Digital proteome: a resource for target expression across 22 tissues in human and preclinical species

Proteins are the primary targets and off-targets of therapeutics such as small molecule and biologics. Thus, a thorough understanding of their expression profiles across various human tissues in health and disease is essential through all stages of drug development.

However, generating such a comprehensive proteome profile for each tissue poses a major analytical challenge due to the large dynamic range of protein abundance within a tissue and the enormous complexity introduced by distinct proteoforms. These proteoforms may arise from genetic variations (e.g. mutations), posttranscriptional events (e.g. isoforms from alternative splicing) or posttranslational modifications, resulting in more than a million protein species from 20,000 protein-coding genes.

Thanks to advancements in sensitivity and speed in the recent years, mass spectrometry (MS)-based proteomics can now be routinely applied to quantify ten thousand of proteins in biological samples. Taking advantage of MS being an unbiased and matrix-agnostic method, we pivoted the generation of a high quality and uniform digital proteome resource of 22 healthy tissues from human, mouse and rat origin. We constructed the atlas in two parts aiming to address the following two aspects respectively: 1) a deep library providing evidence for protein/proteoform presence in each of the 22 tissues; 2) quantitative profiles for cross-species comparison in all tissues.

For the first part, each tissue sample was subjected to an extensive fractionation process followed by MS analysis of each of the fractions. In total, we identified 20"025 protein groups mapped to 17"715 protein-coding genes, providing protein-level evidence for > 90% of all protein-coding genes annotated in the neXtProt repository. Among all identified proteins, 10"755 were detectable across all tissues. Notably, we identified 37 proteins mapping to putative genes.

In part two of the atlas, quantitative expression profiles were generated for 15"772 protein groups using TrueDiscovery DIA-MS across the 22 tissues. These data allow us to investigate ubiquitous vs. tissue-restricted protein expression of various protein classes such as surface receptors or kinases. Moreover, we performed correlation analysis between protein expression and mRNA abundance across 19 tissues, demonstrating that for a significant portion of the proteome, mRNA level-information is not an adequate proxy for protein expression.

Collectively, this MS-based resource alleviates dependency on antibody-based reagents in human and preclinical models. We envision that the tissue atlas can be expanded 1) to additional preclinical model species and 2) to cover various disease indications, providing valuable insights for target discovery and assessment of target/off-target tissue distribution.