An enhancer element harboring variants associated with systemic lupus erythematosus engages the TNFAIP3 promoter to influence A20 expression

Abstract

Functional characterization of causal variants present on risk haplotypes identified through genome-wide association studies (GWAS) is a primary objective of human genetics. In this report, we evaluate the function of a pair of tandem polymorphic dinucleotides, 42 kb downstream of the promoter of TNFAIP3, (rs148314165, rs200820567, collectively referred to as TT>A) recently nominated as causal variants responsible for genetic association of systemic lupus erythematosus (SLE) with tumor necrosis factor alpha inducible protein 3 (TNFAIP3). TNFAIP3 encodes the ubiquitin-editing enzyme, A20, a key negative regulator of NF-κB signaling. A20 expression is reduced in subjects carrying the TT>A risk alleles; however, the underlying functional mechanism by which this occurs is unclear. We used a combination of electrophoretic mobility shift assays (EMSA), mass spectrometry (MS), reporter assays, chromatin immunoprecipitation-PCR (ChIP-PCR) and chromosome conformation capture (3C) EBV transformed lymphoblastoid cell lines (LCL) from individuals carrying risk and non-risk TNFAIP3 haplotypes to characterize the effect of TT>A on A20 expression. Our results demonstrate that the TT>A variants reside in an enhancer element that binds NF-κB and SATB1 enabling physical interaction of the enhancer with the TNFAIP3 promoter through long-range DNA looping. Impaired binding of NF-κB to the TT>A risk alleles or knockdown of SATB1 expression by shRNA, inhibits the looping interaction resulting in reduced A20 expression. Together, these data reveal a novel mechanism of TNFAIP3 transcriptional regulation and establish the functional basis by which the TT>A risk variants attenuate A20 expression through inefficient delivery of NF-κB to the TNFAIP3 promoter. These results provide critical functional evidence supporting a direct causal role for TT>A in the genetic predisposition to SLE.

Figure 1. The TT>A variants result in reduced binding of a nuclear protein complex that contains NF-κB subunits.

a. EMSA was performed with radiolabeled nucleotides containing the non-risk (TT) (40 bp) and risk (-A) (39 bp) polymorphisms. Nuclear extracts were derived from EBV transformed B cell lines at rest or stimulated with P/I and in the presence or absence of the indicated NF-κB subunit antibodies. Rabbit IgG was used as the isotype control. N.E. -nuclear extracts. b. ChIP-pPCR was performed using EBV transformed B cell lines (-A/-A = 4; TT/-A = 8; TT/TT = 12) stimulated with P/I. ChIP was performed with antibodies specific against NF-κB p50, p65, and cRel subunits, followed by qPCR with primers neighboring TT>A polymorphic region. Statistical comparisons were made using one-way ANOVA, * indicates p<0.05.

Figure 2. The regulatory element containing the TT>A variants demonstrates enhancer activity.

Sequences carrying risk (-A) and non-risk (TT) polymorphisms were cloned upstream of a minimal thymidine kinase promoter luciferase construct to measure luciferase activation following transient transfection and stimulation. a. HEK293T cells stimulated with P/I for 48 hours. b. THP1 cells stimulated with LPS for 48 hrs. c. THP1 cells stimulated with P/I for 48 hours. Statistical comparisons were performed using a Student's t-test of three biological independent experiments, * indicates p<0.05.

Figure 3. 3C analysis demonstrates long-distance interactions between the TT>A enhancer and the TNFAIP3 gene region.

a. The track on the top is the location of the TNFAIP3 primers used to identify potential amplified interaction fragments tested by 3C. Primers 6–10, 16 and 24 produced signals and are marked with asterisks. The middle and bottom track shows the genomic region of TNFAIP3 with the location of the promoter CpG island and ENCODE defined transcription factor binding sites. b. Increased relative interaction frequencies (RCF) were detected in three regions of TNFAIP3, the promoter, intron 2 and the 3′ untranslated region. c. Stimulation of THP1 cells with LPS results in increased 3C interactions between the TT>A enhancer and the TNFAIP3 promoter along with a concomitant increase in A20 expression and IκBα phosphorylation. Shown is a representative blot from 3 independent experiments. Statistical differences were determined using Student's t-test.

Figure 4. SATB1 is required for the TT>A enhancer-promoter interaction and TNFAIP3 transcription.

a. Shown is a representative 3C assay from 3 independent experiments in HEK293T cells. Relative crosslinking frequencies (RCF) were normalized to both GAPDH loading control and a TNFAIP3 BAC clone and plotted according to its location to TNFAIP3 gene and the TT>A enhancer. High local interaction frequencies near the TT>A enhancer serve as a positive control (fragments 30–34). b. The top panel shows the protein expression levels of SATB1 and A20 were detected using Western blots with antibodies against SATB1 and A20, respectively. β-actin was used as loading control. The bottom two panels show densitometer quantification of the relative expression of SATB1 (middle panel) and A20 (lower panel) normalized to β-actin. Error bars represent standard error of the mean from 3 independent experiments. Differential protein levels between SATB1 knockdown and non-silencing shRNA control were calculated using Student's t-test.

Figure 5. The TT>A risk variant exhibits reduced enhancer-promoter interaction frequencies and lower expression of TNFAIP3.

a. 3C-qPCR assays were performed in EBV-transformed B cells homozygous for either risk or non-risk alleles in the TT>A enhancer; crosslinking frequencies between the enhancer and the promoter were normalized to GAPDH control. b, c. Protein expression of A20 and phospho-IκBα in homozygous cell lines were determined using Western blot and were normalized to loading control β-actin. Plots represent the relative expression of densitometric measurements for each cell line. d. Allele-specific 3C sequencing results of EBV cell lines heterozygous for the TT>A alleles. The read counts for each paired allele from 3C samples were normalized to the read counts for each allele from parallel analysis of input genomic DNA not subjected to 3C. Comparisons were made between non-risk and risk alleles for each individual. Statistical differences were calculated using paired t-test.