Application of alignment-free bioinformatics methods to identify an oomycete protein with structural and functional...

Diverse plant pathogens export effector proteins to reprogram host cells. One of the most challenging goals in the molecular plant-microbe field is to functionally characterize the complex repertoires of effectors secreted by these pathogens. For bacterial pathogens, the predominant class of effectors is delivered to host cells by Type III secretion. For oomycetes, the predominant class of effectors is defined by a signal peptide that mediates secretion from the oomycete and a conserved RxLR motif. Downy mildew pathogens and Phytophthora species maintain hundreds of candidate RxLR effector genes in their genomes. Although no primary sequence similarity is evident between bacterial Type III effectors (T3Es) and oomycete RXLR effectors, some bacterial and oomycete effectors have convergently evolved to target the same host proteins. Such effectors might have evolved domains that are functionally similar but sequence-unrelated. We reasoned that alignment-free bioinformatics approaches could be useful to identify structural similarities between bacterial and oomycete effectors. To test this approach, we used partial least squares regression, alignment-free bioinformatics methods to identify effector proteins from the genome of the oomycete Hyaloperonospora arabidopsidis that are similar to the well-studied AvrE1 effector from Pseudomonas syringae. This approach identified five RxLR proteins with putative structural similarity to AvrE1. We focused on one, HaRxL23, because it is an experimentally validated effector and it is conserved between distantly related oomycetes. Several experiments indicate that HaRxL23 is functionally similar to AvrE1, including the ability to partially rescue an AvrE1 loss-of-function mutant. This study provides an example of how an alignment-free bioinformatics approach can identify functionally similar effector proteins in the absence of primary sequence similarity. This approach could be useful to identify effectors that have convergently evolved regardless of whether the shared host target is known.