3D reconstruction of lung adenocarcinomas—one module for the development of mathematical multiscale models of lung cancer

Aims The BMBF-funded LungSysII consortium integrates information derived from molecular biology, cell biology, and histology using systems biology approaches to generate integrative multiscale-models of non-small cell lung cancer (NSCLC). In this context, we aim to define the three dimensional spatial relationship of the vascular system and the tumor mass in human pulmonary adenocarcinomas (ADC) as well as adjacent non-tumorous tissues based on histological data. We here report our most recent progress to generate a comprehensive 3-dimensional (3D) picture of ADC. Methods. Material was collected from freshly resected ADC patients and systematically cut into pieces of up to 1 cm in diameter. Samples were processed for optical projection tomography (OPT) scanning, utilizing tissue autofluorescence or specific epitope staining using directly labelled smooth muscle actin (SMA)-specific antibodies. In addition, alternate staining of serial sections derived from tumor samples were performed including H&E-, factor VIII (FVIII)-, and pan-cytokeratin (KL1) staining. Automated whole slide imaging was performed using the Hamamatsu NanoZoomer Digital Pathology system. The resulting 2D information was used to generate a 3D representation of the data by means of a non-linear elastic image registration. Results. Whole tissue OPT-scanning revealed the spatial distribution of bronchial and vasculature structures in the tumor and adjacent non-tumorous lung tissue. The image quality of the 3D vessel structure was improved solving a non-negatively constrained, L2-based reconstruction problem iteratively (MRNSD) on the raw-data produced by the OPT system. To reconstruct the serial sectioning data to a 3D volume, a special non-linear image registration algorithm was developed and applied. Specialization of the algorithm was needed due to cutting artefacts such as shape distortion and staining variation. The optimized non-linear algorithm was successfully applied on the H&E-, FVIII-, and KL1-staining. Conclusions. We here present approaches for 3D reconstruction of the vascular system and tumor mass in ADC as well as bordering healthy tissue. This quantitative information covers the range μm to cm and can be used for computational tissue modelling and for integration in mathematical multiscale models, which are currently under development.