Docosahexaenoic Acid, a Ligand for the Retinoid X Receptor in Mouse Brain

See allHide authors and affiliations

Science  15 Dec 2000:
Vol. 290, Issue 5499, pp. 2140-2144
DOI: 10.1126/science.290.5499.2140


The retinoid X receptor (RXR) is a nuclear receptor that functions as a ligand-activated transcription factor. Little is known about the ligands that activate RXR in vivo. Here, we identified a factor in brain tissue from adult mice that activates RXR in cell-based assays. Purification and analysis of the factor by mass spectrometry revealed that it is docosahexaenoic acid (DHA), a long-chain polyunsaturated fatty acid that is highly enriched in the adult mammalian brain. Previous work has shown that DHA is essential for brain maturation, and deficiency of DHA in both rodents and humans leads to impaired spatial learning and other abnormalities. These data suggest that DHA may influence neural function through activation of an RXR signaling pathway.

The nuclear hormone receptors are ligand-activated transcription factors [reviewed in (1)]. Included in this family are the so-called “orphan receptors,” whose ligands have not been identified [reviewed in (1, 2)]. By screening chemical libraries, progress has been made in identifying ligands for these receptors [for examples, see (3–8)]; however, in most cases endogenous ligands have remained elusive.

RXR is an obligatory component of a large number of nuclear receptor heterodimers and is activated in vitro by the vitamin A metabolite 9-cis retinoic acid (9-cis RA), which binds with high affinity to the RXR ligand binding domain (9,10); however, 9-cis RA acid has been difficult to detect in vivo (11). Nonetheless, because numerous studies have demonstrated striking synergism in biological responses in vitro and in vivo when both partners of the RXR-RAR heterodimer are activated by their respective ligand [for examples, see (12–14)], it seems likely that ligand-induced activation of RXR does occur in vivo. This conclusion has been corroborated by experiments with transgenic mice (15,16).

To search for endogenous ligands that activate retinoid receptors RAR and RXR, we used a cell-based reporter gene assay. Human chorion carcinoma JEG-3 cells were co-transfected with an upstream activating sequence (UAS) luciferase reporter construct and either GAL4-RAR or GAL4-RXR expression vectors encoding the yeast GAL4 DNA-binding domain fused to the ligand-binding domains of human (h) RARα and hRXRα, respectively (17). The cells were then cultured with serum-free media that had been incubated overnight with various tissues from embryonic or adult mice (18). Consistent with previous studies, we found that media conditioned with brain tissue from embryonic day 13.5 (E13.5) embryos activated RAR but not RXR (Fig. 1A) (11). In contrast, media conditioned with adult brain tissue activated RXR but not RAR (Fig. 1A). The RXR-activating factor(s) were present in hippocampus, striatum, motor cortex, and cerebellum, but not in spinal cord (19).

Figure 1

RXR-specific activation by factor(s) enriched in adult brain. (A) JEG-3 cells were transfected with a luciferase reporter plasmid containing four UAS binding sites (UASx4-tk-luc) and either GAL4-RARα or GAL4-RXRα expression vectors, followed by treatment with conditioned medium as depicted. Values are shown as fold induction after normalization to β-galactosidase (β-gal). (B) JEG-3 cells transfected with GAL4-RXR expression vector and UASx4-tk-luc reporter were incubated with brain-conditioned medium (bcm) in the presence or absence of RXR-specific (LG849) or RAR-specific antagonists (RO41-5253). Values are shown as relative light units (RLU) after normalization to β-gal. (C) JEG-3 cells were transfected with GAL4-Nurr1 expression vector alone or in combination with RXR expression vector. The cells were then incubated with brain conditioned medium, with or without an RXR-specific antagonist [see (B)]. Values are shown as RLU after normalization to β-gal.

Two lines of evidence indicated that the adult brain-conditioned medium directly activated RXR rather than an RXR heterodimerizing partner expressed in JEG-3 cells. First, an RXR-specific antagonist (LG849) (15) abolished reporter gene expression, and an RAR-specific antagonist had no effect (Fig. 1B). Second, the brain-conditioned medium activated RXR when RXR heterodimerized with the orphan receptor Nurr1 (20, 21). Thus, when GAL4-Nurr1 was expressed as a bait, potent activation was observed only in cells co-transfected with RXR expression vector (Fig. 1C). Again, activation was abolished by the RXR-specific antagonist LG849.

To identify the RXR-activating factor(s), we extracted the brain-conditioned medium with hexane (22) and found that the activity was present in the hexane extracts, consistent with the possibility that the activating factor is a lipophilic molecule (Fig. 2A). Acidification of the medium with hydrochloric acid increased the yield, indicating that the active substance is a lipophilic, negatively charged molecule (Fig. 2A).

Figure 2

Purification of activating factor from brain-conditioned medium. (A) Brain-conditioned medium was extracted with hexane in the presence (left) or absence (right) of 0.1 M hydrochloric acid (final concentration). Evaporized hexane extracts were redissolved in a small volume of ethanol, and aliquots were added to cultured JEG-3 cells transfected with UASx4-tk-luc reporter and GAL4-Nurr1 and RXR expression vectors. Values are depicted as RLU after normalization to β-gal. (B) Redissolved hexane extract was separated by normal-phase HPLC, and individual fractions were tested on cells transfected as in (A). (C) Active fractions from (B) were pooled and fractionated by reversed-phase HPLC. Individual fractions were tested as in (B). Values in (B) and (C) are shown as fold induction after normalization to β-gal. Time corresponds to time of elution after injection of sample.

We purified the activity to homogeneity by two cycles of high-performance liquid chromatography (HPLC) (Fig. 2, B and C) (22). Redissolved hexane extract was initially fractionated by normal-phase HPLC, and individual fractions were assayed in transfected JEG-3 cells. The fractions containing the activity were then pooled and separated by reversed-phase HPLC. The peak fractions were analyzed by negative-ion nano-electrospray (nano-ES) mass spectrometry (Fig. 3A) (23). The spectrum of the active fraction was dominated by an intense ion at a mass-to-charge ratio (m/z) of 327.2; there were minor peaks at m/z of 283.2, 339.2, and 655.5. From this analysis, we concluded that the molecular formula of the active compound is C22H32O2, similar to that of the polyunsaturated fatty acid (PUFA) docosahexaenoic acid (DHA). Collision-induced dissociation spectra recorded of the ion atm/z 327.2 (Fig. 3B) and the [M-H] ion ofcis-4,7,10,13,16,19-DHA confirmed this conclusion (Fig. 3C). Thus, the major purified component iscis-4,7,10,13,16,19-DHA.

Figure 3

Identification of DHA as a brain-derived RXR-activating factor. (A) Peak fractions from the reversed-phase HPLC were analyzed by mass spectrometry, recording negative-ion nano-ES spectra. The active fraction is dominated by a very intense peak at m/z 327.2. Minor peaks are observed atm/z 283.2, 339.2, and 655.5, the latter consisting of dimers of the m/z 327.2 ion. (B andC) Collision-induced dissociation spectra of the [M-H] ion of the ion at m/z of 327.2 from the active fraction (B) and the reference fatty acidcis-4,7,10,13,16,19-DHA (C) were recorded and compared. (D) DHA has a 22-carbon backbone with six double bonds at carbons 4, 7, 10, 13, 16, and 19.

To confirm that DHA is the active compound, we tested DHA and several other PUFAs for their ability to activate a reporter gene responsive to wild-type RXR in human embryonic kidney 293T cells co-transfected with an expression vector for wild-type hRXRα. In dose-response studies, DHA was the most efficient activator with an EC50 (median effective concentration) at ∼50 to 100 μM (Fig. 4A) (17, 24). Docosatetraenoic acid (C22:4cis7,10,13,16), a PUFA that is structurally related to DHA, induced activation at concentrations approximately 5 to 10 times higher, whereas arachidonic acid (C20:4cis5,8,11,14) and oleic acid (C18:1cis9) induced activation only at the highest concentration tested (330 μM) (Fig. 4A). No activation was induced by erucic acid (C22:1cis13) (Fig. 4A), unsaturated C22, or various C18, C20, C21, C22, and C23 monounsaturated fatty acids, and only weak activation was induced by C18:2, C18:3, C22:3, and C22:5 (25). DHA was specific for RXR; DHA did not activate RAR, the thyroid hormone receptor (TR), or the vitamin D receptor (VDR) (Fig. 4A). Furthermore, activation by DHA was sensitive to mutations that alter the ligand-binding specificity of RXR (26), indicating that DHA activates RXR by a true ligand-binding mechanism (Fig. 4B). We next tested the maximal response compared to that of 9-cis RA using RXRα, RXRβ, and RXRγ. All three isotypes were activated by 150 μM DHA, but activation was only at 40 to 45% the level seen with 0.1 μM 9-cis RA (Fig. 4C). Interestingly, lower concentrations of DHA were 1.5 times more potent than 0.1 μM 9-cis RA in activating GAL4-Nurr1/RXR (Fig. 4D), suggesting that RXR's heterodimerization partner may significantly affect its responsiveness to DHA.

Figure 4

RXR activation by DHA and other FAs. (A) (Left) Various concentrations (1, 3, 10, 30, 100, 200, and 300 μM) of DHA (C22:6) (solid square), docosatetraenoic acid (C22:4) (open square), erucic acid (C22:1) (solid triangle), arachidonic acid (C20:4) (solid circle), and oleic acid (C18:1) (open circle) were titrated on 293T cells transfected with CMX-RXRα expression vector and Apo-A1-tk-luc reporter. Values are shown as fold induction after normalization to β-gal. (Right) 293T cells were transfected with UASx4-tk-luc reporter and either CMX-GAL4-RARα, CMX-GAL4-TRβ, CMX-GAL4-VDR, or CMX-GAL4-RXRα expression vectors, and treated with 100 μM DHA. The results are depicted as fold activation after normalization to β-gal. (B) 293T cells were transfected with Apo-A1-tk-luc reporter and either CMX-RXRα F313I or CMX-RXRα L436F expression vectors, and then treated with 9-cis RA (0.1 μM) or DHA (33 μM). Values are shown as fold activation after normalization to β-gal. (C) 293T cells were transfected with Apo-A1-tk-luc reporter and either CMX-RXRα, CMX-RXRβ, or CMX-RXRγ, followed by treatment with 9-cis RA (0.1 μM) or DHA (150 μM). Values are shown as percent of activation reached by 9-cis RA. (D) Various concentrations (1, 10, 50, 100, 200, and 300 μM) of DHA were titrated on 293T cells transfected with UASx4-tk-luc reporter and GAL4-Nurr1 and RXR expression vectors. Values are shown as percent of activation reached by 0.1 μM 9-cis RA.

To determine whether DHA is a ligand for RXR, we performed experiments with the steroid receptor coactivator 1 (SRC-1) (27). In transfected cells, DHA promoted interaction between RXR and SRC-1 (Fig. 5A). Moreover, DHA bound to the ligand-binding domain of RXR in vitro, as shown by a GST pull-down experiment in which the bacterially expressed ligand-binding domain of RXR interacted with [35S]SRC-1 in the presence but not in the absence of DHA (Fig. 5B) (28). A fusion of GST with the estrogen receptor α LBD (GST-ER) did not interact with SRC-1, either in the presence or absence of DHA (Fig. 5B). Consistent with the activation data, some interaction between RXR and SRC-1 was observed in the presence of docosatetraenoic acid (C22:4) and arachidonic acid (C20:4) (Fig. 5C). Oleic acid (C18:1cis9) and other tested fatty acids (C23:1cis14, C22:1cis13, C21:1cis12, C20:1cis11, C18:1cis15, C18:1cis13, C18:1cis11, and C18:1cis6) induced only very weak or no interaction (Fig. 5C). DHA induced SRC-1 interaction with RXR at concentrations consistent with the activation profile observed in transfected cells (Figs. 4A and 5D). Thus, DHA can activate RXR and bind directly to the ligand-binding domain of this nuclear receptor. Interestingly, it was recently reported that crystals of a mutated derivative of the RXR ligand-binding domain contain oleic acid (29), suggesting a structural basis for how fatty acids and their derivatives may interact with RXR in vivo.

Figure 5

DHA mediates RXR/SRC-1 interaction in vivo and in vitro. (A) 293T cells were transfected with UASx4-tk-luc reporter and either CMX-GAL4-RXR alone or in combination with VP16-SRC-1. The cells were incubated with serum-free medium or medium containing 33 μM DHA. Values are shown as RLU after normalization to β-gal. (B) Bacterially expressed GST fusion proteins of hRXRα and mouse ERα ligand-binding domains were bound to glutathione-Sepharose beads and incubated overnight with [35S]SRC-1 in the presence of vehicle, 1 μM SR11237, DHA, or 1 μM estradiol. (C) Bacterially expressed GST fusion protein of human RXRα ligand-binding domain was bound to glutathione-Sepharose beads and incubated overnight with [35S]SRC-1 protein in the presence of vehicle, 10 μM SR11237, or 100 μM or 1 mM concentrations of the indicated fatty acids. (D) Interaction of bacterially expressed GST fusion RXR and [35S]SRC-1 proteins was analyzed in the presence of DHA. Quantification was performed using a PhosphorImager.

RXR has also been shown to be activated in transfected cells by high concentrations of phytanic acid, a toxic chlorophyll metabolite, but this compound is eliminated from tissues in vivo (30). In contrast, DHA constitutes as much as 30 to 50% of total FAs in the mammalian brain, where it is predominantly associated with membrane phospholipids [reviewed in (31, 32)]. Thus, high levels of DHA are localized in the brain and the amount of free DHA could be significantly increased by release of a minor fraction of the membrane-bound DHA. Indeed, previous studies have demonstrated an autocrine phospholipase A2– and phospholipase C–mediated release of DHA from phospholipids, e.g., in response to neurotransmitters (33) and after brain injury (34). Alternatively, DHA may be released and activate RXR in neighboring cells in a paracrine fashion. Indeed, it has been suggested that brain astrocytes supply neurons with DHA (35).

DHA becomes highly enriched in the mammalian brain during late gestation and early postnatal development [reviewed in (31)] and is essential for brain maturation both in rodents and humans [reviewed in (31, 32, 36)]. High levels of DHA accumulate in the retina, a tissue that develops abnormally in RXRα knock-out mice (37). DHA deficiency in rats and humans results in impaired spatial learning as well as other abnormalities (38–40), and, interestingly, similar learning defects are seen in RXRγ-deficient mice (41). DHA and other PUFAs also exert profound effects on metabolism and energy homeostasis [reviewed in (42)]. DHA is a major constituent of nutrients rich in ω3 PUFAs and has beneficial effects on blood cholesterol levels and insulin sensitivity [reviewed in (42, 43)]. Synthetic RXR ligands have similar influences in mice and humans, suggesting that RXR activation may contribute to effects achieved by dietary intake of ω3 PUFAs (44–46). Several RXR heterodimerization partners such as peroxisome proliferator-activated receptors, the liver X receptors and farnesoid X receptor are essential for regulating energy and nutritional homeostasis in response to their respective ligands [reviewed in (47, 48)]. An intriguing possibility is that DHA modulates these and other regulatory events by binding to the RXR subunit of nuclear receptor heterodimers.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: thomas.perlmann{at}


View Abstract

Navigate This Article