Naturally Occurring Eccentric Cleavage Products of Provitamin A β-Carotene Function as Antagonists of Retinoic Acid Receptors^*

Cell Lines

Cos-1 cells from ATCC (Rockville, MD) were cultured in DMEM supplemented with 10% FBS. Hep G2 cells were cultured in MEM supplemented with 10% FBS. Cells were maintained at 37 °C with 5% CO₂ in air.

Characterization of β-Apocarotenoids

Compounds were characterized by a mix of ¹H (400 MHz) and ¹³C (100 MHz unless noted) NMR spectroscopy, UV, and mass (electrospray ionization) spectrometry and HPLC analysis (Polaris C18 analytical column with 1 ml/min methanol:water of appropriate ratio, all compounds were determined to be at least 94% pure). Essential compound data follow, and procedures for their synthesis are provided below under “Results” and “Discussion.”

β-Cyclocitral

High resolution mass spectrum (HRMS) calculated for C₁₀H₁₆O (M+Na) 175.1099, observed 175.1101.

β-Cyclogeranic Acid

¹H NMR (CD₃COCD₃) δ 1.14 (6H, s), 1.45–1.49 (2H, m), 1.66–1.71 (2H, m), 2.01–2.04 (2H, m), 2.08 (3H, m); UV (ethanol) λ_max 295 nm (ϵ, 7,720); HRMS calculated for C₁₀H₁₆O₂ (M+Na) 191.1048, observed 191.1041.

β-Ionone

HRMS calculated for C₁₀H₁₆O (M+H) 193.1592; observed, 193.1579.

β-Ionylideneacetaldehyde

¹H NMR (CDCl₃), δ, 1.05 (6H, s), 1.46–1.55 (2H, m), 1.60–1.67 (2H, m), 1.74 (3H, s), 2.03–2.10 (2H, m), 2.39 (3H, s), 5.96 (1H, d, J = 8 Hz), 6.23 (1H, d, J = 16 Hz), 6.74 (1H, d, J = 16 Hz), 10.14 (1H, d, J = 8 Hz); UV (ethanol) λ_max, 272 nm (ϵ, 14,800); HRMS calculated for C₁₂H₂₂O (M+H) 219.1749, observed, 219.1737.

β-Ionylideneacetic Acid

¹H NMR (CDCl₃) δ, 1.07 (6H, s), 1.50–1.53 (2H, m), 1.63–1.69 (2H, m), 1.75 (3H, s), 2.06–2.09 (2H, m), 2.36 (3H, s), 5.85 (1H, s), 6.23 (1H, d, J = 16.2 Hz), 6.67 (1H, d, J = 16.2 Hz), 10.14 (1H, d, J = 8 Hz); UV (ethanol) λ_max 296 nm (ϵ, 20,000); HRMS calculated for C₁₂H₂₂O₂ (M+H) 235.1698, observed, 235.1688.

β-apo-13-Carotenone

¹H NMR (CDCl₃) δ 0.98 (6H, s), 1.41–1.44 (2H, m), 1.54–1.60 (2H, m), 1.67 (3H, s), 1.99–2.00 (2H, m), 2.01 (3H, s), 2.25 (3H, s), 6.08–6.15 (3H, m), 6.37 (1H, d, J = 16 Hz), 7.53 (1H, dd, J = 15, 11.9 Hz); ¹³C NMR (CDCl₃; 75 MHz) δ, 13.10, 19.14, 21.72, 27.63, 28.93, 33.14, 34.25, 39.59, 127.67, 129.27, 130.98, 131.27, 136.68, 137.48, 139.30, 145.55, 198.48; UV (ethanol) λ_max, 341 nm (ϵ, 25,300); HRMS calculated for C₁₈H₂₆O (M+Na) 281.1881, observed, 281.1859.

β-Apo-14′-carotenal

¹H NMR (CDCl₃) δ, 1.00 (6H, s), 1.40–1.44 (2H, m), 1.54–1.60 (2H, m), 1.70 (3H, s), 1.93–2.04 (2H, m), 2.00 (3H, s), 2.07 (3H, s), 6.10–6.44 (6H, m), 6.90 (1H, dd, J = 14.9, 11.5 Hz), 7.50 (1H, dd, J = 14.9, 11.5 Hz), 9.58 (1H, d, J = 7.9 Hz); ¹³C NMR (CDCl₃) δ, 13.42, 19.26, 21.61, 27.57, 29.19, 33.55, 34.46, 40.10, 127.59, 128.06, 130.12, 130.95, 133.89, 136.31, 136.63, 137.43, 137.63, 137.99, 143.71, 190.39; UV (ethanol) λ_max 402 nm (ϵ, 55,000); HRMS calculated for C₂₂H₃₀O (M+Na) 333.2194, observed, 333.2190.

β-Apo-14′-carotenoic Acid

¹H NMR (CDCl₃) δ, 1.05 (6H, s), 1.46–1.54 (2H, m), 1.61–1.71 (2H, m), 1.77 (3H, s), 2.06 (3H, s), 2.06–2.10 (2H, m), 2.13 (3H, s), 5.92 (1H, d, J = 15.0 Hz), 6.13–6.32 (4H, m), 6.39 (1H, d, J = 15.1 Hz), 6.90 (1H, dd, J = 15.0, 11.5 Hz), 7.88 (1H, dd, J = 15.1, 11.5 Hz); ¹³C NMR (CDCl₃) δ, 13.65, 19.63, 22.19, 29.38, 33.53, 34.68, 40.01, 119.46, 128.58, 129.57, 130.37, 136.13, 137.81, 138.16, 139.23, 143.19, 146.14, 173.21 UV (ethanol) λ_max, 378 nm (ϵ, 52,200); HRMS calculated for C₂₂H₃₀O₂ (M+Na) 349.2143, observed, 349.2136.

β-Apo-12′-carotenal

¹H NMR (CDCl₃) δ, 1.00 (6H, s), 1.45–1.43 (2H, m), 1.65–1.57 (4H, m), 1.69 (3H, s), 1.85 (3H, s), 1.99 (3H, s), 2.02 (3H, s), 6.14 (3H, m), 6.19 (1H, dd, J = 13.8, 14.2 Hz), 6.36 (1H, d, J = 11.9 Hz), 6.68 (1H, d, J = 11.8 Hz), 6.77 (1H, dd, J = 14.2, 13.8 Hz), 6.92 (1H, dd, J = 13.8, 11.6 Hz), 7.00 (1H, d, J = 11.6 Hz), 9.42 (1H, s); ¹³C NMR CDCl₃) δ, 10.12, 12.74, 13.27, 20.24, 22.18, 25.49, 29.38, 33.53, 34.68, 40.02, 127.31, 127.63, 127.66, 130.47, 130.81, 136.31, 136.62, 137.53, 137.67, 137.89, 138.22, 141.38, 149.47, 194.38; UV (methanol) λ_max, 426 nm (ϵ, 75,600); HRMS calculated for C₂₅H₃₄O (M+H) 351.2682, observed 351.2689.

β-Apo-12′-carotenoic Acid

¹H NMR (CDCl₃) δ, 1.05 (6H, s), 1.47–1.50 (2H, m), 1.60–1.65 (2H, m), 1.74 (3H, s), 2.00–2.03 (11H, m), 6.14–6.22 (3H, m), 6.27 (1H, d, J = 12.2 Hz), 6.36 (1H, d, J = 14.9 Hz), 6.52 (1H, dd, J = 13.7, 12.4 Hz), 6.74 (1H, dd, J = 14.9, 11.5 Hz), 6.93 (1H, d, J = 12.2 Hz), 7.43 (1H, d, J = 11.5 Hz); ¹³C NMR (CDCl₃) δ, 12.84, 13.25, 19.65, 22.19, 29.38, 33.53, 34.68, 40.03, 125.53, 127.51, 127.94, 128.04, 130.11, 130.86, 131.39, 136.92, 137.38, 137.98, 138.24, 141.09, 141.37, 173.92; UV (ethanol) λ_max, 407 nm (ϵ, 67,000); HRMS calculated for C₂₅H₃₄O₂ (M+Na) 389.2457, observed, 389.2463.

β-Apo-10′-carotenal

¹H NMR (CD₃COCD₃) δ, 1.01 (6H, s), 1.46–1.43 (2H, m), 1.62–1.58 (2H, m), 1.68 (3H, s), 1.96 (6H, s), 1.87 (3H, s), 2.02 (2H, m), 6.09 (1H dd, J = 9.4, 7.7 Hz), 6.19 (2H, m), 6.35 (2H, m), 6.74 (2H, m), 6.86 (1H, dd, J = 11.5, 11.5 Hz), 7.27 (1H, d, J = 7.6 Hz), 7.33–7.30 (2H, m), 9.55 (1H, d, J = 7.7 Hz); ¹³C NMR (CD₃COCD₃) δ, 11.95, 12.08, 21.22, 28.54, 28.89, 29.08, 29.27, 29.47, 29.66, 32.86, 39.57, 126.56, 126.85, 127.04, 127.30, 128.15, 129.08, 131.14, 132.00, 135.32, 137.11, 137.98, 140.97, 156.16, 192.91; UV (hexane) λ_max, 438 nm (ϵ, 73,100); HRMS calculated for C₂₇H₃₆O (M+Na) 399.2664, observed, 399.2664.

β-Apo-10′-carotenoic Acid

¹H NMR (CDCl₃) δ, 1.00 (6H, s), 1.42–1.45 (2H, m), 1.56–1.62 (2H, m), 1.69 (3H, s), 1.91 (3H, s), 1.98–2.00 (8H, m), 5.82 (1H, d, J = 15.3 Hz), 6.09–6.19 (3H, m), 6.36 (1H, d, J = 14.9 Hz), 6.22 (1H, d, J = 11.7 Hz), 6.31 (1H, d, J = 14.9 Hz), 6.53 (1H, dd, J = 13.2, 11.9, Hz), 6.70 (1H, dd, J = 12.4, 11.7 Hz), 6.78 (1H, d, J = 12.4 Hz) 7.40 (1H d, J = 15.3 Hz); ¹³C NMR (CDCl₃) δ, 12.98, 13.24, 13.39, 18.81, 19.66, 22.20, 29.39, 31.36, 33.54, 34.69, 40.03, 115.64, 126.99, 127.74, 128.74, 128.94, 130.05, 130.99, 131.97, 133.67, 134.94, 137.13, 137.55, 138.03, 138.26, 139.88, 140.89, 151.46, 173.20; UV (ethanol) λ_max, 407 nm (ϵ, 67,000). HRMS calculated for C₂₇H₃₆O₂ (M+Na) 415.2613, observed, 415.2608.

β-Apo-8′-carotenoic Acid

¹H NMR (CDCl₃) δ, 1.07 (6H, s), 1.49–1.52 (2H, m), 1.60–1.69 (2H, m), 1.76 (3H, s), 1.94 (3H, s), 2.02–2.07 (2H, m), 2.02 (3H, s), 2.04 (3H, s), 6.15–6.23 (3H, m), 6.31 (1H, d, J = 11.6 Hz), 6.40 (1H, dd, J = 14.8 Hz), 6.49 (1H, d, J = 11.6 Hz) 6.64–6.84 (5H, m), 6.98 (1H, dd, J = 10.7 Hz); UV (ethanol) λ_max, 441 nm (ϵ, 108,700); HRMS calculated for C₃₀H₄₀O₂ (M+Na) 455.2926, observed, 455.2944.

Retinoids and Other Materials

All-trans-retinoic acid, retinal, retinyl acetate, and 13-cis-retinoic acid were from Sigma. 9-cis-RA was obtained from Enzo Life Sciences. RARβ/γ selective antagonist CD 2665 was from Tocris Bioscience (Ellisville, MO). All-trans-[³H]RA (50.8 Ci/mmol), and 9-cis-[³H]RA (52.9 Ci/mmol) were purchased from American Radiolabeled Chemicals (St. Louis, MO), and PerkinElmer Life Sciences, respectively. Recombinant proteins including RARα, RARβ, RARγ, and RXRα were from Active Motif (Carlsbad, CA). Probes and primers for TaqMan assays were from Applied Biosystems (Carlsbad, CA).

Transactivation Assays

Cos-1 cells were transfected with plasmids in serum-free medium using Lipofectamine 2000 reagent (Invitrogen). Reporter vectors used were Renilla luciferase under the control of the thymidine kinase promoter and firefly luciferase under the control of RARE. Plasmids with cDNAs for retinoic acid receptors α, β, and γ were cotransfected in individual experiments. Four hours after transfection, cells were treated with test compounds that were dissolved in ethanol or with 0.1% ethanol alone for an additional 24 h. Cell lysates were then assayed for luciferase activities using a GloMax 96 microplate luminometer (Promega) and the Dual-Luciferase reporter assay system (Promega). For each experiment, the firefly luciferase activity (experimental reporter) was normalized to Renilla luciferase (control reporter) activity. The change in normalized firefly luciferase activity was calculated relative to that for cells that were transfected with vehicle (ethanol), which was set as 1. In some experiments, we used the human RARγ stably transfected reporter cell system from Indigo Biosciences, Inc. (State College, PA) according to the manufacturer's directions. These cells show a much more robust response of ATRA-induced reporter gene expression than the transiently transfected cells.

Ligand-binding Assays

Purified human recombinant retinoid receptors were used at 50 fmol with a non-retinoid-binding protein (keyhole limpet hemocyanin) to maintain ∼0.1 mg protein/ml to prevent loss of receptor protein and ligand during the course of binding experiments. For binding analyses, proteins and ligands were incubated at 4 °C for 16 h. For equilibrium saturation binding assays, proteins were incubated with various concentrations of all-trans-[³H]RA for RARs and 9-cis-[³H]RA for RXRα and in the absence or presence of a 100-fold excess of unlabeled compound to determine nonspecific binding. Specific binding was determined by a hydroxyapatite-binding assay (14). The quantity of labeled compound bound was determined by liquid scintillation counting. For competitive binding assays, a constant amount of all-trans-[³H]RA (5 nm) for RARs or 9-cis-[³H]RA (10 nm) for RXRα was incubated with test compounds (from 10⁻⁶ m to 10⁻¹⁰ m). Apparent dissociation constants (K_d) and inhibition constants (K_i) were determined using GraphPad Prism (version 5.0) with non-linear regression.

Gene Expression Assays

Hep G2 cells were treated with test compounds for 4 h, and total RNA was isolated from each well and subjected to reverse transcription. Quantitative PCR was carried out using TaqMan chemistry for human CYP26A1 (Hs01084852_g1), and human RARβ (Hs00977137_m1)) as target genes and human GAPDH as a housekeeping gene. The comparative C_t method (ΔΔC_t) was used to quantify the results obtained by real-time RT-PCR.

Molecular Modeling

Structures were displayed and modeled on a Silicon Graphics O₂ running Sybyl (version 7.1, Tripos). Structures were minimized using Sybyl's Maximin method and docked to the RARβ ligand binding domain crystal structure (Protein Data Bank code 1XAP) using the available FlexX routine (15).

Quantitative HPLC/MS Analysis of β-Apo-13-carotenone in Human Plasma

Blood plasma was purchased from Innovative Research (Novi, MI). Fresh blood plasma was prepared from six individuals over the course of 2 weeks, shipped on dry ice to Ohio State University, and stored at −80 °C before extraction. For each individual, 1 ml of plasma was split among five glass tubes (200 μl per tube). Ethanol (1 ml) was added to each tube, and the samples were sonicated using an Ultrasonic Dismembrator (model 150E, Fisher Scientific). Hexanes (10 ml) were added to each tube, and the mixture was sonicated again. The tubes were centrifuged for 5 min at 300 × g to facilitate phase separation. The upper organic layers from each vial were pooled into a single vial, and the extraction with hexanes was repeated a second time. The pooled extracts were dried under nitrogen gas. The samples were redissolved in 200 μl of 1:1 methyl t-butyl ether/methanol and analyzed using HPLC-MS/MS.

HPLC was conducted on a C30 column (4.6 × 150 mm, s5, YMC) in reversed phase with (A) methanol/0.1% formic acid (80/20 v/v) and (B) methanol/0.1% formic acid/methyl t-butyl ether (20/2/78 v/v/v) as mobile phase solvents at 1.8 ml/min and 35 °C (Agilent 1200 SL, Agilent Technologies, Santa Clara, CA). Eluent was introduced to a triple quadrupole mass spectrometer (QTrap 5500, AB Sciex, Concord, Canada) via an atmospheric pressure chemical ionization probe operated in positive ion mode. Three MS/MS transitions monitored for β-apo-13-carotenone were m/z 259 > 175, 119, and 69 at collision energies of 21, 31, and 27 eV, respectively, with 150-ms dwell times and using nitrogen as the collision-activated dissociation gas. Other MS parameters included the following: 10 p.s.i. curtain gas, 425 °C, heated nebulizer temperature, 5 μA nebulizer current, 60 V declustering potential, 10 V entrance potential, and 11 V exit potential. Calibrating solutions of β-apo-13-carotenone were prepared in 1:1 methyl t-butyl ether/methanol with concentrations based on an extinction of 25,300 at 368 nm in ethanol. The dried residue of serum extract was redissolved in 100 μl of methyl t-butyl ether, and then 100 μl of methanol was added with mixing before centrifuging prior to injection (10 μl). For quantitation, the m/z 259 > 175 transition was used as it had superior signal:noise, and the other two transitions were used for qualitative purposes to confirm peak identity. The limit of detection for β-apo-13-carotenone was 280 pm. We have found that stock solutions of β-apo-13-carotenone in ethanol are stable for 2.5 years at −30° C. The compound is stable in plasmas frozen at −80° C for periods of at least 2 weeks.

Synthesis of β-Apocarotenoids

Although the occurrence and biological activity of selected β-apocarotenoids in mammals has been reported in various systems, progress has been hampered because many of the compounds are not available commercially. To comprehensively assess the activity of β-apocarotenoids, we first undertook to purify or synthesize all of the possible eccentric cleavage products of β-carotene (Fig. 1). Retinal, retinoic acid, and β-apo-8′-carotenal (Fluka) were used as obtained. β-Cyclocitral and β-ionone (both from Sigma-Aldrich) were purified by preparative TLC prior to use. Eleven of the β-apocarotenoids were synthesized by standard organic chemical transformations, including oxidation of aldehydes to acids (16), Wadsworth-Emmons and Wittig homologation, and reduction of esters to acids followed by oxidation to aldehydes (Fig. 2).

β-Apocarotenoids Antagonize Retinoic Acid-induced Expression of Reporter and Endogenous Retinoic Acid-responsive Genes

We first screened all of the compounds for their potential to activate RARα, RARβ, and RARγ, using Cos-1 cells transiently transfected with cDNAs for the individual RARs and with a RARE-luciferase reporter. None of the compounds was as effective as the pan-agonist, ATRA, in activating the RARs, and indeed, most showed no agonist activity at all (Fig. 3). The slight activity of some of the longer chain β-apocarotenoids at high concentration is consistent with previous reports (17–19). We then used the same transactivation assay to screen all of the compounds for their potential to antagonize the ATRA-induced activation of the individual RARs by treating the cells with maximally effective doses of ATRA and equimolar concentrations of the β-apocarotenoids (Fig. 4). Although the shorter products of the eccentric cleavage of β-carotene had little or no effect on ATRA-induced transactivation, β-apo-10′-carotenoic acid and β-apo-12′-carotenoic acid both led to 40–50% inhibition of ATRA-induced activation of all three RAR isoforms. Even more striking inhibition was observed for the products of the “d” cleavage (Fig. 1) of β-carotene (viz. β-apo-14′-carotenal, β-apo-14′-carotenoic acid, and β-apo-13-carotenone), with the greatest inhibitory activity being displayed by β-apo-13-carotenone. Thus, it appeared that these five β-apocarotenoids could be functioning as RAR antagonists.

To characterize the antagonist function β-apo-13-carotenone more quantitatively, we used stably transfected RARγ reporter cells. Cells were treated with ATRA in a concentration range of 0.5 nm to 3 μm in the absence or presence of fixed concentrations (1, 10, or 100 nm) of β-apo-13-carotenone (Fig. 5A). We observed a progressive shift in the ATRA dose-response curve with increasing concentrations of β-apo-13-carotenone in the nanomolar range. Higher concentrations of ATRA were able to overcome inhibition by β-apo-13-carotenone, suggesting direct competition between the two compounds for binding. This suggestion is supported by the results of competitive radioligand binding assays and molecular modeling discussed below.

We then asked whether β-apo-13-carotenone and the β-apo-14′-carotenoids would antagonize the ATRA-induced transcription of endogenous genes. For these experiments, we treated Hep G2 cells in culture with ATRA for 4 h and measured the mRNA levels for RARβ and cytochrome P450-26A1 (CYP26A1). Both of these genes have canonical RAREs in their promoters, and their transcription is directly up-regulated by ATRA treatment (20, 21). As shown in Fig. 5B, treatment with 10 nm ATRA led to 9-fold induction in RARβ mRNA levels and ∼20-fold increases on CYP26A1 mRNA levels. Treatment with 10 nm β-carotene, β-apo-13-carotenone, or other β-apocarotenoids expectedly did not markedly induce expression of either gene (data not shown). However, co-treatment with ATRA and β-apo-13-carotenone or the β-apo-14′-carotenoids led to marked inhibition of the ATRA-induced gene expression (Fig. 5B). Importantly, the inhibition by β-apo-13-carotenone was greater than that for β-apo-14′-carotenal or β-apo-14′-carotenoic acid. This is in keeping with the greater affinity of β-apo-13-carotenone for the RARs than that of the β-apo-14′-carotenoids. Co-treatment of the cells with ATRA and the parent compound (β-carotene) or with a β-apocarotenoid that does not antagonize ATRA-induced transactivation (viz. β-ionylideneacetic acid) had no effect on ATRA-induced gene expression in HepG2 cells.

β-Apocarotenoids Bind to Ligand-binding Site of RARs with High Affinity

Retinoic acid receptors (like other type II nuclear receptors) function in the regulation of endogenous gene expression by binding as heterodimers with retinoid X receptors (7, 22). The heterodimers bind to specific response elements (RAREs and RXREs) in the promoter regions of genes via their respective DNA-binding domains. In the unliganded state, the transcription factor complex binds to nuclear co-repressors, and transcription is repressed. Binding of ATRA (or other agonist) to the ligand-binding domain of RAR induces a conformational change in the RAR (at helix 12), and this leads to co-repressors dissociating from the receptor and the unmasking of a co-activator binding site (22). In the case of the RAR-RXR heterodimer, the binding of an agonist to RXR alone does not lead to activation, but the binding of an RXR agonist in concert with an RAR agonist leads to supra-activation of transcription (22). We wanted to know whether the β-apocarotenoids that demonstrated an antagonist “activity” in the cell-based assays did so by directly binding to the RAR ligand-binding domain and competing for ATRA binding. Thus, we conducted radioligand binding assays using purified recombinant RARα, RARβ, and RARγ and tritium-labeled ATRA in the presence of increasing concentrations of unlabeled ATRA (as a positive control), CD 2665 (a synthetic antagonist of RARβ/γ known to bind to the ligand binding site), retinyl acetate (a retinoid that does not bind and used as a negative control), and the selected β-apocarotenoids (Fig. 6A). For all three receptors, the three β-apocarotenoids with the highest antagonist activity competed for ATRA binding. In fact, β-apo-13-carotenone displayed the same affinity for the RARs as ATRA itself (i.e. 2–6 nm) (Fig. 6B). The affinity of binding for β-apo-14′-carotenal and β-apo-14′-carotenoic acid to RARs was in the 15–60 nm range, whereas those of β-apo-10′- and β-apo-12′-carotenoic acids were >300 nm.

The high affinity binding of β-apo-13-carotenone to the ligand binding site of RARβ was also demonstrated by the results of molecular modeling studies. We displayed the crystal structure of RARβ with the retinoid agonist TTNPB in the binding site (Protein Data Bank code 1XAP), “extracted” the ligand computationally, and then attempted to dock both ATRA and β-apo-13-carotenone into the binding site using FlexX (15). Both molecules docked smoothly (Fig. 6C, left); indeed, the root mean square deviation for β-apo-13-carotenone was very slightly better than for ATRA. Moreover, the crystal conformation of TTNPB and the energy-minimized and then docked conformations of ATRA and β-apo-13-carotenone had nearly identical root mean square deviation values (Fig. 6C, right).

The studies reported here demonstrate that the products of the oxidative eccentric cleavage of β-carotene at the C13–C14 double bond yields products that are antagonists of RARs and that the most active molecule in this regard is β-apo-13-carotenone. We previously showed that this compound was also the most effective β-apocarotenoid in inhibiting the transactivation of RXRα by its agonist, 9-cis-retinoic acid, and molecular modeling studies demonstrated that it could potentially bind to the transcriptionally silent tetramer of RXRα (23). Therefore, we tested the three β-apocarotenoids resulting from cleavage of the C13–C14 double bond of β-carotene for their binding to purified recombinant RXRα (Fig. 7). The β-apo-13-carotenone competed for 9-cis-retinoic acid binding with an affinity (7–8 nm) identical to 9-cis-retinoic acid itself. The affinities of β-apo-14′-carotenal and β-apo-14′-carotenoic acid for RXRα were >250 nm in keeping with their lack of effect on inhibiting RXR transactivation. Given that a number of nuclear receptors form heterodimers with RXR and that ligand binding (either agonist or antagonist) to the RXR leads to modulation of the transcriptional activity of the heterodiomers (22), these eccentric cleavage products of β-carotene could have complex global effects on gene expression.

Most Potent β-Apocarotenoid Antagonist of RARs, β-Apo-13-carotenone, Is Found in Human Plasma at Concentrations That Are Biologically Significant

Although the mechanisms responsible for the formation of the eccentric cleavage products of β-carotene in mammals are not fully known, it is clear that some of the long chain β-apocarotenals (e.g. 8′, 10′, 12′, 14′) are found in the plasma of humans (12) and experimental animals (13) and that these are increased under conditions of oxidative stress and high dietary doses of β-carotene (24). We have also found that all of these β-apocarotenals and, specifically, β-apo-13-carotenone are present in fresh cantaloupe and orange-fleshed melons (11); thus, these compounds may be absorbed directly from the diet.

To further establish the relevance of the potent antagonist activity of β-apo-13-carotenone on retinoid receptors, we developed sensitive HPLC/MS procedures for its detection in human plasma. Fig. 8 shows the analysis of human plasma and authentic standards by HPLC/MS. We used multiple reaction monitoring to ensure the specificity and sensitivity of the assay. We then analyzed the plasmas of six free-living individuals and found the plasma concentration of β-apo-13-carotenone to be 3.8 ± 0.6 nm. Importantly, this is in the range of normal concentrations of retinoic acid in plasma and approximately the same as the binding constant of the compound for the retinoid receptors. This would suggest that β-apo-13-carotenone can function at physiological concentrations as an endogenous modulator of retinoid signaling in humans.

Conclusions and Implications

Our results demonstrate that β-carotene can generate both RAR agonists (ATRA) and RAR antagonists (β-apo-14′-carotenal and β-apo-13-carotenone) depending on the extent of cleavage at the central C15–C15′ double bond or the C13–14 double bond, respectively. These findings may have implications for the unexpected and negative effects of high doses of β-carotene in human clinical trials of cancer prevention (25). An example is the now famous CARET trial, which, based on observational epidemiology, explored whether supplemental β-carotene would decrease incidence of lung cancer in a highly susceptible population, namely smokers and asbestos workers (26, 27). Surprisingly, the supplemented subjects had a higher incidence of disease, and the trial had to be halted early. It was apparent that the doses of β-carotene used in the trial (30 mg/day) were much higher than the range of normal dietary intakes associated with a decreased risk of disease in the observational studies (25). The possible mechanisms involved were explored in elegant studies employing a novel animal model, the smoking ferret (24, 28). These studies revealed that under conditions of high dietary β-carotene and the oxidative stress of smoking, there was a clear increase in preneoplastic lung cancer lesions in the animals. The authors concluded that oxidative stress led to increased eccentric cleavage of β-carotene and that the mixture of cleavage products led to disruption of retinoid signaling via indirect mechanisms. The present work demonstrates that specific β-apocarotenoids exert an anti-vitamin A activity by directly interacting with RARs as high affinity antagonists. Our analyses of both β-carotene-containing animal diets and fruits containing β-carotene suggest that any dietary source of β-carotene also contains β-apocarotenoids. It may also be useful to consider these findings in attempts to alleviate vitamin A deficiency in humans through the biofortification of crops with high levels of β-carotene.