The identified polymorphism is more widely important for specificity of MADS-domain protein–protein interactions. (A) Most abundant motif in the I-region, identified by MEME based on 11966 MIKC-type proteins. Positions 1 and 2 of the sequence logo correspond to amino acid positions 58 and 59 in the MADS-domain proteins (B) Representative Alphafold2 dimers for FUL-AG and FUL1-TM3 showing the location of the AN and ST residues important for dimerization specificity. The proteins are color-coded by probability, from blue (higher probability) to yellow (lower probability). The residues lie in the loop between the M- and I-domains and do not directly contribute to the dimerization interface. Methionine and tyrosine residues putatively important for dimerization stability and/or specificity are shown, with the partner monomer tyrosine residue colored in cyan. Residues are shown as sticks, labeled and color-coded by atom. (C) Alignment of Arabidopsis MADS-domain TFs, showing the part of the protein that contains the identified motif. Proteins that were used to test the importance of the motif are indicated. (D, E) Constructs generated to test the Met63/Tyr70/loop hypothesis. The protein names indicated in boxes represent the wild-type sequence motifs; the others are swaps or modifications. Each column indicates a MADS-domain protein against which interaction was tested with yeast two-hybrid. See for details Supplementary Figures S18–S23. A dark box represents interactions in both directions (AD versus BD and BD versus AD); light boxes interactions in only one direction; white boxes no interaction; ND = not determined. The SEP3 constructs in (E) were only tested in one direction because of autoactivation of BD-SEP3. Where the non-mutated SEP3 was used, we made use of a dSEP3 version, which lacks part of the C-terminus and does not exhibit autoactivation. Number 21 indicates AGL21; 19 indicates AGL19, etc. XAN=AGL12(XAL1).