Figure 1.
Combining predicted protein-protein interaction motifs and modeled protein structures.
(A) Modeled dimer for the Arabidopsis MADS domain protein SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1). Blue indicates the DNA binding helix (in which no protein-protein interaction motifs are present). Residues indicated in spacefill (Ala57, Asn60 and Met61) are part of an experimentally validated interaction motif in the so-called ‘hotspot region’ (see text for details). (B) Crystal structure (PDB 1n6j) of human MADS domain protein MEF2 (grey) in complex with Cabin1 (red). Cabin1 contacts MEF2 via Met62 and a few other amino acid residues. MEF2 Met62 is the equivalent of Met61 in SOC1, with both amino acid residues having comparable positions in the structure. The residues of Cabin1 that contact Met62 (Ser101, Gly104 and Ile106) are shown in red spacefill. Based on the MEF2-Cabin1 structure we hypothesize a similar kind of binding of the α-helix-forming K-box from a SOC1 interacting MADS domain protein on top of the SOC1 MADS/I domain. (C) The black box indicates the predicted interaction motif in the ‘hotspot region’ of the SOC1 protein. The predicted complementary interaction motif (red box) is located in the K-box domain of the MADS domain protein interacting with SOC1.
Figure 2.
Example of the experimental validation approach.
(A) Schematic representation of the SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1) and AGAMOUS LIKE14 (AGL14) MADS domain proteins. Due to the generated mutations one particular interaction motif (green rectangle) is swapped between these two Arabidopsis MADS domain transcription factors. The swapped motif is located in the “hotspot” region between the MADS and I-domain. (B) Part of a sequence alignment of AGL14 and SOC1, including the motif that was selected for mutagenesis (indicated in green in the SOC1 sequence). The mutated residues that were swapped between the two proteins are shown bold/underlined. (C) Bar diagrams showing the number of interaction partners for SOC1, AGL14 and their mutated counterparts SOC1* and AGL14*, respectively; see Table 1. For these two proteins, the F-scores of predicted mutant interactions are 0.63 and 0.71, respectively. (D) Example of a matrix-based yeast two-hybrid screen. Yeast was spotted on medium lacking Leucine, Tryptophan, and Histidine, and supplemented with 1 mM 3-amino 1,2,4-triazole to suppress transcriptional autoactivation. Growth and hence interaction events, was scored after incubation at 20°C for 4 days.
Table 1.
Interaction maps for mutated MADS domain proteins.
Figure 3.
Effect of motif-based mutations on interaction patterns.
(A) Mutations were introduced based on predicted interaction motifs as explained in Figure 2. Different domains in MIKC MADS domain proteins are shown with colored boxes indicating the various regions in which point mutations were introduced. Below these, the various mutant MADS domain proteins that were generated are listed. The descriptions of the mutated proteins are colored based on the domain in which the mutation was generated. The mutated MADS domain proteins are SHORT VEGETATIVE PHASE (SVP1), AGAMOUS LIKE 24 (AGL24), SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1), APETALLA1 (AP1), CAULIFLOWER (CAL), and AGAMOUS (AG). Note that there are two double mutations for which one mutation occurs in the MADS/I domain and one in the K-box. Below each mutated protein, the number of losses and gains of protein-protein interactions in the yeast two-hybrid assay for the mutated proteins in comparison to the native MADS domain proteins is indicated (see Table 1 for interaction partner identities). (B) Histogram of F-scores for prediction of effect of mutants based on randomized input data (see text for details). The arrow indicates the F-score obtained by the predictor trained on experimental input data.
Figure 4.
Interaction motifs and network evolution.
The role of conservation of interaction motifs versus variation of these motifs was investigated. (A) Histogram of occurrences of interaction motifs (black) and SNPs (red) at particular positions in the protein sequences of all Arabidopsis MIKC MADS proteins. Note that there is hardly any overlap between interaction motifs and SNPs. Positions of the M, I, K and C domain are indicated. (B) Histogram of cross-species conservation of interaction motifs (black) and non-motif-sequences (red) in MIKC MADS domain protein sequences. Non-motif sequences are defined at positions in MADS protein sequences where in other MADS sequences a motif is present. (C) Four different scenarios are possible if after duplication of a MADS domain protein sequence an indel occurs in one of the two sequences: (I) indel does not overlap with a predicted interaction motif; (II) both insertion and deletion overlap with a motif; (III) only insertion or (IV) only deletion overlap with a motif. Lines indicate sequences, colored boxes indicate predicted interaction motifs, triangles indicate insertion, and arrows indicate effect of insertion/deletion on motif. As discussed in the text, if an indel overlaps a motif (scenario II-IV), in half of the cases (18% for scenario II vs 9% for each of scenario III and IV) it does not delete but only modifies the motif (illustrated by a change in color for the motif).
Figure 5.
Mechanism of generating protein-protein interaction diversity by shifting intron/exon borders.
(A) After a duplication of a gene or in the case of alternative splicing, a shift of an intron/exon border can modify a protein interaction motif which overlaps or is close to such a border. Top panel, schematic illustration of this process. Line indicates gene sequence, grey bars indicate exons, and colored bars indicate predicted interaction motifs. Bottom panel, part of a protein sequence alignment for the Arabidopsis MADS domain protein SHORT VEGETATIVE PHASE (SVP1) and an identified alternatively spliced SVP form named SVP3. A predicted interaction motif in SVP1 which is almost completely spliced out in SVP3 is shown in red. Two grey bars indicate the two adjacent exons. (B) Predicted interaction motifs can be either close to an intron/exon border (indicated by red motif) or far away from the intron/exon border (green motif). Bars in the graph indicate average number of Arabidopsis MIKC MADS proteins in which predicted interaction motifs occur for two different motif groups: motifs that are located close to the intron/exon border (<3 amino acids distance, red) occur on average in a few proteins only, and motifs that are located far away from the border (> = 3 amino acids distance, green) occur in many proteins.