Loss of Bone Marrow Bcl6 Results in Severe Cholesterol-Dependent Atherosclerosis and Xanthomatous Tendonitis
To examine the impact of Bcl6 on atherogenesis we found that, in contrast to whole body knockouts, mice transplanted with Bcl6 null bone marrow show long term viability (Figures S1A–B
) (Dent et al., 1997
; Ye et al., 1997
), leading us to reconstitute γ-irradiated Ldlr−/−
mice with Bcl6−/−
(BMT-KO) or wild type (BMT-WT) marrow. Multiple cohorts were then exposed to atherogenic diet for up to 16 weeks. First, we analyzed the development of lesions after 8 weeks. While atherosclerosis was minimal in BMT-WT aortas, BMT-KO mice already had plaques along branch points for intercostal arteries and 3-fold more disease (Figures S1C and 1A
). Similarly, aortic root cross-sections demonstrated a nearly 3-fold increase in lesion area in BMT-KO mice compared to controls (). Remarkably, assessment of aortic root sections with Masson trichrome staining and immunostaining revealed massive lesions in BMT-KO mice compared to controls, marked by striking erosions, MØ infiltration, and smooth muscle cells (). Next, we extended our analysis to mice fed atherogenic diet for an additional 5 or 8 weeks, and BMT-KO aortas revealed dramatically more severe atherosclerosis (). Overall, BMT-KO mice developed up to 5-fold more aortic lesion coverage than BMT-WT control mice after 13 or 16 weeks of atherogenic diet feeding and a 1.6-fold increase in aortic root lesion area at 16 weeks (). The modest increase in root lesions compared to entire aortas suggests that BMT-KO aortic root lesions were already near maximal by 8 weeks (). Consistent with their severe disease, aortic roots in BMT-KO mice at later time points revealed noticeably increased lesion complexity, including extensive connective tissue deposition () and significantly larger necrotic cores, resembling advanced human atherosclerosis ().
Figure 1 Bcl6 represses atherogenesis and xanthomatous tendonitis. (A, B) Quantification of en face aorta and aortic root lesions in BMT mice exposed to atherogenic diet. (C) Aortic valve sections from BMT mice fed atherogenic diet for 8 weeks, showing (top) Masson (more ...)
Strikingly, BMT-KO mice also displayed disturbed movement due to abnormal distal limbs and paws () that were nearly 38% larger by volume than those of BMT-WT mice after 12 weeks of exposure to atherogenic diet (). Histological examination revealed xanthomatous tendonitis with surrounding fibrosis ( and S1D–F
) and MOMA-2-positive MØ infiltrates (Figure S1G
). Interestingly, evidence for tendon xanthomas in BMT-KO mice was even apparent by 8 weeks of exposure to atherogenic diet (Figure S1H
). Though rarely seen in mice, this pathology is common in LDLR-deficient patients (familial hypercholesterolemia) and predictive of increased cardiovascular risk (Civeira et al., 2005
; Goldstein et al., 2001
; Ishibashi et al., 1994
). Together, these findings not only demonstrate a critical role for Bcl6 to suppress early vascular lesion development but also atherosclerotic remodeling and tendon xanthoma.
In similar experiments, we performed transplantation of WT or KO bone marrow into wild type C57 mice (C57 BMT-WT and C57 BMT-KO mice). Remarkably, C57 BMT-KO mice on either standard diet for 30 weeks (Figure S2A–G
) or atherogenic diet for 14 weeks (Figure S2J–R
) develop no change in body weight or lipid profile, tendon pathology, or vascular lesions. These results suggest that hypercholesterolemia is critical to elicit BMT-KO-related pathology. Importantly, however, total cholesterol and triglyceride levels as well as lipoprotein fractions were similar between BMT-WT and BMT-KO mice (Ldlr−/−
recipients) on standard or atherogenic diets (Figure S1I-K
). Although overall food consumption did not differ, body weights often diverged after 8 weeks of dietary challenge and were lower in some BMT-KO cohorts by 13–16 weeks (Figures S1I and S1L
). Correspondingly, inguinal white adipose tissue, liver, and total body fat were reduced in some of these groups at 13 or 16 weeks (Figures S1I and S1M
). Overall, these findings indicate a subtle contribution of bone marrow Bcl6 to metabolic homeostasis in Ldlr−/−
mice, with weight loss coincident to severe atherosclerosis and tendon xanthoma in BMT-KO animals exposed to over three months of atherogenic diet.
Bcl6 Deletion De-represses Atherogenic Genes and Sensitizes Macrophages to Oxidized LDL Stimulation
MØs play a central role in atherogenesis, from early fatty streaks to advanced lesions. Previously, we performed genome-wide transcriptome analysis for Bcl6 in WT and KO MØs (Barish et al., 2010
). Further evaluation of this data set revealed not only a collection of Bcl6-repressed chemokines and cytokines, but mitogens, proteases, and clotting factors that collectively promote leukocyte recruitment, activation and proliferation, plaque remodeling and thrombosis. To confirm the microarray findings and explore the impact of atherogenic stimuli on Bcl6 regulated genes, we used quantitative PCR (qPCR) to assess mRNA levels in Bcl6 wild-type (WT), heterozygous (HET), and knockout (KO) cells with or without exposure to mmLDL (, S3A
). In unstimulated MØs, critical chemokines implicated in atherosclerosis were substantially de-repressed in KO cells including Ccl2
, and Cxcl1
(Zernecke and Weber). Moreover, even haploinsufficiency of Bcl6 is sufficient to create a sensitized, pro-inflammatory state as maximal or near-maximal expression changes were seen following exposure of KO and HET cells to mmLDL and correlated closely to Ccl2 and Ccl7 protein levels (, S3A–B
). Csf1 is a pleiotropic MØ growth factor as well as chemoattractant and thus functions as a powerful pro-atherogenic mediator (Glass and Witztum, 2001
levels were fully increased by 30-fold in KO MØs and not further enhanced by mmLDL (Figure S3A
). Notably, WT and KO MØs were comparable in their ability to migrate to a Ccl2 or Ccl3 chemoattractant gradient in vitro
. (Figure S3C
). Yet in a model of sterile peritonitis, transplanted mice devoid of bone marrow Bcl6 generated 2-fold higher leukocyte counts in peritoneal exudates, consistent with the increased elaboration of chemoattractants from KO inflammatory cells (Figure S3D
). Together, these results suggest that Bcl6 is a core regulatory component whose loss promotes disease, in part, by de-repressing a chemotactic network.
Figure 2 Bcl6 regulates pro-atherogenic genes in vitro and in vivo. (A) qPCR assessment of gene expression in WT, HET, and KO MØs ± exposure to mmLDL. (B) Expression analysis of flushed bone marrow from BMT-WT (n = 6) and BMT-KO (n = 7) mice after (more ...)
In addition to inflammatory factors, key plaque remodeling and thrombosis genes were identified as Bcl6 targets. Platelet derived growth factors (PDGFs) are potent smooth muscle mitogens that accelerate vascular stenosis (Andrae et al., 2008
). Notably, Pdgfα
was up-regulated 3-fold in KO MØs either in the presence or absence of mmLDL stimulation (). Tissue remodeling matrix metalloproteinases (MMPs) destabilize atherosclerotic plaques and initiate thrombus formation, and human genetic studies link variants of MMP12 and 13 to vascular disease (Ye, 2006
were strikingly (10 and 17 fold, respectively) de-repressed in Bcl6 KO MØs and, in particular, Mmp13
was further stimulated to levels nearly 40-fold over unstimulated WT cells by mmLDL (). Additionally, the urokinase plasminogen activator (Plau) is expressed in arterial wall MØs and, when overexpressed, causes accelerated atherosclerosis, vascular occlusion, and plaque rupture (Cozen et al., 2004
). Remarkably, Plau transcripts and protein were strongly up-regulated in Bcl6-deficient MØs (, S3E
), and its mRNA expression further increased by mmLDL. The lysosomal proteases cathepsin E (Ctse
) and cathepsin L (Ctsl
), which is associated with necrotic core formation and plaque destabilization (Li et al., 2009
), were also moderately up-regulated in KO cells (Figure S3A
). Finally, tissue factor (F3
), which promotes thrombosis (Toschi et al., 1997
), was more than 20-fold higher in KO compared to WT cells. Additional exposure to mmLDL synergized with Bcl6 deficiency to increase F3
by 60-fold over levels in stimulated control MØs (). Importantly, loss of Bcl6 also potentiated the effect of other pro-atherogenic signals, including TLR-2 and -4 ligands, to stimulate the expression of aforementioned chemokines, cytokines, remodeling enzymes, and clotting factors (Figure S3F
and data not shown). Together, these findings suggest that Bcl6 limits oxidized LDL and TLR-induced inflammatory and tissue remodeling responses by constraining target gene activation.
Bcl6 Controls Atherogenic Gene Expression in vivo
To characterize the Bcl6-regulated inflammatory network in vivo
, we initially examined expression profiles in the bone marrow of transplanted mice. As expected, marrow Bcl6
was virtually undetectable in Bcl6−/−
transplant recipients (). Coordinately, Ccl2, Ccl7, Csf1, F3, Mmp12
were all significantly increased in atherogenic diet-fed Ldlr−/−
mice harboring Bcl6−/−
marrow compared to controls (). Similar expression changes were observed in chow or atherogenic diet fed C57Bl/6J transplanted animals (Figures S2H and S2S
). In white adipose tissue, which contains a significant population of MØs, expression of Ccl7
were also relatively elevated in BMT-KO mice (Figure S1N
). Moreover, Bcl6 regulated gene expression was clearly altered in the aortas of atherogenic diet-fed Ldlr−/−
transplanted mice, as levels of Ccl2, Ccl7, Cxcl1, Il-1α
, Mmp13, Plau, Ctse, Ctsl, Pdgfα
normalized to the MØ marker F4/80
were each significantly up in BMT-KO mice compared to controls () and resembled expression changes observed in KO MØs. Some of these differences were even observed in aortas of C57 BMT-KO mice relative to C57 BMT-WT controls on standard or atherogenic diets (Figures S2I and S2T
). Moreover, Ccl2 and Ccl7 protein were elevated in the plasma of BMT-KO mice (), and immunohistochemistry demonstrated increased Ccl2, Il-1α, Plau, and F3 protein within their aortic valve atherosclerotic lesions (), indicating that loss of bone marrow Bcl6 can promote vascular and systemic inflammation.
Since cholesterol metabolism plays a key role in atherosclerosis, we examined vascular expression of several principal regulators of cellular lipid homeostasis. Importantly, mRNAs of genes involved in MØ cholesterol uptake (Cd36 and Sra1), storage (Acat1), and efflux (Abca1, Abcg1, and Apoe) were similar in BMT-WT and BMT-KO mice (). Interestingly, examination of MØ cholesterol uptake in vitro
revealed a modest decrease in the ability of KO MØs to internalize oxidized or acetylated LDL (Figure S3G
) and a mild reciprocal defect in their HDL or ApoA1-mediated cholesterol efflux (Figure S3H
). These differences appear minor and compensated, as levels of total and esterified cholesterol were similar both in WT and KO foam cells derived in vitro
upon 24-hour exposure to acetylated LDL (Figure S3I
), and in peritoneal MØs from atherogenic diet-fed BMT-WT and BMT-KO mice (Figure S3J
). The minimal impact of Bcl6 on foam cell cholesterol homeostasis (Figure S3G–J
) and circulating lipids (Figure S1I–K
) supports, by exclusion, the key role for inflammation to contribute to the severe atherosclerosis and tendon xanthomas in BMT-KO mice.
Bcl6 Recruits SMRT and NCoR to Repress Atherogenic Inflammation
Although we recently described the Bcl6 cistrome (Barish et al., 2010
), the molecular composition of cofactors on these thousands of binding sites remained unknown. In light of reports suggesting roles for SMRT and NCoR in macrophage inflammatory control (Ghisletti et al., 2009
; Ogawa et al., 2004
), we wondered whether these co-repressors contributed to Bcl6 anti-inflammatory repression. Previous work described the development of a potent, modified therapeutic peptide from the core of SMRT’s BCL6 interaction domain (termed RI-BPI) that activates BCL6 target genes by specifically blocking interaction between BCL6 and the co-repressors NCoR or SMRT to kill lymphoma cells (Cerchietti et al., 2009
). Addition of RI-BPI to bone marrow-derived MØs resulted in rapid, and in many cases dramatic, induction of pro-atherogenic gene transcripts including Ccl2
, Cxcl1, Cxcl2
, Il-1α, Il-6, Mmp-13, Pdgfα
, and Plau
compared to a similar cell-penetrating control peptide (CP) lacking the BCL6 binding domain (). Highly similar expression changes were also observed in primary human bone marrow derived MØs exposed to RI-BPI (), indicating that BCL6 controls a conserved inflammatory network. To better understand the contribution of SMRT and NCoR to Bcl6 repression, we performed genome-wide microarray analysis of mouse MØs exposed for 12 hours to RI-BPI versus a control peptide (CP), which identified over four thousand differentially expressed transcripts. Notably, RI-BPI treatment recapitulated 35% (892/2579 transcripts) of the expression changes incurred by genetic loss of Bcl6 (Barish et al., 2010
). Furthermore, functional classification of RI-BPI regulated transcripts revealed significant enrichment for TLR, cytokine-cytokine receptor interaction, and inflammatory cell signal transduction pathways (). Collectively, these findings indicated that the atherosclerotic phenotype of BMT-KO mice involves loss of Bcl6-dependent NCoR and/or SMRT repression and, conversely, pro-inflammatory changes observed in NCoR or SMRT-deficient MØs could in part be attributable to de-repression of Bcl6.
Figure 3 Bcl6 represses atherogenic genes using SMRT and NCoR. (A, B) Gene expression in wild type mouse (A) and human (B) MØs following 12-hour treatment with 5 μM of control (CP) or Bcl6 inhibitor peptide (RI-BPI). (C) Venn diagram comparing (more ...)
Table 1 Functional analysis of gene expression changes in MØs exposed to RI-BPI peptide (5 μM) for 12 hours and genes annotated with Bcl6-SMRT and/or Bcl6-NCoR in MØs based on ChIP-seq. Top scoring KEGG categories were selectively represented. (more ...)
Next, we sought to determine the individual contributions of SMRT and NCoR to the suppression of Bcl6 target genes. Notably, SMRT and NCoR are highly homologous and, because they interact with many transcription factors, their specificity has previously been difficult to define. ChIP qPCR revealed that both SMRT and NCoR localized to Bcl6 binding sites along pro-atherogenic genes but were reduced along Ccl2, Csf1, Il-1α, and Mmp12 in MØs lacking Bcl6 (Figure S4A
). Then, we performed ChIP-sequencing to uncover the SMRT and NCoR cistromes. Using a false discovery rate of <0.01, we identified 32,435 and 33,522 binding sites for SMRT and NCoR, respectively, in wild type MØs (). Overall, the SMRT and NCoR cistromes overlapped at 14,480 binding sites, reflecting at least 45% redundancy in their genomic localizations (). Motif analysis () revealed putative binding sites for the MØ lineage determining factors Pu.1, Jun, and C/EBP, which mark MØ enhancers (Ghisletti et al., 2010
; Heinz et al., 2010
), as three of the most common consensus sequences near SMRT and NCoR binding peaks. Accordingly, the vast majority (> 90%) of SMRT or NCoR binding was promoter-distal in distribution, although NCoR was almost twice as concentrated on promoters (Figure S4B–C
). Other top-scoring motifs included those for known SMRT and NCoR interaction partners including Runx1 and the nuclear receptor RXR, respectively ().
Next, we determined cistrome sub-sets for SMRT and NCoR that were directly attributable to interactions with Bcl6, using two criteria (Figure S4D
). First, we ChIP-sequenced SMRT and NCoR in Bcl6 KO MØs to normalize the sequencing data obtained in WT cells, thereby identifying SMRT and NCoR sites that were genetically dependent upon Bcl6. Surprisingly, nearly half of SMRT and NCoR binding sites were reduced or lost in Bcl6 KO cells, indicating that loss of Bcl6 may directly or indirectly result in widespread genomic de-repression. Second, we analyzed for direct DNA sites of interaction between Bcl6 and SMRT and NCoR based on co-localization. Previously, we generated the Bcl6 cistrome in MØs by aligning Bcl6 ChIP-seq reads to the mouse genome using the Eland alignment tool (Barish et al., 2010
). More recently, we found the Bowtie aligner to be more efficient, improving ChIP-seq peak calling. Re-analysis of Bcl6 ChIP-seq reads using Bowtie and otherwise unchanged parameters more than doubled the number of identified Bcl6 binding sites in quiescent MØs (from 6,655 to 14,756 sites). ChIP qPCR of randomly selected, additional Bcl6 binding peaks identified from this re-analysis confirmed 90% of these sites (Figure S4E
). Using Bowtie alignments to generate each cistrome and 200 DNA base pair co-localization windows, we identified 3,665 Bcl6-SMRT and 3,664 Bcl6-NCoR complexes across the genome. Remarkably, almost 70% of Bcl6-SMRT and Bcl6-NCoR sites were shared by both co-repressors (), and motif analysis of these sub-cistromes produced the same top-scoring sequences for Pu.1 and Bcl6 (). These results reflect a high degree of overlap between SMRT and NCoR in the repression of Bcl6, which together control 1/3 of the Bcl6 cistrome (4,795/14,756 Bcl6 binding sites).
Importantly, additional analysis of the Bcl6-SMRT and Bcl6-NCoR sub-cistromes revealed dramatic enrichment for pro-inflammatory, atherogenic genes and pathways. Classification of Bcl6-SMRT and/or Bcl6-NCoR annotated genes identified TLR, cytokine-cytokine receptor interaction, and inflammatory cell signaling transduction pathways as the most highly enriched functional groups, highly reminiscent of our functional analysis for RI-BPI-regulated gene expression (). In fact, among genes annotated with one or more Bcl6-SMRT/NCoR binding site, the expression of 87% (2,677/3,076) was impacted by RI-BPI, and 63% of genes whose expression was altered by 12-hour exposure to RI-BPI were directly bound by Bcl6-SMRT and/or Bcl6-NCoR complexes. Moreover, although < 20% of genes annotated with Bcl6 binding sites demonstrate corresponding changes of expression in Bcl6 KO MØs (Barish et al., 2010
and data not shown), the majority (720/1,279) of such transcriptionally-altered, direct target genes contain Bcl6-SMRT/NCoR binding sites. These results suggest that SMRT and NCoR are markers for transcriptionally relevant Bcl6 binding sites, and the Bcl6-SMRT/NCoR sub-cistromes may have dominant functional roles to impact the MØ Bcl6 regulatory program.
Direct examination of sequencing tracks along key Bcl6-repressed atherogenic genes such as Ccl2/Ccl7 and Plau illustrated strong Bcl6-associated recruitment of SMRT and NCoR (). Even SMRT and NCoR peaks that did not co-localize with Bcl6 were in some instances lost in KO cells, presumably due to secondary impacts of Bcl6 de-repression (). At other genomic loci not regulated by Bcl6, such as Slain2, SMRT and NCoR binding was similar in WT and KO cells (Figure S4F
). Furthermore, although SMRT and NCoR binding peaks were highly overlapping, recruitment to some sites was co-repressor-specific, such as NCoR recruitment to the Nos2 (iNos) promoter, consistent with other reports (Ghisletti et al., 2009
) (Figure S4G
). Previously, we found that Bcl6 and NF-κB antagonize the expression of thousands of genes by proximately binding enhancers (Barish et al., 2010
). Given the strong enrichment for Bcl6-SMRT and Bcl6-NCoR at inflammatory genes, we wondered to what extent these sub-cistromes oppose NF-κB. Using our previously described NF-κB cistrome of over 31,000 binding sites in MØs, we found that the Bcl6, SMRT, and NCoR cistromes demonstrate substantial proximal overlap with NF-κB, corresponding to 22, 30, and 40% of their respective binding sites (). Remarkably, however, the Bcl6-SMRT and Bcl6-NCoR sub-cistromes were still further enriched for proximally bound NF-κB, corresponding to nearly 60% of these sites (). Together, these findings define a particular role for Bcl6-SMRT and Bcl6-NCoR complexes in opposing NF-κB driven inflammatory and tissue remodeling genes, which are de-repressed in highly atherosclerosis-prone Bcl6 KO-BMT animals.
While robust acute inflammatory responses are critical for survival, unchecked inflammation underlies a broad array of pathologies, including atherosclerosis. Here, we identify a critical role for Bcl6 to suppress oxidized LDL and TLR-induced inflammation and atherogenesis. Accordingly, loss of transcriptional repression in Bcl6-deficient bone marrow produces amongst the most aggressive atherosclerosis reported in a Ldlr−/− bone marrow transplantation model, along with xanthomatous tendonitis, a complication rarely seen in mice but virtually pathognomonic for familial hypercholesterolemia. Although additional studies will be needed to fully dissect the Bcl6 repression network, our expression and genomic analyses identify SMRT and NCoR as major components in Bcl6 opposition to atherogenic and NF-κB-driven inflammation. The pathological consequences of losing Bcl6 repression raise questions about the basis of chronic inflammatory disease and its treatment. Do such maladies occur through the aberrant signaling of activation pathways such as NF-κB, or could loss of repression play an equal or alternate role? Moreover, could enforcement of Bcl6 repression be leveraged therapeutically? Future investigations of Bcl6 or other repression pathways controlled by SMRT and NCoR may offer insight into these questions.