Comparative analysis of binding patterns of MADS-domain proteins in Arabidopsis thaliana

Abstract

Background: Correct flower formation requires highly specific temporal and spatial regulation of gene expression. In Arabidopsis thaliana the majority of the master regulators that determine flower organ identity belong to the MADS-domain transcription factor family. The canonical DNA binding motif for this transcription factor family is the CArG-box, which has the consensus CC(A/T)₆GG. However, so far, a comprehensive analysis of MADS-domain binding patterns has not yet been performed.

Results: Eight publicly available ChIP-seq datasets of MADS-domain proteins that regulate the floral transition and flower formation were analyzed. Surprisingly, the preferred DNA binding motif of each protein was a CArG-box with an NAA extension. Furthermore, motifs of other transcription factors were found in the vicinity of binding sites of MADS-domain transcription factors, suggesting that interaction of MADS-domain proteins with other transcription factors is important for target gene regulation. Finally, conservation of CArG-boxes between Arabidopsis ecotypes was assessed to obtain information about their evolutionary importance. CArG-boxes that fully matched the consensus were more conserved than other CArG-boxes, suggesting that the perfect CArG-box is evolutionary more important than other CArG-box variants.

Conclusion: Our analysis provides detailed insight into MADS-domain protein binding patterns. The results underline the importance of an extended version of the CArG-box and provide a first view on evolutionary conservation of MADS-domain protein binding sites in Arabidopsis ecotypes.

Keywords: CArG-box; ChIP-seq; MADS-domain proteins; Sequence conservation; Transcription factor binding specificity.

Fig. 1

CArG-box like binding motifs for MADS domain proteins involved in flower formation. Logos represents CArG-box motifs found by MEME. a AG (b) AP1 (c) AP3 (d) FLC (e) PI (f) SEP3 (g) SOC1 (h) SVP

Fig. 2

Enrichment of CArG-box variants in peak centers. A peak center is defined as the 250 bp upstream and downstream of the peak summit. a Frequency of peak centers containing different CArG-box variants divided by the frequency in random 500 bp stretches in the Arabidopsis thaliana genome. Black, relative frequency of CArG-box variant in all Arabidopsis promoters. b Kernel density plot of positions of different CArG-box variants in peak centers of SEP3 relative to peak summits

Fig. 3

Non-CArG box motifs. a Relative enrichment of peak centers containing a secondary motif compared to the promoter background. The frequency of peak centers containing a secondary motif was calculated and divided by a background frequency. A peak center is defined as the 250 bp upstream and downstream of a peak summit. G-box: CACGTG, TCP class I: GGNCCCAC, TCP class II: GGGNCC(A/G)C. b Enrichment of a WRKY-like motif (GTTGACTTT) in SEP3 peaks. c Kernel density plot of positions of the perfect CArG-box and the WRKY-like motif in the peak center compared to the peak summit

Fig. 4

Significantly overrepresented 3′ extensions of the CArG-box core. Enrichment of extensions of 3 nucleotides is calculated as the frequency of the extension after the CC(A/T)₆GGN-core in ChIP-seq peaks divided by the expected frequency of the extension based on the frequencies of nucleotides in the ChIP-seq peaks. All extensions are depicted for which at least one dataset is significant at p < 0.05. For visualization purposes, all extensions that are enriched relative to what is expected from nucleotide frequencies, but are not significant, are set to 1. Note that a similar analysis, but for CArG-box like sequences picked up by MEME-ChIP, is presented in Additional file 15: Table S7

Fig. 5

Conservation of CArG-boxes in ChIP-seq peaks among Arabidopsis thaliana ecotypes. For each position in each CArG-box, mutational entropy was divided by an average background entropy to give a mutation index (blue), averaged over all motif occurrences. This was also done for the subset of 428 perfect CArG-boxes (red). Positions for which the difference in mutation index between perfect CArG-boxes and all CArG boxes was statistically significant are indicated with an asterisk

Fig. 6

Five models that explain the occurrence of different binding motifs in MADS-domain protein ChIP-seq data analyzed in the present study. a The MADS-domain protein binds to a CArG-box. b The MADS-domain protein binds to another transcription factor, which binds DNA at a motif specific for that transcription factor. c Same as (a), but because by chance or as part of an enhanceosome there is a binding site of another transcription factor close by, both the CArG-box and the other motif occur in the ChIP-seq peak. d The MADS-domain protein needs another transcription factor for binding to a motif that is a hybrid between a CArG-box and the motif for the other transcription factor. e The motif is competitively bound by the MADS-domain protein and another protein and is therefore a hybrid between a CArG-box and the motif of the other transcription factor