MSACL 2018 US Abstract

Topic: Microbiology/Virology

Metabolic Phenotyping of Pseudomonas aeruginosa Using Rapid Evaporative Ionization Mass Spectrometry: Strain Characterization and Clinical Applications

Emmanuelle Bardin (Presenter)
Imperial College London

Bio: After graduating in general chemistry from one of the French Grandes Ecoles in 2008, I worked for 4 years as a research engineer at the university and in the industry. My research was mainly applied to the agri-food industry and involved method development using GC-MS. In March 2015, I started my PhD in Zoltan Takats’ research group, at Imperial College London. The objective of my project is to develop a quick and non-invasive diagnostic tool for bacterial infection. At the department of Surgery and Cancer, we are working on metabolomic approaches, using ambient MS for the direct analysis of samples such as sputum, skin secretions and breath.

Authorship: Emmanuelle Bardin (1); Alvaro Perdones-Montero (1); Simon Cameron (1); Frances Bolt (1); Eric Alton (1); Andrew Bush (1, 2); Jane Davies (1, 2); Zoltan Takats (1)
(1) Imperial College London, London, United Kingdom; (2) Department of Paediatric Respiratory Medicine, Royal Brompton and Harefield NHS Foundation Trust, London, United Kingdom

Short Abstract

Rapid evaporative ionization mass spectrometry (REIMS) competes, in terms of performance, cost, and ease of use, with other technologies currently available in clinical laboratories such as matrix assisted laser desorption ionisation mass spectrometry (MALDI). Here, we used REIMS to further study bacterial variability, at the sub-species level within Pseudomonas aeruginosa. REIMS allowed the detection of a highly diversified metabolome with variable levels of virulence-related metabolites, such as quorum sensing molecules and rhamnolipids, thereby supporting its role as a monitoring tool for infection exacerbation. We also demonstrated as a proof of concept that REIMS can achieve a strain classification for MLST types of 84% accuracy.

Long Abstract


Pseudomonas aeruginosa is a ubiquitous and opportunistic bacteria amongst the most commonly-isolated nosocomial pathogens in hospitals, and causes fatal infections both in immunocompromised patients and those with cystic fibrosis (CF) [1]. Indeed, up to 60% of CF patients suffer from a chronic lung infection by the time they reach adulthood, making it the main cause of death in CF patients [2]. P. aeruginosa has one of the largest genomes in the bacterial kingdom and 10% is dedicated to adaptive responses. This leads to a very versatile metabolic profile, allows an increased resistance to environmental stress, and makes it particularly adapted to the predisposed CF lung [3]. Resistance- and virulence-associated mechanisms involve the production of quorum sensing communication molecules [4, 5] and bio-surfactant rhamnolipids [6]. Their secretion is controlled by gene expression and has been shown to be linked with infection progression.

Strain typing is important in understanding pathogen metabolism and evolution, epidemiology, and in control of infections. Yet, current methods to detect, identify, and control bacterial infections often rely upon time-consuming phenotypic assays that lack sensitivity; and costly DNA sequencing which is not always available. The introduction in the clinical laboratory of matrix assisted laser desorption ionisation (MALDI) coupled to time-of-flight (ToF) mass spectrometers have allowed a more timely and reliable classification to species level [7], but still requires preparative steps. Rapid evaporative ionisation mass spectrometry (REIMS) is an ambient ionisation technology that has been developed for the in situ analysis of biological samples in real-time [8]. Thermal stress is applied directly to a tissue, resulting in the vaporization and ionisation of metabolites that are then channelled into the mass spectrometer for detection and identification. This technique can be applied directly onto cultured bacteria whose unique lipid profile, especially phospholipids, allows classification to Gram, genus and species level [9]. Here we show the potential of REIMS to provide at once and directly from bacterial culture, species, strain, and metabolic information regarding bacterial virulence, which may assist clinicians in diagnosing and evaluating risk.


A total of 125 P. aeruginosa clinical isolates were collected at the Royal Brompton Hospital (London, UK) and Charing Cross Hospital (London, UK), and originated from CF sputum samples, bronchiectasis patients and non-respiratory hospital infections. The Microflex LT MALDI-ToF instrument (Bruker Daltonics) was used to confirm species identity. Isolates were grown overnight on Colombia blood agar at 37°C prior to REIMS analysis. Bipolar forceps (Erbe Elektromedizin) were used to sample a small amount of biomass directly from the culture plate and an electrical power of 70W was applied on drawing the probes together. The subsequent aerosol was mixed with isopropanol, and leucine enkaphaline as a lock mass compound, and analysed in negative ion mode with a Q-ToF mass spectrometer (XEVO G2-XS, Waters Corporation). A minimum of three measurements were carried out and averaged per sample.

Each CF-related isolate was re-cultured and re-analysed 10 times in order to study the level of reproducibility of REIMS. Data were collected in the form of mass spectra that underwent pre-processing including background subtraction, mass drift correction, total ion count normalization and 0.1 m/z binning using the Offline Model Builder software (OMB, Waters Corporation). Unsupervised statistical analysis was conducted on the Metaboanalyst 3.0 platform [10] and the Random Forest machine learning algorithm [11] was used for classification. Finally, the 45 CF-related isolates were subjected to whole genome sequencing (WGS) using the HiSeq 2500 platform with 2 x 250 bp sequencing. The relevant gene sequences for multi-locus sequence typing (MLST) were extracted and the PubMLST website used for strain match searches [12].


Phospholipids, including phosphatidylglycerols, phosphatidyletanolamines and phosphatidic acids, form the structure of the bacterial membrane and were the compounds with the highest signal intensity. REIMS analysis also allowed the identification of 18 quorum sensing quinolones and 15 rhamnolipids, providing useful information on bacterial virulence. Clustering analysis was performed on the CF isolates which formed subgroups according to their production of quinolones. Those showed both positive and negative correlations with some particular phospholipids, suggesting an impact of quorum sensing expression on membrane composition. Using the ten replicates of each isolate, the Random Forest algorithm achieved the overall classification of isolates with a precision of 81%. A few isolates demonstrated a similar metabolic profile and it was not possible to classify them accurately.

The MLST types were determined using the WGS data and 35 different strains were identified amongst the 45 isolates, revealing that some of the poorly predicted samples belonged to the same type. When grouping the isolates by strain, the Random Forest algorithm achieved the classification of the strains with 84% accuracy. P. aeruginosa collected from respiratory samples tend to show an increased metabolic diversity, which could be attributed to the complex bacterial adaptation to the lungs.

Conclusions & Discussion

Besides its utility for fungi and bacterial speciation, REIMS also represents a promising tool for metabolic characterization and infection evaluation, through the monitoring of virulence-associated molecules and strain typing. Particularly aggressive strains, such as the Liverpool epidemic strain, are currently under study to evaluate REIMS ability to detect high-risk clones. REIMS analysis is also being tested on sputum samples collected from CF patients with and without verified P. aeruginosa infection, as a direct-from-sample diagnostic tool.

References & Acknowledgements:

1. Sadikot, R.T., et al., Pathogen-host interactions in Pseudomonas aeruginosa pneumonia. Am J Respir Crit Care Med, 2005. 171(11): p. 1209-23.

2. Registry", U.C.F., Annual Data Report 2015: UK Cystic Firbosis Trust. 2016.

3. Schuster, M. and E.P. Greenberg, A network of networks: quorum-sensing gene regulation in Pseudomonas aeruginosa. Int J Med Microbiol, 2006. 296(2-3): p. 73-81.

4. Guina, T., et al., Quantitative proteomic analysis indicates increased synthesis of a quinolone by Pseudomonas aeruginosa isolates from cystic fibrosis airways. Proc Natl Acad Sci U S A, 2003. 100(5): p. 2771-6.

5. Dubern, J.-F. and S.P. Diggle, Quorum sensing by 2-alkyl-4-quinolones in Pseudomonas aeruginosa and other bacterial species. Molecular BioSystems, 2008. 4(9): p. 882.

6. Abdel-Mawgoud, A.M., F. Lepine, and E. Deziel, Rhamnolipids: diversity of structures, microbial origins and roles. Appl Microbiol Biotechnol, 2010. 86(5): p. 1323-36.

7. Clark, A.E., et al., Matrix-assisted laser desorption ionization-time of flight mass spectrometry: a fundamental shift in the routine practice of clinical microbiology. Clin Microbiol Rev, 2013. 26(3): p. 547-603.

8. Schafer, K.C., et al., In vivo, in situ tissue analysis using rapid evaporative ionization mass spectrometry. Angew Chem Int Ed Engl, 2009. 48(44): p. 8240-2.

9. Bolt, F., et al., Automated High-Throughput Identification and Characterization of Clinically Important Bacteria and Fungi using Rapid Evaporative Ionization Mass Spectrometry. Anal Chem, 2016. 88: p. 9419−9426.

10. Xia, J., et al., MetaboAnalyst 3.0--making metabolomics more meaningful. Nucleic Acids Res, 2015. 43(Web server): p. W251-7.

11. Breiman, L., Random Forests. 2001.

12.Jolley, K.A.M., M. C. J., BIGSdb: Scalable analysis of bacterial genome variation at the population level. BMC Bioinformatics, 2010. 11(595): p. 1-11.


Project supported by the Cystic Fibrosis Trust, Strategic Research Centre for P. aeruginosa

Financial Disclosure

GrantsyesWaters Corporation
Board Memberno

IP Royalty: no

Planning to mention or discuss specific products or technology of the company(ies) listed above: