Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Nat Med. Author manuscript; available in PMC 2008 February 6.
Published in final edited form as:
Published online 2007 May 27. doi:  10.1038/nm1587
PMCID: PMC2233696

Retinaldehyde represses adipogenesis and diet-induced obesity


The metabolism of vitamin A and the diverse effects of its metabolites are tightly controlled by distinct retinoid-generating enzymes, retinoid-binding proteins and retinoid-activated nuclear receptors. Retinoic acid regulates differentiation and metabolism by activating the retinoic acid receptor and retinoid X receptor (RXR), indirectly influencing RXR heterodimeric partners. Retinoic acid is formed solely from retinaldehyde (Rald), which in turn is derived from vitamin A. Rald currently has no defined biologic role outside the eye. Here we show that Rald is present in rodent fat, binds retinol-binding proteins (CRBP1, RBP4), inhibits adipogenesis and suppresses peroxisome proliferator-activated receptor-c and RXR responses. In vivo, mice lacking the Rald-catabolizing enzyme retinaldehyde dehydrogenase 1 (Raldh1) resisted diet-induced obesity and insulin resistance and showed increased energy dissipation. In ob/ob mice, administrating Rald or a Raldh inhibitor reduced fat and increased insulin sensitivity. These results identify Rald as a distinct transcriptional regulator of the metabolic responses to a high-fat diet.

Although vitamin A and its metabolite retinoic acid have therapeutic applications, frequent side effects limit their use13. In clinical trials involving β-carotene supplementation, worrisome increases in cardiovascular events and mortality have been noted, despite evidence suggesting possible beneficial vascular effects of this treatment3. These variable responses to retinoids probably derive from the fact that β-carotene and vitamin A (retinol) and their major metabolites—retinaldehyde (Rald) and retinoic acid—regulate diverse cellular responses, including development, immune function and vision4,5. The tight control of retinoid biology is evident in the elaborate system that governs the absorption, formation, transportation and action of these structurally and functionally distinct retinoid metabolites. Despite this, retinoids and their effects remain poorly understood4,5. Of note, retinoid metabolism can vary among tissues, for instance in terms of the expression and activity of specific families of enzymes that govern the transition from retinol to Rald to retinoic acid (Supplementary Fig. 1 online and ref. 5). Alcohol dehydrogenases (Adh) oxidize retinol to Rald, and retinaldehyde dehydrogenases (Raldh) participate in the determination of cellular concentrations of free Rald by oxidizing Rald to retinoic acid5. Differences in the concentrations of specific retinoid metabolites may underlie their effects in various settings.

Retinoic acid, the most studied metabolite in the vitamin A pathway, exerts its broad range of biologic effects in large part by controlling gene expression. Retinoic acid binds to and activates members of the nuclear receptor family, including retinoic acid receptor (RAR) and retinoid X receptor (RXR) - transcription factors that link vitamin A metabolism to the transcriptional regulation of specific gene cassettes68. RXR also controls key metabolic pathways by serving as the obligate heterodimeric partner for several members of the steroid hormone nuclear receptor family, including peroxisome proliferator-activated receptors (PPARs)8. Adipogenesis is a differentiation process regulated by the complex interaction of some 11 RXR heterodimeric partners (Supplementary Fig. 1 and ref. 9). In 3T3-L1 adipocyte differentiation assays, retinoic acid effects vary as a function of the stage of adipogenesis and relative RAR, PPAR-γ and RXR expression9,10. Early in adipogenesis, retinoic acid blocks differentiation, whereas after 48 h of differentiation, it promotes fat cell formation10. Inhibition of endogenous retinoic acid production by the Raldh inhibitor citral reduces weight in vivo in animal models11,12. Moreover, although 9-cis-retinoic acid has been generally accepted as a natural ligand for RXR (refs. 7,8), its role in adipogenesis and its existence in vivo have been challenged by some biochemical and genetic studies5.

Rald is primarily considered to be a precursor for retinoic acid formation5,13. Although Rald (11-cis-Rald) is essential for molecular signaling in vision, a role for Rald outside the eye remains essentially unknown4. We hypothesized that Rald itself might be present in fat in vivo, where it could function as a specific but previously unrecognized regulator of adipogenesis, independent of its conversion to retinoic acid.


Rald is present in rodent fat

Rald is generated by the action of alcohol dehydrogenase-1 (Adh1) on retinol, and its concentration is determined in large part through its subsequent catabolism by retinaldehyde dehydrogenase-1 (Raldh1, the gene for which is officially designated Aldh1a1) to retinoic acid. We found that both Adh1 and Raldh1 were differentially expressed during 3T3-L1 preadipocyte differentiation (Fig. 1a). Whereas preadipocytes expressed mainly Adh1, differentiated 3T3-L1 cells expressed predominantly Raldh1, suggesting specific and temporally regulated Rald production and catabolism in fat. These enzymes were also expressed in white fat from both lean (C56/BL6) and obese (ob/ob) mice (Fig. 1a). Adh1 expression was significantly higher in lean mice than in ob/ob mice, whereas Raldh1 expression was not significantly different between the groups (Fig. 1a, P<0.05). Given the differential regulation between Adh1 and Raldh1 in lean versus genetically obese mice, we tested further for the presence and functional effects of Rald in fat tissue.

Figure 1
Rald and Rald-metabolizing enzymes in rodent fat. (a) Northern blot analysis (left) of Adh1 and Raldh1 expression in 3T3-L1 preadipocytes before and after differentiation; fatty acid binding protein 4 gene (Fabp4), differentiation control. Representative ...

Rald is an unstable molecule, making its detection in tissues challenging14. To counter this, Rald reduction to Rald oximes has been used to generate a stable biochemical Rald signature for purposes of quantification1416. We used this approach to measure the presence of Rald in fat from C57/BL6 mice on either a standard (lean) or a high-fat diet. High-fat fed mice doubled their body weight as compared with those on regular chow (Fig. 1b). We dissected equal amounts of fat tissue from lean and obese mice, immediately reduced the isolated tissues with hydroxylamine, and analyzed the concentrations of retinol and Rald oxime using chromatographic and spectral analysis (for comparison, we used retinol and Rald oxime standards; Supplementary Fig. 2 online). Our analysis revealed that not only did fat contain Rald, but that retinol and Rald were reduced approximately 60% and 82%, respectively, in obese versus lean mice (Fig. 1b). Rald was also present in rabbit fat (data not shown), obviating any species-specific artifact. Mass spectrometric analysis confirmed Rald oxime structures in pooled HPLC-purified fractions from wild-type rodent fat extracts as evidenced by comparison with a Rald oxime standard (Fig. 1c). The presence of Rald oxime in fat was supported by the similar molecular weight of a protonated molecule at 300 mass-to-charge ratio (m/z), the loss of the oxime group at 242 m/z and the overall fragmentation pattern (94–208 m/z). Finally, our analysis showed that Rald concentration in white fat ranged from 100 nM to ~1 μM, based on an analysis of 13 mice on regular chow (data not shown).

The cognate interaction between retinoids and retinol-binding proteins can influence the amounts and effects of retinoids both within and outside the cell4. As retinol is the known binding partner for human cellular retinol-binding protein-1 (CRBP1) and retinol-binding protein 4 (RBP4)17,18, we used standard fluorescence quenching assays to compare Rald versus retinol binding to CRBP1 and RBP4. All-trans-Rald had a similar binding profile as all-trans-retinol to both CRBP1 and RBP4 (Fig. 1d). The existence of mechanisms for Rald binding and transport as well as its presence in nanomolar concentrations in fat supported further investigation into a functional role for Rald in adipogenesis at similar concentrations to those demonstrated in vivo.

Rald inhibits adipogenesis in vitro

To test Rald effects on adipogenesis in vitro, we performed 3T3-L1 mouse preadipocyte differentiation assays using Rald concentrations similar to those found in fat in vivo. Rald concentrations as low as 100 nM suppressed mRNA expression of the adipogenic target genes Cd36, Adipoq (encoding adiponectin) and Fabp4 (also known as aP2); this repression occurred in a concentration-dependent manner in both the presence (Fig. 2a) and the absence (Supplementary Fig. 3a online) of the PPAR-γ agonist BRL49653 (BRL), a potent adipogenic stimulus.

Figure 2
Rald and retinoic acid differentially regulate PPAR-γ–induced adipogenic genes. (a) Northern blotting for Cd36, Adipoq, Fabp4 and Gapdh performed in 3T3-L1 cells differentiated with BRL in the presence of Rald. (b) Adipoq mRNA expression ...

As the effects of retinoic acid on adipogenesis vary as a function of the differentiation stage10, we compared responses to 9-cis-retinoic acid, all-trans-retinoic acid (the most abundant form of retinoic acid) and all-trans-Rald in both the early and late phases of adipocyte differentiation. The expression and release of the adipocytokine adiponectin (encoded by Adipoq) is an indicator of the extent of adipogenesis19,20. Whereas 9-cis-retinoic acid, all-trans-Rald and to a lesser extent all-trans-retinoic acid (~15% inhibition, P<0.05) all inhibited Adipoq expression when added early to differentiating 3T3-L1 preadipocytes (day 0, Fig. 2b), both isomers of retinoic acid had no inhibitory effect when added later (day 2, Fig. 2b) as previously reported10. In contrast to retinoic acid, Rald also inhibited Adipoq expression when added during later stages of differentiation, even at nanomolar concentrations (Fig. 2a,b). In these same experiments, Rald stimulation either early or late in adipocyte differentiation also decreased adiponectin secretion, in a concentration-dependent manner (Fig. 2c). Indeed, nanomolar concentrations of Rald mitigated the sixfold increase in adiponectin induced by BRL stimulation (Fig. 2c). Rald also decreased lipid accumulation during 3T3-L1 preadipocyte differentiation in both the absence and the presence of BRL (Fig. 2d).

Rald regulates distinct nuclear receptor responses

Given the suppression of PPAR-γ agonist–stimulated adipogenesis and adiponectin release by Rald, we tested whether Rald regulates RAR and RXR activity, and, if so, whether it does so in a manner distinct from that of other retinoids. We performed ligand-binding domain (LBD)-GAL4 transfection assays in 3T3-NIH fibroblasts, in the presence and absence of known specific nuclear receptor agonists and Rald. As previously reported21, all-trans-Rald weakly activated the RAR-αLBD, but did not alter activation of RAR-αLBD by its known ligand, 9-cis retinoic acid (Fig. 3a). Rald alone had no effect on RXR-αLBD activation (Fig. 3a), but, in contrast to its effect on RAR-αLBD responses, it significantly inhibited RXR-αLBD activation by 9-cis-retinoic acid (Fig. 3a).

Figure 3
Rald regulates nuclear receptor responses in vitro. (a) RAR-α(left) and RXR-α(right) LBD-GAL4 assays performed in NIH-3T3 cells in the absence (Veh) or presence of 9-cis retinoic acid (9-cis-RA) (300 nM) and increasing Rald concentrations. ...

Given these effects of Rald on RXR-LBD activation, we questioned whether Rald could inhibit the activation of a transfected canonical PPAR response element (PPRE) luciferase construct after transfection of RXR or PPAR-γ and agonist stimulation (9-cis retinoic acid or BRL, respectively). Rald significantly inhibited PPRE activation by each agonist, with the most potent effects seen after PPAR-γ and RXR co-transfection and PPAR-γ agonist stimulation (60% inhibition, Fig. 3b). Given these results, we next investigated the direct interaction between PPAR-γ and all three major Rald isomers (9-cis-, 13-cis- and all-trans-Rald), using cell-free radioligand displacement assays22. All three isomers displaced high affinity PPAR-γ agonists (Kd = 5.9 ± 0.7, 9.7 ± 1.2, and 11.9 ± 1.9 μM, respectively; mean ± s.d.), consistent with direct but weak binding of these molecules to the PPAR-γ–LBD (Fig. 3b). We observed similar effects in cell-based PPAR-γ–LBD assays in the presence of BRL (data not shown). Given Rald’s distinct effects on adipogenesis, and known discrete roles for RXR and PPAR-γ in this process, we evaluated Rald effects in an RXR loss-of-function model, repeating standard 3T3-L1 adipogenesis assays in the presence of Rald, but after decreasing RXR-αand RXR-β using short interfering RNA (siRNA) to the RNA encoding each RXR isotype. These RXR isotypes are expressed early in adipogenesis (48 h), helping to initiate subsequent adipocyte differentiation, as evident in vitro23 and in vivo24. We measured triglyceride accumulation and adiponectin secretion as specific, distinct indicators of adipogenesis20,25. After siRNA exposure, total RXR was undetectable by western blotting (Fig. 3c). As expected, 3T3-L1 adipocytes treated with RXR-αsiRNA and RXR-β siRNA showed decreased triglyceride accumulation, which Rald decreased further, independent of RXR expression. In contrast, whereas RXR-αsiRNA and RXR-β siRNA significantly decreased adiponectin secretion (~92% less), Rald had no further effect (Fig. 3d), consistent with an RXR-dependent Rald effect on adiponectin release.

Rald metabolism controls adipocyte biology

Endogenous concentrations of Rald are dictated by enzymes controlling its production (Adh1) and catabolism (Raldh1)5,14. Raldh1-deficient mice (Raldh1−/−) have been well characterized as a model for Rald overproduction14. These mice have impaired Rald oxidation, as evident in their markedly decreased retinoic acid and increased Rald concentrations while on a vitamin A–containing diet14. Raldh1−/− mice fed a standard chow diet with 4 IU vitamin A per g had twice the plasma Rald of those in age- and sex-matched wild-type controls (8.6 ± 2.6 and 3.8 ± 2.6 nM, respectively). To evaluate whether Raldh1 deficiency also affects fat cell differentiation, we performed adipogenesis assays in primary embryonic fibroblasts isolated from Raldh1−/−and wild-type mice. Adipogenesis was markedly decreased in Raldh1−/− as compared with wild-type cells, as evidenced by oil-redO staining for lipid accumulation (Fig. 4a). Adiponectin secretion was 52% less in Raldh1−/− versus wild-type cells, a difference that was even more pronounced after treatment with BRL at all concentrations tested (63% less in Raldh1−/− versus wild-type cells with 300 nM BRL, Fig. 4b). With higher BRL concentrations (10 μM), these effects of Rald on adiponectin suppression and lipid accumulation approached wild-type levels (Supplementary Fig. 3b).

Figure 4
Raldh1 deficiency is associated with suppressed adipogenesis in vitro and increased Rald and reduced adipocyte size in vivo. (a) Lipid accumulation in differentiated primary fibroblasts from Raldh1−/− as compared with wild-type (WT) embryos. ...

Mass spectroscopy analyses demonstrated Rald in white fat from wild-type and Raldh1−/− mice (242.9 m/z; n = 12 per genotype; Fig. 4c). This spectral pattern was identical to that seen with a Rald oxime standard (Supplementary Fig. 4a online). To study Rald effects on adipogenesis in vivo, we placed wild-type and Raldh1−/− mice on a high-fat diet (45% fat, standard vitamin A 4 IU/g), measuring their weight weekly until tissue analysis at after 6 months of experimental diets (age 8 months). White fat of Raldh1−/− mice had significantly higher retinol (152%) and Rald (206%) as compared with that from wild-type mice (Fig. 4d). Adipocytes from Raldh1−/− fat were half the size of those from wild-type fat (Fig. 4e,f). In these same fat samples, adipocyte size correlated inversely with Rald concentrations (Fig. 4g).

Raldh1 regulates lipid and glucose metabolism

We questioned if Raldh1 deficiency would have systemic metabolic consequences. Indeed, after high-fat feeding (6 months), Raldh1−/− mice gained significantly less weight (93%) than the wild-type mice (weight gain in wild-type = 26.6 ± 1.9 g; in Raldh1−/− = 13.7 ± 3.6 g; Fig. 5a). Weight differences were evident beginning at 1 month of high fat feeding (data not shown). Of note, Raldh1−/− females weighed significantly less than males (relative to wild-type mice of the corresponding sex, 57% versus 41% less, P < 0.001). Given sex differences in metabolic parameters related to fat, we performed additional metabolic analyses in wild-type (n = 5) and Raldh1−/−females (n = 4). DEXA scanning revealed decreased whole-body fat accumulation in Raldh1−/− mice (Supplementary Fig. 4b). The decreased white fat accumulation in Raldh1−/− versus wild-type mice was evident in both subcutaneous and visceral fat pads (76% and 74% less, respectively; Fig. 5b). Raldh1−/− mice also had significantly lower plasma free fatty acids than wild-type mice (0.21 ± 0.1 mmol/l versus 0.53 ± 0.3 mmol/l, respectively; P < 0.04, Wilcoxon rank test). Livers from Raldh1−/− mice also had decreased lipid accumulation compared with wild-type mice, as evident from H&E staining (Fig. 5c) and total liver weight (Supplementary Fig. 4c).

Figure 5
Raldh1−/− mice resist high-fat diet–induced obesity. (a) Weight of wild-type and Raldh1−/− mice after high-fat diet (180 d). *P < 0.001 (Wilcoxon rank test). Representative X-ray images of each genotype. ...

The differences in fat accumulation in Raldh1−/− versus wild-type mice occurred despite similar food and water intake in both groups (Fig. 5d), indicating a shift in total energy balance in Raldh1 deficiency. Indeed, Raldh1−/− mice had a significantly higher metabolic rate (Fig. 5e), respiratory quotient (Fig. 5f), and body temperature (Fig. 5g) compared with wild-type controls. Consistent with the role of thermogenesis in determining body weight26, amounts of uncoupling protein-1 (UCP-1) in brown fat were significantly higher in Raldh1−/− mice than wild-type mice (Fig. 5g).

Fat tissue regulates whole body insulin sensitivity through various mechanisms, including adipokine release2730. As such, we measured changes in adiponectin, leptin and RBP4 in Raldh1−/− mice. Total adiponectin, leptin and RBP4 were significantly decreased in the plasma of Raldh1−/− mice as compared with wild-type mice (Fig. 6a). On high fat diet, Raldh1−/− mice were protected from the increased insulin resistance evident in wild-type mice as seen in glucose and insulin tolerance testing (Fig. 6b). Insulin concentrations did not differ between genotypes (Fig. 6c). To further consider if Rald effects on fat accumulation and insulin resistance were due to Rald itself and not to subsequent generation of retinoic acid or other vitamin A metabolites, we used ob/ob mice that experience progressive weight gain on regular chow28,31 to compare responses to Rald, the Rald parent compound vitamin A, all-trans-retinoic acid and citral, a known inhibitor of Raldh enzymes11. After 3 weeks, we quantified subcutaneous fat mass by magnetic resonance imaging (MRI). The extent of visceral fat accumulation in these mice precluded its accurate quantitative measurement. Mice receiving Rald or citral had significantly less subcutaneous fat relative to total body fat (15.5 ± 0.6% and 14.8 ± 0.6%, respectively), than those receiving vehicle, vitamin A or all-trans-retinoic acid (18.8 ± 1.4%, 17.3 ± 1.3% and 19.1 ± 1.6%) (all P < 0.05; Fig. 6d,e). Rald administration also repressed adipogenesis in preadipocytes isolated from human visceral fat depots (data not shown). RBP4 amounts also varied in ob/ob mice exposed to these different retinoids. Whereas RBP4 amounts did not vary in response to retinoic acid or vehicle, they doubled in response to vitamin A (Fig. 6f). In contrast, both Rald and citral significantly suppressed circulating RBP4 (versus vehicle), recapitulating the RBP4 pattern seen in Raldh1−/− mice (Fig. 6a). Consistent with the changes seen in adiposity and RBP4, both Rald and citral administration improved glucose tolerance in ob/ob mice (Fig. 6g).

Figure 6
Raldh1 deficiency and direct Rald administration protect against insulin resistance. Mice analyzed in ac are the same as in Figure 5. (a) Plasma protein concentrations of adiponectin and leptin and amounts RBP4. *P < 0.048, **P < ...


We demonstrate here that Rald plays a distinct metabolic role in adipocyte differentiation in vitro, and in diet-induced insulin resistance and obesity in vivo. Rald is present in fat at nanomolar concentrations (~1 nmol/g) and can interact with CRBP1 and RBP4, binding proteins involved in intracellular and circulating retinoid transport. Rald suppresses adipogenic gene expression, adipocyte lipid accumulation, RXR-αand PPAR-γ responses, all at concentrations (<1 μM) similar to those in rodent fat in vivo. Despite the widely held assumption that Rald outside of the eye serves mainly as a precursor for retinoic acid4, the differences in the effects of Rald versus those of retinoic acid in vitro and in vivo argue for Rald itself as a distinct mediator in fat. When Rald concentrations were increased in vivo, either in the absence of Raldh1 or after direct Rald administration, fat formation was decreased. Inhibition of Rald catabolism (citral treatment) in vivo had similar effects although again distinct from retinoic acid or vitamin A administration. Together this data identifies Rald as a biologically active metabolite present in fat that may regulate adipogenesis through its action on RXR-αand PPAR-γ responses and in a manner opposite of retinoic acid effects.

Rald’s unique effects on adipokine expression, adipogenesis and body weight focus attention on the parameters regulating the relative cellular concentrations of Rald and retinoic acid. The balance between Rald and retinoic acid is determined by factors such as the concentration of vitamin A in the body, the expression and activity of enzymes that metabolize Rald and retinoic acid, other retinol-modifying enzymes (such as esterases and hydrolases), as well as retinol-binding proteins and the redox status in cells4,5,32. Each of these factors may have functional consequences, as seen with the recently reported association between RBP4 and diabetes in mice and humans29,33. In terms of the enzymes metabolizing Rald, we have demonstrated the genetic absence of Raldh1 results in a metabolic phenotype involving marked alterations in fat accumulation, glucose homeostasis and adipokine production after high-fat feeding. These metabolic changes are probably due at least in part to the increased concentrations of Rald in fat, as demonstrated here, especially as direct administration of Rald reproduced the metabolic profile evident in Raldh1−/− mice. Notably, direct administration of retinoic acid and Rald had different effects in vitro and in vivo, further supporting a role for Rald that is distinct from its function as a retinoic acid precursor.

Rald seems to exert its effects through both RXR-dependent and RXR-independent mechanisms. Rald repressed adiponectin production, but not after RXR expression had been reduced. In contrast, Rald-mediated repression of triglyceride accumulation persisted regardless of RXR expression. Rald inhibited LBD activation and cellular responses to both RXR and PPAR-γ agonists, with effects on early and late adipogenesis that are consistent with the reported temporal expression of PPAR-γ and RXR9,10. Although Rald binding to the RXR- and PPAR-γ–LBDs was weak, it potently suppressed adipogenesis in vitro and in vivo. It has been reported that selective synthetic PPAR and RXR modulators also show a similar divergence between the potency of receptor binding and adipogenic effects34,35. The ability of higher BRL concentrations to overcome the effects of Rald suggests that Rald responses are mediated at least part through PPAR-γ. Various mechanisms may underlie how certain molecules influence nuclear receptor responses independent of receptor-binding potency, including conformational changes in the receptor, salt bridge formation, accessory molecule recruitment and release, and post-translation protein modification36. Rald’s metabolic effects make further studies on Rald derivatives, the factors determining Rald concentrations and Rald transport all of considerable interest. More broadly, these data suggest a regulatory pathway in which a given molecule (vitamin A) can yield both a nuclear receptor agonist (retinoic acid) and a molecule (Rald) capable of inhibiting specific nuclear receptor responses (PPAR-γ, RXR).

These studies provide a link between vitamin A metabolism and responses to a high-fat diet, including pathologic complications such as obesity and insulin resistance. The metabolic changes evident with increased Rald have several potential clinical implications. The protection against diet-mediated obesity and insulin resistance identified in Raldh1−/− mice suggests this inducible enzyme of vitamin A metabolism as a potential candidate for therapeutic targeting and/or a source of body weight variation. Similarly, Rald generation might influence therapeutic responses to vitamin A, retinoic acid or other retinoid-based treatments as well as the effects of PPAR-γ agonists. The presence of Rald in fat and its association with RBP4 suggest that Rald might contribute to the relationship between RBP4 and insulin sensitivity29,33.

In the models studied here, Rald concentrations in fat in vivo correlated tightly with changes in fat accumulation and metabolic responses, providing one probable explanation for the metabolic phenotype of Raldh1-deficient mice. Rald has effects on both visceral and subcutaneous fat, as seen in Raldh1−/− mice. In the absence of Raldh1, energy balance appears shifted toward increased energy dissipation, as suggested by the increased body temperature, metabolic rate42 and UCP-1 expression manifest in these mice. Interestingly, in some animal models, increased UCP-1 has not been associated with an increased respiratory quotient37. Regardless, the changes in energy balance seen with Rald could result from its actions in various Rald-and retinoic acid–sensitive tissues, including brown fat, immune cells and the central nervous system (CNS). Recent work suggests a possible role for Raldh2 in dendritic cells43, although Raldh1 appears the predominant determinant of Rald metabolism in response to vitamin A intake14. The CNS is a particularly important regulator of body weight in general and adipocyte biology specifically; for example, in determining rates of lipolysis38. Our data do not specifically exclude the possibility of Rald exerting some effects via the CNS. Here we focused on demonstrating the presence of Rald and its effects on adipocyte responses. The linear relationship between Rald concentrations and adipocyte size, the lower concentrations of Rald in obese as compared with lean mice, and the impaired adipogenesis seen in Raldh1−/− preadipocytes all suggest that Rald effects in adipose tissue probably contribute to the protection against diet-induced obesity and insulin resistance evident in Rald1-deficient mice. Certainly Rald’s actions in adipose tissue, including the regulation shown here of specific adipokines, can also provide feedback to other systems, including the CNS and the immune system, with subsequent effects on adipose biology.

Taken together, these findings allow Rald to join retinoic acid as a distinct biologically active mediator of energy balance and insulin sensitivity. The integration of these critical metabolic pathways by a natural molecule such as Rald may provide new opportunities for understanding the complex interaction between vitamin A, its meta-bolites and the transcriptional regulation of metabolism.



We obtained reagents and media from Sigma-Aldrich and BioWhit-taker unless otherwise indicated. All media contained amphotericin B, penicillin and streptomycin. BRL49653 (rosiglitazone) was a gift from GlaxoSmithKline. Unless otherwise indicated, retinoids used were all-trans isomers. All diets were from Research Diet, Inc.

Animal studies

C57/BL6 mice (weight: 19.6 ± 2 g) versus ob/ob mice (41.7 ± 2 g), all 12-week-old females on regular chow, were used for gene expression studies and retinoid analysis (n = 3 per genotype). For retinoid MS analysis, we studied pooled subcutaneous and visceral fat from male mice (129S3/SvImJ, 6 months of age, n = 4) and a New Zealand White rabbit. For metabolic studies, we compared age- (8 weeks old) and sex-matched Raldh1−/−14 and wild-type mice (five per sex and genotype). The D12451 high-fat diet contained 45% of calories derived from fat and standard vitamin A (4 IU/g). Water was ad libitum.

For MRI fat distribution studies, we administered retinoids (all-trans-Rald, vitamin A or all-trans-retinoic acid, all 500 nM) or citral (10 μM, equal to 240 nmol/g), all in ethanol:PBS (2:200 μl), by daily intraperitoneal injections for 3 weeks. The Standing Committee on Animals at Harvard Medical School approved all protocols.

Cell culture

We cultured and differentiated mouse 3T3-L1 preadipocytes and primary fibroblasts isolated from 16-d-old embryos using standard adipogenesis and isolation protocols25. Cells were differentiated (7 d) with or without BRL (1 μM). Rald was added at the indicated concentrations in either early (5 h) or late (48 h) stages of adipogenesis, as timed to initiation of differentiation.

Transient transfections

For transient transfection of NIH 3T3 cells (2.3 × 104 cells, 24-well plates), we used pCMX-β-galactosidase and LBD–yeast Gal 4–luciferase constructs and Fugene (Roche) as before39. For siRNA transfections, we administered scrambled control (sequences C, D) or specific RXR-αand RXR-β siRNA sequences (Santa Cruz) to 3T3-L1 cells (90% confluence, antibiotic-free DMEM, 10% calf serum, 24-well plates) using Lipofectamine 2000 (8 h) and OptiMEM medium (Invitrogen). We supplemented the medium with 10% FBS 5 h after transfection. We evaluated RXR amounts by Western blotting 48 h after transfection.

Human recombinant RBP and fluorescence binding assays

Human RBP4 and mouse CRBP1 subcloned in pET expression vectors were expressed and purified as described previously40, but without retinol. We dialyzed refolded protein against binding assay buffer (0.05 M sodium phosphate, 0.15 M NaCl, pH 7.0) overnight, and quantified proteins at 280 nm (RBP4 40,400 M−1cm−1, CRBP1 26,720 M−1 cm−1)17, 41. We measured tryptophan fluorescence (Aminco, Spectronic Unicam) by excitation (285 nm and emission (335 nm), both in 0.05 M sodium phosphate buffer18 which indicated retinoid binding to RBP4 or CRBPI (1 μM).

Scintillation proximity assay

We studied Rald displacement of a 3H2-labeled known synthetic PPAR-γ agonists (nTZD3, Kd = 2.5 nM) from human full-length PPAR-γ2 as before22.

RNA analysis

We determined mouse Adh1, Raldh1 and Actb (β-actin) mRNA (RNeasy, Qiagen) in white fat using semi-quantitative RT-PCR and the following primers:

Adh1: 5′-ATGAGCACTGCGGGAAAAGT-3′, 5′-ACTTTATTGGCCGTGT CTCTAA-3′; Raldh1: 5′-TGGGTTAACTGCTATATCATGTTG-3′, 5′-GGGTG CCTTTATTAAGCTTTGCG-3′. Results were obtained within the linear range for each gene (31 and 29 cycles, respectively) and normalized to Actb (β-actin) expression. We performed northern blotting with HyBond (Amersham) as before39.

Protein and intracellular lipid analysis

We determined triglyceride content in lysed cells (RIPA buffer, complete protease inhibitor cocktail, Roche) using enzymatic colorimetric assay (Wako) and measured adiponectin using ELISA (R&D Systems). For western blotting, we performed a reducing gel separation (10% acrylamide) on cell lysates, plasma and tissue lysates, before hybridization with antibodies to RXR (Santa Cruz), to mouse RBP4 (Alpco) or to UCP-1 (Chemicon).

Liquid chromatography (LC), mass spectrometry (MS)

We generated a Rald oxime standard using reduction of retinaldehyde (100 μM, 22 °C, 2 h under argon protected from light) with hydroxylamine (1 M) and EDTA (25 μM) in PBS as before14,15. White fat (~200 μg) was dissected from relevant animals, immediately reduced with hydroxylamine, purified by solid-phase extraction (Bakerbond amino column) and analyzed for retinol and Rald oxime content using high performance LC (HPLC) analysis (see also legend to Supplementary Fig. 2). Pooled Rald oxime fractions then underwent structural analysis using LC and tandem MS (LC-MS-MS) using atmospheric pressure chemical ionization in positive mode (Bruker Daltonics, Esquire LC) to detect Rald oxime ion molecular ion peak [M+H]+ at 300 m/z. The ionization parameters included capillary voltage, 3,000 V; APCI temperature, 350 °C; source temperature, 300 °C; and scanning ions in the 82 to 306 m/z range.


We embedded fat and liver tissue in paraffin before hematoxylin and eosin (H&E) staining followed by quantification of adipocyte size or liver lipid accumulation (ImageJ software).

Magnetic resonance imaging (MRI)

We obtained MRI scans (1-mm slices) using a Bruker Avance 500 wide-bore spectrometer (11.7 T; 500 MHz for proton) fitted with a gradient amplifier and a 30 mm ‘birdcage’ transmitter/ receiver coil before processing the data (Paravision). The spin-echo parameters for T1-weighted images were as follows: TE = 15 ms, TR = 300 ms, matrix = 256, FOV = 30 mm; for RARE images, TE = 51 ms, TR = 2,500 ms, matrix = 256, FOV = 30 mm. Abdominal fat measurements used axial slices (n = 8) at the level of the left renal pelvis.

Dual-energy X-ray absorptiometry (DEXA)

We used the GE Lunar Corporation PIXImus2 Dexa Scanner, normalizing the data to a quality control plot (Charles River Laboratories).

Metabolic parameters

After mouse acclimation to a powdered high-fat diet (4 d), we measured food and water intake, oxygen consumption and carbon dioxide production in metabolic cages (Ancare, Charles River Laboratories). The calculated metabolic rate (Weir equation) is expressed per g body weight42. We performed insulin (ITT) and glucose tolerance tests (GTT) after fasting (16 h), using intraperitoneal insulin injections (ITT, 0.1 U/ml, 0.005 ml/g body weight) or a single 25% dextrose injection (GTT, 0.004 ml/g body weight), and a glucometer for measurements (Accu-Chek Advantage, Roche).


We thank G. Sukhova (Brigham and Women’s Hospital), J. Kirkland and T. Tchkonia (Boston University), N. Krinsky and R. Russell (Tufts University) for helpful discussions; P. Scherer, S. Kliewer, D. Mangelsdorf (University of Texas), C.H. Lee (Harvard University) and T. Willson for reagents; and E. Shvarz, K. Volz, J. Qin, T. Archibald, N. Sharma, S. Laclair and R. Driscoll for technical support. This research was supported by the Boston Obesity Nutrition Research Center 5P30DK046200 and K12 HD051959-01 NICHD BIRCWH, the American Heart Association SDG 0530101N (O.Z.); the US National Institutes of Health (R01 HL071745 and P01 HL48743) and the Donald W. Reynolds Foundation (J.P.).


Note: Supplementary information is available on the Nature Medicine website.


The authors declare no competing financial interests.

Reprints and permissions information is available online at


1. Paydas S, et al. Vasculitis associated with all trans retinoic acid (ATRA) in a case with acute promyelocytic leukemia. Leuk Lymphoma. 2003;44:547–548. [PubMed]
2. Redlich CA, et al. Effect of long-term β-carotene and vitamin A on serum cholesterol and triglyceride levels among participants in the Carotene and Retinol Efficacy Trial (CARET) Atherosclerosis. 1999;143:427–434. [PubMed]
3. Rapola JM, et al. Randomised trial of α-tocopherol and β-carotene supplements on incidence of major coronary events in men with previous myocardial infarction. Lancet. 1997;349:1715–1720. [PubMed]
4. Napoli JL. Retinoic acid: its biosynthesis and metabolism. Prog Nucleic Acid Res Mol Biol. 1999;63:139–188. [PubMed]
5. Duester G, Mic FA, Molotkov A. Cytosolic retinoid dehydrogenases govern ubiquitous metabolism of retinol to retinaldehyde followed by tissue-specific metabolism to retinoic acid. Chem Biol Interact. 2003;143–144:201–210.
6. Chambon P. A decade of molecular biology of retinoic acid receptors. FASEB J. 1996;10:940–954. [PubMed]
7. Heyman RA, et al. 9-cis retinoic acid is a high affinity ligand for the retinoid X receptor. Cell. 1992;68:397–406. [PubMed]
8. Shulman AI, Mangelsdorf DJ. Retinoid x receptor heterodimers in the metabolic syndrome. N Engl J Med. 2005;353:604–615. [PubMed]
9. Fu M, et al. A nuclear receptor atlas: 3T3–L1 adipogenesis. Mol Endocrinol. 2005;19:2437–2450. [PubMed]
10. Xue JC, Schwarz EJ, Chawla A, Lazar MA. Distinct stages in adipogenesis revealed by retinoid inhibition of differentiation after induction of PPARγ Mol Cell Biol. 1996;16:1567–1575. [PMC free article] [PubMed]
11. Kikonyogo A, Abriola DP, Dryjanski M, Pietruszko R. Mechanism of inhibition of aldehyde dehydrogenase by citral, a retinoid antagonist. Eur J Biochem. 1999;262:704–712. [PubMed]
12. Ress NB, et al. Toxicology and carcinogenesis studies of microencapsulated citral in rats and mice. Toxicol Sci. 2003;71:198–206. [PubMed]
13. Duester G. Families of retinoid dehydrogenases regulating vitamin A function: production of visual pigment and retinoic acid. Eur J Biochem. 2000;267:4315–4324. [PubMed]
14. Molotkov A, Duester G. Genetic evidence that retinaldehyde dehydrogenase Raldh1 (Aldh1a1) functions downstream of alcohol dehydrogenase Adh1 in metabolism of retinol to retinoic acid. J Biol Chem. 2003;278:36085–36090. [PubMed]
15. von Lintig J, Wyss A. Molecular analysis of vitamin A formation: cloning and characterization of β-carotene 15,15′-dioxygenases. Arch Biochem Biophys. 2001;385:47–52. [PubMed]
16. Lakshman MR, Mychkovsky I, Attlesey M. Enzymatic conversion of all-trans-β-carotene to retinal by a cytosolic enzyme from rabbit and rat intestinal mucosa. Proc Natl Acad Sci USA. 1989;86:9124–9128. [PubMed]
17. Vogel S, et al. Characterization of a new member of the fatty acid-binding protein family that binds all-trans-retinol. J Biol Chem. 2001;276:1353–1360. [PubMed]
18. Berni R, Clerici M, Malpeli G, Cleris L, Formelli F. Retinoids: in vitro interaction with retinol-binding protein and influence on plasma retinol. FASEB J. 1993;7:1179–1184. [PubMed]
19. Yu S, et al. Adipocyte-specific gene expression and adipogenic steatosis in the mouse liver due to peroxisome proliferator-activated receptor γ1 (PPARγ1) overexpression. J Biol Chem. 2003;278:498–505. [PubMed]
20. Iwaki M, et al. Induction of adiponectin, a fat-derived antidiabetic and antiatherogenic factor, by nuclear receptors. Diabetes. 2003;52:1655–1663. [PubMed]
21. Repa JJ, Hanson KK, Clagett-Dame M. All-trans-retinol is a ligand for the retinoic acid receptors. Proc Natl Acad Sci USA. 1993;90:7293–7297. [PubMed]
22. Berger JP, et al. Distinct properties and advantages of a novel peroxisome proliferator-activated protein γ selective modulator. Mol Endocrinol. 2003;17:662–676. [PubMed]
23. Canan Koch SS, et al. Synthesis of retinoid X receptor-specific ligands that are potent inducers of adipogenesis in 3T3–L1 cells. J Med Chem. 1999;42:742–750. [PubMed]
24. Imai T, Jiang M, Chambon P, Metzger D. Impaired adipogenesis and lipolysis in the mouse upon selective ablation of the retinoid X receptor αmediated by a tamoxifen-inducible chimeric Cre recombinase (Cre-ERT2) in adipocytes. Proc Natl Acad Sci USA. 2001;98:224–228. [PubMed]
25. Green H, Meuth M. An established pre-adipose cell line and its differentiation in culture. Cell. 1974;3:127–133. [PubMed]
26. Puigserver P, et al. A cold-inducible coactivator of nuclear receptors linked to adaptive thermogenesis. Cell. 1998;92:829–839. [PubMed]
27. He W, et al. Adipose-specific peroxisome proliferator-activated receptor γ knockout causes insulin resistance in fat and liver but not in muscle. Proc Natl Acad Sci USA. 2003;100:15712–15717. [PubMed]
28. Kubota N, et al. Pioglitazone ameliorates insulin resistance and diabetes by both adiponectin dependent and independent pathway. J Biol Chem. 2006;281:8748–8755. [PubMed]
29. Yang Q, et al. Serum retinol binding protein 4 contributes to insulin resistance in obesity and type 2 diabetes. Nature. 2005;436:356–362. [PubMed]
30. Lee S, Bacha F, Gungor N, Arslanian SA. Racial differences in adiponectin in youth: relationship to visceral fat and insulin sensitivity. Diabetes Care. 2006;29:51–56. [PubMed]
31. Nawrocki AR, et al. Mice lacking adiponectin show decreased hepatic insulin sensitivity and reduced responsiveness to peroxisome proliferator-activated receptor γ agonists. J Biol Chem. 2006;281:2654–2660. [PubMed]
32. Maeda K, et al. Adipocyte/macrophage fatty acid binding proteins control integrated metabolic responses in obesity and diabetes. Cell Metab. 2005;1:107–119. [PubMed]
33. Graham TE, et al. Retinol-binding protein 4 and insulin resistance in lean, obese, and diabetic subjects. N Engl J Med. 2006;354:2552–2563. [PubMed]
34. Hallsten K, et al. Rosiglitazone but not metformin enhances insulin- and exercise-stimulated skeletal muscle glucose uptake in patients with newly diagnosed type 2 diabetes. Diabetes. 2002;51:3479–3485. [PubMed]
35. Yamauchi T, et al. Inhibition of RXR and PPARγ ameliorates diet-induced obesity and type 2 diabetes. J Clin Invest. 2001;108:1001–1013. [PMC free article] [PubMed]
36. Cavasotto CN, et al. Determinants of retinoid X receptor transcriptional antagonism. J Med Chem. 2004;47:4360–4372. [PubMed]
37. Sell H, et al. Peroxisome proliferator-activated receptor γ agonism increases the capacity for sympathetically mediated thermogenesis in lean and ob/ob mice. Endocrinology. 2004;145:3925–3934. [PubMed]
38. Fliers E, et al. White adipose tissue: getting nervous. J Neuroendocrinol. 2003;15:1005–1010. [PubMed]
39. Ziouzenkova O, et al. Lipolysis of triglyceride-rich lipoproteins generates PPAR ligands: evidence for an antiinflammatory role for lipoprotein lipase. Proc Natl Acad Sci USA. 2003;100:2730–2735. [PubMed]
40. Xie Y, Lashuel HA, Miroy GJ, Dikler S, Kelly JW. Recombinant human retinol-binding protein refolding, native disulfide formation, and characterization. Protein Expr Purif. 1998;14:31–37. [PubMed]
41. Cogan U, Kopelman M, Mokady S, Shinitzky M. Binding affinities of retinol and related compounds to retinol binding proteins. Eur J Biochem. 1976;65:71–78. [PubMed]
42. Weir JB. New methods for calculating metabolic rate with special reference to protein metabolism. J Physiol (Lond) 1949;109:1–9. [PubMed]
43. Szatmari I, et al. PPARγ controls CD1d expression by turning on retinoic acid synthesis in developing human dendritic cells. J Exp Med. 2006;203:2351–2362. [PMC free article] [PubMed]