(E-165) Optimizing an End-To-End Workflow for Non-Destructive 3D Pathology of Clinical Specimens

Introduction: The burgeoning field of 3D pathology, in conjunction with human and machine-learning analysis, has the potential to provide novel structural and molecular insights that can improve clinical treatment decisions for a variety of diseases [1,2]. Advances in microscopy methods have allowed for improved volumetric imaging of optically cleared (transparent) tissues without the need for destructive serial sectioning. In order to achieve non-destructive 3D pathology of large clinical specimens, various fluorescence staining and tissue-clearing methods can be used. Compared to staining thin tissue sections, achieving bright and uniform staining and robust clearing in large tissues is nontrivial, and methods can suffer from poor consistency as well as be time-consuming and expensive in certain cases. Previously, we have used small-molecule H&E-analog stains for fluorescence imaging of tissues with open-top light-sheet (OTLS) microscopy. While we have had success with smaller pilot studies, achieving consistent data quality (i.e. intensity, uniformity, resolution, and contrast) in large-scale studies has remained challenging, limiting our ability to perform certain human and computational analyses [3,4]. Therefore, we have optimized a H&E analog staining workflow and have developed quality-control methods to ensure consistent high-quality staining of formalin-fixed paraffin-embedded (FFPE) tissues for non-destructive 3D pathology across hundreds to thousands of specimens.

Materials and Methods: Punch biopsies (3-mm diameter) of FFPE prostate were deparaffinized in 60°C xylene for 48 hours to remove the paraffin wax and were subsequently washed in 100% ethanol for 2 hours and partially rehydrated in 70% ethanol for at least 3 hours. The tissues were stained for 48 hours in 70% ethanol with variable NaCl concentrations (0 mM, 10 mM, 50 mM, and 500 mM) and variable eosin (1:40 to 1:400) and TO-PRO-3 concentrations (1:250 to 1:500 diluted from 1 mM stock). The samples were then dehydrated in 100% ethanol for 2 hours and cleared in ethyl cinnamate for 4 hours. The specimens were imaged with OTLS microscopy [5]. Subsequent viewing and signal analysis (quality-control procedure) was performed with ImageJ after the 3D datasets were stitched and fused with the ImageJ plugin, BigStitcher [6]. Qualitative and quantitative assessments of signal strength, uniformity, resolution, and contrast as a function of tissue depth were used to determine optimal tissue-preparation conditions and microscopy parameters.

Results, Conclusions, and Discussions: After extensive optimization of protocols and imaging parameters, 3-mm punch biopsies of prostate generate uniform high-quality images (H&E-analog stain) at depths of at least 0.5 mm when imaged with OTLS microscopy. In terms of uniformity, locally averaged fluorescence intensities (of similar structures) deviate by < 10% across the 3-mm diameter of the punch biopsies. For resolution, sub-nuclear features are resolved at depths of at least 0.5 mm. Signal-to-background ratios (contrast) are also >10 for both the eosin and TO-PRO-3 (nuclear) stains at depths of at least 0.5 mm. A combined TO-PRO-3 (diluted 1:500) and eosin (diluted 1:100) staining method in a 70% ethanol buffer at pH 4 with 10 mM NaCl required less hands-on time than other staining methods and resulted in the best and most-reproducible image quality. The addition of NaCl was found to significantly contribute to the uniformity of the nuclear stain across the 3-mm diameter punch biopsy throughout the depth of the tissue. Higher concentrations of eosin were shown to result in the quenching of the signal at the interior of the tissue. The same protocol has since been successfully applied to various other tissues, including FFPE colorectal cancer and lung specimens. In summary, we have developed an end-to-end workflow for 3D pathology that includes an optimized tissue-preparation protocol, imaging and post-processing pipeline, and quantitative quality control to ensure that our datasets are acceptable for large-scale clinical studies with archived FFPE tissues. Future studies with human observers (pathologists) and/or computational analysis (including deep learning) will aim to demonstrate the value of 3D pathology to improve critical treatment decisions.

Acknowledgements (Optional): This work was supported by funding from the National Institutes of Health (NIH): R01EB031002 and R01CA268207 (J.T.C.L.); the National Science Foundation (NSF): 1934292 HDR: I-DIRSE-FW (J.T.C.L.) and NSF Graduate Research Fellowships DGE-1762114 (K.W.B. and L.A.E.B.); the Department of Defense (DoD) Prostate Cancer Research Program: W81XWH-20-1-0851 (J.T.C.L.) and W81XWH-18-10358 (J.T.C.L.); Prostate Cancer UK: MA-ETNA19-005 (J.T.C.L.), and the Pacific Northwest Prostate Cancer SPORE (P50CA097186). We also acknowledge the Canary Foundation for providing some specimens used in this study.

References (Optional): [1] J.T.C. Liu et al., “Nondestructive 3D pathology with light-sheet fluorescence microscopy for translational research and clinical assays,” Annu. Rev. Anal. Chem. 16, 231-252 (2023)
[2] J.T.C. Liu, et al., “Harnessing non-destructive 3D pathology,” Nature Biomedical Engineering 5, 203-218 (2021)
[3] W. Xie et al., "Prostate cancer risk stratification via non-destructive 3D pathology with deep learning-assisted gland analysis," Cancer Research 82, 334-345 (2022)
[4] R. Serafin et al., "FalseColor-Python: a rapid intensity-leveling and digital-staining package for fluorescence-based slide-free digital pathology," PLoS ONE 15, e0233198 (2020)
[5] A.K. Glaser et al., “A hybrid open-top light-sheet microscope for multi-scale imaging of cleared tissues,” Nature Methods 19, 613-619 (2022)
[6] D. Hörl et al., “BigStitcher: reconstructing high-resolution image datasets of cleared and expanded samples,” Nat. Methods 16, 870–874, (2019)