., 1997; Abu-Abed et al., 1998) and finally excreted. There are two isomers of RA, all-trans-RA and 9-cis-RA, which act by means of different receptors (see below), but whether they have separate enzymatic pathways from all-trans-retinol and 9-cis-retinol or isomerisation takes place as a last step is not really known.

Once RA has been synthesised in the cell, it enters the nucleus and establishes or changes the pattern of gene activity by binding to ligand activated nuclear transcription factors. There are two classes of these transcription factors, the retinoic acid receptors (RARs) and retinoid X receptors (RXRs). In human, rat, and mouse, there are three RARs, α, β, and γ, each of which has multiple isoforms (Kastner et al., 1994) and three RXRs, α, β, and γ, (Kliewer et al., 1994), again each having several isoforms. In the newt, there is a RARα, no RARβ, and a RARδ, which is equivalent to the mammalian RARγ (Ragsdale et al., 1989, 1992a, b). The RARs and RXRs act as heterodimers (e.g., RARα/RXRβ) and recognize consensus sequences known as retinoic acid response elements (RAREs) in the upstream promoter sequences of RA-responsive genes.

Thus, with these properties, RA can readily enter cells and induce novel gene activity. The following examples demonstrate that this novel gene activity includes the regeneration of tissues and organs.


Before describing the examples of RA-induced regeneration, it is worth emphasising that these results reinforce ideas on the relationship between development and regeneration and indeed, in some cases, have come about as a result of developmental studies. It is a common belief that regeneration did not evolve as a separate phenomenon in evolution involving novel gene pathways, but instead is a leftover from development using the same pathways that the organ or tissue developed with in the first place (Muneoka and Bryant, 1982; Bryant and Gardiner, 1992; Imokawa and Yoshizato, 1997). Of course, this belief does not help us understand why some organs such as the amphibian limb can regenerate perfectly, whereas the human limb cannot, but it does narrow down the possibilities when searching for ways of inducing regeneration. RA is a good example of this phenomenon. It is a crucial molecule required for limb development, neurite outgrowth in development, and lung development, and the components of the RA transduction pathway, the synthesising enzymes and the receptors, are all present during the development of these systems. We would, therefore, expect RA and its components to be involved in the regeneration of these organs or that the failure of the regeneration of these organs is due to the failure of one of these RA components. By correcting this failure, regeneration might be induced, and this correction is precisely what happens in the following examples of RA-induced regeneration.


The most perfect example of organ regeneration in the animal kingdom must surely be limb and tail regeneration in urodele amphibians. Axolotls, newts, and many species of salamanders can regenerate perfect replacements of their limbs and tails after amputation even as an adult. In frogs, this ability is lost after metamorphosis, but as tadpoles, limb and tail regeneration is equally efficient.

Because these animals already have this ability to regenerate, we might not expect RA to do anything, but actually it causes “super-regeneration,” that is, the regeneration of extra limb tissue. This super-regeneration was first described in 1978 by Niazi and Saxena by using Bufo andersonii tadpoles which showed that, when the tadpole hindlimbs were amputated through the shank and treated with retinyl palmitate, instead of regenerating what was removed, the regenerates seemed to contain extra elements in the proximodistal axis and often two regenerates appeared instead of one (Niazi and Saxena, 1978). This phenomenon was subsequently more fully characterised using axolotls and Rana temporaria (Maden, 1982, 1983a,b).

In axolotls (Maden, 1982, 1983a), the effect of retinoids, including RA, is to induce a concentration-dependent increase in the amount of tissue regenerated in the proximodistal axis. In control limbs after amputation through, for example, the mid-radius and ulna, the ends of the radius and ulna, carpal, and digits are regenerated (Fig. 1A). After treating the regenerates with a low concentration of retinoids, an extra complete radius and ulna was regenerated (Fig. 1B), and with a higher dose, an extra elbow joint appeared (Fig. 1C). At a higher dose still, a complete limb was regenerated from the amputation plane (Fig. 1D) and this finding also occurred after amputation through the carpals (Fig. 1E). Several other species of amphibian have also been used such as Notophthalmus, Pleurodeles, Triturus, and Xenopus to explore this phenomenon (Thoms and Socum, 1984; Niazi et al., 1985; Scadding and Maden, 1986; Lheureux et al., 1986).

Figure 1.

Retinoic acid (RA) induces super-regeneration in amphibian limbs and tails. Victoria blue–stained whole-mounts show the cartilage patterns. A: Control axolotl forelimb that had been amputated through the mid-radius and ulna (broken line) and regenerated perfectly. h, humerus; ru, radius and ulna; c, carpals; 1,2,3,4, digits. B: Limb amputated through the mid-radius and ulna (broken line) and treated with a low level of retinoids, which regenerated an extra radius and ulna in tandem with the original. C: Limb amputated through the mid-radius and ulna (broken line) and treated with a medium level of retinoids, which regenerated an extra elbow joint and then forearm. D: Limb amputated through the mid-radius and ulna (broken line) and treated with a high level of retinoids, which regenerated a complete limb from the amputation plane. E: Limb amputated through the carpals (broken line) and treated with a high level of retinoids, which regenerated a complete limb from the amputation plane. F: Control Rana temporaria hindlimb, which was amputated through the foot (broken line) and regenerated perfectly. g, girdle; f, femur; tf, tibia and fibula; mt, metatarsals; 1,2,3,4,5, digits. G:Rana hindlimb amputated through the foot (broken line) and treated with a high level of retinoids, which regenerated a complete limb from the amputation plane. H:Rana hindlimb amputated through the foot (broken line) and treated with a high level of retinoids, which regenerated a pair of limbs, including the pelvic girdle from the amputation plane. I: Control Rana temporaria tail, which was amputated (broken line) and which regenerated perfectly. J:Rana temporaria tail amputated through the mid-point and treated with a high level of retinoids, which regenerated two pairs of hindlimbs, including the pelvic girdles (arrow) from the amputation plane.

This effect of RA has been interpreted as a respecification of positional information along the limb axis in a proximal direction. It involves specific components of the retinoid signalling pathway, namely RARδ2 (Pecorino et al., 1996).

In Rana temporaria (Maden, 1983b), an additional super-regeneration effect became apparent. As well as inducing the regeneration of a complete limb from an amputation through the foot level (Fig. 1F,G), at higher concentrations, retinoids induced the regeneration of pairs of limbs, including the pelvic girdle (Fig. 1H). In these cases, not only had the proximodistal axis been reduplicated but so had the anteroposterior axis and the dorsoventral axis as the pairs of limbs were mirror-imaged to each other.

The most dramatic example of the fact that RA can induce complete limbs to form and not only from preexisting limb cells, but cells with a different organ specificity altogether, is the induction of limbs from tails in certain species of amphibians. This was first shown by Mohanti-Hejmadi et al. (Mohanty-Hejmadi et al., 1992) in Uperodon systoma and subsequently by Maden (Maden, 1993) in Rana temporaria. When the tail of a frog tadpole is amputated anywhere along its length, it regenerates perfectly (Fig. 1I); but when it is treated with RA after amputation, then up to nine hindlimbs are regenerated instead (Fig. 1J). Like the effects on limbs described above, the induction of tails into limbs is a concentration-dependent and time-dependent phenomenon and it occurs just before metamorphosis as the levels of circulating thyroid hormone rise (Maden and Corcoran, 1996).

As discussed in the Introduction section, these regeneration and super- regeneration phenomena ought to be reflections of an endogenous role for RA in limb development and limb regeneration in the first place—and that is indeed the case. The presence of endogenous RA has not been looked for in developing amphibian limbs, but when the axolotl limb bud is treated with disulfiram, a compound that inhibits RA synthesis, limb development is temporarily halted (Maden, 1998). RA has been readily detected in the chick and mouse limb bud by high performance liquid chromatography (HPLC; Thaller and Eichele, 1987, 1990; Scott et al., 1994; Maden et al., 1998) and by a reporter cell system (Wagner et al., 1992; Maden et al., 1998; Sonneveld et al., 1998). When RA synthesis is inhibited (Stratford et al., 1996) or RA signalling by the nuclear receptors is prevented (Helms et al., 1996), the chick limb bud does not develop. Similarly, RA has been detected by HPLC in the regenerating amphibian limb (Scadding and Maden, 1994) and by a reporter system (Brockes, 1992; Viviano et al., 1995), and when RA synthesis is inhibited with disulfiram, regeneration is inhibited (Maden, 1998). Presumably then, RA-induced super-regeneration involves the induction of the same gene pathways that are used during the development of the limb in the first place.


The developing CNS is another embryonic system that contains RA and that requires RA for its normal development. In fact, the spinal cord part of the CNS contains the highest levels of endogenous RA in the embryo (Maden et al., 1998). When RA is removed from the embryo to generate RA-deficient quail embryos (Maden et al., 1996) or rat embryos (White et al., 2000), then the CNS shows several severe abnormalities. There are embryonic patterning defects in that the posterior hindbrain fails to develop (Gale et al., 1999) and the dorsoventral organisation of the spinal cord is abnormal (Wilson and Maden, unpublished data), but most importantly from the point of view of the subsequent regeneration of neurons, there is no neurite outgrowth from the spinal cord into the periphery (Maden et al., 1996).

These in vivo data demonstrating that one of the functions of RA in the developing CNS is to permit neurite outgrowth are strongly supported by a large amount of data from studies on embryonal carcinoma cells, teratocarcinoma cells, stem cells, neuroblastoma cells, and dissociated neurons. In all of these studies, it is consistently the case that RA either induces neuronal differentiation in undifferentiated cells (embryonal carcinoma, teratocarcinoma, stem, and neuroblastoma cells) or increases the number and/or length of neurites in dissociated neurones (review Maden, 2001). This it does by inducing a wide range of gene products, including transcription factors, structural proteins, enzymes, cell surface glycoproteins, extracellular proteins, neurotransmitters, neuropeptide hormones, growth factors, and cell surface receptors.

Therefore, RA should be able to induce the regeneration of neurites when axons are damaged. It certainly can do this in cultures of dissociated embryonic neurons, as mentioned above, and in, for example, organ cultures of embryonic mouse spinal cord (Fig. 2A,B). However, similar cultures of adult mouse spinal cord do not respond by extending neurites in response to RA (Fig. 2C,D). We recently have investigated why this is so.

Figure 2.

Retinoic acid (RA) receptor β2 (RARβ2) induces neurite outgrowth in adult mouse spinal cord in vitro. Neurofilament immunoreactivity of whole-mount cultures. A: Control embryonic day 13.5 mouse spinal cord cultured for 5 days in cellagen organ culture in delipidated serum. Some neurites grow out from the cord into the cellagen. B: Embryonic day 13.5 mouse spinal cord cultured for 5 days in cellagen organ culture with added RA (10−6M). Many more neurites have grown out. C: Adult mouse spinal cord cultured for 5 days in cellagen organ culture in delipidated serum. No neurites grow out. D: Adult mouse spinal cord cultured for 5 days in cellagen organ culture with added RA (10−6M). No neurites grow out. E: Adult mouse spinal cord transfected with a lentiviral vector containing the RARβ2 gene. The region of transfection is shown by fluorescein isothiocyanate–labelled anti-Flag immunofluorescence (green). From this region and not from anywhere else, neurites have grown out, as visualised by Texas Red secondary labelling of neurofilament (NF-200) immunostaining (red).

It is possible that during the transition from embryo to adulthood some or all of the transducers of the RA signal might be shut down permanently by transcriptional inactivation. The RA synthesising enzymes would be candidates, but if so, then the provision of excess RA should overcome the requirement for the enzymes and induce neurite outgrowth, which does not happen (Fig. 2D). The receptors would also be candidates, and if such a phenomenon did occur, then excess RA would still not be able to signal to the nucleus because of the absence of receptors. We have shown in studies on dissociated embryonic neurons that the receptor that transduces the RA signal is RARβ2 (Corcoran et al., 2000). This conclusion was arrived at because this receptor is up-regulated after addition of RA and a RARβ agonist (CD2019) induced extensive neurite outgrowth from these neurons, whereas neither a RARα agonist (CD366) nor a RARγ agonist (CD437) behaved in this manner. The role of RARβ2 was then confirmed by showing that it is up-regulated in embryonic spinal cord explants, which respond to RA by extending neurites, but it is not up-regulated in the adult spinal cord explants after addition of RA.

Therefore, although RA is a neurite-inducing molecule, the adult spinal cord cannot respond, because RARβ2 is transcriptionally inactivated, probably postnatally. If this is the case, then transfecting adult spinal cord neurons with RARβ2 should then induce a response to the addition of RA. We have shown recently that this is indeed the case by using a lentiviral vector system to transfect the RARβ2 gene into cultured adult mouse and rat spinal cord (Corcoran et al., 2002). The viral vector carries an antigenic tag so that the transfected cells can be detected. In Figure 2E, a region of the cultured cord can be seen, which is coloured green after antibody staining for the tag. This culture is also stained for neurites with a neurofilament antibody in red and shows that the area which is transfected is the region where neurites have been induced from the spinal cord. In our experiments, the total number of cord cultures transfected was 34 and the number that extended neurites was 24, giving an induction frequency of 70%. We have performed various controls to ensure that it is the RARβ2 gene that is inducing neurites. These strategies include a control virus expressing lacZ,which has told us that approximately 45% of the cells in the cord cultures become transfected, and a virus containing the RARβ2 gene but with a mutated integrase gene so that transfection takes place but the virus does not become integrated into the host genome and transfection with a different isoform of RARβ, namely RARβ4. None of these controls gave any neurite outgrowth from the cultures.

When the cords are transfected with RARβ2, they extend neurites whether or not RA is added to the medium. There is certainly endogenous RA in the adult spinal cord, and this presence could provide a source of ligand for the RARβ2. Alternatively, it is possible that activation could occur in the absence of ligand by the phantom ligand effect, whereby a RAR is activated by a conformation change in its RXR heterodimeric partner (Schulman et al., 1997). We therefore conclude from these experiments that, once again, RA can induce regeneration, this time in the adult mammalian CNS, when allowed to act on the genome through its normal pathway of receptors.


The most clinically advanced example of organ regeneration in response to retinoids is the remarkable model of RA-induced alveolar regeneration. If we consider that the primary function of the lung is to facilitate efficient gas exchange and that this is dependent on the surface area of gas-exchanging tissue (Sa) composed largely of the pulmonary alveoli, then diseases that are characterised by a reduced Sa, such as bronchopulmonary dysplasia and emphysema, could potentially be treated by alveolar regeneration.

The majority of alveoli are formed during a developmentally regulated, postnatal period in the rat, mouse, and human where immature saccules undergo repeated subdivision or septation. Alveolar septation occurs within the first 2 postnatal weeks in the rat and mouse and probably up to 18 months in the human (for review, see Massaro and Massaro, 1996). After this period, there is limited formation of alveoli so that, if alveolar development is disrupted during septation or if alveoli are destroyed after septation, then there is no spontaneous restoration of Sa. This failure of endogenous repair or regeneration leads to the important clinical conditions of bronchopulmonary dysplasia in infants and pulmonary emphysema in adults, where there is insufficient Sa to maintain adequate gas exchange, leading to dyspnea, hypoxia, and often death. These conditions are a massive global health problem. Emphysema together with chronic bronchitis make up the syndrome of chronic obstructive pulmonary disease (COPD), which is predicted to rise from the 12th to the 5th leading cause of disease burden worldwide by 2020 (Lopez and Murray, 1998). Currently, the outlook for affected patients is bleak, there are no effective treatments for this progressive condition, save supplemental oxygen therapy and lung transplantation (Barnes, 2000).

Retinoid signalling has been studied during alveolar formation in both the rat and the mouse, and these studies have demonstrated a requirement for RA for the initial development of alveoli. In the mouse, HPLC results have shown that alveolar septation is preceded by a peak of RA and accompanied by a fall in retinol, suggesting a utilisation of retinoids (Hind et al., 2002a). In the rat, changes in the levels of endogenous retinoids have been described immediately after birth (Geevarghese and Chytil, 1994; McGowan et al., 1995). Two RA synthesising enzymes are present from postnatal day 1 (P1) to P4 in interesting spatial distributions, which correlate with spatial patterns of alveolar proliferation. RALDH1 is expressed in the alveolar parenchyma during the period when there is a dramatic increase in cell proliferation from P4, which is essentially complete by P14. RALDH2 is expressed around the periphery of the lung in the pleural mesothelial cells, which are associated with a low level and peripheral pattern of proliferation (Hind et al., 2002a). When RA synthesis is inhibited by the daily administration of disulfiram to mice from P2 to P14, then alveolar formation is disrupted, large air spaces develop, and the mean chord length (a measure of the average diameter of alveoli) increases (Fig. 3A–C). Alveolar formation is also associated with significant changes in the levels of the retinoid binding proteins CRBP1 and CRABP1 in both the mouse (Hind et al., 2002b) and the rat (Ong and Chytil, 1976; McGowan et al., 1995). The expression of the RARs has been characterised during alveolar formation in the rat (McGowan et al., 1995), and we have characterised the temporal and spatial expression of the RAR isoforms α1, β2, β4, and γ2 in the mouse (Hind et al., 2002b). Knockouts of these receptors have revealed that there are both inhibitory and stimulatory roles for these receptors. Null mutants for the RARγ gene have fewer, larger alveoli, suggesting that RARγ is required for alveolar formation (McGowan et al., 2000). Conversely, mice null mutant for the RARβ gene have smaller, more numerous alveoli, suggesting that RARβ may function as a negative regulator of alveolar formation (Massaro et al., 2000).

Figure 3.

Retinoic acid (RA) is required for alveologenesis and induces alveolar regeneration in adult mouse lungs. Hematoxylin and eosin staining of inflation fixed, wax sectioned tissue; all photomicrographs are at the same magnification. A: Control adult (postnatal day [P] 24) mouse lung, showing typical arrangement of alveoli. B: P24 lung, which had been treated with disulfiram to inhibit RA synthesis during alveologenesis from P2 to P14. Large air spaces caused by inhibition of alveolar development are apparent. C: Quantitation of the mean chord length (average diameter of alveoli) in control (as in A) and disulfiram-treated (as in B) lungs. D: Control adult (P90) mouse lung, showing typical arrangement of alveoli. E: Adult mouse lung (P90), which had been treated with dexamethasone during alveologenesis from P2 to P14. An irreversible inhibition of alveolar development is apparent with highly enlarged air spaces. F: Dexamethasone-treated (P2–P14) mouse, which had received RA from P30 to P42. By P90, alveoli have regenerated and returned the lung to the control condition (compare with D). G: Data on gas exchanging surface area per 100 g body weight for the control animals (as in D), dexamethasone-treated animals (as in E), and RA-treated dexamethasone animals (as in F), showing that RA fully regenerates the lung capacity.

Therefore, if RA is required for the formation of alveoli, will the administration of RA to lung whose alveoli have been lost induce their regeneration? Remarkably, the answer to this question is yes. Massaro and Massaro (1997) reported a landmark study demonstrating that systemic RA treatment can reverse the pathologic features of experimental emphysema in the adult rat. In this experiment, alveolar loss was induced by the instillation of elastase into the trachea, and lung parameters such as lung volume, distance between alveolar walls, and surface area per lung volume were all returned to normal by 12 days of RA treatment. Other groups have generated similar, though less significant data in the elastase-treated rat (Belloni et al., 2000; Tepper et al., 2000). RA also restores alveoli in other models of alveolar destruction such as the tight skin mouse, a genetic cause of spontaneous emphysema, and in models of failed alveolar development such as glucocorticoid treatment of rats (Massaro and Massaro, 2000).

We have used the dexamethasone-treated mouse as a model of alveolar loss. Treatment of mice from P2 to P14 with dexamethasone (dex) results in a dramatic loss of alveoli (Fig. 3D,E). When these dex-treated mice are given a daily injection of RA from P30 to P42 and then grown up to P90, it is clear that their alveoli now look effectively normal (Fig. 3F). After histologic analysis of sections from these experimental lungs, measuring lung volume, distance between alveolar walls, and correcting for body weight, then it is clear that these RA-treated mice have completely regenerated (Fig. 3G). We suggest, therefore, that RA restimulates the same RA-responsive gene pathways that were used during normal alveologenesis. If RA can restore alveoli in humans, it will have a massive impact on COPD as a major global health problem.


We review here the data which demonstrate that RA is a regeneration-inducing molecule in three systems—the regenerating amphibian limb, where it induces super-regeneration; the mammalian CNS, where RA cannot function because RARβ2 is missing, and when RARβ2 is virally transduced into the spinal cord, neurites are induced; and the mammalian lung, where RA induces alveolar regeneration. In each of these systems, we suggest that RA acts to invoke regeneration because that organ developed originally under the influence of RA and those gene pathways used in development are being re-induced for regeneration. If so, then we can identify other organ systems that might be capable of being regenerated by RA by identifying further organ systems whose development depends on RA. One such example is the kidney (Lelievre-Pergorier et al., 1998; Mendelsohn et al., 1999). Perhaps RA can regenerate nephrons. Another example is the pancreas (Stafford and Prince, 2002), and this possibility might mean that RA could regenerate the islets as a treatment for diabetes. There are also potential examples within the CNS. The RA synthesising enzyme RALDH1 is present in a localised region of the ventral mesencephalon during development, which is the forerunner of the neurons of the substantia nigra. This enzyme is localised to these neurons in the adult (McCaffery and Drager, 1994). Perhaps RA may have a role to play in the treatment of Parkinson's disease. It is clear, therefore, that developmental studies of gene pathways have an important role to play in revealing potential therapies for the regeneration of cells and organs.


We thank our colleague Dr. Po-Lin So. The work described here has been supported by the BBSRC, The Wellcome Trust, and Oxford BioMedica plc.