A Spatial Metabolomics Annotation Workflow Leveraging Cyclic Ion Mobility and Machine Learning-Predicted Collision Cross Sections

In
nontargeted spatial metabolomics, accurate annotation is crucial
for understanding metabolites’ biological roles and spatial
patterns. MS2 mass spectrometry imaging (MSI) coverage
is often incomplete or nonexistent, resulting in many unknown features
that represent an untapped source of biological information. Ion mobility-derived
collision cross sections (CCS) have been leveraged as valuable descriptors
for confirming putative metabolite annotations, distinguishing isomers,
and aiding in unknown structural elucidation. In this study, desorption
electrospray ionization cyclic ion mobility mass spectrometry imaging
(DESI-cIM-MSI) data from human renal cell carcinoma (RCC) tissues
is used as a testbed to explore the extent to which CCS measurements
enhance MSI lipid annotation confidence when combined with machine
learning CCS predictions and SIRIUS analysis of MS2 data.
Multipass IM experiments yielded excellent CCS accuracy (<0.4%)
relative to database values for differential lipids found in RCC tissues,
improving the filtering threshold used in previous CCS-based annotation
workflows. High-accuracy multipass CCS measurements enabled the correct
annotation of isobaric lipid database matches, even in the absence
of MS2 data. Additionally, MS2 data from differential
RCC features were uploaded to SIRIUS, and the predicted CCS values
for SIRIUS candidates were compared to experimental CCS data to filter
out unlikely candidates. Finally, CCS measurements contributed to
the annotation of two spatially correlated unknown features, differential
between tumor and control kidney tissues. Both features were assigned
to rocuronium, a surgical muscle relaxant that had not been previously
reported in MSI studies. Overall, these results underscore the potential
of high-accuracy CCS values to enhance metabolite annotations in MSI-based
spatial metabolomics.