Rapid Multi-Omics for Bacterial Identification Using Flow Injection-Ion Mobility-Mass Spectrometry

Abstract

The implementation of mass spectrometry (MS) in clinical microbiology has made a significant improvement in the turnaround time from positive culture to identification, but current protein-based approaches can struggle with species-level identification because of the high degree of homology within a genus. However, other MS-based strategies for bacterial identification that are based on lipids and small molecules have shown promise toward species-level identification and detection of specific phenotypes, including those related to antibiotic resistance. We are leveraging rapid gas-phase ion mobility (IM) separations coupled to MS to simultaneously detect the lipids and metabolites in bacterial pathogens. Using flow-injection (FI) rather than liquid chromatography (LC), we instead rely more directly on the structural separation of the IM dimension to resolve features from different biochemical classes and aid in identification. A head-to-head comparison demonstrates that the FI-IM-MS multiomic strategy performs similarly to LC-IM-MS in its ability to distinguish 24 strains of the high-concern ESKAPE pathogens, while shortening overall analysis time from 17 to 2 min per injection. We demonstrate that the IM dimension has excellent stability and reproducibility, which enables extracted IM peak areas to be used in lieu of chromatographic peak areas. Furthermore, the same features that are important for the discrimination of bacterial species and strains are found within both the FI-IM-MS and HILIC-IM-MS data sets. These results showcase the capabilities of mobility-enabled rapid multiomics and open the possibility to detect subtle strain-level differences and resistance phenotypes in bacterial pathogens by including additional classes of biomolecules.

1

FI-IM-MS (ESI+) data from multiomic mixture containing strain NR-51529. (A) Ion mobility–mass spectrometry conformational space. From the ion mobility dimension, (B) the total 1D ion mobilogram and extracted ion mobilograms for (C) pyocyanin, (D) nicotinamide adenine dinucleotide, and (E) phosphatidylethanolamine (PE) 34:1 are shown.

2

PCA plots of (A) positive and (B) negative mode FI-IM-MS data from 16 injections of a pooled QC and 96 bacteria extracts. RSDs calculated based on (C) 1481 features in ESI+ and (D) 1394 features in ESI– that were detected in all QC replicates with an intensity ≥50.

3

Correlation plots of feature intensities determined by HILIC-IM-MS versus intensities determined by FI-IM-MS. Data were collected in both positive (A–C) and negative (D–F) ionization modes and scaled by natural log (log_e).

4

Abundance of pyocyanin from FI-IM-MS and HILIC-IM-MS of 30 nM solutions with and without bacterial extract. The neat solution was prepared in 2:2:1 ACN/MeOH/H₂O and analyzed in technical triplicate. The spiked solution was prepared by adding pyocyanin into diluted extracts of strain NR-52183 biological replicates.

5

PCA score plots from FI-IM-MS measurements of 24 bacteria strains collected in (A) ESI+ and (B) ESI– modes. Score plots are based on approximately 1740 features in each mode. Colors correspond to the organism, and shading corresponds to strains. (C) Volcano plot of features detected by FI-IM-MS in positive (square) and negative (circle) mode analyses of Gram-positive and Gram-negative bacterial extracts. Identified compounds are colored based on their biomolecular classes. Student’s t-test (two-way, unpaired) p-values were corrected for multiple comparisons with the Benjamini-Hochberg method.

6

(A) PCA scores plot and (B) PCA loadings plot of four strains run in quadruplicate analyzed by FI-IM-MS in negative ionization mode. (C) Bar chart of average abundances of important identified compounds highlighted in the loadings plot. Bars are colored based on their corresponding strain as seen in (A).