Subcellular-Resolution Molecular Pathology by Laser Ablation-Rapid Evaporative Ionization Mass Spectrometry

Abstract

This work demonstrates the combination of ambient laser ablation (LA) with in-source surface-induced declustering, originally developed for rapid evaporative ionization mass spectrometry (REIMS). This combination, termed laser ablation REIMS (LA-REIMS), provides sensitivity, spatial resolution, and chemical coverage comparable to matrix-assisted laser desoprtion ionization (MALDI) but without the requirement for matrix deposition. The atmospheric pressure interface setup was subjected to detailed characterization with regard to geometric and thermal parameters augmented by in-silico flow modeling. The resulting platform was tested using aerosol formed by the infrared laser ablation of tissues. Three different laser systems were successfully employed for ambient mass spectrometric imaging: a carbon dioxide laser (λ = 10.6 μm, τ_L = ∼100 μs), an optical parametric oscillator (OPO; λ = 2.94 μm, τ_L = 8 ns), and an optical parametric amplifier (OPA; λ = 3.0 μm, τ_L = ∼30 ps). Single-cell imaging was achieved using the high-resolving capabilities of the OPA systems, and metabolites and lipids ranging from amino acids through carbohydrates and nuclear bases to complex glycolipids were successfully detected. The technique was also tested as a platform for MS-guided surgery, raising the possibility of using a single technique for generating histological and in vivo data.

1

Schematic representation of the LA-REIMS imaging workflow development. The imaging setup was constructed using mid-infrared lasers, laser optics, and an automated XY stage for raster scanning. The imaging platform was coupled with a mass spectrometer equipped with a REIMS ionization source. (1) Numerical simulations were conducted to gain a better understanding of collision surface mechanics and particle behavior around the impact region. (2) Laser parameters were tested and optimized over a broad range of parameters, including wavelength, pulse width, spot size, and fluence. (3) The system was tested for MSI performance and (4) a proof-of-concept study was performed on breast cancer samples to evaluate the clinical translational capabilities of the technique.

2

Five micrometer high-resolution imaging using the picosecond LA-REIMS setup. A 1 × 1 mm² area of the thalamus region of a mouse brain was analyzed using the high-resolution picosecond laser setup. The distributions of 134.04 m/z (Adenine [M-H]-) (A), 766.54 m/z (PE 38:4 [M-H]-) (B), and 955.72 m/z (unknown) (C) show high spatial matching with the tissue features observed in the optical image postanalysis (D). Individual cell nuclei were identified with numbers from 1 to 5 on a zoomed-in section of the distribution of adenine (E) and the aligned optical image (F).

3

Lipidomic and metabolomic characterization of breast cancer samples using LA-REIMS. Ex vivo human breast samples were analyzed with two different IR lasers using the LA-REIMS method. A: Univariate plots of identified molecules of adenine [M – H]⁻, glutamine [M – H]⁻, glutamate [M – H]⁻, phosphatidylethanolamine PE(36:2) [M – H]⁻, and triglyceride TG(52:2) [M + Cl]⁻ show significant differences between different tissue types. *Statistically significant difference between fat and fibrous tissue. **Statistically significant difference between tumor and fat tissue. ***Statistically significant difference between tumor and fibrous tissue. The tissues contained healthy and cancerous sections, validated by histopathology analysis. Volcano plots generated from the data obtained with the two lasers (B, C) allow us to identify numerous metabolites and phospholipids that show statistically significant fold changes. The breast cancer samples were analyzed using the two lasers at 70 μm raster sizes. Consecutive tissue sections were H&E stained and classified by pathologists (M-CO₂ and Q-OPO). The results of the CO₂ laser can be seen in panels D–G, where the composite image (E) is visualized (red: 744.54 (PE 36:1 [M – H]⁻); green: 217.05 m/z (unknown); and blue: 893.79 m/z (TG 52:2 [M + Cl]⁻)). The results of the OPO imaging can be seen in panels B–E, where the composite RBG image (B) is visualized (red: 744.54 m/z (PE 36:1 [M – H]⁻); green: 195.05 m/z (unknown); and blue: 893.79 m/z (TG 52:2 [M + Cl]⁻)). Individual ion images for different ionic species (O: 744.54 m/z and P: 893.79 m/z for CO₂; S: 744.54 m/z and T: 893.59 m/z for OPO) show good differentiation between different histological status tissues.

4

A PCA model was built from breast healthy and tumor data obtained with all ablation modalities for tissue classification (A). Leave-20%-out cross-validation was performed from the previously built PCA-LDA model; the model yielded good sensitivity (93.65%) and specificity (97.12%) (B).