Transcription Factor MAFF (MAF Basic Leucine Zipper Transcription Factor F) Regulates an Atherosclerosis Relevant Network Connecting Inflammation and Cholesterol Metabolism

Abstract

Background: Coronary artery disease (CAD) is a multifactorial condition with both genetic and exogenous causes. The contribution of tissue-specific functional networks to the development of atherosclerosis remains largely unclear. The aim of this study was to identify and characterize central regulators and networks leading to atherosclerosis.

Methods: Based on several hundred genes known to affect atherosclerosis risk in mouse (as demonstrated in knockout models) and human (as shown by genome-wide association studies), liver gene regulatory networks were modeled. The hierarchical order and regulatory directions of genes within the network were based on Bayesian prediction models, as well as experimental studies including chromatin immunoprecipitation DNA-sequencing, chromatin immunoprecipitation mass spectrometry, overexpression, small interfering RNA knockdown in mouse and human liver cells, and knockout mouse experiments. Bioinformatics and correlation analyses were used to clarify associations between central genes and CAD phenotypes in both human and mouse.

Results: The transcription factor MAFF (MAF basic leucine zipper transcription factor F) interacted as a key driver of a liver network with 3 human genes at CAD genome-wide association studies loci and 11 atherosclerotic murine genes. Most importantly, expression levels of the low-density lipoprotein receptor (LDLR) gene correlated with MAFF in 600 CAD patients undergoing bypass surgery (STARNET [Stockholm-Tartu Atherosclerosis Reverse Network Engineering Task]) and a hybrid mouse diversity panel involving 105 different inbred mouse strains. Molecular mechanisms of MAFF were tested in noninflammatory conditions and showed positive correlation between MAFF and LDLR in vitro and in vivo. Interestingly, after lipopolysaccharide stimulation (inflammatory conditions), an inverse correlation between MAFF and LDLR in vitro and in vivo was observed. Chromatin immunoprecipitation mass spectrometry revealed that the human CAD genome-wide association studies candidate BACH1 (BTB domain and CNC homolog 1) assists MAFF in the presence of lipopolysaccharide stimulation with respective heterodimers binding at the MAF recognition element of the LDLR promoter to transcriptionally downregulate LDLR expression.

Conclusions: The transcription factor MAFF was identified as a novel central regulator of an atherosclerosis/CAD-relevant liver network. MAFF triggered context-specific expression of LDLR and other genes known to affect CAD risk. Our results suggest that MAFF is a missing link between inflammation, lipid and lipoprotein metabolism, and a possible treatment target.

Keywords: atherosclerosis; chromatin immunoprecipitation; coronary artery disease; inflammation; lipopolysaccharides; mafF transcription factor; receptors, LDL.

Figure 1.. Study workflow:

Human and mouse atherosclerosis candidate genes were used to first model liver specific regulatory networks and second decipher key driver genes of gene regulatory networks in both species. Prediction of bioinformatics modeling was validated in human and mouse genetic studies as well as in in vitro and in vivo experiments. CAD: Coronary artery disease; GWAS: Genome wide association study.

Figure 2.. Liver specific regulatory subnetworks and their key driver genes.

The key driver analysis was performed on human and mouse networks respectively and the architecture of the illustrated network is based on both, mouse and human data. Key drivers are depicted as the largest nodes in the networks. All genes highlighted in solid green have already been studied to have a significant effect on atherosclerosis in genetically engineered mouse models. Human CAD GWAS candidate genes are highlighted in magenta. Key driver genes in grey need to be validated. Genes with both colors have an effect on atherosclerosis/CAD in human and mouse. Lower right: The MAFF network is the top ranked key driver gene network based on mouse data and closely connected to other human key driver subnetworks. Directionality between genes was based on the consensus of directional predictions from Bayesian networks constructed from different datasets, with the directionality predicted by the majority of studies shown. Red arrows indicate genes that are predicted to regulate MAFF, whereas green arrows indicate genes that are predicted to be regulated by the transcription factor MAFF. CAD: coronary artery disease; GWAS: Genome wide association study; Human KD: Human key driver gene; Mouse KD: Mouse key driver gene; MAFF: MAF BZIP Transcription Factor F.

Figure 3.. Human data from STARNET.

Correlation of mRNA expression levels of MAFF in human liver samples from the Stockholm-Tartu Atherosclerosis Reverse Network Engineering Task (STARNET) with A: LDLR and B: Sex. ** indicates p<0.01. LDLR: low-density lipoprotein receptor; MAFF: MAF BZIP Transcription Factor F; RPKM: Reads per kilobase million.

Figure 4.. Experimental studies.

A: In vitro results of Ldlr mRNA expression (in the following referred to as expression) after siRNA-knockdown of Maff compared to controls (vehicle) in mouse AML12 liver cells. B: In vitro results of LDLR expression after siRNA-knockdown of MAFF compared to controls (vehicle) in human Hep3b liver cells. C: In vitro results of Ldlr expression cells after Maff overexpression compared to controls (vehicle) in mouse AML12. D: In vivo results of liver Ldlr expression in Maff−/− mice compared to Maff+/− and WT mice. E: In vivo results of liver Maff expression in Maff WT mice 6 hours after LPS stimulation compared to controls (vehicle). F: In vivo results of liver Ldlr expression in Maff WT mice 6 hours after LPS stimulation compared to controls (vehicle). G: In vivo results of plasma Tnfa expression in Maff WT and Maff−/− mice 6 hours after LPS stimulation compared to controls (vehicle). *** indicates p<0.001, ** indicates p<0.01, * indicates p<0.05, ns indicates non-significant. Bonferroni correction was applied for multiple comparison. Ctrl: control group; KD: knockdown; LDLR: low-density lipoprotein receptor; LPS: lipopolysaccharide; MAFF: MAF BZIP Transcription Factor F; OE: overexpression; WT: wildtype.

Figure 5.. MAFF binding motif.

A: ChIP-seq data of human HepG2 cells supports potential binding of MAFF to genes in the MAFF subnetwork. Green edges indicate a binding motif was shared in the selected network genes, whereas red edges indicate that no known shared binding motif was found in the particular network genes. A matching binding motif was found in 20 out of 24 predicted interaction partners of MAFF. B: The matching motif among the MAFF network genes, which agrees with the previously known MAFF binding motif. The matching motif was identified using publically available ChIP-Seq data of the MAFF gene in human HepG2 cells from ENCODE. The height of the letter represents the frequency of the observed nucleotide in that position. C: Presence of the MAFF binding motif (small boxes below) upstream and within the LDLR gene (highlighted in blue).

Figure 6.. Maff binding partners under different conditions.

Volcano plot of the p-values (y-axis) vs. the log2 protein abundance differences (x-axis) of Maff binding partners in AML12 cells identified by ChIP-MS under (A) homeostatic conditions, (B) after Maff siRNA knockdown (which led to a 91% decrease on protein level), (C) LPS stimulation and (D) Maff siRNA knockdown in combination with LPS stimulation. Significant Maff interaction partners were highlighted in blue. Enrichment of binding partners is provided as fold difference compared to negative control (IgG) in panel A and C and compared to control (Maff WT) after siRNA knockdown in panel B and D. C: control (Maff WT); IgG: nonspecific IgG served as negative control.

Figure 7.. The role of inflammation.

LPS stimulation led to a significant increase of BACH1/Bach1 mRNA expression in mouse AML12 cells (A) and human Hep3b cells (B) compared to controls (vehicle). (C) The heat map of z-scored Maff ChIP-MS visualises LFQ intensities of selected Maff interactors in extracts from AML12 cells under homeostatic conditions (Ctrl), LPS stimulation (Ctrl+LPS), after Maff siRNA knockdown (siRNA) and after Maff siRNA knockdown in combination with LPS stimulation (siRNA+LPS). Provided are adjusted p-values. *** indicates p<0.001, ** indicates p<0.01, * indicates p<0.05.

Figure 8.. The role of the transcription factor MAFF in activation or repression of the LDLR is based on heterodimerisation partners and environmental conditions.

MAFF heterodimers bind at the MAF recognition element (MARE) of the LDLR promoter and execute regulation of the LDLR. Under basal conditions 1) MAFF knockdown/knockout led to reduced LDLR mRNA expression, 2) elevated MAFF expression was correlated with higher expression of LDLR in a human CAD cohort (STARNET) and in wildtype mice of the hybrid mouse diversity panel (HMDP), 3) Overexpression of Maff using plasmid DNA transfection led to increased Ldlr expression in vitro. In the presence of LPS stimulation MAFF-BACH1 heterodimers result in downregulation of the LDLR in vivo. HMDP mice on atherogenic background (transgenic expression of human APOE-Leiden and cholesteryl ester transfer protein (CETP)) showed increased inflammation and revealed that elevated Maff expression correlates with lower Ldlr expression. BACH1: BTB domain and CNC Homolog 1; MAFF: MAF BZIP Transcription Factor F; MARE: Maf recognition element; LDLR: low-density lipoprotein receptor; LPS: lipopolysaccharide.