Evidence for network evolution in an Arabidopsis interactome map

Abstract

Plants have unique features that evolved in response to their environments and ecosystems. A full account of the complex cellular networks that underlie plant-specific functions is still missing. We describe a proteome-wide binary protein-protein interaction map for the interactome network of the plant Arabidopsis thaliana containing about 6200 highly reliable interactions between about 2700 proteins. A global organization of plant biological processes emerges from community analyses of the resulting network, together with large numbers of novel hypothetical functional links between proteins and pathways. We observe a dynamic rewiring of interactions following gene duplication events, providing evidence for a model of evolution acting upon interactome networks. This and future plant interactome maps should facilitate systems approaches to better understand plant biology and improve crops.

Fig. 1

Quality of AI-1_MAIN. (A) Fraction of PRS, RRS or AI-1_MAIN sample pairs positive in Y2H or in wNAPPA at a scoring threshold of 1.5. Error bars: standard error of the proportion. P-values: one-sided two-sample t-tests (3). PRS pairs are more often detected than RRS pairs in wNAPPA (P = 2 × 10⁻⁸, one-sided two-sample t-test) and Y2H (P < 2.2 × 10⁻¹⁶, one-sided Fisher’s exact test). (B) The number of literature-curated interactions recovered reflects AI-1_MAIN framework parameters (6). Top: network representations of LCI_BINARY and AI-1_MAIN. Bottom left: data sets are represented by squared Venn diagrams; size is proportional to the number of interactions (3). Bottom right: observed and expected overlap given sensitivity and completeness of AI-1_MAIN (see main text and (3)). PRS pairs were removed from LCI_BINARY multiple evidence for this analysis. Error bars: two standard deviations from the expected counts.

Fig. 1

Quality of AI-1_MAIN. (A) Fraction of PRS, RRS or AI-1_MAIN sample pairs positive in Y2H or in wNAPPA at a scoring threshold of 1.5. Error bars: standard error of the proportion. P-values: one-sided two-sample t-tests (3). PRS pairs are more often detected than RRS pairs in wNAPPA (P = 2 × 10⁻⁸, one-sided two-sample t-test) and Y2H (P < 2.2 × 10⁻¹⁶, one-sided Fisher’s exact test). (B) The number of literature-curated interactions recovered reflects AI-1_MAIN framework parameters (6). Top: network representations of LCI_BINARY and AI-1_MAIN. Bottom left: data sets are represented by squared Venn diagrams; size is proportional to the number of interactions (3). Bottom right: observed and expected overlap given sensitivity and completeness of AI-1_MAIN (see main text and (3)). PRS pairs were removed from LCI_BINARY multiple evidence for this analysis. Error bars: two standard deviations from the expected counts.

Fig. 2

Plant signaling networks in AI-1. (A) Putative ubiquitination subnetwork extracted from LCI_BINARY and AI-1. Bar plot: number of protein-protein interactions between proteins in the ubiquitination cascade in LCI_BINARY and AI-1 (outside and within space 1). (B) Protein-protein interactions in AI-1 suggest a modular assembly of transcriptional hormone-response regulators and support a global regulatory role for TPL.

Fig. 2

Plant signaling networks in AI-1. (A) Putative ubiquitination subnetwork extracted from LCI_BINARY and AI-1. Bar plot: number of protein-protein interactions between proteins in the ubiquitination cascade in LCI_BINARY and AI-1 (outside and within space 1). (B) Protein-protein interactions in AI-1 suggest a modular assembly of transcriptional hormone-response regulators and support a global regulatory role for TPL.

Fig. 3

Communities in AI-1_MAIN (bottom) and in a typical randomized network (top left; fig. S9). Only largest connected component of networks are shown. Colored regions indicate communities enriched in GO annotations summarized by the indicated terms (table S10). Upper right: distribution of randomized networks as a function of the total and GO annotation enriched number of communities they contain; white arrow: position of the shown randomized network, red dot and arrow: position of AI-1_MAIN. GA: gibberellic acid, JA: jasmonic acid, TCA: tricarboxylic acid.

Fig. 4

Evidence for network evolution in AI-1_MAIN. (A) Interaction rewiring over time according to the duplication-divergence model (24). (B) Average fraction of interactors shared between pairs of paralogous proteins with no (n=4), low (n=10), and high (n=3) functional divergence (28). Error bars: standard error of the mean. P-value: one-sided Kendall ranking correlation test (τ= association) (3). (C) Average fraction of shared interactors, corrected for low experimental coverage (3), and average protein sequence identity between pairs of paralogous proteins as a function of the estimated Δ time-since-duplication. Error bars: standard error of the mean (3). Dashed black line: corrected average fraction of shared interactors of non-paralogous pairs. myrs: million years. (D) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins originating from polyploidy events (n=109) as compared to other paralogous protein pairs of similar age (n=147). Error bars: standard error of the mean (3). P-values: Mann-Whitney U-test. (E) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins encoded by gene pairs with high or low co-expression correlation (top and bottom tertile, respectively) as a function of phylogeny-based age group. Error bars: standard error of the mean (3). P < 0.05 (*), < 0.01 (**), < 0.001 (***).

Fig. 4

Evidence for network evolution in AI-1_MAIN. (A) Interaction rewiring over time according to the duplication-divergence model (24). (B) Average fraction of interactors shared between pairs of paralogous proteins with no (n=4), low (n=10), and high (n=3) functional divergence (28). Error bars: standard error of the mean. P-value: one-sided Kendall ranking correlation test (τ= association) (3). (C) Average fraction of shared interactors, corrected for low experimental coverage (3), and average protein sequence identity between pairs of paralogous proteins as a function of the estimated Δ time-since-duplication. Error bars: standard error of the mean (3). Dashed black line: corrected average fraction of shared interactors of non-paralogous pairs. myrs: million years. (D) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins originating from polyploidy events (n=109) as compared to other paralogous protein pairs of similar age (n=147). Error bars: standard error of the mean (3). P-values: Mann-Whitney U-test. (E) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins encoded by gene pairs with high or low co-expression correlation (top and bottom tertile, respectively) as a function of phylogeny-based age group. Error bars: standard error of the mean (3). P < 0.05 (*), < 0.01 (**), < 0.001 (***).

Fig. 4

Evidence for network evolution in AI-1_MAIN. (A) Interaction rewiring over time according to the duplication-divergence model (24). (B) Average fraction of interactors shared between pairs of paralogous proteins with no (n=4), low (n=10), and high (n=3) functional divergence (28). Error bars: standard error of the mean. P-value: one-sided Kendall ranking correlation test (τ= association) (3). (C) Average fraction of shared interactors, corrected for low experimental coverage (3), and average protein sequence identity between pairs of paralogous proteins as a function of the estimated Δ time-since-duplication. Error bars: standard error of the mean (3). Dashed black line: corrected average fraction of shared interactors of non-paralogous pairs. myrs: million years. (D) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins originating from polyploidy events (n=109) as compared to other paralogous protein pairs of similar age (n=147). Error bars: standard error of the mean (3). P-values: Mann-Whitney U-test. (E) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins encoded by gene pairs with high or low co-expression correlation (top and bottom tertile, respectively) as a function of phylogeny-based age group. Error bars: standard error of the mean (3). P < 0.05 (*), < 0.01 (**), < 0.001 (***).

Fig. 4

Evidence for network evolution in AI-1_MAIN. (A) Interaction rewiring over time according to the duplication-divergence model (24). (B) Average fraction of interactors shared between pairs of paralogous proteins with no (n=4), low (n=10), and high (n=3) functional divergence (28). Error bars: standard error of the mean. P-value: one-sided Kendall ranking correlation test (τ= association) (3). (C) Average fraction of shared interactors, corrected for low experimental coverage (3), and average protein sequence identity between pairs of paralogous proteins as a function of the estimated Δ time-since-duplication. Error bars: standard error of the mean (3). Dashed black line: corrected average fraction of shared interactors of non-paralogous pairs. myrs: million years. (D) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins originating from polyploidy events (n=109) as compared to other paralogous protein pairs of similar age (n=147). Error bars: standard error of the mean (3). P-values: Mann-Whitney U-test. (E) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins encoded by gene pairs with high or low co-expression correlation (top and bottom tertile, respectively) as a function of phylogeny-based age group. Error bars: standard error of the mean (3). P < 0.05 (*), < 0.01 (**), < 0.001 (***).

Fig. 4

Evidence for network evolution in AI-1_MAIN. (A) Interaction rewiring over time according to the duplication-divergence model (24). (B) Average fraction of interactors shared between pairs of paralogous proteins with no (n=4), low (n=10), and high (n=3) functional divergence (28). Error bars: standard error of the mean. P-value: one-sided Kendall ranking correlation test (τ= association) (3). (C) Average fraction of shared interactors, corrected for low experimental coverage (3), and average protein sequence identity between pairs of paralogous proteins as a function of the estimated Δ time-since-duplication. Error bars: standard error of the mean (3). Dashed black line: corrected average fraction of shared interactors of non-paralogous pairs. myrs: million years. (D) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins originating from polyploidy events (n=109) as compared to other paralogous protein pairs of similar age (n=147). Error bars: standard error of the mean (3). P-values: Mann-Whitney U-test. (E) Corrected average fraction of shared interactors (3), for pairs of paralogous proteins encoded by gene pairs with high or low co-expression correlation (top and bottom tertile, respectively) as a function of phylogeny-based age group. Error bars: standard error of the mean (3). P < 0.05 (*), < 0.01 (**), < 0.001 (***).