https://orcid.org/0000-0002-4039-5639
Figures
Abstract
Introduction
Early and reliable determination of bacterial strain specificity and antibiotic resistance is critical to improve sepsis treatment. Previous research demonstrated the potential of headspace analysis of volatile organic compounds (VOCs) to differentiate between various microorganisms associated with pulmonary infections in vitro. This study evaluates whether VOC analysis can also discriminate antibiotic sensitive from resistant bacterial strains when cultured on varying growth media.
Methods
Both antibiotic-sensitive and -resistant strains of Pseudomonas aeruginosa, Staphylococcus aureus and Klebsiella pneumonia were cultured on 4 different growth media, i.e. Brain Heart Infusion, Marine Broth, Müller-Hinton and Trypticase Soy Agar. After overnight incubation at 37°C, the headspace air of the cultures was collected on stainless steel desorption tubes and analyzed by gas chromatography time-of-flight mass spectrometry (GC-tof-MS). Statistical analysis was performed using regularized multivariate analysis of variance and cross validation.
Results
The three bacterial species could be correctly recognized based on the differential presence of 14 VOCs (p<0.001). This discrimination was not influenced by the different growth media. Interestingly, a clear discrimination could be made between the antibiotic-resistant and -sensitive variant of Pseudomonas aeruginosa (p<0.001) based on their species-specific VOC signature.
Conclusion
This study demonstrates that isolated microorganisms, including antibiotic-sensitive and -resistant strains of Pseudomonas aeruginosa, could be identified based on their excreted VOCs independent of the applied growth media. These findings suggest that the discriminating volatiles are associated with the microorganisms themselves rather than with their growth medium. This study exemplifies the potential of VOC analysis as diagnostic tool in medical microbiology. However, validation of our results in appropriate in vivo models is critical to improve translation of breath analysis to clinical applications.
Citation: Hintzen KFH, Blanchet L, Smolinska A, Boumans M-L, Stobberingh EE, Dallinga JW, et al. (2024) Volatile organic compounds in headspace characterize isolated bacterial strains independent of growth medium or antibiotic sensitivity. PLoS ONE 19(1): e0297086. https://doi.org/10.1371/journal.pone.0297086
Editor: Awatif Abid Al-Judaibi, University of Jeddah, SAUDI ARABIA
Received: June 14, 2023; Accepted: December 23, 2023; Published: January 26, 2024
Copyright: © 2024 Hintzen et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files.
Funding: T. Lubbers (MLDS) Maag Lever Darm Stichting career development grant CDG16-12 supported by Dutch Digestive Foundation (https://www.mlds.nl/) The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Sepsis is a major health problem worldwide and a severe condition that can lead to septic shock, multiple organ failure and even death [1]. It is defined as a life-threatening organ dysfunction caused by a dysregulated host response to infection of predominantly the respiratory tract, urinary tract or abdomen [2, 3]. Elderly patients with comorbidities, including chronic obstructive pulmonary disease (COPD) and coronary heart diseases, are particularly vulnerable to develop pneumonia upon encountering a respiratory tract infection [4, 5]. Inadequate identification of the causative pathogen, and subsequent inadequate treatment, increases the chance to develop bacteremia and sepsis. The most common causes of pneumonia are respiratory microorganisms including the bacteria Pseudomonas aeruginosa, Staphylococcus aureus and Klebsiella pneumonia [4–6]. For all three forms of pneumonia, being community-acquired pneumonia (CAP), nosocomial pneumonia (NP) or ventilator-associated pneumonia (VAP), bacteremia is associated with prolonged admission at the intensive care unit (ICU) and increased mortality [5]. In the majority of septic patients, the exact causative pathogen is unknown, leading to treatment with broad-spectrum antibiotics which is too unspecific. Both inadequate and excessive antibiotic treatment with broad-spectrum antibiotics are associated with a higher mortality [2, 7]. Additionally, antimicrobial resistance (AMR) is a growing burden in health care [8], warranting the possible misuse of antibiotics to be kept to a minimum. Therefore, early and reliable determination of strain specificity and antibiotic resistance is critical to improve early adequate treatment and is essential to improve patient outcome.
Bacteria in the human body, both commensal as well as pathogenic bacteria, produce volatile organic compounds (VOCs) as part of their species-specific metabolism [9–12]. Both single VOCs [13, 14] and VOC profiles [9, 15, 16] can be used to distinguish different bacteria. VOCs are mostly carbon based molecules that are excreted in, amongst others, feces, urine and exhaled breath [17], thereby offering an opportunity to trace bacterial infections in the human body. The major advantage of exhaled air analysis is that sampling is non-invasive and has a low impact on patients. Exhaled breath can be collected frequently and analyzing the VOC content has a proven diagnostic power related to various respiratory diseases including COPD, asthma and interstitial lung diseases [18–21]. Measuring alterations in exhaled VOCs could therefore be a promising alternative approach to improve the diagnosis of respiratory infections [22]. Studies from our lab as well as other groups have already shown that patients with VAP can be distinguished from non-VAP patients on the ICU based on a distinct VOC profile that could be related to the causative pathogen [23–25].
Consequently, the aim of the current in vitro study is to investigate whether analysis of VOCs can be used to discriminate antibiotic-sensitive and–resistant respiratory pathogens cultured on varying growth media. The primary hypothesis is that these bacteria induce unique VOC profiles independent of their growth media. The second hypothesis is that the cellular mechanisms of antibiotic resistant bacteria result in an altered VOC profile compared to that of their antibiotic sensitive subtype.
Material and methods
Materials and samples
Three different bacterial strains were selected: Klebsiella pneumonia (K. pneumonia, ATCC 700683), Pseudomonas aeruginosa (P. aeruginosa, ATCC 27853 and a clinical isolate) and Staphylococcus aureus (S. aureus, clinical isolate). For all strains, both an antibiotic-sensitive and an antibiotic-resistant strain were included: an extended spectrum β-lactamase (ESBL)-resistant and an ESBL-sensitive for both P. aeruginosa and K. pneumonia and a methicillin-resistant (MRSA, clinical isolate) and a methicillin-sensitive (MSSA, clinical isolate) strain for S. aureus. Antibiotic resistance was tested phenotypically by determining the minimum inhibitory concentration (MIC) of the specific antibiotics, e.g. methicillin and oxacillin for MRSA and MSSA and ESBL for the resistant and sensitive strains of both K. pneumonia and P. aeruginosa. For MRSA and MSSA, the genotype was additionally tested by analyzing the presence of mec A (the gene responsible for methicillin resistance) as well as by performing multi locus sequence typing (MLST) and staphylococcus aureus protein A (Spa typing) to determine whether the two substrains were related. For the two strains of K. pneumonia, the genotype was additionally tested by matching for MLST and RAMAN clustering.
All selected substrains were cultured in 4 different growth media, i.e. Brain Heart Infusion (BHI), Marine Broth (MB), Müller-Hinton (MH) and Trypticase Soy Agar (TSA) with 24 repeats per strain per medium. For P. aeruginosa, 12 repeats of both the ATCC and clinical isolate were performed. At the time of VOC sampling, the average CFU/ml was 109/ml for each genus as determined by serial dilutions of the growth media. No additional media controls were added to the experimental procedure based on our previous study that already revealed methodological standardization as the medium control group was well separated from the different bacteria [9].
Sample collection and analysis
All strains were cultured on Columbia III Agar plates with 5% sheep blood (Becton-Dickinson) and incubated overnight at 37°C. Afterwards, the bacteria were transferred into 4.5 mL sterile Brain Heart Infusion broth (CM0225 Oxoid) and grown for another 4 hours at 37°C under constant agitation. Next, 0.5 mL of the culture was transferred into 100 mL of one of the 4 selected sterile media (BHI CM0225 Oxoid; MH CM0405 Oxoid; TSA CM0131 Oxoid and MB dd15 Difco) in 1L culture flasks.
After overnight incubation at 37°C, overhead air (headspace) was sampled during 15 minutes for VOC analysis as described previously [9]. Briefly, by flushing the flasks with high-grade nitrogen the VOCs present in the headspace of the cultures were transported onto stainless two-bed sorption tubes, packed with carbograph 1TC/Carbopack X (Markes International, Llantrisant, Wales, UK), under standardized conditions.
The VOC content of all tubes was determined using gas chromatography time-of-flight mass spectrometry (GC-tof-MS) as previously published [9, 26]. In short, the desorption tubes were placed inside a thermal desorption unit (Marks Unity desorption unit, Markes International) and heated to 270°C in order to release all trapped VOCs onto the gas chromatography capillary column (RTX-5ms, 30 m x -.25 mm 5% diphenyl, 95% dimethylsiloxane capillary, film thickness 1.0 μm) followed by separation by the GC (ThermoFisher Scientific, Austin, Texas, USA) and detection by a tof-MS (Thermo Electron Tempus Plus time-of-flight mass spectrometer, ThermoFisher Scientific, Austin, Texas, USA). Settings for this separation and detection were similar to those previously published [9, 26].
Data pre-processing
Before the statistical analysis, the raw GC-tof-MS data were preprocessed [27]. Uninformative parts of each chromatogram, (retention time either <1.3 or >23 minutes) were removed. The data quality was improved by denoising (based on wavelets decomposition [28]) and baseline correction. P-splines [29] were used to estimate the shape of the baseline and subtract it from the denoised chromatogram.
A list of chromatographic peaks was established for each chromatogram. These peaks needed to be matched from sample to sample. To this end, the Pearson correlation between the mass spectra, associated to the different chromatographic peaks, were calculated. A high correlation (superior to 85%) indicated that two peaks correspond to the same chemical compound. This information allowed constructing a data table where each line represents a sample and each column a chemical compound. Each value in this table corresponded to the total ion count measured for a specific VOC in a given sample. Finally, to improve the comparison between VOC profiles, the probabilistic quotient normalization was applied [30] to suppress any effect related to different (overall) dilution of the samples.
Multivariate data analysis
The data pre-processing resulted in a data table, reporting semi-quantitatively the concentration of 284 VOCs in 555 samples. Table 1 provides the complete description of all samples allocated to the three main factors influencing their VOC profiles: the four media used to culture the bacteria, the three types of bacteria and the resistance or sensitivity to antibiotic treatments. Some samples were lost due to incorrect sampling or analysis in those groups containing less than 24 samples.
The influence of these three factors was statistically evaluated using Multivariate Analysis of Variance (MANOVA). This analysis was performed using the implementation previously described [31], allowing to more efficiently deal with high number of variables (i.e. high number of detected VOCs). Significance was evaluated using Wilks’ Lambda statistic. As the distribution of Wilks’ Lambda under the null hypothesis was unknown here, a (sequential) permutation test was used to estimate this distribution from the data [32, 33].
Results
In total 555 headspace samples were analyzed by GC-tof-MS. Data pre-processing of all chromatograms resulted in a data-matrix containing 284 VOCs.
Median chromatograms for all strains on each specific medium demonstrate clear differences in the VOC pattern excreted in the headspace by the antibiotic-resistant and–sensitive strains of the 3 selected bacteria cultured on the 4 different media. Fig 1 shows the median chromatograms of the antibiotic-sensitive strains on BHI as an example. All three microorganisms excreted a divergent peak pattern clearly indicating the different volatiles present in the headspace air of S. aureus, P. aeruginosa and K. pneumonia cultures.
Discriminating antibiotic-sensitive strains
MANOVA analysis using the obtained data matrix discriminated the 3 antibiotic-sensitive strains of S. aureus, P. aeruginosa and K. pneumonia, regardless of the used growth medium, based on the selection of 14 significantly different VOCs (p<0.001, Table 2). This separation is visualized in a score plot obtained using the canonical variables (linear combinations of the selected VOCs) of the MANOVA model (Fig 2A). The spread between the data points of the headspace samples of the bacteria indicates there is a reasonable variation within the excreted VOC profiles of each strain (Fig 2A). Most likely, this variation is introduced by the different growth media. However, this does not affect the distinction between the selected strains as can be observed in Fig 2B where recoloring the plot based on growth media did not result in clustering based on media type used.
A. Separation of the antibiotic-sensitive strains of S. aureus, P. aeruginosa and K. pneumonia based on 14 VOCs selected using MANOVA (p<0.001) and visualized with a score plot obtained using the canonical variables of this MANOVA model. B. Same visualization of all samples in the separation model but now recolored based on the applied media. *a.u. arbitrary unit.
The reported change in abundance is with respect to the remaining groups.
The relationship of the 14 identified compounds towards the various microorganisms, and thus their contribution to an adequate identification of a sample, is shown in Table 2. Unlike all other compounds, only methanesulphonylchloride does not display a unique profile for only one of the strains. Instead, the abundance of methanesulphonylchloride is lower in the headspace samples of both K. pneumonia and P. aeruginosa, indicating that both bacteria excrete this volatile to a lesser extent or absorb it to a higher extent. All other significantly different VOCs show an abundance pattern unique for only one of the selected microorganisms, thereby strengthening the separation model visualized in Fig 2A and confirming the lack of influence of the various media depicted in Fig 2B.
To examine the descriptive nature of the selected VOCs with respect to the bacterial species in more detail, the influence of the significantly altered volatiles (as listed in Table 2) on the separation model is visualized in Fig 3. Within this plot, the lines indicate the changes in relative concentrations of the corresponding VOCs whereas the angle between compounds mirrors their correlation, i.e. a small angle means the compounds are positively correlated. For instance, the very small angle between ethanol and acetophenone indicates that the presence of these two volatiles will be influenced in the same way and that they will display similar trends for the three separated groups. Likewise, the very large angle between for instance decanal and pyrrole implicates that their concentrations will be affected in opposite ways in the samples of the different groups as can indeed also be deducted from Table 2. Additionally, the abundance of a compound is elevated if the sign of a new coordinate (i.e. discriminant vector) for samples and compound is the same (i.e. either both positive or both negative). As can be deducted from Fig 3, increased levels of especially decanal and 2-methylbutanal are indicative of the presence of S. aureus whereas elevated levels of mainly ethanol and acetophenone are associated with K. pneumonia. P. aeruginosa is defined by higher levels of particularly dimethylsulfide and isobutanol.
*a.u. arbitrary unit.
Discriminating antibiotic-sensitive from -resistant strains
A significant discrimination between the antibiotic-sensitive and–resistant strains of P. aeruginosa is possible based on the complete VOC signature, consisting of 284 VOCs, of the headspace air samples of their cultures (Table 3). However, a distinction between the antibiotic-sensitive and–resistant strain of S. aureus and K. pneumonia could not be demonstrated. Although the different types of media resulted in significantly different VOCs excreted in the headspace air (p<0.001) of all three bacteria, no interaction of media was found for the discrimination their antibiotic sensitivity. This indicating rather subtle differences in VOC profiles of antibiotic-sensitive and–resistant strains. However, for P-aeruginosa this interaction of the type of media does not influence the ability to discriminate between antibiotic–sensitive and -resistant strains, suggesting more obvious differences independent of the type of media used.
The separation model for the antibiotic-sensitive and–resistant strains of P. aeruginosa is visualized in a score plot using the canonical variables of the MANOVA model (Fig 4). Especially the antibiotic-sensitive strain displays a lot of variation in overall VOC signature excreted in the headspace air.
Separation of the antibiotic-sensitive and–resistant strains of P. aeruginosa based on the overall VOC signature using MANOVA (p<0.001) and visualized with a score plot obtained using the canonical variables of this MANOVA model. *a.u. arbitrary unit.
Discussion
The results of this study demonstrate the potential of VOC analysis to identify the main causative bacteria for pneumonia in vitro. Indeed, all three selected microorganisms can be discriminated based on the relative abundance of unique VOCs in their headspace. Another important finding is that this discriminatory potential is not affected by the use of different growth media. Interestingly, VOC analysis is able to discriminate the antibiotic-resistant variant of P. aeruginosa from its antibiotic-sensitive variant. However, this distinction could not be observed for S. aureus and K. pneumonia.
Metabolic processes of P. aeruginosa, S. aureus and K. pneumonia result in the excretion of certain VOCs. Due to many different possible confounders on the production of VOCs it is possible that the VOCs identified in this study are less clear in other studies or that other VOCs are identified. VOCs in headspace air can be subject to different growth conditions such as the type of media used, the growth fase of the bacteria when collecting the headspace air. The current study indicates that 2-methylbutanal, a known aldehyde, to be produced by S.aureus specifically. This has been previously described in literature where both the isomers 2-methylbutanal and 3-methylbutanal are considered as distinctive compound for S.aureus [10, 12, 16, 34]. Additionally, the aldehyde decanal is also increased for S.aureus specifically. This specific increase of aldehydes can be a result of amino acid degradation by the bacteria [16].
Acetone, which was increased for S.aureus and decreased for P.aeruginosa in the current study, also appears to be increased in vivo in mice infected with S.aureus [35]. However, observed increases or decreases in abundance of compounds for individual bacteria are not always consistent and vary when conditions differ. For instance, in the current study, dimethyldisulfide is increased for P. aeruginosa and decreased for S. aureus, whereas this increase was only the case for S. aureus in our previous study [9]. A possible explanation for this apparent discrepancy is the dynamic process of bacterial growth resulting in changes in the release of different compounds over time. A study, collecting headspace cultures on different time points, shows the presence of dimethyldisulfide earliest after 24h of bacterial growth for P. aeruginosa [16]. However, the growth time was similar in both studies as they cultured the bacteria overnight. An important different is that in the previous study only one type of growth media was used, while the current study used multiple. This could indicate that the type of growth media influences the VOCs produced. Despite the reported differences regarding its presence in the headspace air of P. aeruginosa, several studies have confirmed dimethyldisulfide can be considered a distinctive compound for this microorganism that potentially can be used in combination with other volatiles to identify its presence.
Besides endogenous pathways involved in the formation and excretion of microbial VOCs, exogenous factors such as growth media might also play a role in differences in VOCs produced by micro-organisms. It has been shown that nutrients present in the media can influence the growth pattern of the pathogens or lead to differences in VOCs produced [11, 34]. Some bacteria are able to grow in almost all kinds of growth media whereas others require very specific growth media. Different nutrients present may create a less or more optimal growth environment for the bacteria, which can subsequently be seen in a different growth curve. The nutrients in the growth media are metabolized by the bacteria resulting in different metabolic end-products and thus different volatiles excreted or absorbed. This indicates that compounds identified in studies using only one growth medium might not always be identified in other studies using different growth media. The current study, using 4 different growth media simultaneously, aimed to identify discriminative compounds that are independent of the growth media. For example, 3-hydroxy-2-butanone was identified previously in cultures of S. aureus using Trypticase Soy Broth but not in cultures using BHI or Nutrient Broth (NB) [11]. The fact that this compound was also not found in the current study further underlines the suggestion its presence depends on the growth medium used. Interestingly, dimethyldisulfide is previously identified for both P.aeruginosa and S.aureus in multiple growth media [16, 34] which also corresponds to the findings in the current study. Although the different types of growth media inevitably introduced variation within the excreted VOC profiles of the bacteria, the discriminatory value of the excreted VOCs in this study is still sufficient to enable discrimination between the respective strains. Moreover, the fact that strain-specific VOCs were identified despite the use of different growth media implicates that the discriminative volatiles are endogenously produced or absorbed and thus directly related to the micro-organism itself.
Both an antibiotic-sensitive and an antibiotic-resistant strain of the selected microorganisms were included in the current study. However, only for P. aeruginosa a separation could be made between the antibiotic-sensitive and–resistant type using all information available. Although our previous work did show a significant difference between antibiotic-resistant and–sensitive S. aureus grown on only Mueller-Hinton Broth [9], the current study was not able to repeat these results. The loss of samples due to incorrect sampling or analysis in especially the S. aureus group (n = 11 for antibiotic-resistant and n = 8 for the -sensitive subgroup), could possibly have lowered the statistical power for this analysis leading to a lack of statistical significance. Additionally, the use of multiple types of media created a larger overall variation within the groups which could have masked the differences associated with antibiotic resistance, indicating that those metabolic changes are subtle and thus more difficult to detect in vitro. An alternative approach to detect metabolic differences between antibiotic-sensitive and–resistant microorganisms might be to challenge their cultures with different classes of antibiotics as the genetic changes underlying antibiotic sensitivity might induce a more pronounced difference in response and thus in VOC excretion and absorption. Resistance towards methicillin, such as in S. aureus, is caused by the incorporation of a mec gene into the chromosome. This gene encodes for low affinity penicillin-binding proteins whose structure causes limited binding capacity of methicillin resulting in resistance [36]. The resistant strains of P. aeruginosa and K. pneumonia are able to produce extended spectrum β-lactamase (ESBL) that is able to hydrolyze third-generation cephalosporins by altering the amino acid configuration around the active side of these β-lactamase resulting in resistance [37].
The bacteria applied in the present study differ in their structure as both K. Pneumonia and P. aeruginosa are so-called gram negative bacteria whereas S. aureus is a gram-positive microorganism. The cell envelope of gram-negative bacteria consists of a thin peptidoglycan wall that is surrounded by an outer lipid membrane containing lipopolysaccharide which is important in evading phagocytosis and cell lysis. Gram-positive bacteria have a thick peptidoglycan layer containing teichoic acids and no outer lipid layer. These teichoic acids of gram-positive bacteria may play a role in cell growth, preventing wall breakdown and possible cell lysis and provide much of the wall’s antigenic specificity. As the cell membrane is involved in evading the actions of the immune system of a host and provides a barrier for certain antibiotics, differences in its structure might result in differences in response to environmental changes or to antibiotics [38]. The known differences between gram positive and gram negative bacteria most likely will affect which volatiles these microorganisms can absorb or excrete from specific growth media. However, the current study only studies the metabolic differences of bacterial growth in fixed culture media. These may be different in more complex in vivo situations in which multiple factors can influence the metabolic process which may result in more pronounced differences or the excretion of yet different VOCs. A study comparing both gram-negative and gram-positive bacteria, using metabolites in serum samples of infected mice, shows both a distinction between specific bacteria as well as overlapping characteristics between the different groups. For gram-positive bacteria metabolites such as, glucose, pyruvate, citrate, fumarate, and 2-oxoglutarate were identified, whereas for gram-negative bacteria higher levels of amino acids were found [35, 39].
Previous studies have already shown that VOCs produced in the headspace of in vitro pathogen cultures do not always completely correspond to the VOCs found in the exhaled breath of a diseased host [40]. Compared to standardized in vitro studies, living organisms have a more complex metabolism involving multiple factors such as diet and exercise [41] that confounds the possible interaction between host and pathogen. Although in vitro models cannot be extrapolated to in vivo situations, distinguishing various strains independent of their growth media is a step forward towards the clinical applicability of VOC analysis for quick identification of pathogens. Combining the results of multiple studies may provide insights in potentially overlapping, and thus more robust, VOCs. Additionally, dynamic in vitro sampling procedures can contribute to understanding metabolic changes during bacterial growth and can help identify bacteria-specific compounds independent of growth phases. Additionally, a next step could be the detection of pathogens in more complex models of metabolism, such as standardized animal studies or in clinical studies. As VOC analysis can be done on both in vitro (growth media or patient derived material) or in vivo (exhaled breath) samples, it could be a valuable tool for rapid and non-invasive detection of pathogens. Sampling devices for non-invasive breath collection in in vivo models facilitate in obtaining fundamental insight in the influence of metabolic processed on alterations in VOC composition and contribute to the translation of in vitro to in vivo results [42]. Such studies can contribute to an initial selection of important VOCs for the detection of specific pathogens in vivo allowing for more targeted analysis. Early diagnosis and identification of causative pathogens contributes to early administration of correct and specific antibiotics that would largely impact treatment, thereby increasing the recovery and survival of patients and decreasing their hospital stay and length of sickness.
References
- 1. Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, et al. Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the Global Burden of Disease Study. Lancet. 2020;395(10219):200–11. pmid:31954465
- 2. Rhee C, Kadri SS, Dekker JP, Danner RL, Chen HC, Fram D, et al. Prevalence of Antibiotic-Resistant Pathogens in Culture-Proven Sepsis and Outcomes Associated With Inadequate and Broad-Spectrum Empiric Antibiotic Use. JAMA Netw Open. 2020;3(4):e202899. pmid:32297949
- 3. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315(8):801–10. pmid:26903338
- 4. Ewig S, Torres A. Severe community-acquired pneumonia. Clin Chest Med. 1999;20(3):575–87. pmid:10516905
- 5. Koulenti D, Tsigou E, Rello J. Nosocomial pneumonia in 27 ICUs in Europe: perspectives from the EU-VAP/CAP study. Eur J Clin Microbiol Infect Dis. 2016.
- 6. Parkins MD, Floto RA. Emerging bacterial pathogens and changing concepts of bacterial pathogenesis in cystic fibrosis. J Cyst Fibros. 2015;14(3):293–304. pmid:25881770
- 7. Retamar P, Portillo MM, López-Prieto MD, Rodríguez-López F, de Cueto M, García MV, et al. Impact of inadequate empirical therapy on the mortality of patients with bloodstream infections: a propensity score-based analysis. Antimicrob Agents Chemother. 2012;56(1):472–8. pmid:22005999
- 8. Cassini A, Högberg LD, Plachouras D, Quattrocchi A, Hoxha A, Simonsen GS, et al. Attributable deaths and disability-adjusted life-years caused by infections with antibiotic-resistant bacteria in the EU and the European Economic Area in 2015: a population-level modelling analysis. Lancet Infect Dis. 2019;19(1):56–66. pmid:30409683
- 9. Boots AW, Smolinska A, van Berkel JJ, Fijten RR, Stobberingh EE, Boumans ML, et al. Identification of microorganisms based on headspace analysis of volatile organic compounds by gas chromatography-mass spectrometry. Journal of breath research. 2014;8(2):027106. pmid:24737039
- 10. Bos LD, Sterk PJ, Schultz MJ. Volatile metabolites of pathogens: a systematic review. PLoS Pathog. 2013;9(5):e1003311. pmid:23675295
- 11. Chen J, Tang J, Shi H, Tang C, Zhang R. Characteristics of volatile organic compounds produced from five pathogenic bacteria by headspace-solid phase micro-extraction/gas chromatography-mass spectrometry. J Basic Microbiol. 2017;57(3):228–37. pmid:27874211
- 12. Thorn RM, Reynolds DM, Greenman J. Multivariate analysis of bacterial volatile compound profiles for discrimination between selected species and strains in vitro. J Microbiol Methods. 2011;84(2):258–64. pmid:21167880
- 13. Allardyce RA, Hill AL, Murdoch DR. The rapid evaluation of bacterial growth and antibiotic susceptibility in blood cultures by selected ion flow tube mass spectrometry. Diagn Microbiol Infect Dis. 2006;55(4):255–61. pmid:16678377
- 14. Thorn RM, Greenman J. Microbial volatile compounds in health and disease conditions. J Breath Res. 2012;6(2):024001. pmid:22556190
- 15. Filipiak W, Sponring A, Baur MM, Ager C, Filipiak A, Wiesenhofer H, et al. Characterization of volatile metabolites taken up by or released from Streptococcus pneumoniae and Haemophilus influenzae by using GC-MS. Microbiology. 2012;158(Pt 12):3044–53. pmid:23059976
- 16. Filipiak W, Sponring A, Baur MM, Filipiak A, Ager C, Wiesenhofer H, et al. Molecular analysis of volatile metabolites released specifically by Staphylococcus aureus and Pseudomonas aeruginosa. BMC Microbiol. 2012;12:113. pmid:22716902
- 17. Amann A, Costello Bde L, Miekisch W, Schubert J, Buszewski B, Pleil J, et al. The human volatilome: volatile organic compounds (VOCs) in exhaled breath, skin emanations, urine, feces and saliva. Journal of breath research. 2014;8(3):034001. pmid:24946087
- 18. Dallinga JW, Smolinska A, van Schooten FJ. Analysis of volatile organic compounds in exhaled breath by gas chromatography-mass spectrometry combined with chemometric analysis. Methods Mol Biol. 2014;1198:251–63. pmid:25270934
- 19. Robroeks CM, van Berkel JJ, Jöbsis Q, van Schooten FJ, Dallinga JW, Wouters EF, et al. Exhaled volatile organic compounds predict exacerbations of childhood asthma in a 1-year prospective study. Eur Respir J. 2013;42(1):98–106. pmid:23645402
- 20. Van Berkel JJ, Dallinga JW, Moller GM, Godschalk RW, Moonen EJ, Wouters EF, et al. A profile of volatile organic compounds in breath discriminates COPD patients from controls. Respir Med. 2010;104(4):557–63. pmid:19906520
- 21. Plantier L, Smolinska A, Fijten R, Flamant M, Dallinga J, Mercadier JJ, et al. The use of exhaled air analysis in discriminating interstitial lung diseases: a pilot study. Respir Res. 2022;23(1):12. pmid:35057817
- 22. Ahmed WM, Lawal O, Nijsen TM, Goodacre R, Fowler SJ. Exhaled Volatile Organic Compounds of Infection: A Systematic Review. ACS infectious diseases. 2017;3(10):695–710. pmid:28870074
- 23. Filipiak W, Beer R, Sponring A, Filipiak A, Ager C, Schiefecker A, et al. Breath analysis for in vivo detection of pathogens related to ventilator-associated pneumonia in intensive care patients: a prospective pilot study. Journal of breath research. 2015;9(1):016004. pmid:25557917
- 24. Fowler SJ, Basanta-Sanchez M, Xu Y, Goodacre R, Dark PM. Surveillance for lower airway pathogens in mechanically ventilated patients by metabolomic analysis of exhaled breath: a case-control study. Thorax. 2015;70(4):320–5. pmid:25661115
- 25. Schnabel R, Fijten R, Smolinska A, Dallinga J, Boumans ML, Stobberingh E, et al. Analysis of volatile organic compounds in exhaled breath to diagnose ventilator-associated pneumonia. Sci Rep. 2015;5:17179. pmid:26608483
- 26. Van Berkel JJ, Dallinga JW, Möller GM, Godschalk RW, Moonen E, Wouters EF, et al. Development of accurate classification method based on the analysis of volatile organic compounds from human exhaled air. J Chromatogr B Analyt Technol Biomed Life Sci. 2008;861(1):101–7. pmid:18055279
- 27. Smolinska A, Hauschild AC, Fijten RR, Dallinga JW, Baumbach J, van Schooten FJ. Current breathomics—a review on data pre-processing techniques and machine learning in metabolomics breath analysis. Journal of breath research. 2014;8(2):027105. pmid:24713999
- 28.
Walczak B. Wavelets in Chemistry. Amsterdam: Elsevier Science; 2000.
- 29. Eilers PHC, Marx BD. Flexible smoothing with B-splines and penalties. Statistical Science. 1996;11(2):89–102.
- 30. Aberg KM, Torgrip RJ, Kolmert J, Schuppe-Koistinen I, Lindberg J. Feature detection and alignment of hyphenated chromatographic-mass spectrometric data. Extraction of pure ion chromatograms using Kalman tracking. J Chromatogr A. 2008;1192(1):139–46. pmid:18378252
- 31. Engel J, Blanchet L, Bloemen B, van den Heuvel LP, Engelke UH, Wevers RA, et al. Regularized MANOVA (rMANOVA) in untargeted metabolomics. Anal Chim Acta. 2015;899:1–12. pmid:26547490
- 32. Besag J, Clifford P. Sequential Monto-Carlo p-values. Biometrika. 1991;78(2):301–4.
- 33. Silva IS, Assuncao R, Costa M. Power of the sequential monte carlo test. Sequential Analysis. 2009;28(2):163–74.
- 34. Lawal O, Muhamadali H, Ahmed WM, White IR, Nijsen TME, Goodacre R, et al. Headspace volatile organic compounds from bacteria implicated in ventilator-associated pneumonia analysed by TD-GC/MS. Journal of breath research. 2018;12(2):026002. pmid:28947683
- 35. Hoerr V, Zbytnuik L, Leger C, Tam PP, Kubes P, Vogel HJ. Gram-negative and Gram-positive bacterial infections give rise to a different metabolic response in a mouse model. J Proteome Res. 2012;11(6):3231–45. pmid:22483232
- 36. Lakhundi S, Zhang K. Methicillin-Resistant Staphylococcus aureus: Molecular Characterization, Evolution, and Epidemiology. Clin Microbiol Rev. 2018;31(4). pmid:30209034
- 37. Castanheira M, Simner PJ, Bradford PA. Extended-spectrum β-lactamases: an update on their characteristics, epidemiology and detection. JAC Antimicrob Resist. 2021;3(3):dlab092.
- 38. Silhavy TJ, Kahne D, Walker S. The bacterial cell envelope. Cold Spring Harb Perspect Biol. 2010;2(5):a000414. pmid:20452953
- 39. Hessle CC, Andersson B, Wold AE. Gram-positive and Gram-negative bacteria elicit different patterns of pro-inflammatory cytokines in human monocytes. Cytokine. 2005;30(6):311–8. pmid:15935951
- 40. Zhu J, Bean HD, Wargo MJ, Leclair LW, Hill JE. Detecting bacterial lung infections: in vivo evaluation of in vitro volatile fingerprints. Journal of breath research. 2013;7(1):016003. pmid:23307645
- 41. Blanchet L, Smolinska A, Baranska A, Tigchelaar E, Swertz M, Zhernakova A, et al. Factors that influence the volatile organic compound content in human breath. Journal of breath research. 2017;11(1):016013. pmid:28140379
- 42. Hintzen KFH, Smolinska A, Mommers AGR, Bouvy ND, van Schooten FJ, Lubbers T. Non-invasive breath collection in murine models using a newly developed sampling device(). Journal of breath research. 2022;16(2). pmid:35086080