Mass Spectrometry for the Proteomic Detection and Analysis of Plant Pathogenic Fungi including Soybean Rust

NON-TECHNICAL SUMMARY: The US remains at risk for the intentional and unintentional introduction of exotic plant pathogens. The introduction of a plant pathogen that is not endemic to the US can have severe economic impacts on agriculture industries and may engender food and feed shortage crises. In order to support US quarantine policies and protect US agriculture, the USDA will require advanced and novel pathogen detection capabilities. This project proposal addresses NRI Agricultural and Food Security Issues with the goals of developing alternative technologies that will enable the rapid detection of unknown plant pathogens. To accomplish this, the Cooper lab will develop and use mass spectrometry methods that will enable a reagent-independent method for the detection of proteins from a variety of plant pathogenic fungi. Mass spectrometry technology offers the inherent benefit of providing protein sequence information that can used to specifically identify and characterize proteins. This technology trait can be exploited to identify pathogens.

<P>
APPROACH: Even though PCR and antibody methods are very accurate and sensitive for the detection of pathogens, these technologies are not absolutely suitable for detecting a wide range of unknown pathogen threats. Mass spectrometry offers a complementary and potentially more robust technology for the detection of pathogen proteins since it can provide both accurate mass data and amino acid sequence information from the fragmentation patterns of collisions. These masses and patterns can be used to unambiguously identify a protein that distinguishes a pathogen. To test this, we will first use mass spectrometry to identify proteins from pure cultures of plant pathogenic fungi and determine if these mass spectra match those derived from GenBank or fungal genome/proteome databases. If the mass spectra match, then we will know that this method is suitable for detecting these fungi. If we cannot unambiguously identify a fungal protein with common protein databases, i.e. there is not enough known genome or proteome data for a fungus, then we will use the mass spectra as biomarkers for characterizing a fungus, which will aid in pathogen identification in future assays. Tests will also be performed to establish the lower limits fungal proteins that can be detected. Next, we will infect plants with plant pathogenic fungi and will use mass spectrometry methods and technology for the detection of these pathogens against a plant protein background. We will also mix fungal mycelia and spores into soil to determine if mass spectrometry would be sensitive enough to screen and survey field soil samples. In all cases, we will purify proteins from the samples, digest the proteins into peptides, separate the peptides using HPLC, analyze the peptides with a mass spectrometer and use a computer algorithm to interpret the mass spectra and deduce the sequence of the peptides, ultimately leading to the identification of the progenitor organism of the protein.

<P>
PROGRESS: 2005/03 TO 2008/09 <BR>
OUTPUTS: This project entails developing mass spectrometry methods and technology for the detection of plant pathogens. This project should help accomplish our goal of developing alternative technologies that will enable the rapid detection of pathogens. Activities: We initiated experiments to determine if a prevailing MS workflow was sufficient for identifying proteins of cultured fungal pathogens and thus identify the pathogen. We collected MS spectra for the pathogen proteins and were successful in identifying Ustilago maydis and Fusarium graminearum when matching their spectra to a database of protein sequences but not as successful in identifying Phytophthora sojae and Rhizoctonia solani. The difficulty for the others lay in the amounts of available protein sequence information for them leading to matches to non-target organisms. Consequently, it appeared that pathogens other than P. sojae and R. solani were detected. As a result of these studies we discovered a paradox: The need to identify a wide a range of pathogens requires a diverse protein sequence database, but such a comprehensive database will increase false-positive identifications. Effectively, this proves that the prevalent database search method for spectral interpretation is insufficient for pathogen diagnosis. However, this does not mean MS cannot be used for pathogen detection. Rather than search a protein sequence database, we devised another method that bypasses the protein sequence inferences and compares spectra directly to a spectral reference database. It is well established that comparison between an observed spectrum and reference spectra can make for very accurate diagnosis. This is the industry standard for detecting steroids or drugs in human fluids or pesticides in plants. Using the spectra collected from the pure cultures, we developed software to compound the spectra into reference libraries and then developed software to perform direct spectrum/reference spectrum comparisons. By performing a series of experiments, we were able to show that direct spectral matching is much more accurate and specific for detecting U. maydis, F. graminearum, P. sojae and R. solani than the previous tested method. The false-positive rate was nearly zero. Products: Two software platforms were developed. PTTRNFNDR was used to find sequence patterns in a real protein sequence database and then model and build randomized protein sequence databases that could be used as decoys to estimate the false-positive rate of matching pathogen spectra. This program was published and released to the public. The second platform, PyNISTPL, was used to make reference spectral libraries and compare observed spectra to reference spectra using the National Institute of Standards and Technology dot-product algorithm. The program has been released through a University of Maryland website (http://www.cbcb.umd.edu/~nedwards/research/PyNISTPL.html) As a direct result of the grant, 3 post-doctoral scientists were trained and went on to get jobs of their choices in research at Celgene, Frostburg State University and Microsoft. 1 graduate student at the University of Maryland was also trained. <BR> PARTICIPANTS: Individual, Bret Cooper, PI, led project and defined scope, wrote manuscripts. Individual, Neerav Padliya, Post-doc, performed mass spectrometry and data analysis Partner organization, University of Maryland, College Park, MD Partner organization, Johns Hopkins University, Baltimore, MD Collaborator, Steve Stein, National Institute of Standards and Technology, Gaithersburg, MD Collaborator, David Tabb, Vanderbilt University, Nashville, TN Collaborator, Nathan Edwards, University of Maryland, College Park, MD Training, Jian Feng, post-doc, Johns Hopkins University, Baltimore, MD Training, David Puthoff, post-doc, University of Maryland, College Park, MD Training, Xue Wu, graduate student, University of Maryland, College Park, MD <BR> <BR>
IMPACT: 2005/03 TO 2008/09 <BR>
Change of knowledge: Our experiments comparing spectra from pure cultured fungi to protein sequence databases proved that the database search method for spectral interpretation is insufficient for pathogen diagnosis because the false-positive rate is unpredictable for field samples. Since many other researchers have been developing microbial diagnostic methods that rely on protein sequence databases for spectrum identification, then a broad section of researchers will likely have to reevaluate their MS methods for microbe detection. The development of PTTRNFNDR was an important advance that allowed us to prove that the prevailing method of measuring the false-positive rate of matches between spectra and protein sequences was not statistically sufficient for modeling. These findings mean that a vast majority of members in the proteomics community using database search methods need more accurate false-positive rate modeling methods. The development of PyNISTPL and subsequent experiments proved that direct spectrum matching between observed spectra and reference libraries has a much lower false positive rate than matching to protein sequences. The results suggest that researchers wanting to use MS for pathogen identification should use a direct spectral matching method rather than compare pathogen spectra to protein sequence databases. These findings should have long-term effects and advance the utilization of mass spectrometry for the detection of pathogens.