DNA specificity determinants associate with distinct transcription factor functions

Abstract

To elucidate how genomic sequences build transcriptional control networks, we need to understand the connection between DNA sequence and transcription factor binding and function. Binding predictions based solely on consensus predictions are limited, because a single factor can use degenerate sequence motifs and because related transcription factors often prefer identical sequences. The ETS family transcription factor, ETS1, exemplifies these challenges. Unexpected, redundant occupancy of ETS1 and other ETS proteins is observed at promoters of housekeeping genes in T cells due to common sequence preferences and the presence of strong consensus motifs. However, ETS1 exhibits a specific function in T cell activation; thus, unique transcriptional targets are predicted. To uncover the sequence motifs that mediate specific functions of ETS1, a genome-wide approach, chromatin immunoprecipitation coupled with high-throughput sequencing (ChIP-seq), identified both promoter and enhancer binding events in Jurkat T cells. A comparison with DNase I sensitivity both validated the dataset and also improved accuracy. Redundant occupancy of ETS1 with the ETS protein GABPA occurred primarily in promoters of housekeeping genes, whereas ETS1 specific occupancy occurred in the enhancers of T cell-specific genes. Two routes to ETS1 specificity were identified: an intrinsic preference of ETS1 for a variant of the ETS family consensus sequence and the presence of a composite sequence that can support cooperative binding with a RUNX transcription factor. Genome-wide occupancy of RUNX factors corroborated the importance of this partnership. Furthermore, genome-wide occupancy of co-activator CBP indicated tight co-localization with ETS1 at specific enhancers, but not redundant promoters. The distinct sequences associated with redundant versus specific ETS1 occupancy were predictive of promoter or enhancer location and the ontology of nearby genes. These findings demonstrate that diversity of DNA binding motifs may enable variable transcription factor function at different genomic sites.

Figure 1. Genomic occupancy of ETS1 overlaps with DNase I sensitivity and GABPA occupancy.

(A) Overlap of genomic regions bound by ETS1 in Jurkat T cells and regions found to be DNase I sensitive in CD4⁺ T cells . (B) Overlap of regions occupied by ETS1 and GABPA in Jurkat T cells . Only regions that overlap with DNase I sensitivity were included. (C) Log transformed P values of ETS1 and GABPA occupancy in a scanning 250 bp window mapped using the Integrated Genome Browser (http://igb.bioviz.org/) to regions of the human chromosome (Chr) indicated by chromosomal coordinates (NCBI Build 36.1). Positions of Refseq genes are shown with genes transcribed from left to right above the nucleotide position bar and genes in the opposite orientation below. Vertical lines (right) indicate the ETS/RUNX binding sequences previously tested for function , in the TCRα enhancer (sequence GAGGATGTGGC) or the TCRβ enhancer (sequence CAGGATGTGGT).

Figure 2. Distal ETS1 bound regions are found near T cell–specific genes.

The frequency of neighboring, distal, ETS1 bound regions was compared for three categories of genes; all Refseq genes, CD4⁺ T cell–specific genes, and pancreas specific genes. Tissue specific gene lists were derived from the GNF SymAtlas database and were based on the level of mRNA in T cells or pancreas compared to the median in all surveyed tissues with cutoffs (20-fold higher for T cells, 5-fold higher for pancreas) that returned similar sized lists. The number of genes in each category is indicated in parenthesis. Each ETS1 bound region was matched to a single gene based on the nearest RefSeq TSS. The percent of genes in each category associated with one or more distal ETS1 bound regions (greater than 500 bp from the TSS) was plotted.

Figure 3. Distinct sequence motifs are over-represented in different subsets of ETS1 bound regions.

The indicated subsets of ETS1 bound regions were rank ordered by log transformed binomial P value and the top 250 regions were searched for over-represented sequences by MEME . The most over-represented position weight matrix for each subset is represented (E-values: Motif 1, 9.9×10⁻²⁸¹; Motif 2, 2.8×10⁻²⁸⁷; Motif 3, 8.8×10⁻¹⁰⁵; Motif 4, 1.8×10⁻¹⁴²). The height of each nucleotide indicates conservation at that position. Eight nucleotide positions in ETS binding sites are numbered for reference. ETS bound regions were classified either proximal (center of region within 500 bp of a TSS) or distal (center of region greater than 500 bp from a TSS). ETS1 bound regions were classified as redundant if they overlapped with a GABPA bound region and specific if they did not. ETS1 bound regions were classified as RUNX co-occupied if they overlapped with a RUNX bound region.

Figure 4. Intrinsic DNA binding affinity differences for ETS1 and ELF1.

Full-length recombinant versions of the human ETS proteins ETS1 and ELF1 were purified from bacteria and assayed for affinity to radiolabeled oligonucleotides by gel-shift analysis. Datapoints represent the mean and standard error of the mean of two replicates for ETS1 and three replicates for ELF1. Each K_D was derived by curve fitting by nonlinear least-squares analysis of equilibrium binding curves with fraction of DNA bound  = 1/(1+ K_D/[ETS1]). The K_D of ETS1 and ELF1 for the ETS consensus sequence was 1.7×10⁻⁹ and 1.7×10⁻¹⁰ M, respectively. The fold decrease in affinity due to the A to T change, the C to A change, and the combination of both changes were 1.0, 1.5, and 2.4 for ETS1 and 3.6, 1.9, and 18.3 for ELF1, respectively.

Figure 5. ETS1/RUNX co-occupancy correlates with specific spacing of ETS and RUNX binding sites.

(A) Overlap of ETS1/RUNX and GABPA/RUNX bound regions. The RUNX antibody was raised against the conserved DNA binding domain and does not differentiate between the homologous RUNX1 and RUNX3 proteins present in T cells (N. Speck unpublished observation). (B) Spacing of ETS and RUNX binding sites in ETS1/RUNX co-occupied regions. The 690 ETS1 bound regions that were co-occupied by RUNX were scanned for matches to the in vitro derived position weight matrixes for ETS1 (M00032) and RUNX (M00271) from the Transfac database (http://www.biobase-international.com/index.php?id=transfac). For each ETS1 sequence found, the distance to all forward oriented RUNX sequences in the same region were determined such that a RUNX site 5′ to an ETS site in the orientation CCGGAAGT was negative and 3′ was positive. A similar mapping of RUNX sites in the reverse orientation returned no spacing frequencies higher than five. The prominent peak at a spacing of five bp correlates with the spacing and orientation in the composite sequence CAGGATGTGGT, from Motif 4 (Figure 3).

Figure 6. Distinct properties of promoters and enhancers occupied by ETS1.

(A) Factor or histone modification positions were plotted as a class average across redundant promoters (left) or ETS1 occupied enhancers (right). At redundant promoters, the occupancy profiles of ETS1, GABPA, CBP, H3K4 tri-methyl, and Motif 1 were plotted from the center of each occupied region to the nearest RefSeq TSS. At ETS1-occupied enhancers, the occupancy profiles of RUNX1, CBP, H3K4 mono-methyl, and Motif 4 were plotted with respect to the center of the ETS1 bound region. For each factor, histone modification, or motif, a histogram of 30 bp bins was generated to represent the frequency of occupancy for each distance. The number of occurrences at each distance was normalized to the total number of regions with an occurrence of each factor, histone modification, or motif. A vertical line indicates the zero position of each chart (TSS or center of ETS1 bound region). (B) Protein immunoblot of Jurkat whole cell extracts with the ETS1 antibody. Lane 1, no shRNA; Lane 2, negative control shRNA targeting luciferase; Lanes 3 and 4, two independent shRNAs (A and B) targeting ETS1. The two bands apparent in Lanes 1 and 2 are consistent with the 51 and 42 kDa splicing isoforms of ETS1. (C) ETS1, CBP, and RUNX ChIP enrichment at the TCRβ enhancer and the RPS26 promoter. The shRNAs were expressed in Jurkat T cells prior to ChIP, as indicated. Two independent biological replicates provided similar patterns, but different maximum levels of enrichment. A representative experiment is shown. Primer sequences used are provided in Table S3.