Quantitative 3D histology: Million-cell-scale spatial multi-omics in intact FFPE specimens

Cancer exhibits profound heterogeneity both within and between patients, spanning from individual cell phenotypes to complex microenvironmental niches. Current technologies fail to capture this multiscale complexity while preserving native tissue architecture holistically. Here we present a quantitative 3D histology platform that enables spatially resolved, million-cell-scale multi-omic profiling of intact formalin-fixed paraffin-embedded (FFPE) specimens. It eliminates sampling error and subjectivity in histopathology, and facilitates quantitative phenotyping to guide patient selection in treatments. We applied this platform to profile diverse cancer specimens, uncovering cancer-nerve interactions, cancer-immune cell interactions, and generating quantitative continuous scores for actionable targets, including TROP2 and HER2, through normalised membrane ratios (NMRs) in >0.5 million cells per sample. We also identified missed cancers, pre-cancerous lesions, and lymphovascular invasion in designated normal tissue blocks, as well as decision dilemmas and errors in 2D quantitative digital pathology efforts, due to the fundamental limitations offered by a random specimen cut. In developing this platform, we devised novel chemistry for low-temperature FFPE retrieval to prevent biomolecular damage, followed by custom supramolecular reaction systems that enable deep antibody penetration for up to 28-plex multiplexed immunostaining in > 1,000 μm-thick specimens. Non-destructive tissue clearing, coupled with light-sheet microscopy, then captures the entire specimen in 3D with optical sectioning. We then trained a family of neural networks to segment individual cells in their native 3D positions with precise cellular geometry, allowing computation of distances to tissue boundaries and features (vessels, nerves), as well as spatial relationships between cell types. These 3D masks quantify marker expression across major subcellular compartments, generating continuous molecular scores for each cell. Crucially, tissues remain structurally and molecularly intact throughout processing. Comparative analyses pre- and post-3D profiling show indistinguishable results for H&E, immunohistochemistry, whole-genome sequencing, bulk RNA-seq, laser-capture mass spectrometry proteomics, and 2D spatial transcriptomics. We demonstrate the feasibility of integrating 2D spatial transcriptomic data into the 3D volume for comprehensive single-cell molecular profiling. Moreover, the 3D histology pipeline requires no specialized equipment, is fully automatable, and generalizes across tissue types, enabling rapid clinical deployment for precision oncology applications where spatial context drives therapeutic decisions. Our technology addresses the fundamental limitations of current 2D pathology methods, with broad applicability across multiple cancers.