Differential control of gene expression is an important means by which cells respond to physiological and environmental stimuli. Nuclear receptors comprise a superfamily of ligand regulated, DNA-binding transcription factors that can both activate and repress gene expression [1]. The liver X receptors (LXRs) are type II nuclear receptors, initially identified as orphan nuclear receptors, because their natural ligands were not known [2, 3]. LXRs have been deorphanized or adopted following the discovery of oxysterols (hydroxylated derivatives of cholesterol) as their endogenous ligands [3, 4]. Two LXR isoforms are known, namely LXRα and LXRβ with distinct tissue distributions [5]. LXRα expression is relatively restricted to tissues involved in lipid metabolism, such as liver and intestine [6, 7], whereas LXRβ is ubiquitously expressed. Both LXR isoforms are expressed in the brain [5]. LXRβ expression, in particular is 2-5 fold higher in the brain than in liver [8].

The mechanisms of transcriptional regulation by LXRs involve formation of heterodimers with retinoid X receptor (RXR). RXR is another deorphanized nuclear receptor, activated by 9-cis-retinoic acid (a vitamin A derivative) [9]. Upon heterodimerization, LXR/RXR bind to specific DNA sequences called LXR-responsive elements (LXREs) in the target genes [10]. In the absence of a ligand, LXRs bound to cognate LXREs, are in complex with corepressors such as silencing mediator of retinoic acid and thyroid hormone receptor (SMRT) [11] and nuclear receptor corepressor (N-CoR) [12], consequently, the transcription of target genes is repressed. Receptor ligation induces conformational changes in the ligand binding domain which mechanistically facilitates the release of corepressors and the recruitment of coactivators and histone acetyltransferase (HAT), the enzyme that acetylates lysine amino acids on histone proteins by transferring an acetyl group from acetyl CoA and is generally associated with transcriptional activation [13, 14] (Figure 1). Interestingly, LXR/RXR receptors exhibit the "phantom ligand effect," the ability of ligand-induced allosteric signal transmission by nuclear receptors to activate the unliganded heterodimeric partners [15]. These heterodimers are also referred to as permissive because the complex can be activated by ligands of either partner. These unique features allow multiple ligand-mediated pathways to be integrated into a transcriptional response.

Studies over the last decade have established LXRs as master regulators of lipid metabolism. LXR mediates activation of target genes such as sterol responsive element binding protein 1c (SREBP1c), a master transcription factor that controls the entire fatty acid biosynthetic pathway [16], lipid transporters including members of the superfamily of ATP-binding cassette (ABC) transporters such as ABCA1 [17–19], apolipoproteins (ApoE, ApoD) [20, 21] and lipoprotein modifying enzymes (cholesteryl ester transfer protein (CETP) and phospholipid transfer protein (PLTP) [22, 23]. In addition, primary astrocyte cultures treated with synthetic LXR ligands exhibit increased cholesterol efflux and elevated expression of LXR target genes including ABCA1 and ApoE [8, 24, 25]. LXRα/β knockout mice show a variety of CNS defects including lipid accumulation, astrocyte proliferation and disorganized myelin sheaths [26, 27].

The expression of RXRs has been observed by immunohistochemistry and in situ hybridization in mouse brain [28–30]. There are three isoforms of RXRs: RXRα, RXRβ, RXRγ. RXRα and β are most prevalent in the neocortex and hippocampus while RXRγ expression is restricted to the neocortex [31].

Analysis of data from a large-scale microarray study of postmortem brain specimens obtained from multiple brain regions of elderly patients with varying severity of dementia [32] indicated significant changes in RXR gene expression [33, 34]. Observed changes in gene expression were isoform-specific with more robust Alzheimer's disease (AD)-associated changes observed in RXRα levels. In particular, the dysregulated expression was most obvious in AD vulnerable regions such as inferior temporal gyrus (area 20) and superior temporal gyrus (area 22) and in the hippocampus, but not in primary visual cortex (area 17) which is relatively spared from age related or AD-associated neurodegeneration [35–40]. In addition, changes in LXR/RXR target genes, ABCA1 and ApoE in AD brains as a function of the increasing severity of dementia and neurofibrillary pathology were also observed [33, 34, 41]. The significance of these results is twofold. First, expression levels of nuclear receptors and their target genes have not been previously reported in a single cohort of clinically, neuropsychologically and neuropathologically well-characterized AD and control individuals with minimal mRNA variability, known medical history and absence of protracted agonal state. Second, ABCA1 and ApoE are not only LXR/RXR target genes but also the major determinants of net cholesterol flux. These novel findings provided an impetus to study LXR/RXR gene and protein expression using more quantitative technique. To this end, we analyzed LXRβ and RXRα mRNA expression in the hippocampus and inferior temporal gyrus (area 20) of a large series of cases at different stages of dementia and AD associated neuropathology by quantitative PCR. Because mRNA and protein levels can diverge significantly through post-transcriptional regulation, Western blotting was used to quantify protein levels in the hippocampus of a subset of the postmortem brain specimens used in PCR analyses.

This study explored the expression profile of the master regulators of lipid metabolism in two of the most vulnerable regions of the AD brain. The primary finding was that RXRα levels increased at the very earliest stage of dementia and remained elevated, in general, throughout the course of the disease. In addition, these elevations in gene and protein expression were more strongly associated with the development of AD-associated dementia than with the measures of mature neuropathologic lesions of AD. In order to identify earliest transcriptional changes during the course of AD progression, study individuals were grouped based on dementia severity (CDR score) at the time of death, progression of NFT pathology and the severity of NP pathology. This strategy allowed us to correlate expression of nuclear receptors to cognitive and pathological indices of AD, with an emphasis on individuals presenting earliest signs of AD. These approaches revealed a previously unrecognized transcriptional dysregulation of RXRα in AD. Specifically, alteration in gene expression in both regions was strongly correlated with cognitive impairment. Additionally, we observed highly coordinated upregulation of RXRα protein in AD hippocampus. Along with the identification of altered RXRα expression, this study highlights closely correlated expression of RXRα with the downstream target genes that have been previously implicated in AD pathogenesis including ApoE and ABCA1. One parsimonious interpretation of these findings is that changes in the expression of RXR become evident before the advent of the neuropathological hallmarks of AD and raise the possibility that the upregulation of RXR may be responsible for the changes in subsequent progressive cholesterol dyshomeostasis and AD neuropathology. However, because of the postmortem nature of the current study, the results of this study do not address directly whether the elevated levels of LXR/RXR in dementia are causal, secondary or bystander. It is also important to note that although we interpret transcriptional changes as a function of the severity of dementia or neuropathology to represent disease progression, the cross-sectional nature of this postmortem study, like all postmortem studies, permits only the inference of progression rather than its direct measurement. In addition, the reported results are based on assays of brain tissue homogenates, and therefore cannot inform on the cellular localization of the dysregulated gene expression or the laminar identity of the affected cells.

Although the pathogenesis of AD is not fully understood much direct and circumstantial evidence suggests that many of the genes and pathways involved in cholesterol and/or lipoprotein metabolism in brain are also intimately involved in the pathogenesis of AD. Most notable are ApoE, the principal cholesterol carrier protein in the CNS, and ABCA1, a protein that modulates the efflux of cellular cholesterol and phospholipids to lipid-deficient apolipoprotein acceptors such as ApoE [42–44]. Interestingly, we have observed significant increase in ABCA1 expression in the same cohort and tissues specimens as those described here [33, 34, 41]. Although, in the hippocampus, significant increase in RXRα mRNA level was observed in individuals with moderate to severe dementia, significant increase in RXRα protein expression was evident in individuals at an earlier stage of cognitive decline. Consequently, even small changes in RXRα mRNA expression potentially result in relatively large changes in RXRa protein. Because RXRα expression is strongly correlated with transcriptional changes in both ABCA1 and ApoE, it is conceivable that the variance in RXRα, which is not enough for strong association with CDR to be detected, could nevertheless induce expression of the target genes to the extent that significant correlations with earliest cognitive and neuropathological markers are obvious. These results are particularly intriguing in light of the fact that both ABCA1 and ApoE are under transcriptional regulation of the RXR/LXR signaling pathway. Together, these findings suggest that RXR/LXR-mediated alterations in ApoE and ABCA1 levels can modulate cholesterol metabolism and consequently the risk for AD.

In AD, cholesterol dyshomeostasis has been associated with the processing of amyloid precursor protein (APP) to generate neurotoxic Aβ [45]. Given the critical role played by LXR/RXR in the regulation of cellular cholesterol homeostasis, and their expression in multiple brain types (albeit with significant cell type and brain region expression differences), numerous studies have focused on the ability of LXR/RXRs to regulate amyloidogenic processing of APP. However, contradictory results have been reported on Aβ metabolism following treatment of cultured cells with LXR agonists [24, 46, 47]. The reason for the discrepancies may be related to differences between the expression of APP under physiologic conditions vs. its expression in mutant cells, and the extent to which membrane cholesterol transport was modified under the different in vitro conditions. In vivo studies using a synthetic LXR agonist, TO-901317, were more definitive in suggesting decreased Aβ deposition in the brains of APP transgenic mice [48, 49]. Interestingly, TO-901317 has been shown to inhibit γ-secretase independent of LXR/RXR activation [50]. In addition, TO-901317 is also a farnesoid X receptor and pregnane X receptor agonist [51, 52]. This raises concerns about the specificity of TO-901317 and the interpretation of studies using TO-901317 mediated LXR/RXR activation. Finally, LXR/RXR-mediated cholesterol homeostasis is differentially regulated in mice and humans. These differences are accentuated in studies attributing beneficial outcomes to LXR/RXR-mediated transcriptional regulation. In mice but not in primates, hepatic cholesterol 7α-hydroxylase is upregulated by LXR [4, 53, 54]. Mice do not express CETP. Direct induction of CETP in human cell lines and two CETP-containing animal models, Syrian hamsters and long-tailed macaque monkeys, is accompanied by a significant increase in LDL cholesterol levels that was not previously observed in mice [23, 55, 56]. These findings emphasize the need to exercise caution when extrapolating results from animal studies to humans, especially when significant species-specific differences have been identified in the underlying biological processes. It is noteworthy that in the current systematic analysis, LXR/RXR expression levels were not associated with elevations in NP counts. It may be plausible that change in RXRα expression may influence soluble toxic forms or fractions of Aβ, yet it does not correlate with NP density because RXRα has a monotonic pattern of expression, whereas NPs continue to increase in density.

LXRs are activated by endogenous oxysterols, the most potent of which include 24, 25-epoxycholesterol, 24-OH, and 27-hydroxycholesterol. The role of 24-OH (also known as cerbrosterol) is particularly intriguing because it is primarily produced in the brain by neuron specific CYP46 [57] and its level is increased in plasma and CSF during early stages of AD [58, 59]. 24-OH has been shown to activate LXRs and dramatically elevate ABCA1 expression in both neurons and astrocytes [8, 25, 60]. Therefore increased expression of ABCA1 and ApoE in incipient AD might be a reflection of 24-OH mediated increased LXR/RXR activation. Paradoxically, 24-OH level is reduced in late stages of AD [61], suggesting that sustained increases in the expression of LXR/RXR target genes in relatively late stages is modulated by other factors.

LXRβ heterodimer with RXRα is the only nuclear receptor complex known to date that can be activated in the absence of a ligand, via a mechanism termed "dimerization-induced activation" [62–64]. In this model of LXR transactivation, the interaction of RXR with LXR can allosterically activate LXR by inducing a conformational change in its ligand-binding domain. The relative expression levels of both receptors are therefore likely to regulate signaling via LXRβ and RXRα in a highly complex fashion. Indeed, in transient transfection studies LXR/RXR activation is only observed upon RXRα cotransfection [3, 65, 66], which results in a higher number of LXRβ/RXRα heterodimers for which coregulators would have to compete. Because the activated LXRβ/RXRα heterodimer also exhibits dual ligand permissiveness and synergism [62, 67], its net transcriptional potential depends on the occurrence of dimerization-induced activation and ligand availability. Consistent with this idea, the overexpression of RXRα reported here could allosterically activate LXRβ and consequently increase target gene expression, in addition to that induced by 24-OH. Recently, LRP has been shown to participate in ABCA1 expression by relieving LXR/RXR repression (via cPLA₂ activation) [68]. Strong association of LRP gene expression with that of LXR and RXR suggests that LRP modulates not only LXR/RXR activation but also their transcription. Alternatively, LRP may be an LXR/RXR target gene.

Based on previous studies where a drastic reduction in cholesterol level decreased Aβ production [69–71] and owing to their ability to induce expression of genes mediating cholesterol efflux, LXR/RXR heterodimers have emerged as potential targets for AD therapeutics [72]. However, there are obvious contradicting reports of increased Aβ generation upon lowering brain cholesterol level [73, 74]. More importantly, the hippocampus of AD cases presents a moderate, yet significant, reduction in membrane cholesterol [75]. These latter findings are consistent with increased expression of RXRα reported here. Taken together, increased expression of RXRα and concomitant activation of LXR/RXR can modulate ABCA1 and ApoE gene expression. Increased levels of ABCA1 and ApoE may be the molecular determinants of cholesterol dyshomeostasis and accompanying dementia observed in AD.