MSACL 2025 Abstract

Quality Control and Data Reproducibility for Large-scale MALDI Imaging Studies

Katerina V. Djambazova(1,2), Lukasz G Migas(3), Jamie Allen(2,4), Allison Esselman(1,5), Martin Dufresne(2,4), Melissa A. Farrow(1,2), Raf Van de Plas(2,3,4), and Jeffrey M. Spraggins (1,2,4,5,6)
(1)Department of Cell and Developmental Biology, Vanderbilt University, Nashville, TN, USA (2)Mass Spectrometry Research Center, Vanderbilt University, Nashville, TN, USA (3)Delft Center for Systems and Control, Delft University of Technology, Delft, Netherlands (4)Department of Biochemistry, Vanderbilt University, Nashville, TN, USA (5)Department of Chemistry, Vanderbilt University, Nashville, TN, USA (6)Department of Pathology, Microbiology and Immunology, Vanderbilt University Medical Center, Nashville, TN, USA

Katerina Djambazova, PhD (Presenter) Vanderbilt University

Relevant Financial Disclosures (within past 24 months, reported on Apr 02, 2025) No relevant financial relationship(s) to disclose.

INTRODUCTION:
Matrix-assisted laser desorption ionization imaging mass spectrometry (MALDI IMS) is an established tool for molecular mapping of tissues with widespread biomedical applicability. As such, MALDI IMS has become a key technology in multiple large-scale tissue atlassing projects, such as the Human BioMolecular Atlas Program (HuBMAP), the Kidney Precision Medicine Project (KPMP), and the Human Tumor Atlas Network (HTAN) among others. MS-based imaging assays generate large and often complex datasets. Long-term studies require data collection over months/years, which can increase technical variation, produce outliers, and impede reliable data interpretation. Although standardized quality control (QC) procedures have been established for other -omics approaches, similar standards are yet to be instituted for MS-based imaging technologies. Here, we demonstrate a comprehensive approach to systematically monitor and ensure data reproducibility for large-scale MALDI imaging experiments, allowing for reliable biological data interpretation.

METHODS:
Human kidney samples were obtained from the Central Biorepository, as part of the Kidney Precision Medicine Project, and from the Cooperative Human Tissue Network at Vanderbilt University Medical Center. Samples were cryosectioned at 10µm thickness and mounted onto indium tin oxide-coated slides. The tissue sections were washed 3 times for 45s with chilled isotonic ammonium formate solution to remove salts. An aminated cinnamic acid analog matrix was sublimed onto the tissue surface using an in-house developed sublimation device. MALDI IMS data were acquired in both polarities on small tissue areas 10µm pixel size using a timsTOF FleX mass spectrometer (Bruker Daltonics). Data analysis and visualization were performed using SCiLS and in-house developed software.

RESULTS & DISCUSSION:
This study assessed the reproducibility of MALDI IMS data acquisition and pre-processing workflows on ~200 datasets. Baseline performance was established using ~100 single-patient human kidney lipid datasets. Subsequently, 90 MALDI datasets from kidney biopsies of patients with kidney disease, including acute kidney injury (AKI) and chronic kidney disease (CKD), were acquired over 2.5 years and compared to the initial QC-cohort. Here, we established rigorous MALDI IMS QC protocols, allowing us to reliably mine multimodal molecular data for disease-related molecular trends and alterations.

We standardized QC tuning procedures for the imaging platforms and used data pre-processing steps including mass alignment, calibration, and inter-/intra-experiment normalization to correct for large data variations. Our custom QC software provides information regarding overall signal intensity variation, peak width, mass resolving power, mass error, and number of detected (and identified) peaks. Overall, we observed <10% variability in signal intensity across all data cohorts. In each dataset, we tentatively identified ~500 lipids in negative and ~300 lipids in positive mode, with >50% present in all samples, regardless of disease state. Select diagnostic ions showed minimal variability across mass error (~0.85 ppm), peak width of ~0.016 Da, and mass resolving power (>52,000). These data show that pre-processing steps such as mass alignment, calibration, and inter-/intra-experiment normalization can correct for large data variations, reducing bias in the downstream statistical and biological analyses.

IMS data processed with high rigor and reproducibility allows for reliable multimodal data mining and interpretation. For example, we used histology annotations provided by a board-certified pathologist to generate masks linking spatio-molecular profiles to tissue structures impacted by disease. Here, we generated molecular profiles of interstitial fibrosis and tubular injury (IFTA) in biopsies of patients with AKI. IMS-driven multimodal data mining revealed several ether-linked lipids, including PE O-38:5 (m/z 754.544, negative mode) as potential biomarkers of IFTA. Ether-linked lipids have been shown to regulate lipid raft signaling domains in immune cells. Both the T and B cell receptor signal transduction pathways, as well as signal transduction in innate immune cells, are regulated by these assembly and organization of lipid raft domains. These observations are consistent with the immune-related signaling pathway elucidated by orthogonal transcriptomic studies. Ultimately, these interpretations rely on systematic and reproducible data collection and established QC protocols, which allow us to track and correct deviations in large-scale, long-term MALDI imaging experiments.