A peptide resource for the analysis of Staphylococcus aureus in host-pathogen interaction studies

Abstract

Staphylococcus aureus is an opportunistic human pathogen, which can cause life-threatening disease. Proteome analyses of the bacterium can provide new insights into its pathophysiology and important facets of metabolic adaptation and, thus, aid the recognition of targets for intervention. However, the value of such proteome studies increases with their comprehensiveness. We present an MS-driven, proteome-wide characterization of the strain S. aureus HG001. Combining 144 high precision proteomic data sets, we identified 19 109 peptides from 2088 distinct S. aureus HG001 proteins, which account for 72% of the predicted ORFs. Peptides were further characterized concerning pI, GRAVY, and detectability scores in order to understand the low peptide coverage of 8.7% (19 109 out of 220 245 theoretical peptides). The high quality peptide-centric spectra have been organized into a comprehensive peptide fragmentation library (SpectraST) and used for identification of S. aureus-typic peptides in highly complex host-pathogen interaction experiments, which significantly improved the number of identified S. aureus proteins compared to a MASCOT search. This effort now allows the elucidation of crucial pathophysiological questions in S. aureus-specific host-pathogen interaction studies through comprehensive proteome analysis. The S. aureus-specific spectra resource developed here also represents an important spectral repository for SRM or for data-independent acquisition MS approaches. All MS data have been deposited in the ProteomeXchange with identifier PXD000702 (http://proteomecentral.proteomexchange.org/dataset/PXD000702).

Keywords: Host-pathogen interactions; Mass spectrometry (MS); Microbiology; Spectral library; Staphylococcus aureus.

Figure 1

Experimental workflow. Equal samples, which were expected to cover many proteins of the theoretical proteome of S. aureus HG001, were subjected to three different methods of proteome analysis either with or without fractionation prior to mass spectrometric analysis. Peptides and proteins were identified from the resulting data sets of biological and technical replicates using MASCOT. All data were then combined in a new SpectraST database. Exemplary protein extracts from S9 host cell cultures infected with S. aureus HG001 were processed in a sample-specific way. On comparing the identification of peptides and proteins which resulted from a MASCOT search with those retrieved from a search using the newly generated SpectraST database, the new SpectraST database led to higher identification rates than the traditional MASCOT approach using a database of the theoretical proteins. An improvement of protein identification was especially observed when the identification results of both the MASCOT and the SpectraST database search were combined. exp – exponential growth phase; t0 – entry into stationary growth phase; t4 – 4 h after t0; ABC – ammonium bicarbonate; UT – urea/thiourea.

Figure 2

Detection of peptides and protein identification using three different approaches of proteome analysis. Comparison of detected peptides including peptides with missed cleavages (A) and identified proteins (B) for the three methods of proteome analysis under investigation (three biological replicates per method, without technical replicates). Some peptides were additionally detected in different charge states or in a modified form (e.g. oxidized), but the Venn diagram only shows unique peptide sequences independent of further parameters.

Figure 3

Peptide and protein coverage. The coverage of the annotated proteome of S. aureus HG001 using the three methods of proteome analysis under investigation (A).The Voronoi treemap was created on the basis the TIGRFAMs protein family classification scheme [33] by using HMMER/HMMScan [34]. The small graphs in the upper part display the included functional annotations. Peptides form the lowest level of area subdivision. The area per peptide represents the peptide length (number of amino acids). Therefore, the area per protein correlates with the protein size. Detected peptides of proteins identified by at least one of the three methods applied in this study (three biological replicates, without technical replicates) are colored in shades of orange. The color represents the sequence coverage of the proteins by the detected peptides. Peptides not detected in any set of three biological replicates per method are colored gray. Nevertheless, some of these proteins were identified when technical replicates were included to generate an even more comprehensive new database (SpectraST). Coverage of peptides from an in silico digestion in the real MS data sets (B). Included functional annotations are the same as depicted in Fig. 3A. Additionally, the protein label size correlates with the protein size. White dots indicate the detected peptides from the MS data sets. Coloring was applied to the protein labels: Dark blue labels indicate proteins not identified. Light blue, yellow, and white coloring indicates in this order increasing coverage of identified proteins. The Voronoi treemap contains about 220245 theoretically expected peptides from an in silico digestion. Of these, about 19109 peptides were detected in the MS data sets.

Figure 3

Peptide and protein coverage. The coverage of the annotated proteome of S. aureus HG001 using the three methods of proteome analysis under investigation (A).The Voronoi treemap was created on the basis the TIGRFAMs protein family classification scheme [33] by using HMMER/HMMScan [34]. The small graphs in the upper part display the included functional annotations. Peptides form the lowest level of area subdivision. The area per peptide represents the peptide length (number of amino acids). Therefore, the area per protein correlates with the protein size. Detected peptides of proteins identified by at least one of the three methods applied in this study (three biological replicates, without technical replicates) are colored in shades of orange. The color represents the sequence coverage of the proteins by the detected peptides. Peptides not detected in any set of three biological replicates per method are colored gray. Nevertheless, some of these proteins were identified when technical replicates were included to generate an even more comprehensive new database (SpectraST). Coverage of peptides from an in silico digestion in the real MS data sets (B). Included functional annotations are the same as depicted in Fig. 3A. Additionally, the protein label size correlates with the protein size. White dots indicate the detected peptides from the MS data sets. Coloring was applied to the protein labels: Dark blue labels indicate proteins not identified. Light blue, yellow, and white coloring indicates in this order increasing coverage of identified proteins. The Voronoi treemap contains about 220245 theoretically expected peptides from an in silico digestion. Of these, about 19109 peptides were detected in the MS data sets.

Figure 4

Characterization of the detected peptides and peptides retrieved from an in silico digestion by different scoring methods. The frequency of peptides is depicted in classes of CONSeQuence score [42] values (A), CHEMscore [39] values (B), and Detectability Predictor score [43] values (C). The upper part of the histograms refers to the set of theoretical peptides from the in silico digestion (gray), and the lower part of the histograms displays the data set of detected peptides (orange).

Figure 5

Protein identifications from four data sets of S. aureus HG001 cells after internalization into human bronchial epithelial S9 cells. Two biological replicates and two points in time after infection were analyzed: biological replicate 1 (BR1); biological replicate 2 (BR2); 2.5 h after infection; 4.5 h after infection, respectively. The numbers indicate identified proteins, including proteins with only one detected peptide.