Mapping molecular complexity with spatially-resolved proteomics

Background:

Spatial protein abundance profiles within tissues are a key factor in understanding disease pathology and cellular function. These profiles are typically studied by antibody-based imaging techniques that provide high spatial resolution, but can only probe a limited number of proteins. In order to more precisely define molecular phenotypes in disease, there is a need for unbiased, quantitative technology capable of mapping the expression of many hundreds to thousands of proteins within tissue structures. Laser capture microdissection (LCM) in combination with high-throughput LC-MS/MS-based proteomics is well placed to meet this need.

Methods:

Tissue voxels were isolated from 10 µm-thick sections of a human brain tumour in a gridded pattern as resolutions as low as 40 µm. Spatially variable proteins were detected using spatial auto- and co-correlations. Spatial protein profiles were used for unbiased spatial clustering by affinity network fusion and hierarchical clustering.

Results:

Spatially-variable protein patterns can be detected at a range of resolutions, but the process is sensitive to the handling of missing values. Unbiased spatial clustering generates clusters of samples which co-cluster in space, reflecting the prominent pathology visible in H&E-stained sections along with further nuances such as immune-cell infiltration and oxygen/nutrient gradients, potentially boosting biomarker discovery.

Conclusions:

Spatially-resolved proteomics methods which spatially profile thousands of proteins will push the boundaries of understanding tissue biology and pathology at the molecular level. Recent mass spectrometry technology developments were key to achieving this throughput and depth, and this approach stands to benefit from further technological developments.