Liver X Receptor–Binding DNA Motif Associated With Atherosclerosis‐Specific DNA Methylation Profiles of Alu Elements and Neighboring CpG Islands
Background The signals that determine atherosclerosis‐specific DNA methylation profiles are only partially known. We previously identified a 29‐bp DNA motif (differential methylation motif [DMM]) proximal to CpG islands (CGIs) that undergo demethylation in advanced human atheromas. Those data hinted that the DMM docks modifiers of DNA methylation and transcription.
Methods and Results We sought to functionally characterize the DMM. We showed that the DMM overlaps with the RNA polymerase III–binding B box of Alu short interspersed nuclear elements and contains a DR2 nuclear receptor response element. Pointing to a possible functional role for an Alu DMM, CGIs proximal (<100 bp) to near‐intact DMM–harboring Alu are significantly less methylated relative to CGIs proximal to degenerate DMM–harboring Alu or to DMM‐devoid mammalian‐wide interspersed repeat short interspersed nuclear elements in human arteries. As for DMM‐binding factors, LXRB (liver X receptor β) binds the DMM in a DR2‐dependent fashion, and LXR (liver X receptor) agonists induce significant hypermethylation of the bulk of Alu in THP‐1 cells. Furthermore, we describe 3 intergenic long noncoding RNAs that harbor a DMM, are under transcriptional control by LXR agonists, and are differentially expressed between normal and atherosclerotic human aortas. Notably, CGIs adjacent to those long noncoding RNAs tend to be hypomethylated in symptomatic relative to stable human atheromas.
Conclusions Collectively, the data suggest that a DMM is associated with 2 distinct methylation states: relatively low methylation of in cis CGIs and Alu element hypermethylation. Based on the known atheroprotective role of LXRs, we propose that LXR agonist–induced Alu hypermethylation, a landmark of atherosclerosis, is a compensatory rather than proatherogenic response.
What Is New?
LXR (liver X receptor) regulates local DNA methylation profiles in atherosclerosis by binding to a DR2‐containing DNA motif embedded in Alu short interspersed nuclear elements.
LXR regulates the expression of 3 differential methylation motif−containing long noncoding RNAs that are differentially expressed in atherosclerotic and normal human aortas.
What Are the Clinical Implications?
The findings may help define useful targets for molecular therapies for cardiovascular disease.
A detailed understanding of the DNA methylome of the atherosclerotic lesion is expected to yield novel insights into molecular mechanisms of cardiovascular disease (CVD) and to improve diagnostic and therapeutic strategies.1, 2 Recent microarray‐based and genome sequencing–based work using blood vessel type–matched and donor‐matched samples indicates that the DNA methylome of the atherosclerotic plaque undergoes dynamic changes during disease progression. These include hypermethylation during the stable phase followed by progressive demethylation in the symptomatic postrupture phase.3, 4, 5 The dynamic nature of the atherosclerotic lesion methylome is likely to account for some of the published conflicting reports, particularly the ones pointing to a genome‐wide hypomethylation resembling the cancer methylome.6, 7, 8, 9 In apparent support for the pathological relevance of the initial DNA hypermethylation wave in atherosclerosis, work in mouse models showed that the biochemical inhibition of DNA methyltransferase (DNMT) activity greatly reduces lesion size and inflammatory markers.10, 11, 12 Conversely, overexpression of the DNMT family member DNMT1 exacerbates atherosclerosis.13
A related urgent task is to identify the molecular pathways leading to atherosclerosis‐specific DNA methylation signatures in the vascular tissue. Homocysteine has received significant attention because of its participation in the biochemical machinery that generates the universal methyl donor S‐adenosylmethionine and its role as an independent CVD risk factor; however, the evidence is conflicting.14 Homocysteine induces vascular endothelial cell growth arrest and apoptosis by promoter hypermethylation‐mediated silencing of critical loci, a response that is reversed by biochemical DNMT inhibition.15 Furthermore, homocysteine stimulates cell proliferation through microRNA‐mediated upregulation of the DNMT member DNMT3 in cultured vascular smooth muscle.16, 17 Another explored topic has been the signaling pathways triggered by lipoprotein components, such as triglyceride‐rich lipoproteins and the fatty acid arachidonate. Both induce global hypermethylation in THP‐1 cells, and arachidonate‐induced DNA methylation profiles overlap with atherosclerosis‐specific signatures.18, 19 Hypermethylation is observed both within genic and repetitive Alu elements. Importantly, a causal effect of lipids on DNA methylation profiles in vivo has been demonstrated recently.20
In the present work, we sought to gain further insights into pathways leading to atherosclerosis‐specific DNA methylation signatures. To this end, we functionally characterized a 29‐bp DNA motif that we previously identified by a CpG island (CGI) array‐based survey of human coronary atherosclerotic plaques obtained from ischemic cardiopathy patients undergoing revascularization surgery.8 That motif is present in sequences flanking ≈55% of CGIs that undergo demethylation in those samples and thus was named differential methylation motif (DMM). In this study, we used a combination of molecular and epigenomics approaches to test the hypothesis that a DMM is associated with atherosclerosis‐specific DNA methylation and expression profiles. We discuss our findings in the context of the current knowledge of the epigenomics of atherosclerosis.
Yeast one‐Hybrid Assay
The DMM (5′‐ATCACTTGAGGTCAGGAGTTCGAGACCAG‐3′) was cloned into the pHISi vector (Clontech) upstream to imidazoleglycerol‐phosphate dehydratase (HIS3) between the EcoRI and XbaI restriction sites. The construct contained 2 DMM copies in tandem to achieve a robust HIS3 activation. The DMM‐HIS3 cassette‐containing fragment generated by ApaI digestion was inserted into the genome of the Saccharomyces cerevisiae YM4271 strain (genotype: MATa, ura3‐52, his3‐200, ade2‐101, lys2‐801, leu2‐3, 112, trp1‐901, tyr1‐501, gal4‐D512, gal80‐D538, ade5::hisG) to generate the bait strain. The latter was transformed using 0.5 mg of a Gal4 activation domain (AD)–human aorta cDNA fusion library (prey) in the Matchmaker pACT2 vector (Clontech). HIS3 expression allows cells to grow on medium lacking histidine. Transformed cells were selected in medium lacking leucine, uracil, and histidine, supplemented with the highly stringent 75 mmol/L 3‐aminotriazole, an inhibitor of the HIS3 gene product. These conditions select for strong HIS3 expressants versus revertants. Library clones extracted from growing colonies were sequenced and identified with the GenBank database ( National Center for Biotechnology Information). A subset of those clones was tested for activation specificity by further rounds of transformation into the bait strain and selection in 75 mmol/L 3‐aminotriazole–supplemented medium. To test for sequence binding specificity, mutant bait strains in which the AGGTCA portion of the DMM (positions 9–14) was either scrambled (GATAGC) or deleted were transformed and selected with increasing 3‐aminotriazole doses (0–50 mmol/L).
Human Vascular Samples
Authorization by institutional ethical committees and sample description were reported previously.3
Monocyte THP‐1 cells were cultured in RPMI‐1640 containing l‐glutamate (Life Technologies), 10% (vol/vol) FBS (Life Technologies), 5% CO2, 37°C. For differentiation to macrophages, ≈107 cells were treated with 50 ng/mL PMA (phorbol 12‐myristate 13‐acetate; Sigma‐Aldrich) for 4 hours, followed by 25 ng/mL PMA for 4 days with a medium change on the second day. Lipid‐loaded THP‐1 macrophages (LLMas) were obtained by culture in medium supplemented with 10% heat‐inactivated, hypertrigliceridemic (>200 mg/dL triglyceride) human serum instead of FBS. Those serum samples were in excess of the amount used for routine blood lipid profile tests in the department's clinical laboratory and normally would have been disposed of; lipid loading was assessed by oil red O staining of fixed cells. Stimulation with DMSO‐dissolved GW3965 or T0901317 (1 μmol/L in either case) or DMSO alone (control) was carried out for 24 hours. Final DMSO concentration in the medium was 0.01%. HEK293 cells were cultured in DMEM supplemented with l‐glutamate (Life Technologies), 10% (vol/vol) FBS, 5% CO2, 37°C, at cell density <106 cells/mL.
THP‐1 monocytes were stimulated with 1 μmol/L GW3965 dissolved in DMSO or with DMSO alone (controls) at the same final DMSO concentration (0.01%) for 24 hours. As reported, 20 million cells were processed for each treatment.21 Briefly, sonication settings were 20 seconds at amplitude 90, 15 times. The anti‐LXRB (anti–liver X receptor β) antibody was chromatin immunoprecipitation (ChIP) grade (cb56237; Abcam). Washes and DNA recovery were performed, as described.22 Precipitated DNA was amplified with gene‐specific oligonucleotides indicated in Table S1. An aliquot (5%) of the preimmunoprecipitation extract was amplified in parallel (input control). Polymerase chain reaction (PCR) products were run in 1× Tris/Borate/EDTA buffer (TBE), 2% agarose gels. All experiments were carried out at least in biological and technical triplicate.
Alu Combined Bisulfite Restriction Analysis
DNA extracted from THP‐1 cells was modified with sodium bisulfite with the EZ DNA Methylation‐Gold kit (Zymo Research), as recommended by the manufacturer and, PCR‐amplified with the primers indicated in Table S1. The primers amplify most Alu subfamilies, except FLA, FLAM_A, FLAM_C, Jr, Jr4, Jo, Sq2, Sq10, Sc5, Yf2, and Yh9. The product was digested with TaqI, and fragments were quantified by densitometry, as described.23
Gene‐Specific Methylation Profiling
THP‐1 monocytes and macrophage genomic DNA was treated with sodium bisulfite using the EZ DNA Methylation‐Gold system (Zymo). PCR products obtained with the gene‐specific primers listed in Table S1 were cloned, sequenced, and analyzed with the Kismeth software.24
For whole‐genome bisulfite sequencing data extraction, raw data from a previously published study3 of 1 atherosclerotic (grade VII)/normal human aortic tissue pair was downloaded from GEO data set GSE46327. A random set of 50 million reads from each sample was aligned to the AluS reference using BSMAP version 2.90,25 with default parameters and allowing up to 2 mismatches. Methylation was called from the aligned reads with the methratio.py script provided with BSMAP, without removing duplicates.
The coordinates of Alu elements were obtained from the UCSC Genome Browser, and the sequence of Alu elements was used as published in Repbase (girinst.org/repbase).26 Long noncoding RNA (lncRNA) coordinates were obtained from RNA sequencing libraries available in MiTranscriptome (58 648 lncRNAs; mitranscriptome.org).27 The reference GRCh37/hg19 human genome was used in all cases. The atherosclerotic lesions used in the present study were previously obtained and analyzed by the Infinium Human Methylation 450 BeadChip (450K array).3 The methylation level for each cytosine was expressed as a β value between 0 (unmethylated) and 1 (methylated). Alu flanking sequence methylation was extracted from 450K array data for ±100‐bp or ±2‐kb flanking sequences and averaged every 10 or 50 bp, respectively.
The analysis was performed with the Neighbor‐Joining method in MEGA 7, using 37 Alu subfamilies. The evolutionary distances were computed using the maximum composite likelihood method and are expressed as the number of base substitutions per site.
THP‐1 cells and aorta RNA were extracted with the Trizol method (Life Technologies) and treated with TURBO DNase (Life Technologies). The cDNA was synthesized using 500 ng (THP‐1 cells) or 400 ng (aorta) RNA, with a poly‐dT oligonucleotide, and the reverse transcriptase SuperScript ll (Life Technologies). Platinum (Life Technologies) was used for PCR. For Alu semiquantitative expression, the primers indicated in Table S1 were used. The primers amplify most subfamilies except FLA, FLAM_A, FLAM_C, FAM, FRAM, Jr, Jr4, Jb, Jo, Sp, Yc5, Yg6, Yh9, Ye5, Yf5, and Yk12. For all other transcripts, real‐time PCR assay was performed using SYBR GreenER qPCR SuperMix (Life Technologies) in a StepOne System using the primers indicated in Table S1. Relative expression units were calculated against GAPDH.
lncRNA‐Operated Transcriptional Regulation In Cis
lncACACA and lncSREBF1 cDNAs were amplified with high‐fidelity Platinum Taq DNA polymerase (Invitrogen) and inserted into pTracer‐SV40 (Thermo Fisher Scientific) using the ApaI and SpeI restriction sites. The ACACA and SREBF1 promoters were amplified from human genomic DNA with Platinum Taq DNA polymerase (primer sequence available on request) and inserted upstream to the firefly luciferase CDS in pGL4 (Promega) using the HindIII and XhoI restriction sites. HEK293 cells were first transfected with the lncRNA cDNA (2.5 μg plasmid) using Lipofectamine 3000 (Invitrogen) in 6‐well plates, as recommended by the manufacturer. At 24 hours later, cells were transfected with the promoter‐firefly luciferase construct (0.1 μg) in 96‐well plates. The Renilla luciferase vector (pRL‐TK; Promega) was cotransfected and used as expression control. Normalized expression was calculated as a firefly/Renilla luciferase activity ratio. Luciferase assays were performed in 96‐well chemoluminescence plates using the Dual‐Glo system (Promega), according to the manufacturer's instructions. Each assay was performed in triplicate.
ANOVA with the Scheffé post hoc test was used for expression and methylation data. The χ2 test was used for atherosclerotic/normal donor‐matched aortic sample pairs. Tests were performed using SPSS software for Windows (version 19.0; IBM Corp) and R (version 3.2.0; R Foundation for Statistical Computing).
Genomic and Sequence Features of the DMM
The DMM sequence yielded by the DNA motif search program MEME is partially degenerated.8 However, the reference DMM sequence—consisting of the most represented bases in each position—will hereafter be referred to as DMM if not specified otherwise. A BLAST‐based homology search revealed that DMM is homologous to Alu, an abundant primate‐specific retrotransposon family present in ≈1.1 million copies in the human genome.28, 29 Homology is particularly high with the AluS subfamily, with AluSz bearing an intact DMM. AluS and AluJ members show >93% homology with reference DMM and account for ≈750 000 Alu copies. Conversely, homology was comparatively low with AluY (69–75%; Figure 1A and 1B; Figure S1). In addition, a search for conserved features of DMM revealed the presence of a DR2 nuclear receptor (NR) response element (Figure 1B). DR2 consists of 2 direct RGKTCA repeats (R and K indicate A or G and G or T, respectively) separated by a 2‐nucleotide spacer. The presence of DR2 in Alu has been documented.30, 31 Notably, the Alu DR2 is functionally important because it binds the NR retinoic acid receptor to produce RNA polymerase III transcripts during stem cell differentiation.32 Furthermore, the entire Alu B box, a conserved RNA polymerase III–binding site that regulates Alu transcription, is embedded within DMM (Figure 1B).33
LXRB Binds to the DMM
The aforementioned conserved sequence features hinted that DMM is a putative docking site for DR2‐binding epigenetic modifiers.34, 35 We therefore sought DMM‐binding proteins by screening a Gal4 AD‐human aorta cDNA fusion library (prey) by the yeast 1‐hybrid assay. As bait, we used a DMM‐HIS3 reporter construct integrated into the S. cerevisiae genome (bait strain). The initial screen yielded a total of 45 independent colonies able to grow on histidine‐depleted medium, and from 38 of those, Gal4 AD‐cDNA clones were successfully sequenced. Ten clones were in‐frame (Table S2); of these, 5 encoded LXRB (or NR1H2). Repeated transformations of the Gal4 AD‐LXRB cDNA fusion into the bait strain revealed a specific and robust growth phenotype, pointing to LXRB as a strong DMM‐binding protein candidate. Furthermore, DR2 was necessary for the DMM–LXRB interaction, as either scrambling or deleting the 5′‐most of the 2 DR2 direct repeats (AGGTCA) of DMM abrogated Gal4 AD‐LXRB cDNA fusion‐induced growth on selecting medium (Figure 2A). These results are pathobiologically plausible because LXRs (liver X receptors; LXRA [LXR α] and LXRB members), a family of NRs activated by the endogenous ligands oxysterols, play pivotal roles both in cholesterol efflux from the vascular wall and in the control of atherosclerosis progression.36, 37, 38 Furthermore, the described LXR response elements bear sequence similarities to DMM DR2.39, 40
One of the other growth‐inducing in‐frame Gal4 AD‐cDNA fusion clones encoded the human Peter Pan homolog (Drosophila) (PPAN), a protein involved in ribosome biogenesis, development, and apoptosis regulation.41, 42, 43 Because subsequent experiments revealed a weak phenotype, the DMM–PPAN interaction was not studied further. The remaining 4 in‐frame Gal4 AD‐cDNA fusion clones encoded nonnuclear proteins and thus were discarded.
Next, we sought validation of the DMM–LXRB interaction in human chromatin by ChIP in cultured THP‐1 monocytes. We focused on 3 intergenic DMMs that (1) contain DR2 and are embedded in Alu and (2) map in proximity to known LXR target genes. Although the latter genes do not immediately flank the respective lncRNAs, they might be relevant to understanding LXR's biological activity (Figure 2B). The 3 flanking LXR target genes were ACACA (acetyl CoA carboxylase), FASN (fatty acid synthase), and SREBF1 (sterol regulatory element binding transcription factor 1).44, 45, 46 In addition, the promoters of the 3 LXR target genes are in CGIs, a feature that may help understanding any effect of DMM on CGI methylation in cis.8 The 3 intergenic DMMs also overlapped with 3 lncRNAs: short interspersed nuclear element/unknown lncRNA (chr17:36,810,109–36,810,324), RP13‐516M14.4 (chr17:82,290,046–82,292,814), SMCR2 (chr17:17,674,026–17,677,688; hg38 coordinates in all cases). For simplicity, these will be referred to as lncACACA, lncFASN, and lncSREBF1, respectively (Figure 2B). The DMM of these lncRNAs is within portions homologous to AluSx (lncACACA lncFASN) or AluSq (lncSREBF1). We used the LXR response element in the SREBF1 promoter and the DMM‐devoid GAPDH gene as positive and negative controls, respectively. The ChIP results indicated that LXRB binds to all 3 DMMs (Figure 2B). To test the reproducibility of our ChIP data, we surveyed a publicly available THP‐1 macrophage LXR ChIP sequencing data set in which exhaustive controls with nonimmune immunoglobulin were performed.47 Indeed, each of the 3 DMM‐containing lncRNA loci was specifically precipitated by LXR antibodies in the mentioned study (lncACACA‐, lncFASN‐, and lncSREBF1‐specific reads: 612, 2,708 and 1,555, respectively, of 18,304,931 reads). To assess whether LXR activation modifies the DMM–LXRB interaction, we performed ChIP in THP‐1 monocytes stimulated with the LXR synthetic agonist GW3965. No consistent effect of GW3965 was observed (Figure 2B).
Locus‐Specific and Genome‐Wide Methylation Signatures Associated With Alu DMM and Flanking Sequences in Human Vascular Tissue
DMM was initially identified as a motif proximal (≤2 kb) to CGIs that undergo demethylation in advanced plaques.8 Subsequent more detailed epigenomics showed that CGI demethylation is a widespread phenomenon in the symptomatic atheroma.5 Therefore, if Alu‐embedded DMMs participate in or are the consequence of human atherosclerosis‐specific DNA methylation, the methylation profile within or flanking DMM‐harboring Alu should show associations with the genome‐wide atherosclerosis‐specific signature. To probe for locus‐specific methylation profiles, we asked whether CGI CpGs adjacent to the 3 aforementioned lncRNAs show the expected methylation trend, namely, a relative low‐methylation state in postrupture condition compared with stable atherosclerosis. To this end, we profiled 2 kb up and downstream to the 3 lncRNAs using 450K‐based data. The 450K human atheroma methylation data were of stable aortic (n=15 atherosclerotic/normal tissue pairs) and symptomatic or asymptomatic carotid artery (n=19 each) plaque samples.3, 4, 5 This data set is informative for methylation profiles of Alu‐flanking sequence but not of Alu elements. We identified CpG clusters mapping to a CGI and CGI/shore boundary within 2 kb of lncACACA and lncFASN, respectively, whereas insufficient data were available for lncSREBF1. Both CGIs (neither of which contain any recognizable DMM) showed a significant loss of methylation in the postrupture plaque (P=0.030 and P=0.034, respectively; paired t test; Figure S2).
Next, we asked whether DMM‐harboring Alu shows the previously described atherosclerosis‐specific DNA methylation profile. Previous bulk Alu whole‐genome bisulfite sequencing data showed ≈15% hypermethylation in stable atherosclerotic aortas compared with donor‐matched normal controls (ie, portions of the same aortas adjacent to the atheroma‐occupied segment; 69.9% and 55.1% methylation, respectively, P=4.21×10−18, χ2 test).3 After extracting data from the same whole‐genome bisulfite sequencing set, the same trend was observed of the Alu subfamily AluSx that contains a near‐reference DMM: a significant increase in methylation of atherosclerotic relative to normal aorta by ≈13% (68.3% and 55.0%, respectively; 6384 informative reads on either strand, P=2.78×10−33; χ2 test; Figure S3).
As for Alu‐flanking sequences, we gathered 450K methylation data for ±100‐bp and ±2‐kb flanking sequences of Alu elements to probe for short‐distance and relatively long‐distance effects. Furthermore, we profiled Alu‐flanking non‐CGI and CGI CpGs, based on the rationale that it would enable corroboration of our initial observations that CGIs proximal to DMMs are preferentially demethylated in symptomatic plaques.5, 8
In both atherosclerotic and normal aortas, methylation (β values) abruptly decreases in Alu‐flanking sequences both within and outside CGIs (volcano‐shaped profiles in Figure S4A–S4D). In stable atherosclerosis, this is accompanied by a consistent, albeit small, increase in Δβ (≤0.04), relative to their normal counterparts. This was specific of Alu‐flanking non‐CGI sequences (compare Figure S5E and S5F with Figure S5G and S5H; compare the upward and downward cone‐shaped profiles of β and Δβ, respectively; Figures S4 and S5). Similar trends were observed in symptomatic/asymptomatic carotid plaques (Figure 3E–3H), but the increase in Δβ outside Alu was negligible in both CGIs and non‐CGIs (Figure S5E–S5H). Incidentally, CGI methylation was lower than in non‐CGI counterparts in all cases, as expected (Figure 3; Figure S4).48 In addition, Δβ was generally more extreme in the normal/stable atheroma aorta sample set compared with the asymptomatic/symptomatic carotid counterpart, particularly in the hypermethylated fraction, as reported (Figure S5).3, 5
Next, we compared methylation profiles of sequences flanking DMM Alu and degenerate DMM Alu. The latter were defined as displaying mismatches within and outside DMM DR2 and clustering on a separate branch when Alu was classified for DR2 homology. The elements in question include AluJb, AluJo, FAM, FLA, and FRAM (Figure S6, upper part). As further control, we surveyed sequences flanking mammalian‐wide interspersed repeats (MIR). MIRs are short interspersed nuclear elements that lack any detectable DMM or DR2 and that show an inconsistent methylation profile in atherosclerosis.3, 49 If DMM is functionally relevant, atherosclerosis‐specific DNA methylation profiles should be absent or more relaxed proximal to degenerate DMM–harboring Alu and MIRs. As for any differential association with reference and degenerate DMMs, CGI methylation was significantly lower at short distance from the former (P=1.4×10−6; compare Figure 3C and 3D). A significant difference in Δβ comparing reference DMM with either degenerate DMM or MIR was also observed (P=10−5 and P=3.3×10−4, respectively; compare Figure S5C with Figures S5D and S7B). In addition, a general increase in Δβ variability was apparent in degenerate DMM–harboring Alu‐flanking sequences regardless of Alu distance compared with intact DMM–harboring counterparts, more clearly so in 2‐kb flanking sequences (compare Figure S5A and S5B, Figure S5C and S5D, Figure S5E and S5F, and Figure S5G and S5H). In symptomatic plaques, a higher methylation level was also observed close to degenerate DMM in CGIs in this sample set (P=0.03; compare Figure 3G and 3H). The data suggest that LXR might bind degenerate DMM–harboring Alu to a lesser extent compared with intact DMM–harboring counterparts. We verified this hypothesis in human LXR ChIP sequencing data.47 The number of reads corresponding to intact DMM–Alu was ≈3.7‐fold higher than degenerate DMM counterparts (130 and 35, respectively). This ratio is not significantly different from the observed count ratio (Figure 3; χ2 test), suggesting that LXR binding alone does not explain the differential methylation between sequences flanking the 2 Alu sets. We speculate that LXR must bind an intact DMM to recruit additional essential partners.
LXR Agonists Induce Hypermethylation of the Alu B Box Region
In an effort to mechanistically interpret the Alu hypermethylation observed in the human atherosclerotic aorta, we asked whether LXRB affects internal methylation of the bulk of Alu. To this end, we exploited the serendipitous finding that a previously published combined bisulfite restriction analysis–based Alu methylation assay profiles 2 Alu CpGs present in most Alu subfamilies (see Methods).23 The 2 CpGs are located 29 bp from the B box 5′ end and within the B box and DR2 repeat, respectively (marked by asterisks in Figure 1B). This locus will be referred to as the Alu B box region. As a biological model, we used undifferentiated or differentiated (monocyte and macrophage, respectively) THP‐1 cells, a cellular model of inflammation. In addition, we used LLMas, an approximation of atheroma foam cells. Because of the latter feature, LLMa data can be used to estimate the contribution of lesion macrophages to expression or DNA methylation profiles obtained in the heterogeneous atheroma cell population. The 2 synthetic LXR agonists GW3965 and T0901317 induced a significant increase in methylation of Alu B box region (P<0.05 in all cases; ANOVA and Bonferroni post hoc test; Figure 4). The response was particularly prominent in macrophages and LLMas stimulated with GW3965 (15–17% increase). LXR agonist–induced Alu B box region methylation increase is relatively small but comparable with the overall Alu hypermethylation observed in atherosclerosis. As for Alu expression, inconsistent effects were observed in our experimental conditions in THP‐1 cells and in donor‐matched normal/atherosclerotic aortic tissue pairs (data not shown).
Methylation Profile of DMM‐Harboring lncRNAs
Next, we detailed the impact of LXR agonists on methylation profiles within and in cis to specific DMM/Alu‐harboring transcripts. We focused on the aforementioned lncACACA, lncFASN, and lncSREBF1 lncRNAs and the respective flanking CGIs. Data were obtained in monocyte and macrophage THP‐1 stimulated with the LXR agonist GW3965 and controls. LncRNA DMMs were highly methylated (ie, ≈50–100%), and LXR agonists did not exert any notable effect (Figure S8). Similarly, LXR agonists failed to induce any consistent pattern on LXR target gene promoter CGI methylation, except relatively small (≈8%) changes in the ACACA promoter. Methylation levels were generally low, as expected for CGIs.48
Expression of DMM‐Harboring lncRNAs and Flanking Genes in LXR Agonist‐Stimulated Cells and Human Atheromas
The largely negative results of methylation profiling described in the previous paragraph do not rule out that LXR agonists exert methylation‐independent effects on lncRNA expression. Expression of all 3 lncRNAs tended to increase in response to LXR agonist stimulation, although the response in LLMas was relatively blunted (Figure 5A through 5C). As for the respective lncRNA‐flanking genes, LXR agonists exerted similar effects (ie, an expression increase; Figure S9A through S9C).
To gain pathobiologically relevant insights, we obtained expression data in human atherosclerotic aortas and compared them with expression profiles in THP‐1 LLMas, a proxy of plaque foam cells, relative to monocytes. In human aortic atheromas, lncACACA and lncSREBF1 expression was increased ≈2‐ and ≈4‐fold, respectively, compared with unaffected matched controls, whereas lncFASN showed the opposite profile (Figure 5D). Baseline (unstimulated cell) expression of lncACACA and lncSREBF1 was significantly increased in LLMas in comparison with monocyte counterparts, whereas lncFASN followed the opposite trend (P<0.05 in all cases; Figure 5A through 5C, the significant difference is not shown in the figure for simplicity). Thus, the expression trend of the 3 lncRNAs in THP‐1 LLMas compared with monocytes recapitulated the corresponding trend in human atherosclerotic aortas compared with normal tissue.
In contrast, expression of lncRNA‐flanking genes was significantly decreased in atherosclerotic aortas compared with controls in all cases (Figure S9D). Comparing the trend in THP‐1 monocytes and LLMas, only ACACA displayed a significant baseline expression difference that mirrored the atheroma profile (compare Figure S9A and S9D).
In cis Transcriptional Effect of lncACACA, lncFASN, and lncSREBF1
Functionally, it is possible that the surveyed lncRNAs belong to the well‐known in cis transcription regulating noncoding RNAs.50 In that case, lncRNA and respective flanking gene expression patterns should be consistently correlated, either positively or negatively. The data revealed a generally concordant effect of LXR agonists on lncRNA and respective flanking gene expression in THP‐1 cells. In contrast, expression profiles of lncRNAs and flanking genes were opposite for lncACACA and lncSREBF1 but concordant for lncFASN in human aortas. To interpret this ambiguity, we directly tested the ability of lncACACA and lncSREBF1 overexpression to affect the transcription of ACACA and SREBF1 promoter‐luciferase reporter constructs, respectively. In the case of lncFASN, cloning of the experimental and control constructs could not be successfully completed. Assays were conducted in HEK293 cells given higher transfection efficiency in comparison with THP‐1s. Furthermore, HEK293 cells express LXRB (http://www.proteinatlas.org/ENSG00000131408-NR1H2/cell). The data failed to show any clear in cis effect of lncACACA or lncSREBF1 overexpression (data not shown).
We provide evidence that the DMM DNA motif, previously proposed as a candidate binding site for factors that induce a low methylation state of CGIs in symptomatic atherosclerotic plaques, is embedded in Alu. DMM overlaps with the Alu RNA polymerase III–binding B box and contains a DR2 direct repeat that is necessary for LXRB binding. The presence of NR‐binding motifs such as DR2 in Alu is well documented, and rules for NR‐type binding specificity have been proposed. In contrast with our data, at least 2 independent human studies showed that LXRs bind mainly to DR4—namely, 2 direct RGKTCA repeats separated by a 4‐nucleotide spacer—in peripheral blood monocyte‐derived macrophages and in THP‐1 macrophages, although a weaker LXR–DR2 interaction was detected in the former model.40, 47 The same authors and a mouse liver–based study showed that the DR2 specifically recognizes the NR peroxisome proliferator‐activated receptor γ.51, 52 It is possible that binding to Alu DR2 is a comparatively minor LXRB activity with specialized functions. At any rate, LXR binding to Alu has been documented in the MPO promoter40 and in the autoregulation of LXRA gene expression.53 Interestingly, although it is widely accepted that the LXR binds to its response elements as obligate dimer with RXR (retinoid X receptor), LXRB binds to DR2 in the 1‐hybrid assay despite the absence of RXR in yeast.54, 55 Further studies are necessary to understand the frequency and significance of RXR‐independent LXRB binding sites in the genome. In addition, S. cerevisiae does not synthesize the physiological LXR agonist oxysterol from cholesterol.56 Yet homologs of mammalian oxysterol‐binding proteins that are involved in sterol transport have been described in yeast.57 To our knowledge, whether LXR and yeast oxysterol‐binding proteins share any ligand is unknown. Even if S. cerevisiae was devoid of any LXR ligands, specific agonist‐independent LXR DNA binding was detected in human cells and might explain our 1‐hybrid data.47
By epigenomics analysis, we gathered evidence supporting an association between DMM and CGI low‐methylation state.5, 8 CGIs proximal to near‐reference DMM‐harboring Alu are significantly less methylated than CGIs proximal to degenerate DMM–harboring Alu or to the DMM‐devoid MIR short interspersed nuclear element in human atherosclerotic arteries. Accordingly, methylation of CGIs adjacent to DMM‐harboring lncRNA is significantly reduced in symptomatic versus stable plaques.
The presence of DMM in Alu prompted us to dissect the methylation profile within Alu elements and in flanking sequences in atherosclerosis and THP‐1 cells. In addition, we tested whether LXR agonists establish any of those profiles. By extrapolation, the observation that LXR agonists hypermethylate the Alu B box region suggests an NR‐mediated mechanism for Alu hypermethylation, a landmark epigenetic signature of atherosclerosis. Even among other short interspersed nuclear elements, Alu is the only repeated sequence that is consistently hypermethylated in the stable atheroma.3 Furthermore, similar data were obtained in peripheral blood of CVD patients.58 The effects of LXR agonists are not accompanied by any change in expression of the bulk of Alu, in line with evidence that DNA methylation does regulate Alu transcription.59 The observation that the activation of the atheroprotective LXR reinforces an alleged genomic landmark of CVD is unexpected.37 Our results question any proatherogenic, functional role for Alu hypermethylation in atherosclerosis. Rather, the data suggest a compensatory mechanism aimed at further limiting retrotransposition above constraints already imposed by physiological Alu hypermethylation. This line of thinking is supported by the comparison of methylation within Alu and in flanking sequences. Alu is slightly hypomethylated relative to immediate flanking DNA in control aortas. The latter difference is reduced in atherosclerosis, as Alu gains methylation to reach the level of immediate flanking DNA, where minimal differences between atherosclerotic aortas and controls are observed. Thus, Alu hypermethylation in atherosclerosis results from spreading of the local DNA methylation profile of the normal aorta into Alu. Moreover, the data show that with the exception of the CGIs mentioned, Alu maps to regions with very little or no differential methylation between atherosclerotic and control aortas. At any rate, Alu hypermethylation remains a potentially useful candidate CVD marker.
As for specific Alu‐ and DMM‐harboring transcripts, we show that LXR binds to 3 lncRNA loci mapping in cis to known LXR target genes. In addition, LXR agonists increase the expression of all 3 lncRNAs in THP‐1 cells. The data echo a recent report that LXR agonists activate 1 enhancer RNA in the same cellular model.60 In all 3 cases, a consistent direction of expression change was observed between donor‐matched normal and atherosclerotic human aortas and between THP‐1 monocytes and LLMas. This suggests that the observed expression profiles are mainly contributed to by the lesion lipid‐loaded macrophages, although the participation of other cell types in the highly heterogeneous atheroma cannot be ruled out. Functionally, direct molecular experiments and correlative expression analysis failed to support a role for the surveyed lncRNAs as in cis transcriptional regulators.
A pending issue is whether the other LXR family member, LXRA, participates in the cellular responses described. LXRA is not present in the human aorta cDNA library used in the 1‐hybrid assay, as judged by reverse transcription PCR. In THP‐1 cells, we detected the LXRA transcript (not shown), but a recent study failed to detect the protein by immunoblotting in naïve THP‐1 macrophages.60 Therefore, at least in the latter cell type, the observed responses to LXR agonists are likely to be mediated by LXRB. At any rate, other factors may bind DMM, since significant binding site sharing has been demonstrated for NRs.52
From an evolutionary standpoint, the DMM shows a relatively high homology with phylogenetically old Alu subfamilies—FLAM, AluJ, AluS—whereas the recently diverged AluY cluster away from the DMM.61 Remarkably, LXR was shown to induce active enhancer histone marks in loci specifically enriched in AluJ and AluS in THP‐1 cells.60 These data are in accordance with the evolutionary importance of Alu NR‐binding sites in primates. Accordingly, we previously showed that the portion of the DMM that is homologous to mouse sequences overlaps with the DR2, suggesting biological relevance for the LXR–DMM interaction beyond primates, possibly within the Alu‐homolog mouse B1 elements.8, 62
In summary, we provide evidence for an association between the DMM and methylation state of selected CGIs. In addition, we show that LXRs modulate Alu methylation and the expression of 3 Alu‐harboring lncRNAs that are differentially expressed in atherosclerosis. The data do not support Alu hypermethylation as a therapeutic target in atherosclerosis.
Sources of Funding
This work was funded by the Mexican National Council for Research and Technology (CONACyT) “Ciencia Básica” grant no. 134631 to Zaina. Studentships from CONACyT supported Tristán‐Flores, Silva‐Martinez, de la Rocha (PhD Studentships), Ortega‐Kermedy and Cruz‐Torres (MSc Studentships).
Table S1. Primer Sequences
Table S2. Candidate Differential Methylation Motif Ligand‐Encoding Human Aorta cDNA Clones Yielded by the 1‐Hybrid Assay
Figure S1. Homology between differential methylation motif and AluY subfamily members.
Figure S2. Methylation profile (450K array data) of long noncoding RNA (lncRNA)–flanking CpG island (CGI) CpGs. The ±2‐kb flanking sequences were surveyed. Gray boxes and arrows in the x‐axes indicate lncRNA size/position and transcription direction. Black boxes indicate CGIs. Each dot indicates the methylation Δβ of individual CpGs. Black and gray dots, stable atherosclerotic/normal aortic and symptomatic/asymptomatic carotid samples, respectively. Solid and open dots, hypermethylated and hypomethylated fraction, respectively. Genomic coordinates (hg38) are chr17:36 808 109–36 812 324 (lncACACA), chr17:82 288 046–82 294 814 (lncFASN) and chr17:17 672 026–17 679 688 (lncSREBF1). The scale bar applies to all 3 graphs.
Figure S3. AluSx CpG methylation profile (WGBS data) in stable atheroma.
Figure S4. Methylation profile (450K array data) of Alu‐flanking non–CpG island (non‐CGI) and CGI CpGs in human atherosclerosis.
Figure S5. Methylation profile (450K array data) of Alu‐flanking non–CpG island (non‐CGI) and CGI CpGs.
Figure S6. DR2 homology‐based clustering analysis of differential methylation motif and Alu.
Figure S7. Methylation profile (450K array data) of MIR‐flanking non–CpG island (non‐CGI) and CGI CpGs.
Figure S8. Methylation status of long noncoding RNA differential methylation motif and flanking gene promoter CpG island in THP‐1 cells.
Figure S9. Long noncoding RNA–flanking gene expression in LXR (liver X receptor) agonist–stimulated THP‐1 cells and in atheromas. Quantitative reverse transcription polymerase chain reaction analysis. A through C, Expression in THP‐1 cells. Open, gray, and solid bars indicate unstimulated, T0901317‐, and GW3965‐stimulated cells, respectively. D, Expression in normal/atherosclerotic paired aortic samples (N and A, respectively). In all cases, significant differences with respective controls or normal aortas are shown. *P<0.05; **P<0.01 (ANOVA in A–C; paired t test in D). A indicates atherosclerotic portion of the human aorta; LLMa, lipid‐loaded TH‐1 macrophage; Ma, macrophage; Mo, monocyte; N, normal portion of the human aorta.
- ↵Ordovás J, Smith C. Epigenetics and cardiovascular disease. Nat Rev Cardiol. 2010;29:46–57.
- ↵Voelter‐Mahlknecht S. Epigenetic associations in relation to cardiovascular prevention and therapeutics. Clin Epigenetics. 2016;8:4.
- ↵Zaina S, Heyn H, Carmona FJ, Varol N, Sayols S, Condom E, Ramirez‐Ruz J, Gomez A, Goncalves I, Moran S, Esteller M. DNA methylation map of human atherosclerosis. Circ Cardiovasc Genet. 2014;7:692–700.
- ↵Valencia‐Morales Mdel P, Zaina S, Heyn H, Carmona FJ, Varol N, Sayols S, Condom E, Ramírez‐Ruz J, Gomez A, Moran S, Lund G, Rodríguez‐Ríos D, López‐González G, Ramírez‐Nava M, de la Rocha C, Sanchez‐Flores A, Esteller M. The DNA methylation drift of the atherosclerotic aorta increases with lesion progression. BMC Med Genomics. 2015;8:7.
- ↵Zaina S, Gonçalves I, Carmona FJ, Gomez A, Heyn H, Mollet IG, Moran S, Varol N, Esteller M. DNA methylation dynamics in human carotid plaques after cerebrovascular events. Arterioscler Thromb Vasc Biol. 2015;35:1835–1842.
- ↵Lund G, Andersson L, Lauria M, Lindholm M, Fraga MF, Villar‐Garea A, Ballestar E, Esteller M, Zaina S. DNA methylation polymorphisms precede any histological sign of atherosclerosis in mice lacking apolipoprotein E. J Biol Chem. 2004;279:29147–29154.
- ↵Aavik E, Lumivuori H, Leppänen O, Wirth T, Häkkinen S‐K, Bräsen J‐H, Beschorner U, Zeller T, Braspenning M, van Criekinge W, Mäkinen K, Ylä‐Herttuala S. Global DNA methylation analysis of human atherosclerotic plaques reveals extensive genomic hypomethylation and reactivation at imprinted locus 14q32 involving induction of a miRNA cluster. Eur Heart J. 2014;36:993–1000.
- ↵Zhuang J, Luan P, Li H, Wang K, Zhang P, Xu Y, Peng W. The Yin‐Yang dynamics of DNA methylation is the key regulator for smooth muscle cell phenotype switch and vascular remodeling. Arterioscler Thromb Vasc Biol. 2017;37:84–97.
- ↵Yu J, Qiu Y, Yang J, Bian S, Chen G, Deng M, Kang H, Huang L. DNMT1‐PPARγ pathway in macrophages regulates chronic inflammation and atherosclerosis development in mice. Sci Rep. 2016;6:30053.
- ↵Bostom AG, Shemin D, Verhoef P, Nadeau MR, Jacques PF, Selhub J, Dworkin L, Rosenberg IH. Elevated fasting total plasma homocysteine levels and cardiovascular disease outcomes in maintenance dialysis patients. A prospective study. Arterioscler Thromb Vasc Biol. 1997;17:2554–2558.
- ↵Chang PY, Lu SC, Lee CM, Chen YJ, Dugan TA, Huang WH, Chang SF, Liao WSL, Chen CH, Lee YT. Homocysteine inhibits arterial endothelial cell growth through transcriptional downregulation of fibroblast growth factor‐2 involving G protein and DNA methylation. Circ Res. 2008;102:933–941.
- ↵Cao C, Zhang H, Zhao L, Zhou L, Zhang M, Xu H, Han X, Li G, Yang X, Jiang Y. miR‐125b targets DNMT3b and mediates p53 DNA methylation involving in the vascular smooth muscle cells proliferation induced by homocysteine. Exp Cell Res. 2016;347:95–104.
- ↵Zhang H‐P, Wang Y‐H, Cao C‐J, Yang X‐M, Ma S‐C, Han X‐B, Yang X‐L, Yang A‐N, Tian J, Xu H, Zhang M‐H, Jiang Y‐D. A regulatory circuit involving miR‐143 and DNMT3a mediates vascular smooth muscle cell proliferation induced by homocysteine. Mol Med Rep. 2016;13:483–490.
- ↵Rangel‐Salazar R, Wickström‐Lindholm M, Aguilar‐Salinas CA, Alvarado‐Caudillo Y, Døssing KBV, Esteller M, Labourier E, Lund G, Nielsen FC, Rodríguez‐Ríos D, Solís‐Martínez MO, Wrobel K, Wrobel K, Zaina S. Human native lipoprotein‐induced de novo DNA methylation is associated with repression of inflammatory genes in THP‐1 macrophages. BMC Genomics. 2011;12:582.
- ↵Silva‐Martínez GA, Rodríguez‐Ríos D, Alvarado‐Caudillo Y, Vaquero A, Esteller M, Carmona FJ, Moran S, Nielsen FC, Wickström‐Lindholm M, Wrobel K, Wrobel K, Barbosa‐Sabanero G, Zaina S, Lund G. Arachidonic and oleic acid exert distinct effects on the DNA methylome. Epigenetics. 2016;11:321–334.
- ↵Dekkers KF, van Iterson M, Slieker RC, Moed MH, Bonder MJ, van Galen M, Mei H, Zhernakova DV, van den Berg LH, Deelen J, van Dongen J, van Heemst D, Hofman A, Hottenga JJ, van der Kallen CJ, Schalkwijk CG, Stehouwer CD, Tigchelaar EF, Uitterlinden AG, Willemsen G, Zhernakova A, Franke L, ‘t Hoen PA, Jansen R, van Meurs J, Boomsma DI, van Duijn CM, van Greevenbroek MM, Veldink JH, Wijmenga C; BIOS Consortium , van Zwet EW, Slagboom PE, Jukema JW, Heijmans BT. Blood lipids influence DNA methylation in circulating cells. Genome Biol. 2016;17:138.
- ↵Carey MF, Peterson CL, Smale ST. Chromatin immunoprecipitation (ChIP). Cold Spring Harb Protoc. 2009;2009:pdb.prot5279.
- Stockert J
- Espada J
- Blázquez‐Castro A
- ↵Sirivanichsuntorn P, Keelawat S, Danuthai K, Mutirangura A, Subbalekha K, Kitkumthorn N. LINE‐1 and Alu hypomethylation in mucoepidermoid carcinoma. BMC Clin Pathol. 2013;13:10.
- ↵Iyer M, Niknafs Y, Malik R, Singhal U, Sahu A, Hosono Y, Barrette T, Prensner J, Evans J, Zhao S, Poliakov A, Cao X, Dhanasekaran S, Wu Y‐M, Robinson D, Beer D, Feng F, Iyer H, Chinnaiyan A. The landscape of long noncoding RNAs in the human transcriptome. Nat Genet. 2015;47:199–208.
- ↵Lander ES, Linton LM, Birren B, Nusbaum C, Zody MC, Baldwin J, Devon K, Dewar K, Doyle M, FitzHugh W, Funke R, Gage D, Harris K, Heaford A, Howland J, Kann L, Lehoczky J, LeVine R, McEwan P, McKernan K, Meldrim J, Mesirov JP, Miranda C, Morris W, Naylor J, Raymond C, Rosetti M, Santos R, Sheridan A, Sougnez C, Stange‐Thomann Y, Stojanovic N, Subramanian A, Wyman D, Rogers J, Sulston J, Ainscough R, Beck S, Bentley D, Burton J, Clee C, Carter N, Coulson A, Deadman R, Deloukas P, Dunham A, Dunham I, Durbin R, French L, Grafham D, Gregory S, Hubbard T, Humphray S, Hunt A, Jones M, Lloyd C, McMurray A, Matthews L, Mercer S, Milne S, Mullikin JC, Mungall A, Plumb R, Ross M, Shownkeen R, Sims S, Waterston RH, Wilson RK, Hillier LW, McPherson JD, Marra MA, Mardis ER, Fulton LA, Chinwalla AT, Pepin KH, Gish WR, Chissoe SL, Wendl MC, Delehaunty KD, Miner TL, Delehaunty A, Kramer JB, Cook LL, Fulton RS, Johnson DL, Minx PJ, Clifton SW, Hawkins T, Branscomb E, Predki P, Richardson P, Wenning S, Slezak T, Doggett N, Cheng JF, Olsen A, Lucas S, Elkin C, Uberbacher E, Frazier M, Gibbs RA, Muzny DM, Scherer SE, Bouck JB, Sodergren EJ, Worley KC, Rives CM, Gorrell JH, Metzker ML, Naylor SL, Kucherlapati RS, Nelson DL, Weinstock GM, Sakaki Y, Fujiyama A, Hattori M, Yada T, Toyoda A, Itoh T, Kawagoe C, Watanabe H, Totoki Y, Taylor T, Weissenbach J, Heilig R, Saurin W, Artiguenave F, Brottier P, Bruls T, Pelletier E, Robert C, Wincker P, Smith DR, Doucette‐Stamm L, Rubenfield M, Weinstock K, Lee HM, Dubois J, Rosenthal A, Platzer M, Nyakatura G, Taudien S, Rump A, Yang H, Yu J, Wang J, Huang G, Gu J, Hood L, Rowen L, Madan A, Qin S, Davis RW, Federspiel NA, Abola AP, Proctor MJ, Myers RM, Schmutz J, Dickson M, Grimwood J, Cox DR, Olson MV, Kaul R, Raymond C, Shimizu N, Kawasaki K, Minoshima S, Evans GA, Athanasiou M, Schultz R, Roe BA, Chen F, Pan H, Ramser J, Lehrach H, Reinhardt R, McCombie WR, de la Bastide M, Dedhia N, Blöcker H, Hornischer K, Nordsiek G, Agarwala R, Aravind L, Bailey JA, Bateman A, Batzoglou S, Birney E, Bork P, Brown DG, Burge CB, Cerutti L, Chen HC, Church D, Clamp M, Copley RR, Doerks T, Eddy SR, Eichler EE, Furey TS, Galagan J, Gilbert JG, Harmon C, Hayashizaki Y, Haussler D, Hermjakob H, Hokamp K, Jang W, Johnson LS, Jones TA, Kasif S, Kaspryzk A, Kennedy S, Kent WJ, Kitts P, Koonin E V, Korf I, Kulp D, Lancet D, Lowe TM, McLysaght A, Mikkelsen T, Moran J V, Mulder N, Pollara VJ, Ponting CP, Schuler G, Schultz J, Slater G, Smit AF, Stupka E, Szustakowki J, Thierry‐Mieg D, Thierry‐Mieg J, Wagner L, Wallis J, Wheeler R, Williams A, Wolf YI, Wolfe KH, Yang SP, Yeh RF, Collins F, Guyer MS, Peterson J, Felsenfeld A, Wetterstrand KA, Patrinos A, Morgan MJ, de Jong P, Catanese JJ, Osoegawa K, Shizuya H, Choi S, Chen YJ, Szustakowki J; International Human Genome Sequencing Consortium . Initial sequencing and analysis of the human genome. Nature. 2001;409:860–921.
- ↵Blattler A, Farnham PJ. Cross‐talk between site‐specific transcription factors and DNA methylation states. J Biol Chem. 2013;288:34287–34294.
- ↵Claudel T, Leibowitz MD, Fiévet C, Tailleux A, Wagner B, Repa JJ, Torpier G, Lobaccaro JM, Paterniti JR, Mangelsdorf DJ, Heyman RA, Auwerx J. Reduction of atherosclerosis in apolipoprotein E knockout mice by activation of the retinoid X receptor. Proc Natl Acad Sci USA. 2001;98:2610–2615.
- ↵Mak P, Kast‐Woelbern H, Anisfeld A, Edwards P. Identification of PLTP as an LXR target gene and apoE as an FXR target gene reveals overlapping targets for the two nuclear receptors. J Lipid Res. 2002;43:2037–2041.
- ↵Migeon JC, Garfinkel MS, Edgar BA. Cloning and characterization of peter pan, a novel Drosophila gene required for larval growth. Mol Biol Cell. 1999;10:1733–1744.
- ↵Bugner V, Tecza A, Gessert S, Kühl M. Peter Pan functions independently of its role in ribosome biogenesis during early eye and craniofacial cartilage development in Xenopus laevis. Development. 2011;138:2369–2378.
- ↵Pfister AS, Keil M, Kühl M. The Wnt target protein peter pan defines a novel p53‐independent nucleolar stress‐response pathway. J Biol Chem. 2015;290:10905–10918.
- ↵Zhang Y, Yin L, Hillgartner FB. Thyroid hormone stimulates acetyl‐coA carboxylase‐alpha transcription in hepatocytes by modulating the composition of nuclear receptor complexes bound to a thyroid hormone response element. J Biol Chem. 2001;276:974–983.
- ↵Joseph SB, Laffitte BA, Patel PH, Watson MA, Matsukuma KE, Walczak R, Collins JL, Osborne TF, Tontonoz P. Direct and indirect mechanisms for regulation of fatty acid synthase gene expression by liver X receptors. J Biol Chem. 2002;277:11019–11025.
- ↵Repa J, Liang G, Ou J, Bashmakov Y, Lobaccaro J, Shimomura I, Shan B, Brown M, Goldstein J, Mangelsdorf D. Regulation of mouse sterol regulatory element‐binding protein‐1c gene (SREBP‐1c) by oxysterol receptors, LXRalpha and LXRbeta. Genes Dev. 2000;14:2819–2830.
- ↵Kumar AP, Piedrafita FJ, Reynolds WF. Peroxisome proliferator‐activated receptor gamma ligands regulate myeloperoxidase expression in macrophages by an estrogen‐dependent mechanism involving the ‐463GA promoter polymorphism. J Biol Chem. 2004;279:8300–8315.
- ↵Boergesen M, Pedersen TÅ, Gross B, van Heeringen SJ, Hagenbeek D, Bindesbøll C, Caron S, Lalloyer F, Steffensen KR, Nebb HI, Gustafsson J‐Å, Stunnenberg HG, Staels B, Mandrup S. Genome‐wide profiling of liver X receptor, retinoid X receptor, and peroxisome proliferator‐activated receptor α in mouse liver reveals extensive sharing of binding sites. Mol Cell Biol. 2012;32:852–867.
- ↵Willy PJ, Umesono K, Ong ES, Evans RM, Heyman RA, Mangelsdorf DJ. LXR, a nuclear receptor that defines a distinct retinoid response pathway. Genes Dev. 1995;9:1033–1045.
- ↵Heery DM, Zacharewski T, Pierrat B, Gronemeyer H, Chambon P, Losson R. Efficient transactivation by retinoic acid receptors in yeast requires retinoid X receptors. Proc Natl Acad Sci USA. 1993;90:4281–4285.
- ↵Price AL, Eskin E, Pevzner PA. Whole‐genome analysis of Alu repeat elements reveals complex evolutionary history. Genome Res. 2004;14:2245–2252.