This study provides the first genome-wide view of LXR binding patterns in a human cell line using ChIP-Seq assays. We performed this study in PMA-differentiated THP-1 cells, a macrophage-type cellular system, which increases our understanding not only of the well-known role of LXR in lipid metabolism and transport, but also the receptor's assumed role in innate immunity and other physiological processes.
The highly stringent analysis of the ChIP-Seq peaks obtained from PMA-treated THP-1 cells after a 60 min stimulation with the synthetic LXR ligand T09 or its vehicle DMSO yielded a surprisingly low number of 202 genome-wide binding locations of the receptor. However, from the 151 LXR sites observed in the presence of T09, 57% (86) are also bound in the absence of ligand. This supports the canonical model for nuclear receptor binding being valid for most members of the superfamily . According to this model the receptor binds genomic DNA already in the absence of ligand, probably in a complex with co-repressor and histone deacetylase proteins, and locally represses the chromatin structure. The addition of ligand induces a conformational change in the ligand-binding domain of the receptor, which then leads to dissociation of the co-repressor proteins and the recruitment of co-activators that open chromatin structure. Alternatively, co-activators act as mediators building a bridge to the basal transcription machinery, which leads to the activation of RNA polymerase II and gene transcription. According to the data presented here, this model seems to apply for a number of known LXR binding sites close to target genes, such as ABCA1.
A lower stringency in the detection of LXR ChIP-Seq peaks increased the number of sites substantially, to a total of 1357 (FDR < 1%) or even 8139 (FDR < 5%). However, with this increase of putative LXR binding sites the percentage of overlapping sites reduced to 28.1 and 19.4%, respectively, while the percentage of apparent binding sites in the absence of ligand increased to 61.2 and 65.4%, respectively. We assume that most of the latter sites are not genomic locations from which LXR initiates gene activation, but rather unspecific contact points of the genome with no functional impact. However, the tendency that ligand stimulation reduces the number of sites to less than half (from 1212 to 526 and from 6903 to 2817) suggests that in its ligand-activated state LXR makes a more focused selection of genomic targets.
Surprisingly, only 7.3% of the LXR peak summit sequences contain a DR4-type RE, which, based on in vitro studies, is the only high affinity DNA binding site for LXR-RXR heterodimers. This fits with the observations of the two presently published mouse LXR ChIP-Seq studies, where only 6.3%  or up to 8%  of the LXR peaks contained a DR4-type RE. In a very recent ChIP-on-chip report, 2035 LXRβ-RXR binding sites were identified within promoter regions of normal human epidermal keratinocytes . Although 1666 (82%) of these rather large genomic fragments contain a kind of DR4-type RE, only 142 (7.0%) of them are highly scored, resulting in a very similar percentage as in our human and the mouse ChIP-Seq studies. For comparison, in our ChIP-Seq study on genomic vitamin D receptor (VDR) locations in undifferentiated THP-1 cells , we found only 31% of the VDR peak summits containing a VDR-specific DR3-type RE. We obtained the same result also by a re-analysis of VDR ChIP-Seq data from human lymphoblast cell lines [42, 43]. These observations indicate that either the binding specificity of LXR (and VDR) is in the chromatin context far different than in in vitro assays (maybe due to cooperative binding with other transcription factors, such as PPARγ) or LXR is not directly contacting DNA, but sits "piggyback" on another DNA-binding transcription factor. Interestingly, in our T09 microarray the PPARG gene is 1.4-fold up-regulated, a value that was also confirmed by qPCR with two different LXR ligands. In parallel, we found for 16.5% of all LXR peaks in the T09-treated sample a PPARγ binding site. This suggests that PPARγ may be involved in the regulation of a subset of the LXR target genes. Boergesen et al. reported very recently that in mouse liver many genomic binding sites of LXR are also bound by PPARα, which is the predominant PPAR subtype in liver . They suggest that PPARα helps LXR in recognizing its preferred genomic locations, a concept that may apply also for PPARγ in macrophages. Furthermore, Boergesen et al. also found a higher percentage of DR1-type REs below their LXR ChIP-Seq peaks than DR4-type REs, but in sum this can explain only less than 25% of all LXR locations. Therefore, they assume that both LXR and PPAR are far more promiscuous in the recognition of their binding sites as suggested by in vitro studies.
Nevertheless, we found DR4-type REs as functional LXR binding sites in a number of prominent and highly ligand-responsive LXR target genes, such as ABCA1, ABCG1, NR1H3 and SCD. Moreover, in a reasonable number of gene regulatory scenarios two or more genomic LXR binding sites seem to work together in the regulation of a gene cluster (for example, NACA, GPR137 and their respective neighboring genes) or individual genes (for example, ABCA1, SLC3A2 and SMPDL3A). These multiple RE arrangements confirm observations from single gene analyses with other nuclear receptors, such as the VDR [44–46].
More important than the actual number of individual LXR binding sites may be their spatial clustering along the genome. We identified 112 genomic regions of an average size of 1.7 Mb that contain clusters of highly stringent LXR binding sites and up to 11 T09 target genes. In total we identified 432 T09 target genes within these genomic LXR hotspots. Although each of these regions displays a rather individual arrangement of LXR binding sites in relation to up- and down-regulated genes, they seem to represent the core of the genome-wide activities of LXR. However, 13 of the 112 regions do not have any T09 target gene, suggesting that these stringent LXR locations may have effects on more distant target genes.
In total our microarray analysis of the 4 h T09-stimulated human macrophage-type cells lists 1713 genes, 73% of which are up-regulated, i.e. more than double as many genes are up-regulated than down-regulated. Out of these T09 target genes, 432 (25%) are found within the 112 genomic LXR hotspots, 814 genes (48%) have an LXR binding site within 100 kb distance from the respective gene's TSS and for a further 249 genes (15%) an LXR location within 1 Mb of the TSS could be identified. This rate is similar to other reports comparing ChIP-Seq and microarray results . Nevertheless, this raises the question about the regulation of the 650 T09 target genes without an LXR peak. These genes may be secondary LXR targets. As only 15% of them are down-regulated, a trans-repression mechanism is not very likely.
Our study confirmed a number of known primary LXR target genes, such as ABCA1, ABCG1, MYLIP, NR1H3 and SCD, but we identified also a number of previously unknown, novel LXR targets. The most up-regulated LXR target gene, NACA, encodes for the nascent polypeptide associated complex alpha protein, which is associated to translation and protein folding related processes and when depleted is responsible of triggering endoplasmic reticulum stress-driven apoptosis in hypoxic cells . The PTGES3 gene, which is co-located with the NACA gene, encodes for the co-chaperone protein prostaglandin E synthase 3 and is required together with the primary chaperone proteins for proper folding and functioning of the glucocorticoid receptor and other steroid receptors . The most down-regulated gene, SLC3A2, is related to various processes, for instance in the migration of leukocytes from blood to the central nervous system , but also to serum cholesterol levels . Also the SMPDL3A gene is an interesting new LXR target, but it has not yet been well characterized as there is only one report linking SMPDL3A to bladder tumorigenesis .
LXR actions have been related to a number of autoimmune diseases, such as multiple sclerosis [52, 53], rheumatoid arthritis [54, 55] and experimental autoimmune encephalomyelitis . Interestingly, the association of LXR with autoimmune and metabolic diseases is also one of the major results of the annotation analysis of the 1063 T09 responding genes that we consider as true LXR target genes. The analysis showed the expected relation to lipid metabolism and transport genes, but did not provide any significant link to genes related to innate immunity. However, the LXR regulation of inflammatory cytokines is generally observed in experimental settings, where lipopolysaccharide is used for stimulation and where longer LXR ligand treatment times are allowed . Under these conditions inflammatory cytokines are induced via the transcription factors AP1 and NF-κB and can be inhibited via tethered LXR. In this study, we used PMA for the differentiation of THP-1 cells and T09 treatment of only 4 h. Interestingly, under our experiment conditions, we observed a strong association with apoptosis as more than 70 genes within the 1063 candidates are related to this physiological process. The link of LXRs to apoptosis has is already been reported, not only in macrophages , but also in pancreatic β-cells  and different cancer cells [60, 61].