LXR–cofactor complexes activate the gene expression program responsible for cholesterol efflux in macrophages. Inflammation antagonizes this program, resulting in foam cell formation and atherosclerosis; however, the molecular mechanisms underlying this antagonism remain to be fully elucidated. We use promoter enrichment‐quantitative mass spectrometry (PE‐QMS) to characterize the composition of gene regulatory complexes assembled at the promoter of the lipid transporter Abca1 following downregulation of its expression. We identify a subset of proteins that show LXR ligand‐ and binding‐dependent association with the Abca1 promoter and demonstrate they differentially control Abca1 expression. We determine that NCOA5 is linked to inflammatory Toll‐like receptor (TLR) signaling and establish that NCOA5 functions as an LXR corepressor to attenuate Abca1 expression. Importantly, TLR3–LXR signal crosstalk promotes recruitment of NCOA5 to the Abca1 promoter together with loss of RNA polymerase II and reduced cholesterol efflux. Together, these data significantly expand our knowledge of regulatory inputs impinging on the Abca1 promoter and indicate a central role for NCOA5 in mediating crosstalk between pro‐inflammatory and anti‐inflammatory pathways that results in repression of macrophage cholesterol efflux.
NCOA5 is a key target of the crosstalk between pro‐inflammatory TLR3 signaling and the anti‐inflammatory LXR pathway. Here, NCOA5 functions as an LXR corepressor to attenuate expression of the lipid transporter Abca1, thereby providing a mechanism whereby inflammatory signaling leads to repression of macrophage cholesterol efflux.
Promoter enrichment‐quantitative mass spectrometry (PE‐QMS) maps the dynamic assembly of protein complexes at the Abca1 promoter.
PE‐QMS identifies the LXR ligand‐stimulated and LXR binding‐induced occupancy of NCOA5 at the Abca1 promoter.
In response to TLR3‐LXR signal crosstalk, NCOA5 is recruited to the Abca1 promoter, where it functions as an LXR corepressor.
NCOA5 inhibits Abca1 gene expression by disrupting RNA Polymerase II recruitment and function.
Macrophages have a critical role in maintaining systemic lipid homeostasis. They promote the efflux of excess intracellular cholesterol to high‐density lipoproteins (HDL), thereby facilitating its return to the liver through the reverse cholesterol transport system. Macrophage cholesterol efflux is dependent on membrane‐associated lipid transporters, such as ATP‐binding cassette, sub‐family A, member 1 (ABCA1) (Oram et al, 2000; Vaisman et al, 2001; Wang et al, 2007; Yvan‐Charvet et al, 2007). The importance of ABCA1 function to this process is illustrated in Tangier disease, where mutations in the Abca1 gene disrupt cholesterol efflux resulting in the formation of lipid‐filled macrophage foam cells (Bodzioch et al, 1999; Brooks‐Wilson et al, 1999; Rust et al, 1999; Orso et al, 2000).
Macrophage foam cells are highly prevalent and significant contributors to the pathology of the chronic inflammatory disease atherosclerosis, a leading cause of morbidity and mortality worldwide (Moore et al, 2013; http://www.who.int/mediacentre/factsheets/fs317/en/). During the development of atherosclerosis, extrinsic inflammatory stimuli activate pro‐inflammatory macrophage Toll‐like receptors (TLRs), which repress expression of lipid transporters, and therefore impede cholesterol efflux (Castrillo et al, 2003; Spann et al, 2012). Notably, in mice, the absence of Abca1 expression in macrophages results in increased foam cell formation and atherosclerosis (Aiello et al, 2002; van Eck et al, 2002), while in humans, loss of macrophage cholesterol efflux capacity is a significant risk factor for atherosclerosis (Khera et al, 2011). In fact, therapeutic efforts are now focused on enhancing macrophage cholesterol efflux capacity in atherosclerosis (Rader & Tall, 2012), signifying the importance of understanding the molecular regulation of lipid transporter expression.
The central regulators of lipid transporter expression are the sterol sensing liver X receptor (LXR) transcription factors LXRα (Nr1h3) and LXRβ (Nr1h2). LXRs heterodimerize with retinoid X receptors (RXRs), and upon binding to regulatory sites for the lipid transporters Abca1 and Abcg1, the LXR–RXR complex induces transcription in a sterol ligand‐dependent manner (Costet et al, 2000; Repa et al, 2000; Venkateswaran et al, 2000a,b). As a result, sterol ligand stimulation of LXRs inhibits atherosclerotic plaque formation in mice (Joseph et al, 2002). The specificity of this sterol‐mediated gene induction is dependent on LXR–cofactor interactions, where corepressor complexes are replaced with coactivator complexes (Wagner et al, 2003; Huuskonen et al, 2004, 2005; Lee et al, 2008; Jakobsson et al, 2009). However, the full compendium of these signal‐dependent interactions at the regulatory regions of lipid transport genes remains to be elucidated.
Interestingly, the anti‐inflammatory LXRs and pro‐inflammatory TLRs antagonize each other in macrophages (Castrillo et al, 2003; Joseph et al, 2003; Ogawa et al, 2005). LXRs prevent induction of pro‐inflammatory gene expression through sumoylation‐dependent transrepression (Ogawa et al, 2005; Ghisletti et al, 2007; Huang et al, 2011). Conversely, TLR3‐4 signaling represses LXR‐dependent gene activation (Castrillo et al, 2003). In atherosclerosis, the pro‐inflammatory pathway appears dominant over the anti‐inflammatory LXR pathway, resulting in reduced macrophage cholesterol efflux and foam cell formation. While these observations indicate the importance of signal crosstalk, much remains to be understood about the mechanisms governing these interactions and their effects on lipid transporter expression.
We hypothesized that activation of the TLR3 pathway downregulates LXR‐mediated cholesterol efflux by recruiting corepressors to, or maintaining corepressors on, the promoters of lipid transporter genes. To identify such factors, we employed promoter enrichment‐quantitative mass spectrometry (PE‐QMS) to characterize the composition of gene regulatory complexes at the Abca1 promoter (Ranish et al, 2003, 2004; Kim et al, 2007; Mittler et al, 2009; Foulds et al, 2013; Mirzaei et al, 2013; Viturawong et al, 2013). We performed a systems‐level analysis of proteins that associate with Abca1 regulatory DNA in macrophages at a time when Abca1 expression was being downregulated, as these are likely to reflect corepressor interactions. Using quantitative mass spectrometry, we simultaneously identified a compendium of LXR ligand‐stimulated and LXR binding‐dependent regulators, including the nuclear receptor cofactor NCOA5. We further demonstrate that NCOA5 is recruited to the Abca1 promoter in vivo in response to both lipid and inflammatory signals where it functions as an LXR corepressor to suppress Abca1 transcription and cholesterol efflux.
Identification of LXR‐dependent transcriptional regulators using promoter enrichment‐quantitative mass spectrometry (PE‐QMS)
To investigate the mechanisms controlling Abca1 transcription, we began by examining its expression dynamics in response to the synthetic LXR ligand T0901317 (Schultz et al, 2000). Using isolated C57BL/6 bone marrow‐derived primary macrophages (BMMs), we observed an initial ligand‐stimulated increase in both Abca1 mRNA (Fig 1A) and ABCA1 protein (Supplementary Fig S1A and B). This response was dependent on LXR, as Nr1h3−/−:Nr1h2−/− (LXR−/−) BMMs failed to induce ligand‐stimulated expression (Fig 1A). We also observed minimal changes in Nr1h3 and Nr1h2 mRNA expression during this response (Supplementary Fig S1C and D).
Notably, Abca1 ligand‐stimulated mRNA expression peaked at 8–16 h and was subsequently followed by a decline in expression in LXR+/+ BMMs (Fig 1A). We observed a similar pattern in ABCA1 protein expression (Supplementary Fig S1A). While current evidence suggests LXR coactivators are responsible for ligand‐stimulated transcriptional induction (Calkin & Tontonoz, 2012), a mechanism to explain the subsequent attenuation of the Abca1 transcriptional response following ligand treatment remains unknown (Fig 1A).
To test the hypothesis that LXR cofactor complexes are responsible for attenuating Abca1 expression in macrophages, we undertook a multifaceted approach to compare Abca1 promoter‐associated proteins using promoter enrichment‐quantitative mass spectrometry (PE‐QMS) (Fig 1B). PE‐QMS uses in vitro biotinylated regulatory DNA to enrich for gene regulatory complexes, which are subsequently identified and quantified by mass spectrometry. Notably, this technology requires no a priori knowledge of transcription factor binding. We chose a 321‐bp genomic regulatory sequence flanking the proximal direct repeat 4 (DR4) LXR response element (LXRE) of Abca1, which contains the TATA box and is responsive to LXR ligand stimulation (Supplementary Fig S2A). Importantly, this sequence is within a DNaseI hypersensitive region and is devoid of nucleosomes, as measured by DNase‐Seq and acetylated histone H4 ChIP‐seq, respectively (Supplementary Fig S2B) (Ramsey et al, 2010; Gold et al, 2012).
To isolate and identify transcriptional regulators which bind to this region in an LXR ligand‐stimulated manner, nuclear extracts from RAW 264.7 macrophages stimulated for 18 h with LXR ligand or vehicle control were incubated with the biotinylated Abca1 regulatory sequence, and stably bound proteins were eluted and analyzed by quantitative mass spectrometry (Fig 1B). These macrophages displayed a similar response to the synthetic LXR ligand T0901317 as primary BMMs (Supplementary Fig S3). To determine which of these associations were dependent on LXR promoter binding, we performed another PE‐QMS experiment in which we compared the compendium of proteins bound to the wild‐type LXRE in the Abca1 promoter with those bound to an LXRE that we engineered to contain mutations that completely abolished LXR binding and the ligand‐stimulated transcriptional response of Abca1 (Fig 1B; Supplementary Figs S4 and S5).
After removing any proteins encoded by genes that were not expressed in primary macrophages (Ramsey et al, 2008; Gold et al, 2012), we identified 79 potential Abca1 promoter‐associated proteins (FDR < 1%; Fig 1C; Supplementary Table S1A and B). Using a 1.5‐fold change in binding response as a cutoff, which we chose after examining the binding response of LXR and RXR, we identified ligand‐stimulated and LXR‐dependent subsets of interactions (Fig 1C; Supplementary Table S1C). Notably, these subsets were enriched for gene ontology (GO) terms related to transcription (Supplementary Table S2A–C).
We detected 19 proteins that were reproducibly enriched in the ligand‐stimulated experiments, of which seven were dependent on LXR promoter binding, ten were independent, and two were undetermined (Fig 1C; Supplementary Table S1C). Interestingly, of these 19 ligand‐stimulated associations, 15 were the products of genes whose RNA levels in macrophages did not change more than 1.5‐fold in response to T0901317 (Supplementary Fig S6). Furthermore, all seven LXR‐dependent associations, along with five of the LXR‐independent associations, were from proteins annotated as transcription factors (Supplementary Table S1C) (Ashburner et al, 2000; Roach et al, 2007).
Functional validation of the ligand‐stimulated and LXR‐dependent candidate regulators
We next sought to assess the function of these LXR‐dependent putative regulators identified by PE‐QMS. We began by generating a protein interaction network in silico. This analysis indicated that both NCOA5 and NHP2 were connected to TLR‐dependent inflammation through physical protein–protein interactions in mouse or human (Fig 2A). Moreover, reporter assays using the same regulatory sequence as that used for the PE‐QMS enrichment indicated that while SND1, SART1, and HMBOX1 induced Abca1 transcription almost twofold (P < 0.05; Fig 2B), NCOA5 expression significantly repressed the Abca1 transcriptional response threefold (P = 0.00005; Fig 2C). These results suggest a regulatory balance between activators and repressors functions to coordinate Abca1 expression.
Interestingly, NCOA5 has previously been described as a cofactor for a small subset of nuclear receptors, including estrogen receptor 1 (ESR1); however, its role in transcription seems to be context dependent (Sauve et al, 2001; Jiang et al, 2004; Gao et al, 2013; Sarachana & Hu, 2013). In addition, polymorphisms in Ncoa5 are associated with both chronic inflammatory disease and metabolic disease (Bento et al, 2008; Lewis et al, 2010; Zervou et al, 2011). To our knowledge, NCOA5 has never previously been implicated in LXR‐regulated gene expression, cholesterol efflux, or atherosclerosis. This, together with its repression of Abca1 transcription and its potential involvement in TLR signaling, makes NCOA5 an ideal candidate for further investigation.
NCOA5 functions as an LXR corepressor
To further investigate the role of NCOA5 in Abca1 transcription, we performed reporter assays using NCOA5 and Abca1 mutants. Deletion of the NH2‐terminus (∆1–280aa), which contains a putative repressor domain (Sauve et al, 2001), abolished the repressive effect of NCOA5 on the reporter (Fig 2C). However, NCOA5 retained repressive activity after mutation of the LXRE (Fig 2C), suggesting it can still function in the absence of LXR binding. In addition, over‐expression of NCOA5 in primary BMMs confirmed its ability to repress Abca1 expression (Fig 2D; Supplementary Fig S7A).
To determine whether NCOA5 directly interacts with LXR, we performed in vitro pulldown assays. Full‐length in vitro translated NCOA5 protein specifically bound to GST‐LXRα in a ligand‐independent manner (Fig 3A). Notably, deletion of the NH2‐terminus (∆1–280aa) prevented this interaction, while the NH2‐terminal fragment strongly bound GST‐LXRα (Fig 3A). Given the nuclear receptor interaction motif on NCOA5 is found within its COOH‐terminus (Sauve et al, 2001), these results indicate NCOA5 directly interacts with LXR through a non‐canonical domain in the NH2‐terminus.
To confirm that NCOA5 functions as an LXR cofactor, we examined their interaction in macrophage nuclear extracts. Indeed, endogenous NCOA5 immunoprecipitated with a Protein C‐tagged version of LXRα expressed in RAW 264.7 macrophages (Fig 3B). Similar to the in vitro pulldowns, this in vivo interaction also occurred both in the presence and absence of LXR ligand. Together, these results indicate that NCOA5 functions as an LXR corepressor, and its mode of interaction appears to be distinct from its mode of interaction with ESR1 (Sauve et al, 2001).
To determine whether NCOA5 localizes to the Abca1 promoter in vivo, we performed chromatin IP (ChIP) assays in primary BMMs in the presence or absence of LXR ligand stimulation. Using primers spanning the proximal LXRE just upstream of the Abca1 transcriptional start site (TSS), which is contained within the same region analyzed by PE‐QMS, we detected ligand‐dependent occupancy of NCOA5 on the Abca1 promoter at 18 h (Fig 3C). We also assayed an upstream region lacking an LXRE as a negative control and failed to detect significant NCOA5 binding at 8 h or 18 h post‐ligand treatment (Supplementary Fig S7B), confirming the specificity of our results. These data also indicate that despite interacting directly with LXR in the presence or absence of ligand, NCOA5 only localizes to the Abca1 promoter in vivo following ligand stimulation.
To assess the requirement of LXR for NCOA5 binding at the Abca1 promoter, we performed ChIP assays in LXR−/− BMMs. Notably, we detected ligand‐independent occupancy of the Abca1 promoter by NCOA5 in the absence of LXR (Fig 3D), suggesting an additional mechanism exists for NCOA5 recruitment under these conditions. Taken together, these results identify NCOA5 as an LXR corepressor, which localizes to the Abca1 promoter in vivo following LXR ligand treatment to repress Abca1 expression. In the absence of LXR, NCOA5 can still occupy the Abca1 promoter and repress its expression through an undetermined mechanism.
NCOA5 attenuates Abca1 expression
To investigate the dynamic role of NCOA5 as an LXR corepressor of Abca1, we silenced its expression in primary BMMs using shRNA and performed a time course of LXR ligand stimulation. With a non‐silencing control shRNA, we observed an Abca1 mRNA expression profile that was similar to earlier experiments (see Fig 1A), with expression peaking at 16 h post‐ligand stimulation (Fig 4A). Moreover, we did not detect any significant change in Ncoa5 expression in response to LXR ligand (Supplementary Fig S7C). Following administration of Ncoa5 shRNA, which persistently and substantially lowered Ncoa5 expression at both the mRNA and protein levels (Supplementary Fig S7C and D), we observed a similar initial induction of Abca1 expression (Fig 4A). However, with Ncoa5 silencing, Abca1 expression remained elevated even at 30 h post‐ligand stimulation (Fig 4A). This effect was recapitulated at the protein level, where ABCA1 abundance was greater at 30 h post‐ligand stimulation in Ncoa5 knockdown BMMs (Fig 4B; Supplementary Fig S7E). Consistent with its proposed role as an LXR corepressor, these results demonstrate that NCOA5 is required for attenuation of Abca1 expression following LXR ligand stimulation.
TLR3 signals to NCOA5 to suppress Abca1 expression
The TICAM1/TRIF‐dependent inflammatory receptors TLR3 and TLR4 have a profound antagonistic effect on ligand‐induced LXR target gene expression (Castrillo et al, 2003). To investigate whether NCOA5 is involved in signal crosstalk between these pro‐inflammatory TLRs and the anti‐inflammatory LXR pathway, we silenced Ncoa5 expression in primary BMMs and stimulated with LXR ligand, the TLR3 agonist polyinosinic‐polycytidylic acid (PolyIC), and the TLR4 agonist bacterial lipopolysaccharide (LPS). In non‐silencing shRNA controls, we detected an LXR ligand‐dependent increase in Abca1 expression at 4 h (Fig 5A and B). The addition of PolyIC or LPS, either in the presence (Fig 5A and B) or absence (Fig 5C and D) of LXR ligand, significantly reduced Abca1 expression, together with some elevation of Ncoa5 mRNA (Supplementary Fig S8A–D). Strikingly, knockdown of Ncoa5 (Supplementary Fig S8A–D) abolished the ability of PolyIC but not LPS to attenuate LXR ligand‐dependent Abca1 expression (Fig 5A and B). Moreover, loss of NCOA5 failed to prevent either PolyIC‐ or LPS‐mediated reduction of basal Abca1 expression in the absence of LXR ligand in LXR+/+ BMMs (Fig 5C and D).
To test the hypothesis that TLR3–LXR signal crosstalk recruits NCOA5 to the Abca1 promoter to repress its activity, we performed NCOA5 ChIP assays in primary BMMs at 3 h post‐stimulation. In macrophages lacking TLR stimulation, we failed to detect binding of NCOA5 near the proximal Abca1 LXRE in response to LXR ligand or control (Fig 5E). However, following the addition of PolyIC to LXR ligand‐stimulated BMMs, we observed a threefold increase in occupancy of the Abca1 promoter by NCOA5 (Fig 5E). In addition, administration of PolyIC in the absence of LXR ligand failed to recruit NCOA5 to the Abca1 promoter (Fig 5F). These results further indicate the responsiveness of NCOA5 to TLR3 signals only when co‐stimulated with LXR ligand.
To explore the biological significance for the role of NCOA5 in mediating the TLR3‐dependent repression of LXR ligand‐induced Abca1, we performed sterol efflux assays in control and Ncoa5 knockdown primary BMMs. In non‐silencing shRNA control macrophages, we detected diminished cholesterol efflux from ABCA1 to its lipid acceptor apolipoprotein A–I (APOA1) upon addition of T0901317 and PolyIC, compared to LXR ligand alone (Fig 5G). Notably, following shRNA silencing of Ncoa5, LXR ligand‐induced ABCA1 efflux to APOA1 failed to be impeded by TLR3 stimulation (Fig 5G). These results indicate that NCOA5 has an important role in attenuating Abca1 expression and function in response to TLR3–LXR signal crosstalk.
To investigate the mechanism by which NCOA5 suppresses Abca1 transcription, we examined RNA polymerase II (RNAPII) recruitment and activation. Using ChIP assays in primary BMMs, we observed increased occupancy of both unmodified/pSer5 and pSer2 RNAPII at the Abca1 TSS in response to LXR ligand (Figs 6A and B). Notably, we detected reduced RNAPII levels when PolyIC was combined with LXR ligand (Fig 6A and B), indicating that NCOA5‐mediated attenuation of Abca1 expression in response to TLR3–LXR signal crosstalk correlates with a defect in RNAPII function, likely through recruitment.
To ascertain the role of LXR in this RNAPII defect, we performed the above experiments in LXR−/− BMMs. We did not detect an LXR ligand‐dependent increase in occupancy of either unmodified/pSer5 or pSer2 RNAPII at the Abca1 TSS (Fig 6C and D). However, we still observed a reduction in RNAPII occupancy following treatment with T0901317 and PolyIC (Fig 6C and D). In support of this, we also detected decreased Abca1 mRNA expression in response to both treatments in LXR−/− BMMs (Fig 6E; Supplementary Fig S8E), further suggesting that NCOA5 can function even in the absence of LXR.
To determine whether TLR3–LXR signal crosstalk prevents RNAPII function through NCOA5, we performed ChIP assays following silencing of Ncoa5 expression. With a non‐silencing control, we observed increased RNAPII pSer2 occupancy at the Abca1 TSS in response to LXR ligand (Fig 6F). The addition of both LXR and TLR3 ligands returned RNAPII pSer2 occupancy to basal levels (P = 0.33 versus Veh; Fig 6F). Following Ncoa5 gene silencing, RNAPII pSer2 binding remained elevated, failing to return to basal levels, in response to both LXR and TLR3 ligands (P = 0.004 versus Veh; Fig 6F). Taken together, these data establish NCOA5 as a critical downstream mediator of the crosstalk between the pro‐inflammatory TLR3 and anti‐inflammatory LXR pathways, and in response to these signals, NCOA5 represses macrophage cholesterol efflux through inhibition of RNAPII function at the Abca1 gene locus (Fig 7).
Inappropriate and sustained signaling events interfere with homeostasis by perturbing regulatory network function, resulting in chronic disease. In atherosclerosis, these signals arise from chronic inflammation and inhibit the macrophage homeostatic response to hyperlipidemia, specifically the cholesterol efflux pathway, which results in foam cell formation and plaque progression. While LXRs directly induce transcription of the transporters responsible for cholesterol efflux, it currently remains unknown whether additional transcriptional regulatory proteins are recruited to these promoters by inflammatory signals to block LXR‐dependent gene regulation, and the mechanisms by which this might occur.
We hypothesized that LXR cofactors receive these inflammatory signals from TLRs to directly modulate lipid‐mediated gene expression. We therefore characterized the occupancy of the Abca1 promoter by transcriptional regulators using promoter enrichment‐quantitative mass spectrometry (PE‐QMS). Notably, PE‐QMS provides a DNA‐context‐dependent method of identifying transcriptional regulatory complexes that requires no a priori knowledge of transcription factor identity (Ranish et al, 2003, 2004). Moreover, it can identify cofactor and combinatorial interactions, which are not readily achieved using epigenomic and chromatin accessibility information (Ramsey et al, 2010; Neph et al, 2012; Sherwood et al, 2014). A unique aspect of our PE‐QMS design was assessing not only the effect of LXR ligand stimulation on gene regulatory complex composition, but also the subset of that response which was dependent on LXR promoter binding. Identification of these subsets would otherwise be unattainable using protein‐, informatics‐, or omics‐based methods.
We identified five previously unknown LXR‐dependent regulators of Abca1 and demonstrated that four of these proteins control Abca1 expression in response to LXR ligand stimulation. We have established for the first time, to our knowledge, that NCOA5 functions as an LXR corepressor to attenuate Abca1 expression. In our PE‐QMS experiments, NCOA5 binding was twofold lower but not undetectable in the absence of LXR ligand stimulation. This may explain the reduction in Abca1‐luciferase expression by NCOA5 even in the absence of ligand (Fig 2C). Notably, our primary BMM ChIP and expression analyses indicate a ligand‐dependent NCOA5 occupancy and function at the Abca1 locus when LXR is present.
TLRs recognizing endogenous ligands transmit inflammatory signals in atherosclerosis; however, their role is complicated by their expression and function in multiple cell types. TLR2 functions in endothelial cells to recruit monocytes/macrophages to plaques to facilitate inflammation (Mullick et al, 2005, 2008; Monaco et al, 2009). TLR4 also promotes inflammation leading to atherosclerosis, albeit through a macrophage centric mechanism (Xu et al, 2001; Michelsen et al, 2004; Stewart et al, 2010).
Recent studies have also identified a role for TLR3 signaling in promoting atherosclerosis (Zimmer et al, 2011). TLR3 can sense RNA released from necrotic cells as a result of the inflammatory process (Kariko et al, 2004; Cavassani et al, 2008; Ahmad et al, 2010; Baiersdorfer et al, 2010). Loss of TICAM1/TRIF function, the downstream adaptor for TLR3, either systemically or specific to the hematopoietic system, alleviates inflammation and reduces atherosclerosis (Lundberg et al, 2013; Richards et al, 2013). Moreover, bone marrow‐specific deletion of Tlr3 produces a similar phenotype (Lundberg et al, 2013), suggesting that the macrophage TLR3 signaling pathway could be an important mediator of inflammation in atherosclerosis. Conversely, global Tlr3 deletion exacerbates lesion size, indicating an additional role for non‐bone marrow‐derived cells in atherosclerosis (Cole et al, 2011; Richards et al, 2013).
Signal crosstalk between the pro‐inflammatory TLR3 and anti‐inflammatory LXR pathways has previously been reported to induce IRF3 expression, which competes with LXR for the coactivator EP300, resulting in reduced Abca1 transcription (Castrillo et al, 2003). Our results demonstrate that in addition to competition for a shared coactivator, TLR3–LXR signal crosstalk actively recruits the corepressor NCOA5 to an LXR‐regulated promoter to repress Abca1 gene expression.
Notably, the repression of Abca1 following PolyIC treatment alone is independent of NCOA5. Moreover, LPS‐mediated repression of Abca1 is completely independent of NCOA5. One possible explanation for this is a regulatory event, such as a post‐translational modification of NCOA5, which specifically occurs when both the TLR3 and LXR pathways, but not the TLR4 and LXR pathways, are simultaneously activated (Fig 7). TLR4 induction results in widespread phosphoproteome changes (Weintz et al, 2010), so it is reasonable to hypothesize that phosphoproteome changes would occur downstream of TLR3. Alternatively, there could be distinct but complementary events mediated by the TLR3 and LXR pathways that target NCOA5 function. For instance, activation of TLR3 signaling could result in modification of NCOA5, while ligand binding to LXR could permit binding of modified NCOA5 to LXR. Castrillo et al (2003) observed the reduction in Abca1 expression by these signals was independent of IFNαβ, which may further restrict potential candidates. Future quantitative mass spectrometry studies will be critical in delineating these mechanisms.
We also discovered that NCOA5 binds the Abca1 promoter and retains a repressive function in the absence of LXR. An intriguing hypothesis is these proteins are part of a larger transcriptional complex and that NCOA5 might associate with LXR and another constituent. This additional interaction would be sufficient to recruit NCOA5 in the absence of LXR, such as in LXR−/− BMMs. Future experiments will focus on identifying this factor, using our list of candidate regulatory proteins identified by PE‐QMS.
Several LXR cofactors reported to modulate its activity also regulate chromatin structure, including p300, SMARCA4, NCOA6, and the NCOR complex (Castrillo et al, 2003; Wagner et al, 2003; Huuskonen et al, 2004, 2005; Lee et al, 2008; Jakobsson et al, 2009). We detected LXR ligand‐stimulated binding of the chromatin regulators SMARCC2, INO80, and HIRA to the Abca1 promoter by PE‐QMS (Fig 1C). Despite the fact we targeted a region of the Abca1 locus devoid of nucleosomes for PE‐QMS, nucleosomes are present adjacent to this region (Supplementary Fig S2), suggesting the binding of SMARCC2, INO80, and HIRA may be important for remodeling these regions or perhaps for long‐distance interactions. However, we cannot discount the possibility of a non‐chromatin‐related role for these proteins, such as that described for the NCOR complex component HDAC3 (Sun et al, 2013).
PE‐QMS is not without its limitations. Chromatin structure is an important regulator of gene expression and can affect transcription factor binding. We controlled for this by using an open chromatin region, as well as confirming NCOA5 genome localization by ChIP. However, recent studies are beginning to show the feasibility of capturing and identifying chromatin‐associated proteins in their native context (Dejardin & Kingston, 2009; Pourfarzad et al, 2013; Alabert et al, 2014), and it will be attractive to apply these approaches to the analysis of low copy number, dynamic, chromatin‐associated complexes in the future.
Another limitation of this study is we cannot be completely certain the full compendium of proteins bound to the Abca1 promoter was identified. For instance, MED15 is part of the large multi‐subunit Mediator complex. The lack of detection of other subunits in this complex may reflect their instability on this promoter under our experimental conditions, or duty cycle limitations of the data‐dependent mass spectrometry strategy employed. Another recent study using a similar enrichment strategy identified four Mediator subunits bound to their promoter sequences (Foulds et al, 2013). Recent advances in mass spectrometry‐based protein identification technologies with improved sensitivity and reproducibility, such as targeted (Mirzaei et al, 2013) and data‐independent peptide identification approaches (Gillet et al, 2012; Egertson et al, 2013; Lambert et al, 2013), hold great promise for overcoming these limitations and are currently being explored.
Whether aberrant targeting of NCOA5 to the Abca1 promoter to repress its expression exacerbates disease progression remains an important question with therapeutic relevance. Impairment of cholesterol efflux in vivo promotes foam cell formation and enhances atherosclerosis (van Eck et al, 2002; Tangirala et al, 2002; Yvan‐Charvet et al, 2007). Moreover, macrophages display an enhanced inflammatory response to TLR signals in the absence of cholesterol transporter gene expression (van Eck et al, 2002; Yvan‐Charvet et al, 2008; Zhu et al, 2008). Taken together, this suggests that NCOA5 activity could potentiate the chronic inflammatory process responsible for atherosclerosis.
Materials and Methods
Macrophage isolation and cell culture
All experiments were performed in accordance with the Institute for Systems Biology and Seattle Biomedical Research Institute Institutional Animal Care and Use Committees. Mice were euthanized by CO2 asphyxiation. C57BL6, Nr1h3−/−, Nr1h2−/− mice were obtained from The Jackson Laboratory, and the latter two strains were crossed in house to generate LXR−/− double knockouts. BMMs were isolated from 8‐ to 12‐week‐old females as previously described (Gilchrist et al, 2006). Bone marrow was collected from femurs with complete RPMI 1640 media (10% FBS, 100 U/ml penicillin and streptomycin, 2 mM L‐glutamine) (Corning Cellgro; Life Technologies) supplemented with 50 ng/ml recombinant human macrophage colony‐stimulating factor (rhM‐CSF; Peprotech). Collected bone marrow was grown on non‐tissue culture‐treated dishes for 6 days, then plated on tissue culture dishes, and stimulated with T0901317 (CAS 293754‐55‐9; Cayman Chemical), PolyIC (6 μg/ml; GE Healthcare), and/or LPS (10 ng/ml; Salmonella minnesota; List Biologicals) such that BMMs were harvested on day 7. All ligands were administered simultaneously for combined treatments. RAW 264.7 macrophages (ATCC TIB‐71) and Phoenix cells (Kinsella & Nolan, 1996) were cultured in complete RPMI without rhM‐CSF. Cell lines were tested for the absence of mycoplasma contamination.
Promoter enrichment‐quantitative mass spectrometry (PE‐QMS)
A promoter region of Abca1 (chr4:53,172,714‐53,173,035; NCBI Build 37; mm9) was used to enrich for associated gene regulatory complexes from RAW 264.7 macrophage nuclear extracts (Dignam et al, 1983). The promoter region was PCR amplified with a biotinylated forward primer and bound to NeutrAvidin–agarose beads. Following blocking, immobilized promoters were incubated with nuclear extracts in 20 mM Hepes, pH 7.4, 10% glycerol, 60 mM NaCl, 1 mM EDTA, 5 mM MgCl2, 2 mM DTT, and 0.05% NP‐40 for 2 h. Promoter–protein complexes were washed three times with the above buffer and then twice with the above buffer lacking NP‐40 and then eluted twice with 15 mM Tris, pH 8.3, 600 mM NaCl, 5.5 M urea, and 1 mM EDTA for 30 min at 37°C each. Eluted proteins were digested with trypsin and purified using C18 reversed‐phase chromatography. Orthogonal label‐free and isotope labeling strategies were separately employed for protein quantification. Labeled peptides were subsequently fractionated by Partisphere strong cation exchange chromatography. Purified peptides were separated by online reversed‐phase HPLC over an increasing gradient of acetonitrile and analyzed by mass spectrometry (LTQ‐Orbitrap Velos and LTQ‐Orbitrap XL). Raw output files were processed with GeneData Expressionist and Mascot (label free) or the Trans‐Proteomic Pipeline and X!Tandem (labeled). A reverse sequence database was used to assess FDR. Data were manually inspected to ensure accuracy of identifications and quantifications. Data were filtered to remove single peptide identifications, along with proteins not satisfying a 1% FDR cutoff. In addition, proteins corresponding to genes not expressed in primary BMMs were removed (Ramsey et al, 2008; Gold et al, 2012). Using correlated quantitative information, we identified 79 promoter‐associated proteins in the ligand‐stimulated experiments. We used a 1.5‐fold cutoff as described in the Results to subdivide proteins by their ligand dependency and their LXR dependency. A detailed description of this methodology is available in the Supplementary Materials and Methods.
GO enrichment and network analysis
GO enrichment analysis was performed using DAVID (P < 0.01) (Huang da et al, 2009). All expressed transcripts from unstimulated mouse primary BMMs (Ramsey et al, 2008; Gold et al, 2012) were used as background to calculate enrichment. Protein–protein network analysis was performed using ingenuity pathway analysis (Ingenuity) and GeneMANIA (Mostafavi et al, 2008), filtering for only physical associations. The network was filtered to include only LXR‐dependent regulators and their associations with the immune system.
Total RNA was isolated using TRIzol Reagent (Life Technologies) according to manufacturers’ protocol. RNA was reverse transcribed using random primers and Superscript II (Life Technologies) according to manufacturers’ protocol. cDNA was analyzed by real‐time PCR using Taqman Gene Expression assays (Life Technologies) (Supplementary Table S3). Data were acquired using a 7900HT Fast Real‐Time PCR System (Life Technologies) and were normalized to Eef1a1 transcript expression within individual samples.
Abca1‐luciferase was constructed by amplifying the 53,173,035–53,172,714 regulatory sequence (same sequence used for PE‐QMS) from mouse gDNA and cloning it into pGL3‐basic (Promega). Ncoa5‐∆NT was designed by removing the NH2‐terminal 840 nt from the full‐length cDNA (Thermo Scientific). RAW 264.7 macrophages were transfected using Lipofectamine 2000 (Life Technologies) according to manufacturers’ protocol. Cells were lysed with 1× passive lysis buffer, and luciferase assays were analyzed on a GloMax 96 Microplate Luminometer (Promega). All luciferase activity was normalized to co‐transfected Renilla luciferase and to the pGL3‐basic empty reporter construct.
Chromatin immunoprecipitation (ChIP) assays
1.5 × 107 BMMs were crosslinked with 1% formaldehyde in PBS for 10 min at room temperature, quenched with 125 mM glycine for 5 min at room temperature, and then washed 3× with ice cold PBS. Cells were scraped, pelleted, and lysed in RIPA buffer (10 mM Tris, pH 7.4, 140 mM NaCl, 0.1% SDS, 1% Triton X‐100, 1% Na‐deoxycholate). Extracts were sonicated 5 × 1 min using an Ultrasonic Processor 130 W at 3 W and 35% output. NCOA5 (A300‐790A; Bethyl Laboratories), RNAPII unmodified/pSer5 (05‐623; Millipore), RNAPII pSer2 (ab5095; Abcam), rabbit IgG (sc2027; Santa Cruz Biotechnology), or mouse IgG antibodies (sc2025; Santa Cruz Biotechnology) were pre‐conjugated to Protein G Dynabeads (Life Technologies) in 0.5% BSA in PBS and used to immunoprecipitate sheared chromatin complexes overnight. Complexes were washed twice with Wash Buffer I (20 mM Tris, pH 7.4, 150 mM NaCl, 0.1% SDS, 1% Triton X‐100, 2 mM EDTA), once with Wash Buffer II (10 mM Tris, pH 7.4, 250 mM LiCl, 0.5% NP‐40, 0.5% Na‐deoxycholate, 1 mM EDTA), and once with TE (10 mM Tris, pH 8.0, 1 mM EDTA) or washed three times with RIPA buffer and once with TE. Complexes were eluted twice with 1% SDS in TE at 65°C for 15 min, and eluates were combined. Protein was digested, and crosslinks reversed (40 mM Tris, pH 8.0, 10 mM EDTA, 240 mM NaCl, 25 μg Proteinase K) at 55°C for 2.5 h and then 65°C overnight. DNA was PCR‐purified (Qiagen) prior to qPCR analysis using SYBR Green (Life Technologies) (Supplementary Table S3).
Pulldowns, immunoprecipitation, and immunoblots
Recombinant GST and GST‐LXRα proteins were expressed from the pGEX4T vector (GE Healthcare) and purified from isopropyl‐β‐D‐thiogalactoside (IPTG; 2 mM)‐induced E. coli BL21(DE3)pLysS using glutathione‐Sepharose beads (GE Healthcare). Full‐length and mutant Ncoa5 cDNAs were generated by PCR, cloned into pEF6‐V5‐His (Life Technologies), and transcribed/translated in vitro using the TNT T7 Quick Coupled System (Promega) according to manufacturer's protocol. In vitro translated proteins (5 μl) were then incubated with 2.5 μg GST proteins pre‐bound to glutathione‐Sepharose beads for 2 h at 4°C. Beads were washed 3× with NETN Buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 0.1% NP‐40, 1 mM EDTA) and boiled in LDS Sample Buffer (Life Technologies) to elute complexes prior to Western blot analysis. For IPs, RAW 264.7 macrophages stably expressing Protein C‐tagged LXRα were maintained in complete RPMI plus 10 μg/ml blasticidin. The Protein C tag construct was generated and kindly provided by Adrian Ozinsky. Co‐IPs were performed from nuclear extracts. Due to the Protein C tag, EDTA was omitted from all buffers except elution buffer. Protein C‐agarose affinity matrix (Clone HPC4; Roche Applied Science) was used for IPs. Protein was eluted using 5 mM EDTA and analyzed by Western blotting. Antibodies used for Western immunoblotting were NCOA5 (A300‐790A; Bethyl Laboratories), LXR (sc1000; Santa Cruz Biotechnology), ABCA1 (ab18180; Abcam), V5‐Tag (MCA1360; AbD Serotec), and tubulin (Clone DM1A; Sigma‐Aldrich). Blots were imaged using a CCD camera (FluorChem E; Protein Simple) and quantified using AlphaView SA software.
miR‐30‐based shRNAs were designed as described (Paddison et al, 2004), based on siRNA sequences from Thermo Scientific, and cloned into the LMP retroviral vector (Dickins et al, 2005) (Supplementary Table S3). A non‐silencing negative control sequence was used to ensure specificity (Thermo Scientific). Retrovirus was prepared by transfecting the Phoenix Ecotropic packaging cell line (Kinsella & Nolan, 1996) with the above constructs using Lipofectamine 2000 (Life Technologies). After 48 h, viral supernatant was removed, 0.44 μm filtered, and used to infect day 2 BMMs. Infections were supplemented with 6 μg/ml polybrene (Millipore) and 50 ng/ml rhM‐CSF and performed for 2 h at 1,800 rpm at 32°C, then for another 2 h at 37°C. Viral media were replaced with complete RPMI media for another 72 h, and then, infected cells were selected with 5 μg/ml puromycin (Invivogen) for 5 days. For overexpression studies, Ncoa5 cDNA was cloned into the LMP retroviral vector and infections were performed in day 4 BMMs as indicated above. Infected cells were harvested on day 7.
Cholesterol efflux assays
Macrophages were labeled with 1 μCi/ml [1,2,3H(N)]‐cholesterol (Perkin Elmer) in RPMI containing 0.2% BSA, 50 ng/ml MCSF, 5 μg/ml puromycin, and 2 μg/ml acyl‐coenzyme A cholesterol acyltransferase (ACAT) inhibitor (Sandoz 58‐035; Sigma) for 24 h. Macrophages were washed and equilibrated in RPMI containing 0.2% BSA, 5 μg/ml puromycin, and 2 μg/ml ACAT inhibitor for 6 h. Stimulations with vehicle, 1 μM T0901317, or 1 μM T0901317 + 6 μg/ml PolyIC were performed simultaneously with equilibration. Efflux was measured following wash and incubation for 4 h with equilibration media containing 10 μg/ml mouse APOA1 protein (IgG Fc tagged; Life Technologies) or IgG protein (Meridian Life Science) as a control. Supernatants were clarified by centrifugation, and one‐fifth measured by liquid scintillation counting. Cells were lysed in 0.1 N NaOH + 0.2% SDS and 1/20 measured by liquid scintillation counting. After adjusting for dilutions, percent efflux was calculated by dividing the 3H‐cholesterol effluxed to the media by the total 3H‐cholesterol present in the cells and media. Specific efflux to APOA1 was calculated in response to each agonist by subtracting out any background efflux to IgG.
All error bars are presented as standard error of the means. Statistical significance of means was calculated using the two‐tailed Student's t‐test, using a cutoff of P < 0.05. The one exception to this was the GO enrichment analysis, where we used a cutoff of P < 0.01 in a modified version of the Fisher's exact test called the EASE score (Huang da et al, 2009).
Raw data deposition
All raw proteomics data were deposited in Peptide Atlas under the identifiers PASS00515, PASS00516, and PASS00517.
MAG designed, performed, and analyzed experiments; SAR performed network analysis; IP assisted with shRNA experiments; AA and JAR supervised the studies; and MAG, ESG, AA, and JAR conceived experiments and wrote the manuscript.
Conflict of interest
The authors declare that they have no conflict of interest.
Supplementary Figure S1
Supplementary Figure S2
Supplementary Figure S3
Supplementary Figure S4
Supplementary Figure S5
Supplementary Figure S6
Supplementary Figure S7
Supplementary Figure S8
Supplementary Materials and Methods
Supplementary Tables S1 and S2
Supplementary Table S3
We thank Peter Askovich and David Shteynberg for helpful assistance with mass spectrometry analysis software, Jay Heinecke and Angela Irwin for helpful assistance with the sterol efflux protocol, and Alan Diercks for critical reading of the manuscript. Research reported in this publication was supported by the Canadian Institutes of Health Research Fellowship MFE‐112984 (to M.A.G.), the National Heart, Lung, and Blood Institute Award Number HL098807 (to S.A.R.), the National Institute of Allergy and Infectious Diseases Award Numbers R01 AI032972, R01 AI025032, and U19 AI100627 (to A.A.), and the National Institute of General Medical Sciences Center for Systems Biology Award Number 2P50 GM076547 (J.A.R.).
FundingCanadian Institutes of Health Research Fellowship MFE‐112984
- © 2015 The Authors