Multiparametric cellular and spatial organization in cancer tissue lesions with a streamlined pipeline.

Abstract

Multiplex immunostaining analysis remains fragmented, underperforming and labour intensive despite tissue proteomic methodologies achieving ever-increasing marker complexity. Here we propose an open-source, user-guided automated pipeline that streamlines start-to-finish, single-cell resolution analysis of whole-slide tissue, named multiplex-imaging analysis, registration, quantification and overlaying (MARQO). MARQO integrates elastic image registration, iterative nuclear segmentation, unsupervised clustering with mini-batch k-means and user-guided cell classification through a graphical interface. We compare and validate MARQO using multiplexed immunohistochemical consecutive staining on a single slide using human tumour and adjacent normal tissue samples. Performance is compared with manually curated pathologist determinations and quantification of multiple markers. We optimize MARQO to analyse diverse tissue sizes from whole tissue, biopsy, and tissue microarray and staining approaches, such as singleplex immunohistochemistry and 20-colour multiplex immunofluorescence, to determine marker co-expression patterns in multiple human solid cancer types. Lastly, we validate CD8⁺ T cell enrichment in hepatocellular carcinoma responders to neoadjuvant cemiplimab in a phase 2 clinical trial, further showing the ability of MARQO to identify spatially resolved in situ mechanisms by providing multiplex whole-slide single-cell resolution data.