A comparative hidden Markov model analysis pipeline identifies proteins characteristic of cereal-infecting fungi

Abstract

Background: Fungal pathogens cause devastating losses in economically important cereal crops by utilisingpathogen proteins to infect host plants. Secreted pathogen proteins are referred to as effectors and have thus farbeen identified by selecting small, cysteine-rich peptides from the secretome despite increasing evidence that notall effectors share these attributes.Results: We take advantage of the availability of sequenced fungal genomes and present an unbiased method forfinding putative pathogen proteins and secreted effectors in a query genome via comparative hidden Markov modelanalyses followed by unsupervised protein clustering. Our method returns experimentally validated fungal effectors inStagonospora nodorum and Fusarium oxysporum as well as the N-terminal Y/F/WxC-motif from the barley powderymildew pathogen. Application to the cereal pathogen Fusarium graminearum reveals a secreted phosphorylcholinephosphatase that is characteristic of hemibiotrophic and necrotrophic cereal pathogens and shares an ancient selectionprocess with bacterial plant pathogens. Three F. graminearum protein clusters are found with an enriched secretionsignal. One of these putative effector clusters contains proteins that share a [SG]-P-C-[KR]-P sequence motif in theN-terminal and show features not commonly associated with fungal effectors. This motif is conserved in secretedpathogenic Fusarium proteins and a prime candidate for functional testing.Conclusions: Our pipeline has successfully uncovered conservation patterns, putative effectors and motifs offungal pathogens that would have been overlooked by existing approaches that identify effectors as small, secreted,cysteine-rich peptides. It can be applied to any pathogenic proteome data, such as microbial pathogen data of plantsand other organisms.