Protein interaction evolution from promiscuity to specificity with reduced flexibility in an increasingly complex network

Abstract

A key question regarding protein evolution is how proteins adapt to the dynamic environment in which they function and how in turn their evolution shapes the protein interaction network. We used extant and resurrected ancestral plant MADS-domain transcription factors to understand how SEPALLATA3, a protein with hub and glue properties, evolved and takes part in network organization. Although the density of dimeric interactions was saturated in the network, many new interactions became mediated by SEPALLATA3 after a whole genome triplication event. By swapping SEPALLATA3 and its ancestors between dimeric networks of different ages, we found that the protein lost the capacity of promiscuous interaction and acquired specificity in evolution. This was accompanied with constraints on conformations through proline residue accumulation, which made the protein less flexible. SHORT VEGETATIVE PHASE on the other hand (non-hub) was able to gain protein-protein interactions due to a C-terminal domain insertion, allowing for a larger interaction interface. These findings illustrate that protein interaction evolution occurs at the level of conformational dynamics, when the binding mechanism concerns an induced fit or conformational selection. Proteins can evolve towards increased specificity with reduced flexibility when the complexity of the protein interaction network requires specificity.

Figure 1. SEP3 structural information.

(A) Left: MIKC-type subdomains and their reported functions, right: possible dimer conformations based on available MADS-domain protein 3D structure of SRF and MEF2 homology modeling. (B) Resolved X-ray crystal structure of the SEP3 K-domain, with its subdomain K1, K2 and K3 and their role in oligomerization.

Figure 2. Evolution of SEP3 as a network hub.

(A) Yeast three-hybrid (Y3H) assays of reconstructed ancestral and extant networks with (pSEP3, gSEP3, SEP3 and TM5) mediating ternary complexes formation. Black and green lines indicate conserved and novel gained ternary interactions respectively through comparison from pre-gamma PIN to post-gamma PIN and from post-gamma PIN to AraPIN and to Sol-PIN. An example is presented in the gray box where PI-AP3 and SEP3-SOC1 interactions are mediated by SEP3, which is represented by a line in network, and where AG-AG and SOC1-SOC1 interactions are mediated by SEP3, which is represented by a looped line in network. (B) Increase of network densities in yeast-three hybrid compared to yeast-two hybrid networks, total number of edges is shown above the graph line (orange: Y3H, blue: Y2H). (C) Total amount of conserved, gained and lost interactions in addition to the gain and loss rate (per edge per mya) between networks (gain and loss rate was defined as gained or lost interactions divided by number of potential interactions in networks times the divergence time). (D) Multiple sequence alignment of SEP3 and its resurrected ancestors in addition to its homolog in tomato TM5. Red boxes indicate proline accumulation in the I-domain. For Miller units of each interaction see Supp. Table.1.

Figure 3

Yeast two-hybrid assays assessing the interaction of proteins in a network of a different age (swapping between networks) (A) SEP3, pSEP3 and ancE. (B) AP3, pAP3 and ancB. (C) SVP, pSVP24. (D) Three-hybrid interaction mediated by SEP3 vs ancE. (E) Two-hybrid assay results of mutant SEP3 (P78Q, P80T, P83S), showing the gain of total protein-protein interactions (green) from (8 to 13). For Miller units of each interaction see supp. Table. 2.

Figure 4. Molecular dynamics simulations of SEP3 K-domain tetramer (black lines) and monomer (red lines).

(A) Complete K-domain RMSD calculations. (B) RMSD calculations for K1 subdomain residues 98–111. (C) RMSD calculations for K2 subdomain residues 129–146. (D) RMSD calculations for K3 subdomain residues 150–1171. (E) RMSD calculations for loop region between K1-K2 subdomains residues 117–123. (F) Solvent accessible surface area (SASA) calculations for one SEP3 K-domain in bound -tetramer complex- (black line) and free monomer (red line).

Figure 5

The role of SEP3 loop structure on PPI: (A) SEP3 loop sequence (GEDLGPL 117–123) located between the end of K1-subdomain and the beginning of K2-subdomain, underlined residues indicate 4Ala and 3Ala mutations while the star shows P122, and K-subdomain partial Crystal structure. (B) Y2H assay for native SEP3 and three SEP3 mutants (SEP3(P122A), SEP3(G121A/P122A/L123A) and SEP3(G117A/D118A/E119A/L120A)). Protein interactions were shown by the straight lines. Relative interaction strength was illustrated by different colors (black, orange and grey) based on the values of Miller Units for all interactions. wt: wild type; m: mutant. (C) EMSA results show the binding affinities of native SEP3 and three SEP3 mutants with AG by SEP3 and SEP3 mutants titrations. Black triangles represent the concentration gradient of SEP3 and SEP3 mutants from highest to lowest comparing with constant concentration of AG in each group. The concentration ratios between AG and SEP3 or SEP3 mutants were demonstrated inside the black bars (1:1, 1:0.75, 1:0.50 and 1:0.25). Cartoons on the left side along the bands show the supposed quartet complex forming (top) and the dimerized forming (bottom) binding on the probe. The probe was a SEP3 promoter fragment containing two CArG boxes (see Methods) which was illustrated as double straight or bend lines harboring white and black bars in cartoons. (D) Loop dynamics, loop region in dimer conformation in resolved crystal structure (PDB: 4Ox0). And at 48 ns of MDS run of, (E) Native SEP3 K-domain (F) P122A mutant, (G) GEDLAAA mutant, (H) AAAAGPL mutant. Color codes (blue: K2 subdomain, red: K1 subdomain, green: loop region, magenta: K2 extended in GEDLAAA mutant, dark blue lines: represent possible hydrogen or polar bonds numbers next to it shows distance in angstrom). (I) RMSD of K1 residues 98–111 and K2 residues 129–146 as function of MDS time of native and loop mutants in monomer and tetramer states.

Figure 6. SVP site-directed mutagenesis Y2H assay, and C-domain insertion de novo modeling.

(A) Multiple sequence alignment of pSVP24, SVP and AGL24, red squares indicate mutation sites. (B) Mutant SVP two-hybrid relative to wtSVP two-hybrid (red base line 1 = 100%). (C) Top de novo models of SVP C-domain insertion using Quark and CαβS servers.