In vitro analysis of DNA-protein interactions by proximity ligation

Abstract

Protein-binding DNA sequence elements encode a variety of regulated functions of genomes. Information about such elements is currently in a state of rapid growth, but improved methods are required to characterize the sequence specificity of DNA-binding proteins. We have established an in vitro method for specific and sensitive solution-phase analysis of interactions between proteins and nucleic acids in nuclear extracts, based on the proximity ligation assay. The reagent consumption is very low, and the excellent sensitivity of the assay enables analysis of as few as 1-10 cells. We show that our results are highly reproducible, quantitative, and in good agreement with both EMSA and predictions obtained by using a motif finding software. This assay can be a valuable tool to characterize in-depth the sequence specificity of DNA-binding proteins and to evaluate effects of polymorphisms in known transcription factor binding sites.

Fig. 1.

Analysis of DNA–protein interactions by proximity ligation. A DNA-binding protein can be investigated by using two affinity probes, each carrying an oligonucleotide extension. One of the probes would be an antibody directed against a particular DNA-binding protein, whereas the other would consist of a partially double-stranded DNA sequence potentially recognized by the same DNA-binding protein. If the protein were simultaneously bound by both affinity probes (A and B), the ends of their appended oligonucleotides would be brought sufficiently close so that they could hybridize together to a subsequently added connector oligonucleotide, allowing them to be joined by enzymatic ligation (C). The ligated DNA sequence, which would serve as a specific DNA representation of the binding event between the protein and the investigated recognition sequence, would be subsequently amplified and detected by real-time PCR (D).

Fig. 2.

Analysis of the DNA binding specificity of p53. (A) Three sequences previously reported to bind p53 were analyzed by PLA along with specificity controls in nuclear lysates prepared from MCF-7 cells. Positive controls: Pos I was a p53 consensus sequence published by Kastan et al. (23), Pos II (24) was a focal adhesion kinase promoter region shown to interact with p53, and Pos III was a polymorphic microsatellite that mediates induction of p53-inducible gene 3 by p53 (25). Negative controls: Pos I mut was a probe in which four nucleotides within the consensus binding site had been altered, and Pos I inhib control, a 1,000-fold excess of the consensus p53 DNA probe without the proximity probe extension, was added to the reaction together with the full-length consensus probe. The negative control was a probe positive for HNF-4α binding but with no known affinity for p53, and in the no lysate control, no nuclear lysate was added to the reaction. The S/N is shown on the y axis. (B) Dilutions of MCF-7 nuclear lysate were analyzed with the Pos I (filled bar) and the Pos I Mut probe (open bar) to determine the limit of detection of p53 in the extract.

Fig. 3.

PLA analysis and EMSA of tentative HNF-4α-binding sequences. (A) Three sequences identified by ChIP-chip analysis were analyzed by PLA for interactions with HNF-4α in HepG2 nuclear extracts together with a negative control (a p53 consensus oligonucleotide). The nuclear extract was diluted as indicated along the x axis to investigate the lowest dilution with detectable amounts of binding protein. S/N values are shown on the y axis. (B) EMSAs of sequences identified by ChIP-chip (Pos) along with negative control sequences (Neg), competitor oligonucleotides having the same (Self Comp) or an unrelated sequence (Unr.comp), and supershift reactions with HNF-4α antibodies (α-HNF-4α and with irrelevant AP2α antibodies (α-AP2α.

Fig. 4.

HNF-4α consensus motives as identified by PLA and TRANSFAC. The 50 sequences shown in Fig. 4 were divided in groups based on their S/N values. The values defining each group were selected so that similar numbers of sequences were assigned to each of them. Subsequently, each of the group of sequences was analyzed by using the motif-finding program BioProspector (30), which identifies the most common motifs found among such sequences. The consensus motif found for each of the groups is illustrated with the program WebLogo (33). The consensus for each of the sequence groups represents the overall results obtained when BioProspector was run several times for each of the group, and very similar results were obtained in all cases. The previously known binding consensus sequence for HNF-4α as found in TRANSFAC (32) is indicated at the bottom.

Fig. 5.

DNA sequence specificity of USF1. (A) Four potential USF1 binding sequences identified by ChIP-chip analysis (Pos, positive control; Seq 1, sequence 1; Seq 2, sequence 2; and Seq 3, sequence 3), as well as a negative control (Neg; an HNF-4α consensus oligonucleotide) were analyzed with PLA for interactions with USF1 in 10-ng HepG2 nuclear extracts. S/N values are shown on the y axis. (B) EMSAs of three of the positive sequences as identified by ChIP-chip along with negative controls, competitive probes with the same (SelfComp) or an unrelated (Unr.comp) binding sequence, and supershift reactions with USF1 (α-USF1) and Sp1 (α-Sp1) antibodies.