Secreted papain-like Cys proteases are important players in plant immunity. We previously reported that the C14 protease of tomato is targeted by cystatin-like EPIC proteins that are secreted by the oomycete pathogen Phytophthora infestans (Pinf) during infection. C14 has been under diversifying selection in wild potato species coevolving with Pinf and reduced C14 levels result in enhanced susceptibility for Pinf. Here, we investigated the role C14-EPIC-like interactions in the natural pathosystem of Arabidopsis with the oomycete pathogen Hyaloperonospora arabidopsidis (Hpa). In contrast to the Pinf-solanaceae pathosystem, the C14 orthologous protease of Arabidopsis, RD21, does not evolve under diversifying selection in Arabidopsis, and rd21 null mutants do not show phenotypes upon compatible and incompatible Hpa interactions, despite the evident lack of a major leaf protease. Hpa isolates express highly conserved EPIC-like proteins during infections, but it is unknown if these HpaEPICs can inhibit RD21 and one of these HpaEPICs even lacks the canonical cystatin motifs. The rd21 mutants are unaffected in compatible and incompatible interactions with Pseudomonas syringae pv. tomato, but are significantly more susceptible for the necrotrophic fungal pathogen Botrytis cinerea, demonstrating that RD21 provides immunity to a necrotrophic pathogen.

Rcr3 and Pip1 are paralogous secreted papain-like proteases of tomato. Both proteases are inhibited by Avr2 from the fungal pathogen Cladosporium fulvum, but only Rcr3 acts as a co-receptor for Avr2 recognition by the tomato Cf-2 immune receptor. Here, we show that Pip1-depleted tomato plants are hyper-susceptible to fungal, bacterial, and oomycete plant pathogens, demonstrating that Pip1 is an important broad-range immune protease. By contrast, in the absence of Cf-2, Rcr3 depletion does not affect fungal and bacterial infection levels but causes increased susceptibility only to the oomycete pathogen Phytophthora infestans. Rcr3 and Pip1 reside on a genetic locus that evolved over 36 million years ago. These proteins differ in surface-exposed residues outside the substrate-binding groove, and Pip1 is 5- to 10-fold more abundant than Rcr3. We propose a model in which Rcr3 and Pip1 diverged functionally upon gene duplication, possibly driven by an arms race with pathogen-derived inhibitors or by coevolution with the Cf-2 immune receptor detecting inhibitors of Rcr3, but not of Pip1.