Photodecomposition of Vitamin A and Photobiological Implications for the Skin


  • Current address: Department of Biotechnology, Hung Kuang University, Sha-lu, Taichung, Taiwan, ROC.

*Corresponding author email: (Peter P. Fu)


Vitamin A (retinol), an essential human nutrient, plays an important role in cellular differentiation, regulation of epidermal cell growth and normal cell maintenance. In addition to these physiological roles, vitamin A has a rich photochemistry. Photoisomerization of vitamin A, involved in signal transduction for vision, has been extensively investigated. The biological effects of light-induced degradation of vitamin A and formation of reactive species are less understood and may be important for light-exposed tissues, such as the skin. Photochemical studies have demonstrated that excitation of retinol or its esters with UV light generates a number of reactive species including singlet oxygen and superoxide radical anion. These reactive oxygen species have been shown to damage a number of cellular targets, including lipids and DNA. Consistent with the potential for damaging DNA, retinyl palmitate has been shown to be photomutagenic in an in vitro test system. The results of mechanistic studies were consistent with mutagenesis through oxidative damage. Vitamin A in the skin resides in a complex environment that in many ways is very different from the chemical environment in solution and in in vitro test systems. Relevant clinical studies or studies in animal models are therefore needed to establish whether the pro-oxidant activity of photoexcited vitamin A is observed in vivo, and to assess the related risks.


Vitamin A (all-trans-retinol; retinol), its metabolites, analogues and derivatives, are referred to as retinoids (Fig. 1). Vitamin A is an essential human nutrient that plays an important role in critical physiological functions including vision, reproduction and developmental morphogenesis (1,2). In the skin, vitamin A is known to influence a wide range of biological processes, including sebaceous gland activity, cell-mediated immune responses and epidermal cell growth, differentiation and maintenance (3). Retinyl esters, including retinyl linoleate, retinyl myristate, retinyl oleate, retinyl palmitate and retinyl stearate, have been reported to account for more than 70% of the endogenous vitamin A found in human skin (4–7). The storage of retinol in human skin depends entirely on the capacity of cells within the skin layers to transform retinol, derived from dietary sources, to retinyl esters. Skin cells contain lecithin:retinol acyltransferase and acyl CoA:acyltransferase, which catalyze retinyl ester synthesis (6). The esterification activity of the acyltransferases in human skin in vivo is four-fold greater in keratinocytes in the basal layer of the epidermis than in keratinocytes in the upper layers, suggesting that retinoid levels are higher in the lower epidermis (8). A current hypothesis is that during epithelial differentiation, the stored retinyl esters may provide immature keratinocytes in the lower cell layers with a source of retinol as they differentiate and migrate to the upper cell layers of the epidermis. The predominance of retinyl esters in skin and their connection to physiological processes in the epidermis reflect their importance in normal skin function.

Figure 1.

 Structures of selected retinoids found in skin.

Vitamin A in the skin is predominantly derived from the diet. Increasingly, however, skin care products containing various retinoids are becoming an important source of cutaneous vitamin A for some populations. Many of these products are marketed to reduce the appearance of skin aging and photo-aging. As they are more chemically and thermally stable than retinol, retinyl esters have frequently been the preferred retinoid used in skin care products (9). Retinyl palmitate is the most commonly used retinyl ester in cosmetics and has been reported to occur in 702 products in 2004 (10). An increased interest in the cutaneous photochemistry and photobiology of retinyl esters, such as retinyl palmitate, is justified both because of the physiological importance of retinyl esters and their widespread use in skin care products applied to sun-exposed skin.

In sun-exposed skin, vitamin A, whether derived from the diet or topical application, is irradiated with light having wavelengths expected to drive their light-induced alteration. As the maximum absorbance for retinol and its esters is around 325 nm, both UVA and UVB radiation are efficiently absorbed. While both UVA and UVB radiation in incident sunlight penetrate into the skin, the effects of UVA on photochemical alteration of vitamin A may be quantitatively more important because of the relative abundance of UVA radiation in sunlight compared with UVB radiation and the greater penetration of UVA radiation into the skin. For example, about one-third of UVA radiation having a wavelength of 400 nm penetrates into the epidermis to a depth of 0.1 mm in vivo, whereas only about 1% of UVB radiation having a wavelength of 300 nm penetrates into the epidermis beyond a depth of 0.03 mm (11). Insights into the effects of UV exposure on cutaneous vitamin A are provided by studies of retinol and retinyl esters irradiated in various solvents and by studies in animal models. It has been reported that retinyl esters, dissolved in a range of solvents, readily photodegrade and are more photolabile than retinol (12,13). Photodegradation of cutaneous retinol and retinyl esters has been observed following a single exposure of excised human skin to sunlight, rabbits to narrow band radiation in the UVA and UVB spectral ranges, and mice to UVA or UVB light (14–17). These studies suggest that in skin, as in solution, retinyl esters are more photolabile than retinol. The biological consequences of photodegradation of cutaneous retinol and retinyl esters are not known. Potential effects could include light-induced vitamin A deficiency in the skin or toxicity resulting from photoproducts of vitamin A. This review describes studies that provide insights into the biological effects of photodegradation of cutaneous retinol and retinyl esters. In addition, the described studies help to frame questions that may be addressed in future investigations on the photodegradation of cutaneous vitamin A.

Photophysics And Transient Intermediates Following Photoexcitation Of Retinol And Its Esters

Because of the importance of retinal as the chromophore in the visual pigment rhodopsin, the photophysics of retinal and related retinoids has been extensively investigated. These investigations have been addressed in several reviews (e.g. 18–20). Only those aspects of the photophysics of retinol and its esters that are particularly relevant for understanding phototransformations of retinoids in the skin will be presented here.

Retinol and its esters have an absorption maximum in the UV spectral region at approximately 325 nm (Fig. 2). As a result of this broad absorption centered around 325 nm, these retinoids can be efficiently photoexcited by sunlight in both the UVA and UVB spectral regions. Absorption of a UVA or UVB photon is accompanied by a π→π* transition to produce a singlet excited state that has been the object of many experimental and theoretical investigations. Several “anomalous” experimental observations have been reported for photoexcited retinol and retinyl esters (18,21), the most noteworthy being their unexpectedly long fluorescence lifetimes. The intrinsic fluorescence lifetimes of several retinoids, including retinol and retinyl acetate, have been reported to be more than 100-times longer than predicted (21–23). The explanation for these long fluorescence lifetimes is the involvement of two distinct singlet excited states (18–20). After absorption of a single photon, the predicted allowed π→π* transition for substituted polyenes, such as retinol, is a transition from the ground state to a singlet excited state having a point group symmetry of 1Bu+. However, both experimental measurements and theoretical analyses suggest the additional involvement of a singlet excited state having a point group symmetry of 1Ag and having an energy similar to, or slightly lower than, the 1Bu+ excited state (18–20). While direct population of the 1Ag excited state is not expected because of the selection rules for one photon excitations, there is experimental evidence for the involvement of this electronic state after single photon absorption. This evidence includes observation of a weak absorbance attributable to excitation of ground state retinoids to the 1Ag state (24) and population of the 1Ag state by two photon absorption (25). The photoprocess resulting in fluorescence has been described as photoexcitation of the retinoid in its ground state to give the 1Bu+ excited state, followed by relaxation of the 1Bu+ excited state to the 1Ag excited state. Fluorescence results from the radiative transition of the 1Ag excited state to the ground state. It is fluorescence emission from the 1Ag excited state that results in the observed lengthening of the fluorescence lifetimes for retinol and its esters. In addition, this longer lived excited singlet state increases the probability of photochemical transformations involving the excited singlet state.

Figure 2.

 UV-visible absorption spectrum of 1.7 × 10−5 M retinol in ethanol.

Characterization of the electron distribution and bonding in the excited singlet state of retinol and retinyl esters has been the subject of intense investigation. This interest is because of the central role of the excited singlet state in the photochemistry of these retinoids. Both theoretical and experimental evidence suggest that the electronic configuration of the excited singlet state is significantly different from that of the ground state. Theoretical studies on simple polyenes, such as butadiene and hexatriene, have established that their lowest excited single states are highly polar and can be described as having a zwitterionic charge distribution (26). Similarly, theoretical investigations on retinoids predict that photoexcitation leads to a highly polar excited singlet state (18). These theoretical predictions have been confirmed experimentally. Mathies and Stryer (27) have studied the effects of an intense electric field on the absorption spectra of several retinyl polyenes. These studies demonstrated that a large increase in dipole moment accompanies photoexcitation and indicated that the excited state directly resulting from photon absorption (1Bu+) is polar relative to the ground state. An additional experimental demonstration of the polar nature of, as well as the photochemical importance of, the excited singlet state is the ionic photodissociation of retinol and its esters, first observed by Rosenfeld et al. (28). In laser flash photolysis studies of retinol and retinyl acetate in polar solvents, these investigators found that photoexcitation resulted in the appearance of a transient species absorbing at 590 nm and a concomitant increase in the conductivity of the solution. The transient species was identical to the spectrum of the retinyl cation (Fig. 3), which had been previously observed by Blatz and Pippert (29) following addition of retinyl acetate to strongly acidic solutions. Subsequent triplet energy transfer studies have conclusively established that the excited singlet state, rather than the triplet state, is the progenitor of the retinyl cation (30). The fate of the retinyl cation is largely determined by the solvent. In nonpolar solvents, recombination is favored, yielding the original or an isomerized retinol. In polar solvents, while isomerization is also observed, the loss of a proton can result in the formation of anhydroretinol or interaction with the solvent can result in solvolysis products (30) (Fig. 3). The ionic nature of the excited singlet state results in redistribution of electron density over the retinyl polyene and alteration of the bond order to facilitate isomerization. Furthermore, because of the zwitterionic character of the excited singlet state, elimination of poor leaving groups, such as hydroxide and carboxylate anions, is facilitated.

Figure 3.

 Ionic photodissociation of retinyl palmitate in a polar solvent (methanol). Absorption of a photon (hv) results in the formation of a highly polar singlet excited state which can dissociate to form a retinyl cation. The retinyl cation can recombine to form the original or isomerized retinyl polyene. Alternatively, the cation can lose a proton to yield anhydroretinol, or react with the solvent to yield solvolysis products.

In addition to the excited singlet state, the importance of the triplet state in the photophysics and photochemistry of retinoids has been extensively investigated. Three distinct mechanisms have been described for generating the lowest triplet excited state of retinol and retinyl esters. The first mechanism is photoexcitation of the retinoid to produce the singlet excited state, which undergoes intersystem crossing to yield the triplet excited state. For retinoids such as retinal, intersystem crossing from the excited singlet state is relatively efficient. Reported quantum yields for intersystem crossing for all-trans-retinal range from 0.3 to 0.7 (18,20,31) and depend strongly on the solvent. Compared with retinal, intersystem crossing for retinol and retinyl esters is approximately one order of magnitude less efficient. Generation of the triplet state by direct excitation of retinol was first detected by Rosenfeld et al. (27) and was characterized by a transient absorption at 405 nm. For retinol dissolved in N2-saturated hexane, the estimated quantum yield to form a triplet state was 0.06 (27). Subsequently, quantum yields for the formation of the triplet state by direct excitation of retinol or retinyl acetate have been reported by other investigators and usually range from 0.02 to 0.06 in nonpolar solvents (18,32–34). In polar solvents, an approximately 10-fold reduction in the quantum yield for forming the triplet state via direct excitation of retinol or retinyl esters is observed (18,32). The dramatically higher quantum yield observed for intersystem crossing for retinal, compared with retinol, has been attributed to spin-orbit coupling and electronic interactions involving the n→π* transition in retinal (18). The second mechanism identified for population of the triplet state of retinol and its esters is O2 assisted intersystem crossing (32,33). This phenomenon, also termed O2 catalyzed intersystem crossing, has been additionally described for other chromophores (35). As previously mentioned, because of the involvement of the 1Ag excited state, retinol has a relatively long excited singlet state lifetime. This long excited singlet state lifetime makes retinol more susceptible to interactions with O2 and facilitates O2 assisted intersystem crossing. Rosenfeld et al. (33) have reported that the triplet state quantum yield for retinol or retinyl acetate is 10-fold higher in air saturated hexane solutions compared to deaerated hexane. For O2 saturated hexane solutions of retinol or retinyl acetate, the quantum yield for the triplet state of retinol or retinyl acetate increases approximately 25 times compared with that observed in deaerated solutions (33). Oxygen assisted intersystem crossing has also been observed in polar solvents such as methanol (33). Quantum yields for O2 assisted intersystem crossing have been reported to be 0.39 ± 0.06 in benzene and 0.47 ± 0.08 in methanol (32). In addition to formation of the triplet state of the retinoid, O2 assisted intersystem crossing has been shown to produce both singlet oxygen (1O2) and triplet O2 in an excited state (32). This enhancement of intersystem crossing by O2 and production of reactive oxygen species, such as 1O2, may also be important in aerated biological systems. A third mechanism for generation of the excited triplet state is energy transfer to the retinoid from an excited triplet sensitizer. Sykes and Truscott (34,36) first reported that the triplet states of hydrocarbon photosensitizers, such as benz[a] anthracene, were efficiently quenched by retinol. Flash photolysis studies demonstrated that this quenching was accompanied by the appearance of a triplet state attributable to retinol and characterized by a triplet-triplet absorption spectrum having maxima at about 382 nm and 402 nm (34,37). Systematic investigations involving triplet energy transfer from a series of sensitizers with triplet energies between 88 and 171 kJ mol−1 have demonstrated that the lowest triplet state energy of retinol is 150–160 kJ mol−1 (30,37,38). Using the wavelength for the onset of retinol’s fluorescence (∼ 420 nm) and the energy of the lowest triplet state derived from triplet energy transfer studies (∼ 145 kJ mol−1), the energy gap between the lowest excited singlet state and the lowest excited triplet state of retinol can be estimated to be ∼ 140 kJ mol−1 (Fig. 4). It is noteworthy that the chemical milieu (e.g. partial pressure of O2, polarity of environment, presence of triplet sensitizers) plays a crucial role in moderating the population of the excited triplet state of retinol and its esters. This observation is particularly relevant when considering the photophysics of retinol and its esters in biological systems.

Figure 4.

 Diagram of estimated energies for retinol’s excited singlet state and excited triplet state.

Photosensitized Formation Of Reactive Intermediates And Photodegradation Of Retinyl Palmitate

It has been observed that important characteristics of a good photodetector are having a high probability for absorption of a photon and a correspondingly low probability for emission (23). In this context, retinol and its esters, which efficiently absorb UV light and have low quantum yields for fluorescence and phosphorescence, are examples of good photodetectors. The light absorbed by retinoids is known to drive a number of photochemical transformations including isomerization, generation of reactive/radical species, and formation of degradation products. The importance of retinoid photoisomerization for signal transduction in vision and in light-sensing by microorganisms is well established (20,39). Less is known about the biological importance of retinoid-photosensitized formation of reactive species and light-induced degradation of retinoids. These photochemical reactions sensitized by retinoids are being increasingly investigated due to their potentially destructive effects on light-exposed tissues such as the eye and skin.

Photosensitized formation of reactive species

Generally, the photosensitized formation of reactive species proceeds through two possible mechanisms, and in many cases leads to oxidative damage. In Type I photosensitized reactions, the photoexcited sensitizer reacts directly with the substrate or solvent resulting in either H-atom or electron transfer. The radicals so produced may initiate free radical chain reactions. Type II photosensitized oxidations involve exchange of energy from the excited photosensitizer to O2 yielding 1O2, which readily oxidizes a wide range of substrates (40). It has also been proposed that electron transfer from the photoexcited sensitizer to O2 to yield superoxide radical anion (O2−.) be considered a Type II photosensitization reaction (40). Both Type I and Type II photosensitization reactions are energetically feasible for photoexcited retinol and its esters. Type I reactions photosensitized by retinol (or its esters) could be initiated by electron or H-atom transfer from a substrate to the excited singlet or triplet state of retinol. While there are experimental determinations of the potential for electron transfer to photoexcited retinal (41), no analogous data are available for retinol. However, Bhattacharyya et al. (42) have estimated that based on the energy of retinol’s excited singlet state (∼ 285 kJ mol−1) and the one-electron reduction potential of retinol, which has been determined to be -2.19 V versus the standard calomel electrode (43), substrates with one-electron oxidation potentials (Eox) less than approximately 1.00 V could be oxidized by photoexcited retinol. This estimate correctly predicts the experimentally observed photooxidation of N,N-dimethylaniline (Eox = 0.68 V) sensitized by both retinol and retinyl acetate (42), and suggests that many biologically significant substrates could be photooxidized by retinol through a Type I mechanism (44). Type II photosensitization reactions, involving the formation of 1O2, are energetically feasible for both the excited singlet and triplet states of retinol. As shown in Fig. 4, transitions from either of these excited states of retinol can provide the energy required to form 1O2 in the 1Δg state (94 kJ mol−1) (44). Also, as previously mentioned, quenching of the excited singlet state of retinyl polyenes by O2 is enhanced by their uncharacteristically long lifetimes.

Experimental studies have demonstrated that photoexcitation of retinol and retinyl esters results in the formation reactive species via both Type I and Type II photosensitization reactions (45). In studies to investigate the role of charge separation in photoreception, Grady and Borg (46) irradiated glasses formed by cooling solutions of retinol or retinal in acetone to −196°C. After exposure to visible light, radicals were detected by electron spin resonance spectroscopy (ESR). However, the radicals could not be identified. More recently, Dillon et al. (47) have used the spin trap, nitrone 5,5-dimethylpyrroline-N-oxide (DMPO), to demonstrate the formation of carbon-centered radicals when retinal, retinol, or retinyl palmitate were irradiated with broad band UV light (λ > 300 nm). Irradiation of retinal or retinol, dissolved in methanol, resulted in spin-trapped hydroxymethyl radicals formed by hydrogen abstraction from the solvent by the photoexcited retinoid. Similarly, irradiation of retinyl palmitate dissolved in dimethylformamide resulted in the formation of a carbon-centered radical derived from the solvent. No information could be obtained about the retinyl radicals formed. Indeed, at this time there is little known about the nature of the retinyl radicals formed during electron and/or H-atom transfer between photoexcited retinol and different substrates. Studies on retinol in the electronic ground state have shown that, under certain conditions, the retinol radical cation or retinol radical anion can be formed through electron transfer (42,48,49). However, much remains unknown about the excited states involved, the nature of retinoid radicals formed, and the influence of the chemical environment on Type I photosensitization by retinol and its esters.

Several authors have reported that retinol and its esters can photosensitize Type II reactions, resulting in the formation of 1O2 or O2−.. Using a pulsed laser emitting at 264 nm, Smith has observed the formation of 1O2 after excitation of a solution of retinol dissolved in hexane (50). The formation of 1O2 was measured using 1,3-diphenylisobenzofuran as a trapping agent for 1O2. The quantum yield for the formation of 1O2 was determined to be 0.25 ± 0.05. Bhattacharyya and Das (32) have shown that, following excitation with 337.1 nm laser pulses, 1O2 is formed through O2 quenching of both the excited singlet and triplet states of retinol. These investigators also used 1,3-diphenylisobenzofuran as a trapping agent for determining 1O2. Quantum yields for 1O2 formation via O2 quenching of the excited singlet state were estimated to be 0.42 ± 0.06 in benzene and 0.30 ± 0.05 in methanol. Higher quantum yields for O2 quenching of the excited triplet state of retinol were observed. A quantum yield of 0.75 ± 0.15 was observed in benzene and 0.78 ± 0.11 in methanol. Dillion et al. (47) used both steady-state and time-resolved measurements of the near infrared phosphorescence of 1O2 to determine quantum yields for the photosensitized formation of 1O2. A medium pressure Hg lamp, filtered with an interference filter (λ = 366 nm) or cutoff filters (λ > 350 nm) was used for steady-state measurements, while a pulsed dye laser (λ = 360 nm) was used for time-resolved measurements. When methanol was the solvent and steady-state illumination was used, quantum yields were determined to be 0.05, 0.03, and <0.01 for photosensitization by retinal, retinol, and retinyl palmitate, respectively. Similar quantum yields were obtained using time-resolved measurements. When carbon tetrachloride was the solvent, steady-state photoexcitation of retinal, retinol, and retinyl palmitate resulted in quantum yields of 0.25, 0.006 and 0.003, respectively. The authors note that earlier studies, using 1,3-diphenylisobenzofuran as a trapping agent for 1O2 , may have overestimated the production of 1O2 due to the reaction of the trapping agent with other radicals formed during photosensitization. In addition, because of the low quantum yields observed for the retinoid-photosensitized formation of 1O2, the authors suggest that Type I photosensitization, involving free radicals, plays a larger role than Type II photosensitization, involving 1O2 (47).

In a number of reports, spin trapping techniques and ESR have been used to demonstrate that irradiation of retinyl palmitate results in the formation of 1O2 and O2−.. Xia et al. (51) showed that these reactive oxygen species (ROS) were formed when solutions of retinyl palmitate in 70% ethanol/water were irradiated at 320 nm. It was found that the photoexcitation of retinyl palmitate in the presence of 2, 2, 6, 6-tetramethylpiperidine (TEMP), a specific probe for 1O2 (52,53), resulted in the formation of 2, 2, 6, 6-tetramethylpiperidine-N-oxyl radical (TEMPO), providing evidence that 1O2 was generated (Fig. 5). Additional spin traps were used to investigate the role of O2−. (51). Both 5,5-dimethyl N-oxide pyrroline (DMPO) and 5-tert-butoxycarboxycarbonyl 5-methyl-1-pyrroline N-oxide (BMPO) have been found to be useful probes for detecting O2−. (54). When retinyl palmitate was irradiated in the presence of the DMPO or BMPO, ESR signals characteristic for adducts with O2−. (i.e. DMPO-OOH or BMPO-OOH, respectively) were observed. Consistent with the involvement of O2−., the ESR signal was quenched in the presence of superoxide dismutase (SOD) (Fig. 6). These results demonstrated unambiguously that retinyl palmitate can photosensitize the formation of O2−.. Thus both 1O2 and O2−. are formed following photoexcitation of retinyl palmitate.

Figure 5.

 Time-dependent changes in the steady-state ESR spectra of TEMPO formed from reaction of 4 mg mL−1 retinyl palmitate in 70% ethanol in water with 200 mM spin trap TEMP after being irradiated with UVA light at 320 nm. Conventional ESR spectra were obtained with a Varian E-109 X-band spectrometer. ESR signals were recorded with 15 mW incident microwave and 100 kHz field modulation of 1.25 G. All measurements were performed at room temperature (25 °C).

Figure 6.

 ESR spectrum observed when a solution containing 0.25 mg mL−1 retinyl palmitate and 200 mM spin trap BMPO in 70% ethanol in water are without irradiation (a) and irradiated at 320 nm for 8 min (b). Inhibition of the ESR signal by SOD demonstrates the involvement of O2 (d). The spectrum in (b) is similar to that obtained when O2 is formed by xanthine and xanthine oxidase in the presence of BMPO (c). Conventional ESR spectra were obtained with a Varian E-109 X-band spectrometer. ESR signals were recorded with 15 mW incident microwave and 100 kHz field modulation of 1.25 G. All measurements were performed at room temperature (25 °C).

The described studies show that irradiation of retinol and its esters can result in formation of free radicals and ROS. Potential biological targets for these reactive species have also been studied. Polyunsaturated fatty acids are especially susceptible to oxidative damage elicited by reactive species (44). Cherng et al. have shown that retinyl palmitate photosensitizes the formation of lipid hydroperoxides (55). These investigators used a broad band radiation source emitting 98.9% UVA (315–400 nm), 1.1% UVB (280–315 nm) and <0.0001% UVC (250–280 nm). Following irradiation of a solution containing 1 mM retinyl palmitate and 100 mM methyl linoleate in ethanol, the extent of lipid peroxidation was assessed by HPLC analysis, monitoring the eluate at 235 nm. A UV dose-dependent formation of lipid peroxides was observed (Fig. 7). The lipid peroxidation photosensitized by retinyl palmitate was inhibited by dithiothreitol (DTT) and NaN3 (Fig. 7a). These inhibitors are known to quench a wide range of reactive species (56). The observed inhibition suggested the involvement of free radicals, 1O2, or other intermediates. Lipid peroxidation was also inhibited by SOD, indicating that O2−. plays a role in the lipid peroxidation photosensitized by retinyl palmitate (57). Since 1O2 has a significantly longer half-life in deuterated water (D2O) than in H2O, enhancement of product formation in D2O is often used as a diagnostic test for the involvement of 1O2 (58). As seen in Fig. 7b, lipid peroxidation was significantly greater when measured in 10% D2O compared to 10% H2O. These studies are consistent with the involvement of 1O2, O2−. and possibly other reactive intermediates in lipid peroxidation photosensitized by retinyl palmitate. Similar results have been reported using a light source emitting higher levels of UVB light (59).

Figure 7.

 Peroxidation of methyl linoleate (ML) initiated by irradiating retinyl palmitate with UVA light. Panel a shows the inhibitory effects of NaN3, DTT and SOD, and panel b shows the enhancing effect of deuterium oxide on peroxidation. The levels of peroxidation were measured by HPLC analysis monitoring the elution at 235 nm.

Isolated DNA has also been investigated as a target for damage elicited by reactive intermediates photosensitized by retinyl palmitate (60). Solutions containing supercoiled ΦX174 phage DNA and 0.1 or 1 mM retinyl palmitate were exposed to broad band radiation comprised of predominantly UVA light. Light-dependent DNA strand cleavage was observed. Strand cleavage was partially inhibited by NaN3, indicating the involvement of radicals or ROS such as 1O2 and/or O2−.. This observation is consistent with evidence that retinyl palmitate can photosensitize damage to DNA in in vitro photobiological systems (vide infra).

Photodegradation of retinol and retinyl palmitate

Because of the commercial importance of vitamin A in a diverse array of products, including foods, drugs and cosmetics, the chemical stability and photostability of vitamin A has been investigated extensively. Many of these studies have focused on retinyl palmitate due to its widespread commercial use. In addition, interest in the photodegradation of retinyl palmitate has recently increased due to its potential importance in understanding degenerative diseases of the eye (61,62). Information on degradation and photodegradation of retinol and its esters is also potentially important for understanding retinoid-photosensitized effects in sun-exposed skin.

Vitamin A has limited chemical and photochemical stability. The chemical stability of retinol and its esters is strongly dependent on environmental factors such as solvent, temperature and availability of oxygen (63,64). Multiple pathways for chemical decomposition of all-trans-vitamin A have been observed, including thermal isomerization yielding geometric isomers, dehydration producing anhydroretinol, and oxidation resulting in complex mixtures of products (48,63–67). Retinyl esters are widely used in many products because they are more chemically stable than retinol (63). However, it has been shown that retinyl esters, such as retinyl palmitate, are less photostable than retinol (68). The photochemical mechanisms behind this difference in photolability are not fully understood.

While a number of investigators have examined the effects of different chemical environments on photodegradation of retinol and its esters, relatively few have characterized the photoproducts formed (69). Tatariunas and Matsumoto studied the photoirradiation of retinyl palmitate in methanol with a 365 nm wavelength multiband mineral light UVGL-58 lamp (61). Samples were irradiated for up to 15 min. The four photodecomposition products were identified: palmitic acid, anhydroretinol, 4,5-dihydro-5-methoxy-anhydroretinol and a dihydro-methoxy-anhydroretinol isomer that contained three conjugated double bonds (Fig. 8). Cherng et al. studied the photolysis of retinyl palmitate dissolved in ethanol and exposed to a broad band UV light source emitting 98.9% UVA (315–400 nm), 1.1% UVB (280–315 nm) and <0.0001% UVC (250–280 nm) (55). Samples were exposed to 14 J/cm2 UVA. The resulting photodecomposition products were separated by reversed-phase HPLC (Fig. 9) and the structures were identified spectroscopically. A total of 14 photodecomposition products was identified, including 5,6-epoxyretinyl palmitate, 4-keto-retinyl palmitate, 11-ethoxy-12-hydroxy-retinyl palmitate, 13-ethoxy-14-hydroxyretinyl palmitate, all-trans-anhydroretinol, 6Z-cis-anhydroretinol, 8Z-cis-anhydroretinol, 12Z-cis-anhydroretinol, palmitic acid, ethyl palmitate and four tentatively assigned cis- and trans isomeric 15-ethoxyanhydroretinols (Fig. 10). When retinyl palmitate in an oil-in-water emulsion was exposed to UVA light, many of the same photodecomposition products were formed (51). All the photodecomposition products were found to be photochemically unstable. Upon prolonged irradiation, the photodecomposition products further decomposed into multiple products (51). Thus, there was a near complete and time-dependent photodecomposition of retinyl palmitate and its photodecomposition products. These results are consistent with a free radical mechanism. The observation that NaN3 inhibited the formation of photoproducts provided further evidence for the involvement of free radicals. In addition, the authors noted that a similar product distribution was obtained when retinyl palmitate was reacted with 2,2′-azobis(2,4-dimethylmaleronitrile), a free radical initiator. It was noteworthy, however, that this free radical initiator, when incubated with retinyl palmitate, did not elicit the formation of products related to anhydroretinol. The authors concluded that both ionic dissociation, resulting in the formation of anhydroretinol analogues (Fig. 11), and free radical reactions, resulting in photooxidation products (Fig. 12), play roles in the photolysis of retinyl palmitate by UVA. As previously mentioned, there is evidence from ESR studies that retinyl palmitate, when exposed to UV light, produces both 1O2 and O2−.. Similarly, ESR studies have shown that photoexcitation of retinyl palmitate’s photoproducts, 5,6-epoxy-retinyl palmitate and anhydroretinol, results in the formation of 1O2 and O2−. (51).

Figure 8.

 The photodecomposition products following exposure of retinyl palmitate to UV light (365 nm) for up to 15 min.

Figure 9.

 Reversed-phased HPLC profile of the photodecomposition products of retinyl palmitate (0.5% in ethanol) after irradiation with 14 J cm−2 of UVA light. HPLC analysis was conducted on a Prodigy 5 µm ODS column (4.6 × 250 mm) eluted isocratically with methylene chloride in methanol (1/9; v/v) at 1 mL/min.

Figure 10.

 Photoirradiation of retinyl palmitate in ethanol with UVA light (14 J cm−2) resulted in 14 identified photodecomposition products and formation of reactive oxygen species.

Figure 11.

 Photodecomposition products of retinyl palmitate formed through an ionic photodissociation pathway.

Figure 12.

 Proposed pathways for photodecomposition of retinyl palmitate by UVA light. The photoproducts shown can be generated through a free radical mechanism.

A study of the photodecomposition of retinyl palmitate using a light source containing significant amounts of UVB light has been reported (59). Photoexcitation of retinyl palmitate in ethanol with a UV source emitting 46% UVA, 53.6% UVB and 0.5% UVC light resulted in photoproducts similar to those seen when using a source emitting predominantly UVA light (51).

Crank and Pardijanto have studied the photolysis of retinol and retinyl palmitate, dissolved in ethanol, using a single wavelength (254 nm) UVC radiation source. Oxygenated samples were irradiated for 24 h. Photolysis of retinol resulted in the formation of retinal and 3 epoxy-retinols (Fig. 13). The authors found anhydroretinol, palmitic acid and 2-butenyl palmitate as the primary photoproducts formed during photolysis of retinyl palmitate (Fig. 14). The photoproducts were drastically altered when a photosensitizer, such as rose bengal, riboflavin or chlorophyll, was present. Photolysis of retinyl palmitate in the presence of rose bengal yielded two isomeric aldehydes (Fig. 15). When riboflavin and chlorophyll were used as photosensitizers, each produced at least four different palmitate esters (Fig. 15). These results illustrate that the presence of a photosensitizer can dramatically change photoproducts resulting from irradiation of retinyl palmitate.

Figure 13.

 The photodecomposition products of retinol in ethanol exposed to UVC light (254 nm) for 24 h.

Figure 14.

 The photodecomposition products of retinyl palmitate in ethanol exposed to UVC light (254 nm) for 24 h.

Figure 15.

 The photosensitized oxidation products of retinyl palmitate in ethanol exposed to UVC light in the presence of the photosensitizer chlorophyll, rose bengal or riboflavin (254 nm). Exposures to UVC light were variable and lasted up to 4 h.

Photobiological Effects With Potential Relevance For The Skin

The studies described above demonstrate that retinol and its esters have a rich photochemistry involving the formation of free radicals, ROS and multiple photoproducts. The potential impact of this photochemistry has been examined in a number of biological systems.

In vitro photobiology

In vitro test systems, employing prokaryotic or eukaryotic cells, have proven to be useful experimental models for examining the connections between the photochemistry and photobiology of vitamin A. Most mammalian cells, grown in culture, require vitamin A, which is usually contained in the serum used to supplement media. Concentrations of vitamin A in commonly used media supplemented with 10% fetal bovine serum have been reported to be around 0.04 μM; about 2% of that in human plasma (70). Increasing the levels of vitamin A in culture media can have dramatic cellular effects, particularly when studying cell lines sensitive to vitamin A such as keratinocytes (70).

The photocytotoxicity of vitamin A, and concomitant damage to cellular components, has been studied by a number of investigators. Klamt et al. (71) have examined the photocytotoxic effects of retinol using Sertoli cells. Cells were treated with 7 μM retinol for up to 48 h and then exposed to 1 J cm−1 of UVC light (256 nm). A 20–25% reduction in viability was noted along with fragmentation of genomic DNA, generation of ROS and peroxidation of mitochondrial lipids. Mitigation of damage was observed when chelators were added prior to irradiation, implicating metal catalyzed enhancement of UV-damaging effects. This and previous work by these investigators suggested that Fe+2-catalyzed destruction of H2O2, to yield hydroxyl radicals, contributed to the observed photocytotoxicity and cellular damage (71–73). Yan et al. (60) have observed photocytotoxicity after irradiation of Jurkat T-cells which had been pre-incubated with retinyl palmitate, anhydroretinol or 5,6-epoxy-retinol. Cells exposed to 150 µM of each retinoid and a combination of 3.5 J cm−2 UVA and 6.3 J cm−2 visible light exhibited significant toxicity (Fig. 16). Using the single-cell electrophoresis (i.e. Comet) assay, these authors also assessed the amount of DNA damage concomitant with photocytotoxicity. Pre-incubation of cells with retinyl palmitate, 5,6-epoxy-retinyl palmitate or anhydroretinol (50, 100, 150 or 200 µM) and exposure to 3.5  J cm−2 UVA and 6.3  J cm−2 visible light resulted in significant fragmentation of cellular DNA (Fig. 17). Wamer et al. (74) have reported a reduction in the viability of human dermal fibroblasts exposed to 20 μM retinyl acetate and 1–5  J cm−2 UVA light, emitted from a broad band source. In addition, flash photolysis experiments on fibroblasts, preincubated with 20 μM retinyl acetate, provided evidence for the intracellular formation of the retinyl cation. Subsequent studies demonstrated that quenchers, such as NaN3, inhibited both photocytotoxicity and the formation of the retinyl cation. This work suggested that the retinyl cation may play a role in the photocytotoxicity elicited by retinyl acetate. These studies demonstrate that retinol and its esters can be photocytotoxic and photosensitize damage to cellular components such as lipids and DNA. Photooxidative damage would appear to play a significant role in the observed in vitro effects elicited by photoexcited vitamin A.

Figure 16.

 Photocytotoxicity of retinyl palmitate, anhydroretinol and 5,6-epoxyretinyl palmitate irradiated with a combination of 3.5 J cm−2 UVA and 6.3 J cm−2 visible light in human Jurkat T-cells.

Figure 17.

 DNA damage measured by the comet tail moment (top panel) and the percent of tail DNA (bottom panel) for human Jurkat T-cells in the presence of retinyl palmitate, anhydroretinol or 5,6-epoxyretinyl palmitate. Following pre-incubation with retinoid, cells were irradiated with a combination of 3.5 J cm−2 UVA and 6.3 J cm−2 visible light.

Several investigators have examined the biological significance of DNA damage photosensitized by vitamin A. Cherng et al. (55) have reported that, based on the Ames mutagenicity assay in Salmonella typhimurium TA102, retinyl palmitate and its photodecomposition products, anhydroretinol and 5,6-epoxy-retinyl palmitate, are neither mutagenic nor photomutagenic. In addition, they reported that when these retinoids were incubated with calf thymus DNA, no DNA-adducts are detected using a P32 post-labeling method (55). Mei et al. (75) have used L5178Y/Tk+/− mouse lymphoma cells to examine the photomutagenicity of retinyl palmitate. While retinyl palmitate alone was not mutagenic, it was photomutagenic in a dose-dependent manner when the cells were treated with 1–25 µg/mL retinyl palmitate and exposed to UVA light (82.8 mJ cm−2 min−1 for 30 min) (Fig. 18). To elucidate the underlying mechanism for photomutagenicity, Mei et al. (75) evaluated the loss of heterozygosity (LOH) in the Tk mutants at four microsatellite loci spanning the entire chromosome 11 on which the Tk gene is located. While the mutational spectrum for the retinyl palmitate + UVA light treatment was significantly different from the control, it was not significantly different from mutatations elicited by UVA exposure alone. Ninety four percent of the mutants from the samples concomitantly exposed to retinyl palmitate and UVA light lost the Tk+ allele, and 91% of the deleted sequences extended more than 6 cM (centimorgan) in chromosome length, indicating clastogenic events that affected a large segment of the chromosome (Fig. 19). These results suggested that retinyl palmitate in combination with UVA light is photomutagenic in mouse lymphoma cells and that the mode of action is via a clastogenic mechanism (75). The positive photomutagenicity results for retinyl palmitate when tested in mouse lymphoma assay contrast with the negative photomutagenicity results obtained in S. typhimurium TA102 (55). This inconsistency is most likely related to the fact that combined treatment with retinyl palmitate and UVA light causes clastogenicity. Compounds acting primarily by a clastogenic mechanism induce detectable mutagenicity in the mouse lymphoma assay but are not, or only weakly, mutagenic in the microbial mutagenesis assays (76). UVA light alone can induce genetic damage in cells via an oxidative mechanism (77–79). The mutational spectra induced by UVA light and by retinyl palmitate in combination with UVA light were not significantly different, which suggests that retinyl palmitate in combination with UVA light induce mutations through the same mechanism as UVA light alone, i.e. mainly oxidative DNA damage. However, further investigation is necessary to confirm this hypothesis.

Figure 18.

 Mutant frequencies in the Tk gene of mouse lymphoma cells treated with retinyl palmitate and UVA light. The cell suspensions were irradiated with 1.38 mW cm−2 UVA for 30 min during the 4 h incubation.

Figure 19.

 Comparison of percentage of mutational types produced in mouse lymphoma cells treated with vehicle (control), UVA light and a combination of UVA light and retinyl palmitate. The numbers indicate different types of mutations relative to loss of heterozygosity (LOH): 1, Non-LOH; 2, LOH at Tk locus only; 3, LOH extending to D11Mit42 (about 6 cM); 4, LOH extending to D11Mit29 (about 38 cM); and 5, LOH extending to the top of chromosome 11 (see Ref. [75] for details).

In light of the described studies of vitamin A’s photophysics, photochemistry and in vitro photobiology, it is possible to propose a pathway through which photoexcitation of vitamin A can elicit damage in cells (Fig. 20). The formation of radicals and ROS through photoexcitation of retinyl palmitate or its photoproducts is central to this pathway. Major cellular components, such as DNA and lipids, may be damaged by these reactive intermediates. Damage to DNA may result from direct interaction between these reactive intermediates, or from indirect interactions between products of lipid peroxidation and DNA (80–82). Damage to DNA, if not eliminated by high fidelity DNA repair, can result in mutations (Fig. 20).

Figure 20.

 Potential pathways for cellular damage following photoexcitation of retinyl palmitate. The formation of radicals and ROS by photoexcitation of retinyl palmitate, anhydroretinol or 5,6-epoxy-retinyl palmitate is central to these pathways. Major cellular components, such as DNA and lipids, may be damaged by these reactive intermediates. Damage to DNA may result from direct interaction between these reactive intermediates, or from indirect interactions between products of lipid peroxidation and DNA. Damage to DNA, if not eliminated by high fidelity DNA repair, can result in mutations.

Photobiology in the skin

While results from photophysical, photochemical and in vitro photobiological studies have revealed much about potential damage resulting from photoexcitation of vitamin A, uncertainties remain about the direct relevance of these results for human skin. Clearly, cutaneous vitamin A resides in a complex environment. In the skin, vitamin A is predominately derived from the diet. The reported vitamin A content of human epidermis varies widely but averages 335 ± 163 ng g−1 tissue wet weight (83). Vitamin A in the papillary dermis and reticular dermis has been found to be 186 and 393  ng g−1, respectively (4).The predominant chemical forms of vitamin A in the skin are shown in Fig. 1. Although retinol and its esters are usually the predominant retinoids, 3,4-didehydroretinol (vitamin A2) and its esters can represent 10% to 30% of the retinoids in human epidermis (83). Since studies to date have focused on retinol and its esters, little is known about the photophysics, photochemistry and photobiology of 3,4-didehydroretinol and its esters. Long chain fatty acid esters of retinol and 3,4-didehydroretinol constitute 43% to 60% of the epidermal retinoids and are commonly viewed as the storage form of vitamin A in the skin. Unesterified vitamin A in the skin occurs both free and bound to cellular retinol-binding protein (CRBP). The relative abundance of free and bound retinol has been debated and may depend on the cutaneous levels of the inducible CRBP (83). CRBP is critically important for modulating biochemical transformations of retinol such as esterification and oxidation, and may also serve to protect retinol against non-enzymatic oxidation (83). Factors such as location of vitamin A in the skin and binding to cellular proteins may have substantial effects on the photosensitizing activity of vitamin A. It is also noteworthy that the partial pressure of O2 in the skin, reported to be approximately 30 mm Hg in the papillary dermis (84), is expected to be lower than that in fully aerated chemical and in vitro test systems. Therefore, generation of ROS, though predicted in studies conducted in solution or in vitro, may be moderated in the skin. In addition, the skin possesses substantial constitutive and inducible defenses against oxidative damage that might result from photoexcitation of retinol and its esters (85). The differences between vitamin A’s environment in experimental test systems and in the skin must be considered when evaluating the biological significance of reactive intermediates generated following photoexcitation of vitamin A.

While there are differences between vitamin A’s environment in the skin and in solution or in vitro models, animal studies have shown similarities in vitamin A’s photochemical behavior. Analogous to results obtained in studies of solutions containing vitamin A, light-induced degradation of retinol and its esters has been observed in human and animal epidermis. Tang et al. (14) exposed human skin explants to sunlight for 4 h. For winter exposures, a loss of approximately 87% of the retinyl esters in the epidermis and approximately 53% of retinyl esters in the dermis was observed. Retinol degradation was less. No significant degradation of retinol was observed in the epidermis and approximately 50% of dermal retinol had degraded. Berne et al. (15) have obtained similar results using rabbits. Rabbits were exposed to 3 J cm−2 of narrow band radiation at 313, 334, 365 or 405 nm. Maximum degradation was observed for sites receiving exposures to 334 nm radiation. Levels of retinol decreased by approximately 75% in both the epidermis and dermis. The levels of retinol in the epidermis and dermis returned to near normal levels after 1 week. These studies indicate that in skin, as in solution, retinol and its esters are susceptible to light-induced degradation. At present, the photoproducts of vitamin A in the skin are unknown. In addition, it is unknown if oxidative stress is induced in the skin due to photodegradation of vitamin A.

Topical application of skin care products containing vitamin A is increasingly a source of cutaneous vitamin A in some populations. However, relatively little is known about the effects of topically applied vitamin A on the skin’s responses to UV light. Topical application of retinol or its esters has been reported to dramatically increase cutaneous levels of vitamin A (3). Clinical and experimental studies have shown that topically applied vitamin A can act as a sunscreen, providing protection against UV light-induced erythema. Antille et al. (86) have reported that human skin, treated once daily for 3 days with 2% retinyl palmitate, was protected against UV light-induced erythema similar to sites receiving a sunscreen with a sun protection factor of approximately 20. In addition, application of retinyl palmitate protected against UV light-induced DNA damage. Subsequent studies showed that absorption of UV light, rather than effects on retinoid responsive genes, gave rise to the protective effects (87). These results indicate that vitamin A, located on the skin’s surface after topical application, can be photoprotective. In addition, these results emphasize that vitamin A’s cutaneous environment is important when considering the biological effects of photoexcited vitamin A. Clearly, clinical studies or studies in relevant animal models would provide valuable information on the biological significance of the reactive intermediates generated following photoexcitation of vitamin A.


Vitamin A is commonly viewed as an antioxidant vitamin. Studies suggest that vitamin A is a chain-breaking antioxidant that is particularly effective in lipid environments (88,89). In some systems, the antioxidant action of vitamin A is reported to be comparable to, or even exceed, the effectiveness of α-tocopherol (48). The antioxidant activity of vitamin A is usually attributed to trapping chain reaction-carrying peroxyl radicals rather than donating hydrogen atoms (48). The studies described in this review demonstrate that under some conditions vitamin A can have prooxidant activity. This dual character has also been observed for vitamin C, vitamin E, and carotenoids (90–96). The cross-over conditions, under which an antioxidant becomes a prooxidant, have been shown to involve such factors as the concentration of the antioxidant, the rate of radical generation, the partial pressure of oxygen, the presence of metal ions, and the availability of ancillary antioxidant systems to remove species formed during free radical chain-breaking events (90,91,95,97). Tesoriere et al. (97) have reported that retinol behaves as a more effective antioxidant at low oxygen partial pressure, low retinol concentrations and high radical flux. The studies reported here clearly identify UV-exposure as another factor that affects the cross-over between antioxidant and prooxidant activity. Vitamin C and vitamin E have also been shown to be prooxidants when placed in test systems irradiated with UV light (90,92,95). Further studies are need to better understand the dual antioxidant and prooxidant characters of vitamin A in sun-exposed skin.

Acknowledgement— We thank Dr. Frederick A. Beland for critical review of this manuscript.