Exploring 6-aza-2-Thiothymine as a MALDI-MSI Matrix for Spatial Lipidomics of Formalin-Fixed Paraffin-Embedded Clinical Samples

Abstract

Background/Objectives: In recent years, lipids have emerged as critical regulators of different disease processes, being involved in cancer pathogenesis, progression, and outcome. Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry Imaging (MALDI-MSI) has significantly expanded the technology's reach, enabling spatially resolved profiling of lipids directly from tissue, including formalin-fixed paraffin-embedded (FFPE) specimens. In this context, MALDI matrix selection is crucial for lipid extraction and ionization, influencing key aspects such as molecular coverage and sensitivity, especially in such specimens with already depleted lipid content. Thus, in this work, we aim to explore the feasibility of mapping lipid species in FFPE clinical samples with MALDI-MSI using 6-aza-2-thiothymine (ATT) as a matrix of choice. Methods: To do so, ATT performances were first compared to those two other matrices commonly used for lipidomic analyses, 2',5'-dihydroxybenzoic acid (DHB) and Norharmane (NOR), on lipid standards. Results: As a proof-of-concept, we then assessed ATT's performance for the MALDI-MSI analysis of lipids in FFPE brain sections, both in positive and negative ion modes, comparing results with those obtained from other commonly used dual-polarity matrices. In this context, ATT enabled the putative annotation of 98 lipids while maintaining a well-balanced detection of glycerophospholipids (60.2%) and sphingolipids (32.7%) in positive ion mode. It outperformed both DHB and NOR in the identification of glycolipids (3%) and fatty acids (4%). Additionally, ATT exceeded DHB in terms of total lipid count (62 vs. 21) and class diversity and demonstrated performance comparable to NOR in negative ion mode. Moreover, ATT was applied to a FFPE glioblastoma tissue microarray (TMA) evaluating the ability of this matrix to reveal biologically relevant lipid features capable of distinguishing normal brain tissue from glioblastoma regions. Conclusions: Altogether, the results presented in this work suggest that ATT is a suitable matrix for pathology imaging applications, even at higher lateral resolutions of 20 μm, not only for proteomic but also for lipidomic analysis. This could enable the use of the same matrix type for the analysis of both lipids and peptides on the same tissue section, offering a unique strategic advantage for multi-omics studies, while also supporting acquisition in both positive and negative ionization modes.

Keywords: 6-aza-2-thiothymine; FFPE; MALDI-MSI; spatial lipidomics.

Figure 1

Lipid class distribution across different matrix and ionization mode conditions. Lipid abbreviations correspond to standard nomenclature (e.g., PI = phosphatidylinositol, SM = sphingomyelin, DG = diacylglycerol). Donut charts display the relative abundance of lipid classes identified with the three MALDI matrices—ATT, DHB, and NOR—analyzed in both positive and negative ion modes. Each lipid species is categorized into one of five major lipid classes: glycerophospholipids (blue), sphingolipids (green), glycerolipids (orange), fatty acids (red), and sterol lipids (purple), as indicated by the legend. Compared to DHB and NOR, ATT shows consistent performances across diverse ion modes, with a prominent detection of sphingolipids (e.g., Cer, SHexCer) in negative ion mode.

Figure 2

(a) Molecular images of three lipid species (SM 20:0; PC 34:1; PS 38:4) detected in positive ion mode with ATT, DHB, and NOR matrices; and (b) molecular images of three lipid species (PI 38:4; SHexCer 36:1;2O; SHexCer 44:1;3O) detected in negative ion mode with ATT, DHB, and NOR matrices. A viridis color map is reported, indicating the relative intensity of each ion from 0 (purple) to 100% (yellow).

Figure 3

Spatial segmentation of analytical replicates of mouse brain sections analyzed with MALDI-MSI in positive and negative ion mode using ATT, DHB, and NOR matrices.

Figure 4

(a) Average lipid profiles of glioblastoma (n = 34, red) and normal brain tissue (n = 5, blue); (b) hierarchical clustering performed with 171 putatively annotated lipids enables the separation of glioblastoma cores (red) and normal brain tissue cores (blue). The top 10 features are shown; (c) Student’s t-test highlighted a consistent downregulation of sphingomyelin and ceramides in glioblastoma cores. Here, representative MALDI-MS images of three tumor cores and three healthy cores display the downregulation of a sphingomyelin (SM 38:1;O2) and a ceramide (Cer 42:2;O2) in tumor cores (red) compared to healthy cores (blue). An intensity box plot is also displayed, reporting mean (yellow dot) and medial value, to represent the statistically significant downregulation of these lipid species in glioblastoma, along with a receiver operating characteristic (ROC) curve demonstrating a statistically significant AUC value.