# Integrative spatial profiling pipeline for determining TME architectures in archival clinical specimens using CmTSA superplex technology

## TL;DR

A new method called CmTSA superplex technology enables detailed analysis of tumor microenvironments in archived clinical samples, improving understanding of cancer progression and treatment response.

## Contribution

The novel hybrid optochemical fluorescence depletion (HOC-FD) technology and RNN-based spatial profiling pipeline enable high-resolution, cost-efficient TME analysis in FFPE archival tissues.

## Key findings

- HOC-FD combined with CmTSA allows 30–60 biomarker labeling in FFPE tissues with high signal-to-noise ratios.
- A deep learning pipeline enables accurate cellular segmentation and phenotype classification in TME imaging.
- Radius-constrained neighborhood networks (RNNs) reliably define functional niches with biological and prognostic relevance.

## Abstract

The tumor microenvironment (TME) comprises diverse cellular components that spatially interact to form distinct functional niches (FNs). Profiling these TME spatial features has proven to be a critical approach for correlating tumor progression and therapeutic response. However, RNA stability limitations constrain the broad clinical implementation of spatial transcriptomics, while high background noise and low signal resolution compromise the accuracy of direct labeling-based spatial proteomic approaches in clinical specimens. To overcome these constraints in archival samples, we developed hybrid optochemical fluorescence depletion (HOC-FD) technology that integrates autofluorescence quenching with cyclic multiplex tyramide signal amplification (CmTSA) for formalin-fixed paraffin-embedded (FFPE) tissues. This unified platform enables the concurrent labeling of 30–60 biomarkers with ultrahigh signal-to-noise ratios while maintaining cost efficiency and compatibility with high-throughput processing of archival FFPE specimens. While superplex imaging captures multidimensional TME data, extracting spatial features from raw pixel-level outputs remains technically challenging. To resolve this problem, we implemented a computer vision pipeline beginning with deep learning-based cellular segmentation and phenotype classification under predefined biomarker annotation rules. Using single-cell spatial mapping of human colon and cervical cancer specimens, we systematically evaluated and selected radius-constrained neighborhood network (RNN) analysis to define functional niches, validating their accuracy and reliability in generating spatially coherent FNs with biological and prognostic relevance. In summary, the CmTSA platform combined with RNN-based spatial profiling provides an integrated framework for visualizing and quantifying multicellular functional states within architectures of the TME, potentially enhancing tumor immunology investigations and precision immunotherapies.