Strain-Level Discrimination of Bacteria by Liquid Chromatography and Paper Spray Ion Mobility Mass Spectrometry

Determining bacterialidentity at the strain level is criticalfor public health to enable proper medical treatments and reduce antibioticresistance. Herein, we used liquid chromatography, ion mobility, andtandem MS (LC-IM-MS/MS) to distinguish Escherichiacoli (E. coli)strains. Numerical multivariate statistics (principal component analysis,followed by linear discriminant analysis) showed the capability ofthis method to perform strain-level discrimination with predictionrates of 96.1% and 100% utilizing the negative and positive ion information,respectively. The tandem MS and LC separation proved effective indiscriminating diagnostic lipid isomers in the negative mode, whileIM separation was more effective in resolving lipid conformationalbiomarkers in the positive ion mode. Because of the clinical importanceof early detection for rapid medical intervention, a faster technique,paper spray (PS)-IM-MS/MS, was used to discriminate the E. coli strains. The achieved prediction ratesof the analysis of E. coli strainsby PS-IM-MS/MS were 62.5% and 73.5% in the negative and positive ionmodes, respectively. The strategy of numerical data fusion of negativeand positive ion data increased the classification rates of PS-IM-MS/MSto 80.5%. Lipid isomers and conformers were detected, which servedas strain-indicating biomarkers. The two complementary multidimensionaltechniques revealed biochemical differences between the E. coli strains confirming the results obtainedfrom comparative genomic analysis. Moreover, the results suggest thatPS-IM-MS/MS is a rapid, highly selective, and sensitive method fordiscriminating bacterial strains in environmental and food samples.

Automated summary

Determining bacterial identity at the strain level using LC-IM-MS/MS is effective in distinguishing E. coli strains, with prediction rates of 96.1% and 100% for negative and positive ion information, respectively. The PS-IM-MS/MS technique, a faster method, achieved prediction rates of 62.5% and 73.5% in negative and positive ion modes, respectively. The fusion of negative and positive ion data increased the classification rates to 80.5%.