Clinical Chemistry 63:8 1398–1408 (2017)
Proteomics and Protein Markers
A Genoproteomic Approach to Detect Peptide Markers of Bacterial Respiratory Pathogens Honghui Wang,1 Steven K. Drake,1 Chen Yong,2 Marjan Gucek,2 Matthew A. Lyes,1 Avi Z. Rosenberg,3 Erik Soderblom,5 M. Arthur Moseley,5 John P. Dekker,4 and Anthony F. Suffredini1*
BACKGROUND: Rapid identification of respiratory pathogens may facilitate targeted antimicrobial therapy. Direct identification of bacteria in bronchoalveolar lavage (BAL) by matrix-assisted laser desorption/ ionization time-of-flight mass spectrometry is confounded by interfering substances. We describe a method to identify unique peptide markers of 5 gramnegative bacteria by liquid chromatography–tandem mass spectrometry (LC-MS/MS) for direct pathogen identification in BAL. METHODS: In silico translation and digestion were performed on 14 –25 whole genomes representing strains of Acinetobacter baumannii, Moraxella catarrhalis, Pseudomonas aeruginosa, Stenotrophomonas maltophilia, and Klebsiella pneumoniae. Peptides constituting theoretical core peptidomes in each were identified. Rapid tryptic digestion was performed; peptides were analyzed by LCMS/MS and compared with the theoretical core peptidomes. High-confidence core peptides (false discovery rate ⬍1%) were identified and analyzed with the lowest common ancestor search to yield potential speciesspecific peptide markers. The species specificity of each peptide was verified with protein BLAST. Further, 1 or 2 pathogens were serially diluted into pooled inflamed BAL, and a targeted LC-MS/MS assay was used to detect 25 peptides simultaneously. RESULTS: Five unique peptides with the highest abundance for each pathogen distinguished these pathogens with varied detection sensitivities. Peptide markers for A. baumannii and P. aeruginosa, when spiked simultaneously into inflamed BAL, were detected with as few as 3.6 (0.2) ⫻ 103 and 2.2 (0.6) ⫻ 103 colony-forming units, respectively, by targeted LC-MS/MS.
1
Critical Care Medicine Department, Clinical Center, National Institutes of Health, Bethesda, MD; 2 Proteomic Core Facility, National Heart, Lung and Blood Institute, National Institutes of Health, Bethesda, MD; 3 Kidney Disease Section, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD; 4 Microbiology Service, Department of Laboratory Medicine, Clinical Center, National Institutes of Health, Bethesda, MD; 5 Proteomics and Metabolomics Shared Resource, Duke University School of Medicine, Durham, NC. * Address correspondence to this author at: 10/2C145, Critical Care Medicine Department, Clinical Center, National Institutes of Health, 10 Center Drive, Bethesda, MD 20892. Fax 301-402-1213; e-mail
[email protected].
1398
CONCLUSIONS: This proof-of-concept study shows the feasibility of identifying unique peptides in BAL for 5 gram-negative bacterial pathogens, and it may provide a novel approach for rapid direct identification of bacterial pathogens in BAL.
© 2017 American Association for Clinical Chemistry
Rapid identification of bacteria causing hospitalacquired pneumonia may improve patient outcomes by providing a basis for specific rather than empiric therapy (1 ). This is particularly relevant in immunocompromised patients where analysis of bronchoalveolar lavage (BAL)6 by microbiologic stains, cultures, and molecular assays is used to determine the etiology of pneumonias (2, 3 ). Although tests based on RNA or DNA amplification are employed to detect viruses and some bacteria, there are few rapid, culture-independent methods that can be used directly on primary BAL specimens for microbiologic diagnosis of bacterial pneumonias (1, 4, 5 ). Species-level identification requires culture results that may be available only after 24 –72 h (6 ). In facilities where multidrug-resistant gram-negative bacteria are prevalent, initial empiric therapy may require antimicrobials that have intrinsic toxicity (i.e., colistin-related nephrotoxicity or neurotoxicity) or additive toxicity to other drugs (7, 8 ). Thus, rapid diagnosis of a specific bacterial pathogen may limit patient exposure to unnecessary antimicrobials and provide the basis for targeted antimicrobial therapy with a shorter interval to treatment (3, 9 ). Proteomic identification of microorganisms has rapidly evolved during the past decade (10 ). Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) analysis of clinical isolates
Disclaimer: The opinions expressed in this paper are those of the authors and do not necessarily reflect those of the Department of Health and Human Services or the National Institutes of Health. Received December 8, 2016; accepted May 2, 2017. Previously published online at DOI: 10.1373/clinchem.2016.269647 © 2017 American Association for Clinical Chemistry 6 Nonstandard abbreviations: BAL, bronchoalveolar lavage; CFU, colony-forming unit; LCA, lowest common ancestor; MALDI-TOF, matrix-assisted laser desorption/ionization– time of flight mass spectrometry; NCBI, National Center for Biotechnology Information.
Genoproteomic Detection of Bacterial Pathogens
Table 1. Bacterial isolates used for protein identification.
Species
A. baumannii
a
Isolate
Strain
MLST (Oxford)
Protein FASTA
ABNIH2
GenBank
AFTA00000000.1
Culture plate
ABNIH1
strain A
ST281
ABNIH2
strain B
ST348
Sheep Blood Agar
Sheep Blood Agar
ABNIH3
strain C
ST353
Sheep Blood Agar
ABNIH6
strain D
ST231
Sheep Blood Agar
ATCC 19606a
strain E
ST931
Sheep Blood Agar
OIFC11
strain F
ST128
Sheep Blood Agar
P. aeruginosa
ATCC 27853
NCGM2.S1
GCA_000284555.1
Sheep Blood Agar
M. catarrhalis
ATCC 8176
BBH18
GCA_000092265.1
Chocolate Agar
S. maltophilia
ATCC 51331
K279a
GCA_000072485.1
Sheep Blood Agar
K. pneumoniae
ATCC BAA-1705
BAA-1705
GCA_000349265.2
Sheep Blood Agar
For ATCC 19606, the annotation protein FASTA file containing 3652 protein sequences was used in this study. The current annotation contains 3677 protein sequences.
has become a standard means to identify microorganisms based on intact protein profiling of cultured bacterial colonies combined with a matrix (11 ). The spectrum of the resulting ions separated by their m/z is compared with a database, and the pathogen’s final identification is based on a match with a species-specific score. The direct application of MALDI-TOF MS to clinical specimens has been described after removal of leukocytes and erythrocytes (12, 13 ). However, direct application of MALDITOF MS to urine specimens may be limited by defensins that suppress the bacterial spectra and generate additional peaks that interfere with automated pathogen identification (14 ). Mass spectrometry-based approaches using the “bottom-up” analysis of peptides derived from bacterial protein digests have accelerated the discovery of unique bacterial peptide markers. We recently described a rapid genoproteomic method to identify strain-specific peptide markers based on the liquid chromatography–tandem mass spectrometry (LC-MS/MS) profiles of digested peptides and peptidomic analysis (15 ), an approach combining the generation of a theoretical peptidome from in silico translation and digestion of a bacterial whole genome sequence with bioinformatic confirmation of peptide specificity. Experimental detection of the peptide as unique for a specific bacterial strain verifies its usefulness as a marker (15 ). Here, we applied this approach to identify unique peptides for 5 major gram-negative pathogens that are associated with pneumonia (3, 16 –18 ), with the goal of developing a proof-of-concept method that may provide the basis for the development of a multiplexed assay for rapid and direct pathogen identification in BAL by LC-MS/MS.
Materials and Methods BACTERIAL ISOLATES
Bacterial isolates of Acinetobacter baumannii, Moraxella catarrhalis, Pseudomonas aeruginosa, Stenotrophomonas maltophilia, and Klebsiella pneumoniae were grown either on blood agar or chocolate agar plates (Remel) for 18 –24 h at 35 °C with 5% CO2 (Table 1). Cell suspensions were made and diluted as previously described (15 ) (see Methods in the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/ content/vol63/issue8). SPIKED BAL SPECIMENS
Spiked specimens for study were made by taking stored acellular BAL from 7 healthy volunteers previously challenged with subsegmental saline (normal) or endotoxin (inflammatory) in an NIH investigational review boardapproved study of lung inflammation (protocol 92-CC0141). One milliliter (1 mL) of pooled aliquots (see Table 1 in the online Data Supplement) were mixed with 100 L of 0.1⫻, 0.01⫻, 0.001⫻, and 0.0001⫻ stocks of A. baumannii (ABNIH2) and/or P. aeruginosa (ATCC 27853) or M. catarrhalis (ATCC 8176), S. maltophilia (ATCC 51331), and K. pneumoniae (ATCC BAA-1705) (Table 1 and see Table 1 in the online Data Supplement) to produce a dilution series of spiked BAL specimens. ANALYSIS OF CLINICAL BAL SAMPLE
To show the feasibility of applying these methods to clinical samples, a frozen BAL aliquot without any preprocessing was analyzed (kindly provided by J. Kovacs, MD and acquired under research protocol NCT01212042). The clinical sample was mixed with Mucolyse (ProClinical Chemistry 63:8 (2017) 1399
Laboratory Diagnostics) for BAL (1:10 v:v) and frozen within 1 h of acquisition. The gram stain showed mixed oral flora and the subsequent culture revealed heavy growth of P. aeruginosa identified by MALDI-TOF MS. RAPID TRYPTIC DIGESTION AT 55°C
The sample processing and digestion have been previously described (15 ). Rapid tryptic digestion was performed at 55 °C for 15 min using a Discover microwave system (CEM Corporation) at 50 W or in a water bath. Both methods have previously been shown to be comparable in terms of the number of peptides and proteins identified (15 ). The intact bacterial cells were digested with trypsin without cell lysis, protein denaturation, disulfide reduction, or alkylation. LC-MS/MS
Samples were analyzed on the following instruments during the study: selected digests were sent to Duke Proteomic Facility for protein identification using a hybrid quadrupole-orbitrap mass spectrometer (Thermo Q-Exactive LC-MS) and multiple reaction monitoring LC-MS was acquired with Xevo TQ-S (Waters). Additional protein identification and targeted LCMS/MS were performed using the Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific) in the National Heart, Lung and Blood Institute Core Facility as previously described (15 ). Targeted LC-MS/MS was run in scheduled targeted MS2 (tMS2) mode with a retention time window of 4 min. Synthetic labeled peptides (500 fmol each) were used for determine the retention times for each peptide and as a spectral library for comparison. The maximum acquisition time and gain for each targeted MS/MS spectrum was 100 ms and 1 ⫻ 105 except for clinical frozen BAL for which the acquisition time and gain were set to 200 ms and 2 ⫻ 105, respectively. A detailed protocol is provided in the online Data Supplement. Skyline 3.5 (MacCross Lab) was used for tMS2 data processing. Gradients and mobile phases used were similar across the 4 platforms (see online Data Supplemental Methods). PROTEIN IDENTIFICATION
Protein search and identification were performed using the Proteome Discoverer 1.4 (Thermo Fisher Scientific) and the raw data set from both Q-Exactive and Orbitrap Fusion LC-MS/MS against a custom-built FASTA database. The FASTA database consists of 5 individual protein FASTA files from the 5 bacterial species studied (A. baumannii ABNIH2, M. catarrhalis BBH18, P. aeruginosa NCGM2.S1, S. maltophilia K279a, and K. pneumoniae ATCC BAA-1705). The protein FASTA files were downloaded from the National Center for Biotechnology Information (NCBI) website (19 ). The search algorithm Sequest HT was used with the following pa1400
Clinical Chemistry 63:8 (2017)
rameters: precursor mass tolerance: 20 ppm; fragment mass tolerance: 0.05 Da; and dynamic modifications of oxidation on M and deamination on N and Q. The following parameters were included in the Decoy database search criteria: target false discovery rate (strict) for highconfidence peptides ⫽ 1% and target false discovery rate (relaxed) for medium-confidence peptides ⫽ 5%. CORE PEPTIDOME CREATION
The workflow used to develop the bacterial species’ panpeptidome (collection of all peptides in a group of selected isolates), core-peptidome (common peptides that are present in all isolates in the constructed panpeptidome), and genome-specific peptidome (peptides that are present only in a given genome) is summarized in Fig. 1 in the online Data Supplement. The GenBank accession number for the isolates used to construct the pan-peptidomes, core peptidomes, and genome-specific peptidomes is summarized in Table 2 in the online Data Supplement. The protein FASTA files were downloaded from the NCBI website (20 ). The selected set of protein FASTA files was uploaded to the web-based program Peptidome Analysis (21 ) to create the core peptidomes. The resulting text file was processed with UltraEdit (IDM Computer Solutions, Inc.) to filter out peptides with amino-acid lengths ⬍6 and ⱖ25. The core peptides were imported to an Excel spreadsheet (Microsoft) for comparison with detected high-confidence peptides using the MATCH function. PEPTIDOME ANALYSIS
A Unipept web tool, Peptide Finder (21 ), was used to analyze the peptidome similarity among the selected protein FASTA databases. The protein FASTA databases for the selected bacterial isolates were uploaded to the website and peptidome-based clustering was performed on the website. LOWEST COMMON ANCESTOR ANALYSIS
Another Unipept web tool, pept2lca (21 ), was used to perform lowest common ancestor (LCA) analysis for every peptide as described previously (22, 23 ). Peptides were defined as species-specific if they were not found in the genomes of any different species by means of the LCA analysis. The set of identified core peptides that were found by LCA to be species-specific was selected as potential unique peptide markers. Further details regarding the LCA analysis have been previously described (15 ). Results The effects of human inflamed lavage on bacteria identification by MALDI-TOF are shown in Fig. 1. The characteristic spectra of P. aeruginosa was completely suppressed when analyzed in the presence of inflamed BAL
Genoproteomic Detection of Bacterial Pathogens
Fig. 1. MALDI-TOF MS spectra of a single clone culture of P. aeruginosa without (A) and with BAL (B). Approximately 1 × 108 bacterial cells were spiked into 1 mL of BAL, spun, decanted, washed (ethanol), and then lysed (formic acid, acetonitrile). Two microliters of the 2 lysates were spotted on the MALDI target. Intense peaks at (3375, 3446, and 3490 Da) in the BAL matrix depressed the P. aeruginosa spectrum.
matrix, with the only major peaks found at m/z 3375, 3446, and 3490 Da. Using LC-MS/MS, these peaks have been previously identified as human neutrophil defensins 2, 1, and 3, respectively, (14, 24 ), and were confirmed in our laboratory (data not shown). To develop a rapid method of identification of gram-negative bacteria in BAL, we chose 5 gram-negative bacteria known to cause serious pneumonia in immunocompromised hosts. Peptidomic analysis using the Unipept web tool demonstrated that the peptidome similarities among these 5 pathogens ranged from 0.62–1.61% (see Fig. 2 in the online Data Supplement), suggesting that differences in peptide sequences might be used to identify these pathogens (22 ). The genome assemblies that were used to create core peptidomes for A. baumannii, M. catarrhalis, P. aerugi-
nosa, S. maltophilia, and K. pneumoniae are summarized in Table 2 in the online Data Supplement. In total, 23 A. baumannii genome assemblies were chosen to span the dendrogram of 1059 A. baumannii genome assemblies (19 ). To analyze the effect of a cluster of more closely related strains, we included 4 A. baumannii strains that have been extensively studied (ABNIH1, ABNIH2, ABNIH3, and ABNIH6) and were representative of the A. baumannii strains detected during a 2007 hospital outbreak of A. baumannii at our facility (25, 26 ). The number of genome assemblies used to create core peptidomes for M. catarrhalis, P. aeruginosa, S. maltophilia, and K. pneumoniae were 14, 25, 15, and 21, respectively. To detect the most conserved and robust core peptides specific to a given pathogen species by LC-MS/MS data, 2 strategies were evaluated. In the first strategy, Clinical Chemistry 63:8 (2017) 1401
conserved, well-performing peptides were identified by assessing tryptic digests from different strains of the same species to identify core peptides that were common across strains. In the second strategy, the effects of varying quantities of the same isolate were evaluated to identify core peptides that were detected in both low and high cell density (by nephelometer) samples reflecting those peptides with high detection sensitivity and/or high abundance. Both methods succeeded in producing a list of readily detected core peptides. For A. baumannii, we used 6 different strains to detect the core peptides. For the other 4 bacteria studied, we used the samples with higher and lower cell densities from the same isolate (see Table 3 in the online Data Supplement). The detected peptides from both strategies were compared with the theoretical core peptidomes of the tested species. To assess the specificity of each peptide, we used the web tool Tryptic Peptide Analysis, which analyzes each peptide to find its LCA on the basis of UniProtKB records along with a complete taxonomic lineage derived from the NCBI taxonomy (22, 23 ). High-confidence core peptides were defined as those detected during the protein identification and search using Thermo Proteome Discoverer with a target false discovery rate of 1%. The species-specific peptides were selected as potential peptide markers and underwent further confirmation using the Protein BLAST (27 ). The number of high-confidence peptides detected by LC-MS/MS ranged from 189 for K. pneumoniae to 6921 for sample AB-D of A. baumannii (see Table 3 in the online Data Supplement). For M. catarrhalis, we detected fewer peptides with sample MC-H (with higher cell density) than with sample MC-L (with lower cell density), possibly because of overloading the column with either sample or ion suppression. For A. baumannii with 6 known strains, 919 –2406 high-confidence core peptides were detected, of which only 39, 68, 76, 105, 43, and 35 core peptides were found to be specific to A. baumannii for samples with strains AB-A through AB-F, respectively. This was ⬃4% of the detected highconfidence core peptides. In all 6 samples, 26 A. baumannii-specific core peptides were detected. In the analysis of the LC-MS/MS data of K. pneumoniae (KP), 51 and 1276 high-confidence core peptides were detected for samples of KP-low (KP-L) and KP-high (KP-H), respectively. However, the LCA search yielded no species-specific peptides. Of 1276 high-confidence core peptides in the sample KP-H, the LCA of 1072 core peptides was classified as bacteria (superkingdom). To understand the underlying reason, we manually inspected the LCA search results for the 51 highconfidence core peptides that were detected in the sample of KP-L and 300 high-confidence core peptides (with at least 2 spectra) in the sample of KP-H. Fig. 3 in the online Data Supplement shows the taxonomic lineage of 1402
Clinical Chemistry 63:8 (2017)
peptide IGSEAYNQQLSEK by LCA search. This peptide was detected in both samples of KP-L and KP-H. IGSEAYNQQLSEK was also found in Escherichia coli ISC56. The taxonomic lineage is linked back to Enterobacteriaceae (family). Notably, out of the large number of E. coli genome assemblies in the public genome assemblies (4503, accessed on April 20, 2016), IGSEAYNQQLSEK is not commonly found in sequenced E. coli species. Therefore, this peptide remains highly specific to Klebsiella at a genus level. To further determine the specificity of IGSEAYNQQLSEK, we used Protein BLAST to check its specificity. It matched 100% in sequence to ⬎700 records in the Unipept database related to strains of K. pneumoniae and a few strains of Klebsiella variicola and Klebsiella oxytoca. Thus, with manual inspection of the LCA analysis for each peptide, the Protein BLAST analysis, and using abundance in the LC-MS/MS spectra as a criterion, 5 unique peptides with relatively high abundance for each pathogen were chosen as potential peptide markers (see Table 5 in the online Data Supplement). The LCA analysis of the high-abundance peptide SALVNEYNVDASR was found to be specific to the genus Acinetobacter and highly specific to A. baumannii or its complex with few exceptions (see Fig. 4 in the online Data Supplement). To determine if this proof-of-concept approach could detect the potential peptide markers of the selected pathogens in BAL (see Table 4 in the online Data Supplement), we spiked 1 or 2 pathogens into the BAL specimens. Spiked inflamed BAL specimens with decreasing colony-forming unit (CFU) counts of each pathogen or mixtures of A. baumannii and P. aeruginosa were prepared and digested with rapid 15 min of digestion at 55 °C. The samples were analyzed by targeted LCMS/MS analysis by Thermo Orbitrap Fusion. Fig. 2 shows the relative intensities (Skyline 3.5) of 5 detected peptide markers for 5 pathogens in 27 pathogen-spiked BAL specimens (see online Data Supplement). The library represents the detected transitions for each peptide by bottom-up protein identification. Each color line represents 1 transition and its relative intensity among other transitions. On the x axis, A, K, M, P, and S represent A. baumannii, K. pneumoniae, M. catarrhalis, P. aeruginosa, and S. maltophilia, respectively, with the number following the letter representing the relative level of the CFU. The most sensitive markers are listed and based on serial dilution; the minimum CFUs required to be detected were estimated to be 3.5 (0.2) ⫻ 103, 2.2 (0.6) ⫻ 103, 7.6 (0.6) ⫻ 104, 3.6 (0.4) ⫻ 104, and 1.9 (0.5) ⫻ 104 for A. baumannii, P. aeruginosa, M. catarrhalis, S. maltophilia, and K. pneumoniae, respectively (Fig. 3). Other peptides had various detection sensitivities (Table 2). In the clinical BAL, VNAVGYGESRPVADNATAEGR was detected with low signal intensity and
Genoproteomic Detection of Bacterial Pathogens
Fig. 2. Relative quantification of peptide markers in spiked inflamed bronchoalveolar lavage (BAL) specimens for 5 bacteria as either 1 or 2 pathogens. Actual CFU estimates are found in Table 1 in the online Data Supplement.
QAGAEIVSFVR was not detected (data not shown). To confirm the positive identification, the analysis was repeated with the addition of 40 fmol of labeled peptides and the digests desalted with a C18 ZipTip. Both peptide markers, VNAVGYGESRPVADNATAEGR and QAGAEIVSFVR, were detected as shown in Fig. 4. The retention times, mass error, and the transition rank order of the detected native peptides matched well with the labeled peptide, confirming the identification of P. aeruginosa. The ratio dot product (rdotp, Skyline 3.5) is the normalized dot product of the light transition peak areas with the heavy transition peak areas and is a measure of whether the transition peak areas in the 2 label types are in the same ratio. The rdotp for both VNAVGYGESRPVADNATAEGR and QAGAEIVSFVR was 0.94 and 0.99, respectively. To show the reproducibility of the results using a different analytical platform, we spiked a series of 2-bacterial mixtures (A. baumanni and P. aeruginosa) using labeled peptides and multiple reaction monitoring into BAL specimens in the way similar to that described
for Orbitrap tMS2 and analyzed the mixtures by Xevo TQ-S (Waters). Methods and results are shown in Table 3 and Figs. 5 and 6 in the online Data Supplement. These results were comparable with those obtained with the Orbitrap tMS2. Discussion To facilitate the rapid identification of respiratory pathogens, we analyzed cultured P. aeruginosa alone and in the presence of inflamed BAL by MALDI-TOF MS. Our results showed that the characteristic MALDI-TOF MS spectrum of these bacteria was limited by the presence of defensins. Other studies using MALDI-TOF MS for direct identification of bacteria in urine found that defensins suppress the intensity of bacterial protein peaks and compromise the database matching for bacterial identification (14 ). As an alternative approach for rapid pathogen identification, we developed a genoproteomic method for discovering bacterial peptide markers that can be used to identify specific bacterial pathogens. As a Clinical Chemistry 63:8 (2017) 1403
Fig. 3. Representative LC-MS/MS chromatograms of peptide markers from 5 bacterial pathogens. On the left side of the figure, the transition rank order of the labeled peptide is shown. The native peptides detected with decreasing bacteria CFU are shown from the left to the right in each row.
proof-of-concept, we showed that this method provided the basis to identify 5 distinct bacterial species and simultaneously detect 2 pathogens (A. baumannii and P. aeruginosa) in spiked acellular inflamed BAL. Rapid digestion was achieved without cell lysis, protein denaturation, or reduction/alkylation. Although reduction and alkylation are required when a high level of protein coverage is needed, the rapid protein digestion approach provided the means to discover unique peptide markers that could be quickly and routinely identified with LCMS/MS. Moreover, as a proof of concept, we could detect multiple pathogens within a mixture using this workflow. Using spiked BAL specimens, we showed that LC-MS/MS can detect some peptides (i.e., IGSEAYNQQLSEK, which is specific to K. pneumoniae) with as few as 1.9 (0.5) ⫻ 103 CFU. We believe that this method has potential implications for culture-free, rapid detection of multiple pathogens directly from patient BAL samples without extensive bottom-up proteomic sample processing. We acknowledge several potential limitations of this approach. The successful clinical application of this method will depend on several factors. The availability and quality of genome assemblies are fundamental for building peptidome sets to guide the search for unique peptide markers. Although the number of bacterial genomes that have been sequenced has steadily increased, 1404
Clinical Chemistry 63:8 (2017)
quality control measures on the bacterial genome sequence data in public databases have not been consistent (28 ). Any inaccurate information can confound and compromise downstream analysis. Particularly troublesome are contaminated genome assemblies and taxonomic misclassifications. The revised RefSeq prokaryotic genome addresses inconsistencies in the prokaryotic data set gene annotation and enhances the identification of unique prokaryotic protein sequences (29 ). Future work may require refinements in the definition of core peptides and modification of the stringency criteria (e.g., instead of requiring a given core peptide to be present in 100% of isolates, the criterion may be relaxed). For LCA searches, an algorithm to identify or remove some outliers may help to detect species or genus-specific peptides. We evaluated a limited repertoire of bacterial pathogens in a BAL matrix. Expansion of the database with additional unique peptide markers from other bacteria in combination with improvements in analytical detection sensitivity may provide a means of identifying other potential pathogens in BAL and provide the basis for targeted antimicrobial therapy. Validation of this approach will require analysis of a broad array of clinical BAL samples. Further, prospective comparison with other microbiologic approaches (e.g., culture, RNA/DNA amplifica-
Genoproteomic Detection of Bacterial Pathogens
Table 2. Detected peptides by targeted LC-MS/MS using Thermo Orbitrap Fusion. Dilution
1
1:10
1:100
1:1000
× 105
× 104
× 103
× 102
Sample ID
A6
A5
A4
A3
SALVNEYNVDASR
ⴙ
ⴙ
ⴙ
−
SGTTGNIEAATK
ⴙ
ⴙ
−
−
AGNGTTINPEAVQK
ⴙ
ⴙ
−
−
ATGTNVANFVR
ⴙ
ⴙ
−
−
IVEGEQLAIYK
ⴙ
−
−
−
Sample ID
P6
P5
P4
P3
VNAVGYGESRPVADNATAEGR
ⴙ
ⴙ
ⴙ
−
Species (Injected CFU) A. baumannii (3.6 ± 0.2 × 10x)
P. aeruginosa (2.2 ± 0.6 × 10x)
TIGLDNMQGPVAGK
ⴙ
−
−
−
MIAPVLDEVAR
ⴙ
ⴙ
−
−
QAGAEIVSFVR
ⴙ
ⴙ
−
−
TGELNEAQFDEKa
−
−
−
−
M. catarrhalis (7.6 ± 0.6 × 10x) Sample ID
M6
M5
M4
GITINTSHIEYDTAAR
ⴙ
ⴙ
−
GTFLVDPDGVIK
ⴙ
ⴙ
−
AQYDITQNAGTER
ⴙ
ⴙ
−
ALAVASIVETNSAQAKPIADR
ⴙ
−
−
LNAVGYGFDRPIAPNTTAEGK
ⴙ
−
−
Sample ID
S6
S5
S4
ADFAEDALKa
−
−
−
TDANGNVDVEDIR
ⴙ
−
−
VVEDYAAEVAK
ⴙ
ⴙ
−
LADGQSIDVTAK
ⴙ
ⴙ
−
NVTSDTYDISVR
ⴙ
−
−
Sample ID
K6
K5
K4
IGSEAYNQQLSEK
ⴙ
ⴙ
−
−
QAPFNFALPYNPADIQPNAR
ⴙ
−
−
−
S. maltophilia (3.6 ± 0.4 × 10x)
K. pneumoniae (1.9 ± 0.5 × 10x) K3
WQDGDSVTEEDIR
ⴙ
−
−
−
GPQGQTVTWYQLR
ⴙ
ⴙ
−
−
LATEYAEQYASPEVK
ⴙ
−
−
−
Mixture of A. baumannii and P. aeruginosa Species (Injected CFU) A. baumannii (3.6 ± 0.2 × 10x) Sample ID P. aeruginosa (2.2 ± 0.6 × 10x) Sample ID
× 105
× 105
× 105
× 105
× 104
× 103
× 102
A6
A6
A6
A6
A5
A4
A3
× 105
× 104
× 103
× 102
× 105
× 105
× 105
P6
P5
P4
P3
P6
P6
P6
Continued on page 1406
Clinical Chemistry 63:8 (2017) 1405
Table 2. Detected peptides by targeted LC-MS/MS using Thermo Orbitrap Fusion. (Continued from page 1405) Dilution
1
1:10
1:100
1:1000
A6 + P6
A6 + P5
A6 + P4
A6 + P3
A5 + P6
A4 + P6
SALVNEYNVDASR
ⴙ
ⴙ
ⴙ
ⴙ
ⴙ
ⴙ
−
SGTTGNIEAATK
ⴙ
ⴙ
ⴙ
ⴙ
ⴙ
−
−
Mixture Sample ID
A3 + P6
A. baumannii
AGNGTTINPEAVQK
ⴙ
ⴙ
ⴙ
ⴙ
−
−
−
ATGTNVANFVR
ⴙ
ⴙ
ⴙ
ⴙ
ⴙ
−
−
IVEGEQLAIYK
ⴙ
ⴙ
ⴙ
ⴙ
−
−
−
P. aeruginosa
a
VNAVGYGESRPVADNATAEGR
ⴙ
ⴙ
ⴙ
−
−
−
−
TIGLDNMQGPVAGK
ⴙ
−
−
−
ⴙ
ⴙ
ⴙ
MIAPVLDEVAR
ⴙ
ⴙ
−
−
ⴙ
ⴙ
ⴙ
QAGAEIVSFVR
ⴙ
ⴙ
−
−
ⴙ
ⴙ
ⴙ
TGELNEAQFDEKa
−
−
−
−
−
−
−
Strong interfering peaks were observed preventing positive identification.
tion) to assess true positives and negatives will be needed to determine its clinical value. Clinical BAL samples are complex protein mixtures that are affected by the dilution of the lung lining fluid by the amount of BAL instilled and returned, the presence of inflammatory cells and proteins, and the varying amounts of blood and mucus. A potential limiting factor
in the use of a targeted peptide approach is the large amount of protein from the aforementioned sources and the limited loading capacity of nano-LC-MS columns. Although we were successful in identifying the presence of unique peptide markers for P. aeruginosa in frozen BAL without any preprocessing, ⬍10% of the available digests were injected into the LC-MS because of its limited loading
Fig. 4. LC-MS/MS chromatograms of 2 peptide markers from frozen clinical BAL sample known to be culture-positive for P. aeruginosa. Retention times and transition rank order of the native (top row) and spiked labeled peptide (bottom row) are well matched. The rdotp scores determine the spectra similarity between the analyte precursor and isotope-labeled precursor.
1406
Clinical Chemistry 63:8 (2017)
Genoproteomic Detection of Bacterial Pathogens
capacity. Whether this limitation can be overcome with peptide antibody enrichment or fractionation to enhance peptide detection will require further study (30 –32 ). The rapid trypsin digestion of intact bacterial cells is limited in scope and is not meant to provide a comprehensive proteomic analysis of a potential pathogen. It detected high-abundance peptides that were detectable by our approach. The outer membranes of bacteria may provide equal or greater discriminatory power compared with whole cell lysates (33 ). Trypsin digestion generates data that will be biased against proteins with limited accessibility to digestion (e.g., transmembrane proteins) or proteins with a low number of arginine–lysine residues and peptides with low ionization potential (10 ). Peptides containing N, Q, or M may be subject to deamination and oxidation and may affect sample stability. However, the peptides identified in our study were not substantially affected by amino-acid modification. Given the ability to identify genome-specific peptides with this approach (15 ), rapid, culture-free, clone-level tracking from primary specimens may be possible in the context of nosocomial outbreaks of resistant pathogens by providing rapid microbial identification to the clinician. When linked to the proteomic analysis of additional bacterial strains and expression of plasmid proteins, this approach could potentially decrease the time to pathogen identification and enhance the ability to direct targeted antimicrobial therapy (15, 34 ).
Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article. Authors’ Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest: Employment or Leadership: None declared. Consultant or Advisory Role: None declared. Stock Ownership: None declared. Honoraria: None declared. Research Funding: The Intramural Research Program of the Clinical Center and National Heart, Lung and Blood Institute, National Institutes of Health. Expert Testimony: None declared. Patents: None declared. Other Remuneration: S.K. Drake, WASPaLM Congress 2015. Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript. Acknowledgments: We thank COL Lesho, Dr. Patrick Timothy McGann, and the MRSN surveillance team at WRAIR for providing A. baumannii AC OIFC111 (strain F) and Dr. J Kovacs (Critical Care Medicine, NIH) for providing the clinical frozen BAL sample for this study.
References 1. Endimiani A, Hujer KM, Hujer AM, Kurz S, Jacobs MR, Perlin DS, Bonomo RA. Are we ready for novel detection methods to treat respiratory pathogens in hospitalacquired pneumonia? Clin Infect Dis 2011;52(Suppl 4): S373– 83. 2. Maschmeyer G, Carratala J, Buchheidt D, Hamprecht A, Heussel CP, Kahl C, et al. Diagnosis and antimicrobial therapy of lung infiltrates in febrile neutropenic patients (allogeneic SCT excluded): updated guidelines of the Infectious Diseases Working Party (AGIHO) of the German Society of Hematology and Medical Oncology (DGHO). Ann Oncol 2015;26:21–33. 3. Mikulska M, Del Bono V, Viscoli C. Bacterial infections in hematopoietic stem cell transplantation recipients. Curr Opin Hematol 2014;21:451– 8. 4. Huttner A, Emonet S, Harbarth S, Renzi G, Kaiser L, Schrenzel J. Polymerase-chain reaction/electrospray ionization-mass spectrometry for the detection of bacteria and fungi in bronchoalveolar lavage fluids: a prospective observational study. Clin Microbiol Infect 2014;20:O1059 – 66. 5. Koncan R, Parisato M, Sakarikou C, Stringari G, Fontana C, Favuzzi V, et al. Direct identification of major gramnegative pathogens in respiratory specimens by respiFISH® HAP Gram (-) panel, a beacon-based FISH methodology. Eur J Clin Microbiol Infect Dis 2015;34:2097–102. 6. Mok JH, Eom JS, Jo EJ, Kim MH, Lee K, Kim KU, et al. Clinical utility of rapid pathogen identification using matrix-assisted laser desorption/ionization time-offlight mass spectrometry in ventilated patients with pneumonia: a pilot study. Respirology 2016;21: 321– 8.
7. Ordooei Javan A, Shokouhi S, Sahraei Z. A review on colistin nephrotoxicity. Eur J Clin Pharmacol 2015;71: 801–10. 8. Kadri SS, Hohmann SF, Orav EJ, Bonne SL, Moffa MA, Timpone JG, et al. Tracking colistin-treated patients to monitor the incidence and outcome of carbapenemresistant gram-negative infections. Clin Infect Dis 2015; 60:79 – 87. 9. Averbuch D, Orasch C, Cordonnier C, Livermore DM, Mikulska M, Viscoli C, et al. European guidelines for empirical antibacterial therapy for febrile neutropenic patients in the era of growing resistance: summary of the 2011 4th European Conference on Infections in Leukemia. Haematologica 2013;98:1826 –35. 10. Karlsson R, Gonzales-Siles L, Boulund F, SvenssonStadler L, Skovbjerg S, Karlsson A, et al. Proteotyping: proteomic characterization, classification and identification of microorganisms–a prospectus. Syst Appl Microbiol 2015;38:246 –57. 11. Patel R. MALDI-TOF MS for the diagnosis of infectious diseases. Clin Chem 2015;61:100 –11. 12. Ferreira L, Sanchez-Juanes F, Gonzalez-Avila M, Cembrero-Fucinos D, Herrero-Hernandez A, GonzalezBuitrago JM, Munoz-Bellido JL. Direct identification of urinary tract pathogens from urine samples by matrixassisted laser desorption ionization-time of flight mass spectrometry. J Clin Microbiol 2010;48:2110 –5. 13. Segawa S, Sawai S, Murata S, Nishimura M, Beppu M, Sogawa K, et al. Direct application of MALDI-TOF mass spectrometry to cerebrospinal fluid for rapid pathogen identification in a patient with bacterial meningitis. Clin Chim Acta 2014;435:59 – 61.
14. Kohling HL, Bittner A, Muller KD, Buer J, Becker M, Rubben H, et al. Direct identification of bacteria in urine samples by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry and relevance of defensins as interfering factors. J Med Microbiol 2012; 61:339 – 44. 15. Wang H, Drake SK, Yong C, Gucek M, Tropea M, Rosenberg AZ, et al. A novel peptidomic approach to strain typing of clinical Acinetobacter baumannii isolates using mass spectrometry. Clin Chem 2016;62: 866 –75. 16. Tada K, Kurosawa S, Hiramoto N, Okinaka K, Ueno N, Asakura Y, et al. Stenotrophomonas maltophilia infection in hematopoietic SCT recipients: high mortality due to pulmonary hemorrhage. Bone Marrow Transplant 2013;48:74 –9. 17. Potron A, Poirel L, Nordmann P. Emerging broadspectrum resistance in Pseudomonas aeruginosa and Acinetobacter baumannii: mechanisms and epidemiology. Int J Antimicrob Agents 2015;45:568 – 85. 18. Munoz-Price LS, Poirel L, Bonomo RA, Schwaber MJ, Daikos GL, Cormican M, et al. Clinical epidemiology of the global expansion of Klebsiella pneumoniae carbapenemases. Lancet Infect Dis 2013;13:785–96. 19. National Center for Biotechnology Information. http:// www.ncbi.nlm.nih.gov/genome (Accessed between December 2013 and January 2014). 20. National Center for Biotechnology Information. http:// www.ncbi.nlm.nih.gov/genome (Accessed between December 2014 and January 2015). 21. Mesuere B, Willems T, Van der Jeugt F, Devreese B, Vandamme P, Dawyndt P. Unipept web services for
Clinical Chemistry 63:8 (2017) 1407
22.
23.
24.
25.
26.
metaproteomics analysis. Bioinformatics 2016;32: 1746 – 8. Mesuere B, Devreese B, Debyser G, Aerts M, Vandamme P, Dawyndt P. Unipept: tryptic peptide-based biodiversity analysis of metaproteome samples. J Proteome Res 2012;11:5773– 80. Mesuere B, Debyser G, Aerts M, Devreese B, Vandamme P, Dawyndt P. The Unipept metaproteomics analysis pipeline. Proteomics 2015;15:1437– 42. Nelsestuen GL, Martinez MB, Hertz MI, Savik K, Wendt CH. Proteomic identification of human neutrophil alpha-defensins in chronic lung allograft rejection. Proteomics 2005;5:1705–13. Snitkin ES, Zelazny AM, Montero CI, Stock F, Mijares L, Program NCS, et al. Genome-wide recombination drives diversification of epidemic strains of Acinetobacter baumannii. Proc Natl Acad Sci USA 2011;108: 13758 – 63. Snitkin ES, Zelazny AM, Gupta J, Program NCS, Palmore TN, Murray PR, Segre JA. Genomic insights into the fate of colistin resistance and Acinetobacter bau-
1408
27.
28.
29.
30.
31.
Clinical Chemistry 63:8 (2017)
mannii during patient treatment. Genome Res 2013; 23:1155– 62. National Center for Biotechnology Information. http:// blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastp& PAGE_TYPE=BlastSearch&LINK_LOC=blasthome (Accessed between December 2013 – January 2015. Schmedes SE, King JL, Budowle B. Correcting inconsistencies and errors in bacterial genome metadata using an automated curation tool in Excel (AutoCurE). Front Bioeng Biotechnol 2015;3:138. O’Leary NA, Wright MW, Brister JR, Ciufo S, Haddad D, McVeigh R, et al. Reference sequence (RefSeq) database at NCBI: current status, taxonomic expansion, and functional annotation. Nucleic Acids Res 2016;44: D733– 45. Shi T, Fillmore TL, Sun X, Zhao R, Schepmoes AA, Hossain M, et al. Antibody-free, targeted massspectrometric approach for quantification of proteins at low picogram per milliliter levels in human plasma/ serum. Proc Natl Acad Sci USA 2012;109:15395– 400. Razavi M, Leigh Anderson N, Pope ME, Yip R, Pearson
TW. High precision quantification of human plasma proteins using the automated SISCAPA Immuno-MS workflow. N Biotechnol 2016;33:494 –502. 32. Whiteaker JR, Zhao L, Anderson L, Paulovich AG. An automated and multiplexed method for high throughput peptide immunoaffinity enrichment and multiple reaction monitoring mass spectrometry-based quantification of protein biomarkers. Mol Cell Proteomics 2010;9:184 –96. 33. Jabbour RE, Wade MM, Deshpande SV, Stanford MF, Wick CH, Zulich AW, Snyder AP. Identification of Yersinia pestis and Escherichia coli strains by whole cell and outer membrane protein extracts with mass spectrometry-based proteomics. J Proteome Res 2010; 9:3647–55. 34. Lau AF, Wang H, Weingarten RA, Drake SK, Suffredini AF, Garfield MK, et al. A rapid matrix-assisted laser desorption ionization-time of flight mass spectrometrybased method for single-plasmid tracking in an outbreak of carbapenem-resistant Enterobacteriaceae. J Clin Microbiol 2014;52:2804 –12.
