Convergent targeting of a common host protein-network by pathogen effectors from three kingdoms of life

Abstract

While conceptual principles governing plant immunity are becoming clear, its systems-level organization and the evolutionary dynamic of the host-pathogen interface are still obscure. We generated a systematic protein-protein interaction network of virulence effectors from the ascomycete pathogen Golovinomyces orontii and Arabidopsis thaliana host proteins. We combined this data set with corresponding data for the eubacterial pathogen Pseudomonas syringae and the oomycete pathogen Hyaloperonospora arabidopsidis. The resulting network identifies host proteins onto which intraspecies and interspecies pathogen effectors converge. Phenotyping of 124 Arabidopsis effector-interactor mutants revealed a correlation between intraspecies and interspecies convergence and several altered immune response phenotypes. Several effectors and the most heavily targeted host protein colocalized in subnuclear foci. Products of adaptively selected Arabidopsis genes are enriched for interactions with effector targets. Our data suggest the existence of a molecular host-pathogen interface that is conserved across Arabidopsis accessions, while evolutionary adaptation occurs in the immediate network neighborhood of effector targets.

Figure 1. Gor effector identification and interactome mapping

A. Golovinomyces orontii is a pathogenic ascomycete that diverged approximately 2.7 and 1.5 billion years ago (Gya), respectively, from the other Arabidopsis pathogens (Kemen and Jones, 2012; Markow, 2005). B. Effector identification pipeline and family relationships of identified and cloned Gor effector candidates (OEC) and the presence of homologs in the powdery mildews Blumeria graminis f. sp. hordei (green dots) and Erysiphe pisi (red dots). C. Our Y2H pipeline consists of three interrogation steps: screening, phenotyping and four-fold verification, resulting in the indicated number of interactions. D. Degree distribution of Arabidopsis proteins interacting with OECs. Asterisks indicate 8k_space proteins. E. Degree distribution of OECs interacting with Arabidopsis proteins in the 8k_space (light green) and 12K_space (dark green). See also Figure S1, Table S1.

Figure 2. Network integration

A. The integrated PPIN-2_{8k_sys} network of host proteins interacting with Gor, Hpa, and Psy effectors and physical interactions among host proteins derived from AI-1_MAIN in the 8k_space. B. Random and convergent interaction of Arabidopsis proteins (green) with effectors (red) can be distinguished by degree-preserving random network rewiring and simulation. C. Random interactors observed in degree-preserving network rewiring simulations of Gor effector-host protein interactions vs. observed value. D. As in C, but for Hpa effector-target interactions. E. As C for Psy effector-host protein interactions. F. Venn diagram showing observed overlap between effector-interactors from the three pathogens. G. Simulated random and observed overlap between Gor and Hpa effector-interactors. H. As in G but between Hpa and Psy effector-interactors. I. As in G but between Gor and Psy effector-interactors. J. As in G but between Gor, Hpa, and Psy effector-interactors. In C–E and G–J, random simulations are shown by black lines and observed values are highlighted by red arrows. See also Figure S2 and Table S1.

Figure 3. Phenotypic characterization of effector-interactor mutants

A. Heat-map summarizing the outcome of phenotypic analyses of mutants in genes encoding the indicated effector-interactors in infection assays with the noted pathogens and developmental stages. Host proteins are sorted by the number of pathogens interacting with them, then by number of observed phenotypes and performed assays. Mutant lines for 59 proteins interacting with effectors from a single pathogen did not show any disease phenotype and are not shown. Refer to Table S3 for raw data for all phenotyped loci and Figure S3 for complete results for all tested lines. B. Fraction of mutant lines for proteins interacting with effectors from the indicated number of pathogens that exhibited an edr, eds or divergent phenotypes across the assays. C. Pie chart representation of the phenotype density; the number of observed phenotypes relative to individual assays performed for that group. Each pie displays data for proteins that interacted with effectors from the number of pathogens given in the center. D. Fraction of mutant lines for proteins targeted by the indicated number of Gor effectors for which edr or eds phenotypes were observed. Numbers above bars indicate the number of targets in that class. E. As in D, but for proteins targeted by the indicated number of Hpa effectors for which edr or eds phenotypes were observed. Numbers above bars indicate the number of effector-interactors in the class.

Figure 4. TCP14 re-localizes effectors to sub-nuclear foci

A. Technical control demonstrating that YFP- and RFP channels do not leak into each other. The images show localization of TCP14-RFP and CRN4b-YFP; image data for YFP and RFP channels were collected for both. The same settings were then applied to all assays below. Note that TCP14-YFP forms sub-nuclear foci. B and C. The lower panel exhibits an enlarged view of a representative nucleus boxed in the upper panel. The histogram illustrates the intensity of fluorescent signal across the path indicated by the red arrow. All confocal pictures were taken 40–48 h after infiltration of Agrobacterium strains expressing the different fluorophore–tagged proteins. B. Negative control: TCP14 does not re-localize YFP. C. TCP14 re-localizes effectors from Psy (HopBB1), Hpa (HaRxL45) and Gor (OEC45) to sub-nuclear foci. D. TCP14 is co-immunoprecipitated by HopBB1, HaRxL45, and OEC45. All proteins were expressed from the CaMV 35S promoter in N. benthamiana leaves. ‘P.S’ in D denotes Ponceau S staining. See also Figure S4 and Table S4.

Figure 5. Proteins with high natural genetic variation interact with effector-interactors

A. Schematic illustration of the analysis in B–E: in the AI-1_MAIN network (left) the effector-interactors directly interacting with top Dθ-ranking gene products are counted and compared to the distribution of counts observed in 1,000 randomly rewired networks (single example shown). Effectors are shown for illustration only and not included in the analysis. B. Analysis as described in A. Plotted along the Y-axis are cumulative counts of effector-interactors interacting with proteins encoded by the top Dθ-ranking × genes. Data from AI-1_MAIN are shown as red dots, the black line shows the median of 1,000 randomly rewired networks, grey shaded areas show the 25^th and 75^th percentiles of values from rewiring controls. The lower panel shows the corresponding experimental P values (* 0.05; ** 0.005). The steep rise in the simulations at x = 56 is caused by a high-degree protein (NLM1, AT4G19030) at that position; the many rewired interactions for this protein increase the count of random interactors in all categories. C. As in B but counting proteins interacting with effectors from two or three pathogens. D. As in B but counting proteins interacting with effectors from three pathogens. E. As in B but counting proteins whose mutation caused altered immune phenotypes. F. Among the 13 interaction partners of the five most selected proteins are eleven effector-interactors, including the five most targeted proteins. The tables show for all top Dθ-ranking proteins and effector-interactors the relative combined rank, D_T, θ_W and the count of AAPs. The non-effector-interactor interactors of the variable proteins are ^§ AT1G51580 and ^§§ ZPF7. See also Figure S5 and Table S5.