Molecular Networking and On-Tissue Chemical Derivatization for Enhanced Identification and Visualization of Steroid Glycosides by MALDI Mass Spectrometry Imaging

Spatial metabolomics describes the spatially resolved analysis of interconnected pathways, biochemical reactions, and transport processes of small molecules in the spatial context of tissues and cells. However, a broad range of metabolite classes (e.g., steroids) show low intrinsic ionization efficiencies in mass spectrometry imaging (MSI) experiments, thus restricting the spatial characterization of metabolic networks. Additionally, decomposing complex metabolite networks into chemical compound classes and molecular annotations remains a major bottleneck due to the absence of repository-scaled databases. Here, we describe a multimodal mass-spectrometry-based method combining computational metabolome mining tools and high-resolution on-tissue chemical derivatization (OTCD) MSI for the spatially resolved analysis of metabolic networks at the low micrometer scale. Applied to plant toxin sequestration in Danaus plexippus as a model system, we first utilized liquid chromatography (LC)–MS-based molecular networking in combination with artificial intelligence (AI)-driven chemical characterization to facilitate the structural elucidation and molecular identification of 32 different steroidal glycosides for the host-plant Asclepias curassavica. These comprehensive metabolite annotations guided the subsequent matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI MSI) analysis of cardiac-glycoside sequestration in D. plexippus. We developed a spatial-context-preserving OTCD protocol, which improved cardiac glycoside ion yields by at least 1 order of magnitude compared to results with untreated samples. To illustrate the potential of this method, we visualized previously inaccessible (sub)cellular distributions (2 and 5 μm pixel size) of steroidal glycosides in D. plexippus, thereby providing a novel insight into the sequestration of toxic metabolites and guiding future metabolomics research of other complex sample systems.


S5
Exactive) at m/z 200. Internal lock-mass calibration was performed by using the DHB matrix cluster ion signal at m/z 716.12461 ([5DHB−4H2O+NH4] + ), resulting in a high mass accuracy of ±1 ppm. For desorption/ionization, 50 laser pulses per pixel at a wavelength of 343 nm were focused perpendicularly to the sample surface. The step size of the XYZ sample stage was set to the desired pixel size. The full-pixel mode was used for MALDI MSI experiments conducted with 25 µm step size. In full-pixel mode, the pixel area is ablated by multiple laser pulses by a meandering movement to improve the ion signal intensities of the MSI experiment.
The scan speed for all MSI experiments was 1.6 pixel/s. The acceleration voltage was set to 3 kV. The ion injection time was set to 500 ms. The capillary temperature was 250 °C, and the S-lens level was set to 100 arbitrary units.

Supplementary Note 3: Experimental parameters for HPLC-MS analysis
The injection volume was 15 µL, and the column compartment was set to 50 °C.

Supplementary Note 4: LC-MS data pre-processing using MZmine 2
The mzXML files were imported and following parameters were applied: an initial threshold of 1E6 for MS 1 spectra and 1E4 for MS 2 spectra was used. For feature-detection, the ADAP chromatogram builder was used with a minimum signal intensity and group intensity threshold of 3E6 and with ± 3 ppm mass tolerance. A minimum mass appearance over 5 consecutive scans was set. The extracted ion chromatograms were deconvoluted using the local minimum search algorithm with a chromatographic threshold of 10%, search minimum in RT range of 0.3 min, minimum relative height of 10%, minimum absolute height of 3E6, minimum ratio of peak top/edge 0.5 and peak duration between 0.02 min and 3 min. For MS 1 -MS 2 pairing, a S6 mass range of 0.01 Da for median m/z centre calculation and RT range of 0.2 min was used.
Isotope signals were grouped with m/z tolerance of ± 3 ppm and RT tolerance of 0.1 min.

Supplementary Note 5: Feature-based molecular networking from the GNPS analysis infrastructure
The .mgf files were uploaded into the GNPS ecosystem and for FBMN following parameters were applied: precursor ion mass tolerance: 0.01 Da, fragment ion mass tolerance: 0.01 Da.
The minimum cosine score between a pair of MS2 spectra in order to form an edge in the molecular network was 0.7. The maximum amount of neighbour nodes from a single node was 10. The minimum number of fragment ions that were shared between pairs of related MS 2 spectra was 5. The maximum precursor ion mass difference between two nodes was 600 Da.
For spectral library annotations, the MS 2 spectra contained a minimum number of 6 matched fragment ions. Additionally, the minimum cosine score after spectral matching of experimental MS 2 spectra with spectral library MS 2 spectra was 0.7. The molecular networking results were imported into Cytoscape (v.3.8.0). For optimal visualization, H + -adducts were exclusively shown.

Supplementary Note 6: Method development for high-resolution OTCD MALDI MSI of cardiac glycosides
We optimized the sample preparation protocol to achieve high sensitivity while retaining spatial information. In summary, the solvent composition of methanol/water (7:3) v/v enabled high GirT-concentration (15 mg/mL), which subsequently allowed us to use a low spray volume (35 µL) combined with a low flowrate (7 µl/min) to prevent analyte delocalization and washing effects (as shown in Figure S5 for 50 µL spray volume and 10 µL/min flow rate). Between OTCD and matrix application, the sample tissue was transferred into a desiccator for two hours at room temperature to increase reaction yield ( Figure S6) and to prevent spatial artefacts due to hygroscopic properties of the GirT reagent. Any further incubation step that included increased humidity and temperature resulted in analyte delocalization and washing effects ( Figure S7) and was thus avoided. Next, OTCD was quenched by matrix application, which also provides chemical preservation and inhibits oxidation of derivatized cardiac glycosides ( Figure S6). Using an ultrafine pneumatic spraying protocol for DHB as a matrix with methanol/water (1:1) v/v as solvent showed excellent results regarding homogenous matrix crystallization (crystal sizes ≤ 10 µm) for different surface characteristics of D. plexippus tissue sections, including fat body tissue, digested plant material and integument ( Figure S8 and Figure S9 for comparison with DHB matrix layer without OTCD).