Retinoid X receptor-α mediates (R )-flurbiprofen’s effect on the levels of Alzheimer’s β-amyloid

Alzheimer’s disease (AD), the most common form of dementia in the elderly population, is characterized by extracellular neuritic plaques composed of fibrillar β-amyloid (Aβ) peptide and intracellular neurofibrillary tangles containing hyperphosphorylated Tau (Lee et al. 2001; Hardy and Selkoe 2002). Multiple lines of evidence suggest that overproduction/aggregation of Aβ in the brain is a major cause of AD pathogenesis (Hardy and Selkoe 2002). Aβ is derived from its precursor protein β-amyloid precursor protein (APP) by sequential cleavages by β-secretase (BACE1) and γ-secretase (Cole and Vassar 2007; Zhang and Xu 2007). γ-cleavage of APP generates Aβ40 and Aβ42 peptides, of which the latter is more amyloidogenic and neurotoxic (Hardy and Selkoe 2002).

Epidemiological studies indicate that long-term use of non-steroidal anti-inflammatory drugs (NSAIDs) is associated with a reduced risk of AD development (Weggen et al. 2001; Eriksen et al. 2003). Studies in both cells and animals suggest that certain NSAIDs can specifically reduce Aβ42 levels independent of cyclooxygenase (COX) activity (Weggen et al. 2001; Eriksen et al. 2003), probably through direct interaction with, and inhibition of, the γ-secretase complex (Kukar and Golde 2008). In addition, NSAIDs may suppress the inflammatory reaction (i.e. microglia and astrocyte activation) associated with Aβ plaques (Lim et al. 2000; Sastre et al. 2008). For example, indomethacin, a COX-1 and COX-2 nonselective inhibitor, can significantly reduce microglia activation surrounding the intraparenchymal Aβ deposit that develops after an Aβ infusion in the lateral ventricle of rats (Netland et al. 1998).

Among NSAIDs, flurbiprofen raises particular interest because of its multiple interactions with key AD hallmarks: the R-enantiomer of flurbiprofen racemate, (R)-flurbiprofen, and its nitric oxide-releasing derivatives, HCT1026 and NCX2216, have been found effective in reducing AD amyloid pathology. Moreover, HCT1026 and NCX2216 can influence the cellular component of neuroinflammation (i.e., microglia activation) in some experimental settings (Gasparini et al. 2005). Chronic administration of (R)-flurbiprofen has also been found to attenuate learning impairments in AD mouse models (Eriksen et al. 2003; Kukar et al. 2007). However, the result from a phase III clinical trial using (R)-flurbiprofen for AD treatment by Myriad Genetics has recently turned out negative. The detailed mechanism underlying the (R)-flurbiprofen–mediated effect on Aβ generation/AD risk remains largely undetermined.

Several NSAIDs, including indomethacin and ibuprofen, can activate peroxisome proliferator-activated receptor γ (PPARγ), whose activation has been proposed to repress some microglial functions and contribute to the delay of AD progression (Bernardo et al. 2005). This suggests that PPARγ and related nuclear receptors might be involved in an NSAID-mediated beneficial effect on AD. Indeed, it has been found that activation of liver X receptor (LXR) can stimulate Aβ degradation (Xiong et al. 2008). The levels of several nuclear receptors are altered in AD brains and their agonists can reduce β- and γ-secretase activities responsible for Aβ generation (Jiang et al. 2008). However, while HCT1026, a flurbiprofen derivative that releases nitric oxide, activates PPARγ, flurbiprofen itself does not possess such PPARγ agonistic properties (Bernardo et al. 2005). Recently we found that the R-enantiomer of one NSAID, (R)-etodolac, can bind to the retinoid X receptor RXRα, another nuclear receptor family member (Kolluri et al. 2005). Hence, in the present study, we investigated whether RXRα is involved in (R)-flurbiprofen –mediated Aβ generation.

Materials and methods

Results

(R)-flurbiprofen reduces Aβ secretion but increases the level of intracellular Aβ

We first studied the effects of (R)-flurbiprofen treatment on Aβ generation. As expected and consistent with previous studies, we found that in cells treated with (R)-flurbiprofen, the levels of secreted (extracellular) Aβ species (including Aβ40, 42 and 38) were all significantly reduced compared to those in untreated cells (Fig. 1a). Surprisingly, we found that (R)-flurbiprofen treatment significantly increased the levels of intracellular Aβ species (Fig. 1a). Immunoprecipitation-western blot analyses revealed that (R)-flurbiprofen treatment significantly reduced the level of extracellular Aβ but increased the level of intracellular Aβ (Fig. 1b). In addition, cells treated with (R)-flurbiprofen showed much stronger immunostaining of Aβ40 within the cell body than the control (Fig. 1c). Biotinylation analyses showed that cell surface levels of APP were not affected by (R)-flurbiprofen treatment (Fig. 1b), implying that the effects of (R)-flurbiprofen treatment on extracellular Aβ and intracellular Aβ levels were not attributable to altered trafficking of APP to the cell surface.

RXRα regulates Aβ secretion

To study whether RXRα regulates Aβ generation, we either over-expressed or down-regulated (by RNAi) the level of RXRα in N2a695 cells. As shown in Fig. 2(a), down-regulation of RXRα significantly increased the level of secreted (extracellular) Aβ, whereas over-expression of RXRα slightly reduced Aβ secretion (Fig. 2a). However, the level of cell surface APP was not affected by RXRα modulation (Fig. 2a). In addition, RXRα had no effect on the levels of total APP, BACE1, and PS1, an important component of the γ-secretase complex (Fig. 2a). We also found that RXRα did not affect the in vitro activity of BACE1 (Fig. 2b). As Notch is another major substrate of γ-secretase, we investigated Notch cleavage and found that the cleavage of Notch by γ-secretase for Notch intracellular domain generation was not affected by RXRα (Fig. 2c). Moreover, we found that RXRα had no effect on degradation of extracellular Aβ (Fig. 2d).

RXRα is required for (R)-flurbiprofen-mediated Aβ generation

We next studied whether RXRα was involved in (R)-flurbiprofen-mediated Aβ generation. We found that (R)-flurbiprofen treatment dramatically reduced the levels of extracellular Aβ in both control cells (Fig. 3a, lanes 3 vs. 4) and RXRα-over-expressing cells (Fig. 3a, lanes 5 vs. 6). However, (R)-flurbiprofen treatment did not reduce Aβ levels in RXRα-knock-down cells (Fig. 3a, lanes 1 vs. 2). ELISA results showed that the levels of extracellular Aβ40 and 42 were significantly reduced, and the levels of intracellular Aβ40 and 42 were significantly increased by (R)-flurbiprofen treatment in the presence of RXRα (control and over-expressing), but not when RXRα was knocked down (Fig. 3b and c). To exclude the possibility that (R)-flurbiprofen still reduces Aβ secretion in RXRα knock-down cells but the effect is overridden by dramatically increased Aβ secretion upon RXRα knock-down, we also treated cells with TNFα/IFNγ, which has been found to increase Aβ secretion (Blasko et al. 1999). We found that although TNFα/IFNγ treatment significantly increased the levels of Aβ40 and 42 secretion, beyond that of RXRα knock-down cells, additional (R)-flurbiprofen treatment was still able to reduce secreted Aβ40 and 42 levels in TNFα/IFNγ-treated cells (Fig. 3b).

(R)-flurbiprofen interferes with the binding of 9-cis-RA to RXRα

To investigate the correlation between (R)-flurbiprofen and RXRα, we carried out an RXRα ligand competition assay with [³H]9-cis-RA, which is known to directly bind the RXRα homodimer LBD (Zhang et al. 1992; Kolluri et al. 2005). Our results showed that the binding between [³H]9-cis-RA and RXRα was competitively inhibited by both unlabeled (R)-flurbiprofen and 9-cis-RA. (R)-flurbiprofen inhibited the binding of [³H]9-cis-RA to RXRα LBD with an IC₅₀ value of ≈75 μM (Fig. 4).

9-cis-RA interferes with (R)-flurbiprofen’s reduction of Aβ secretion

Having found that (R)-flurbiprofen can competitively inhibit the binding of 9-cis-RA to RXRα, we asked whether 9-cis-RA affects (R)-flurbiprofen’s reduction of Aβ. A previous study showed that treatment with 9-cis-RA at 10 μM concentration could significantly reduce β- and γ-secretase activities (Xiong et al. 2008). However, we found that treatment with 9-cis-RA at a lower concentration (100 nM) slightly increased Aβ40 and 42 secretion (Fig. 5). Therefore, we used a 100 nM concentration in our experiments. In control cells, (R)-flurbiprofen treatment significantly reduced extracellular Aβ40 and 42 levels by approximately 40% and 50%, respectively. However, in the presence of 9-cis-RA, (R)-flurbiprofen treatment only reduced extracellular Aβ40 and 42 levels by approximately 20% and 23%, respectively, suggesting that 9-cis-RA interferes with (R)-flurbiprofen’s reduction of Aβ secretion.

Discussion

Several reports have suggested that NSAIDs, including (R)-flurbiprofen, can reduce Aβ generation (Weggen et al. 2001; Eriksen et al. 2003). However, these studies either investigated only extracellular (secreted) Aβ, or did not differentiate between changes in extracellular and intracellular Aβ. As the level of extracellular Aβ is dramatically higher than that of intracellular Aβ, any change in intracellular Aβ will be very likely masked by the change in extracellular Aβ. Therefore, we compared the changes in extracellular and intracellular Aβ resulting from (R)-flurbiprofen treatment. In contrast with the significant reductions in extracellular Aβ levels found in previous studies, we found that (R)-flurbiprofen treatment resulted in a significant increase of intracellular Aβ levels (Fig. 1). (R)-flurbiprofen has been tested for its potential in the treatment of AD but a recent phase III clinical trial by Myriad Genetics failed to show any beneficial effects in AD patients. Although there may be multiple reasons for the failure of the (R)-flurbiprofen clinical trial, our results provide a possible explanation: i.e., an increase in intracellular Aβ, which is believed to be more closely correlated to AD progression (Gouras et al. 2000), may counteract any beneficial effect from a reduction of total Aβ levels.

Retinoid X receptor α is a ligand-activated nuclear receptor that acts as a transcription factor in multiple important cellular processes, including memory (Chen et al. 1998; Zetterstrom et al. 1999; Luria and Furlow 2004). Some RXRα ligands have been found to rescue memory deficits in rats treated with scopolamine (Shudo et al. 2004). docosahexaenoic acid, which is essential for brain maturation and spatial learning, may also influence neural function through activation of an RXR signaling pathway (de Urquiza et al. 2000). Moreover, RXRα gene variants have been recently linked to AD risk: a three marker haplotype (TGC) consisting of rs3118570, rs1536475 and rs3132293 in the RXRα gene decreases the risk of AD (Kolsch et al. 2009), suggesting that altered RXRα function may be involved in AD pathogenesis.

We found that modulation of RXRα levels directly affects Aβ generation and down-regulation of RXRα dramatically increased the level of extracellular Aβ (Fig. 2a). Interestingly, over-expression of RXRα only slightly reduced Aβ secretion, implying that the effect of RXRα on Aβ secretion is saturable.

Our results showed that the protein level and in vitro activity of BACE1, the protein level of PS1 and Notch cleavage by the PS1/γ-secretease, and the degradation of extracellular Aβ were not affected by RXRα. Therefore, how exactly RXRα regulates Aβ generation and contributes to AD pathology remains speculative. One theory is that RXRα can affect cholesterol metabolism and thus substrate accessibility of APP by secretases, because of the fact that cholesterol metabolism has a well established link to AD (Xiong et al. 2008; Kolsch et al. 2009) and cholesterol-composed lipid rafts may be a major site for Aβ generation (Ma 2007). RXRs have been found to be key regulators of cholesterol metabolism through ATP-binding cassette transporter A1 and apolipoproteins (ApoE) that control the release of cellular cholesterol into plasma high-density lipoprotein (Ghose et al. 2004; Wang et al. 2005). Altered cholesterol levels may also affect APP trafficking and thus Aβ production. Although our results showed that cell surface level of APP is not affected by RXRα (Fig. 2a), we can not exclude the possibility that intracellular trafficking of APP to major sites for Aβ generation, such as the Golgi/TGN and endosomes (Zhang and Xu 2007), is affected by RXRα. In addition, expression of ATP-binding cassette transporter A1 is directly stimulated by LXR/retinoid X receptor (RXR) activation (Nishimaki-Mogami et al. 2008). The LXR/RXR heterodimer is also considered a transcription factor for ApoE, which is closely linked to AD (Ripolles Piquer et al. 2006). One recent study also showed that ApoE can activate LXR to stimulate Aβ degradation (Jiang et al. 2008). Another possibility is RXRα’s anti-inflammatory features. Retinoids have been demonstrated to suppress pro-inflammatory cytokine production by immune cells in vitro and these effects are induced by ligand-mediated activation of the retinoid receptors, including retinoic acid receptor and RXR (Mou et al. 2004).

In addition, we found that although both down-regulation of RXRα and TNFα/IFNγ treatment significantly increased the levels of secreted Aβ, co-treatments with (R)-flurbiprofen only rescued the levels of extracellular Aβ in TNFα/IFNγ-treated cells but not in RXRα knock-down cells (Fig. 3b). These results suggest that RXRα is involved in (R)-flurbiprofen-mediated Aβ secretion. As we found that (R)-flurbiprofen can interfere with the binding of 9-cis-RA to RXRα (Fig. 4) and previously showed that another NSAID, R-etodolac, can bind to RXRα (Kolluri et al. 2005), one possibility is that (R)-flurbiprofen also directly binds to RXRα and affects the function and stability of RXRα in favor of Aβ reduction. This notion may be supported by the finding that a low concentration of 9-cis-RA (100 nM) can attenuate (R)-flurbiprofen’s reduction of Aβ secretion (Fig. 5), implying competition between 9-cis-RA and (R)-flurbiprofen for RXRα binding. However, although the low concentration of 9-cis-RA (100 nM) in our experiments had no effect on Aβ secretion, it has previously been found that 9-cis-RA can significantly reduce β- and γ-secretase activities when used at a high concentration (10 μM) (Xiong et al. 2008). Therefore, one possibility is that both binding of RXRα to (R)-flurbiprofen and to 9-cis-RA can reduce Aβ generation but that the effect of (R)-flurbiprofen is more profound than that of 9-cis-RA, so that the net effect of competition of 9-cis-RA for (R)-flurbiprofen-RXRα binding is an attenuation of the reduction in Aβ secretion. Together, our results demonstrate a novel finding that (R)-flurbiprofen treatment reduces the level of extracellular Aβ but increases that of intracellular Aβ. They also suggest that RXRα regulates Aβ generation and that the reduction of Aβ secretion by (R)-flurbiprofen requires the involvement of RXRα.