In depth profiling of the cancer proteome from the flowthrough of standard RNA- preparation kits for precision oncology

Background

Precision medicine is revolutionizing cancer treatments, where cancer patients are matched with targeted drug treatments based on whole genome and transcriptome sequencing of their tumours. However, it is proteins, the functional building blocks of cells translated from RNA, which most drugs target. In addition, DNA and RNA data do not reliably predict to protein function. Huge efforts are therefore ongoing to integrate an additional layer of omics analysis, proteomics, into translational research and patient care. Most proteomics analyses are performed on formalin-fixed paraffin-embedded tumour tissue routinely prepared during histological analyses; however, the QIAGEN AllPrep kit is widely used for DNA and RNA extraction.

In this preprint the authors investigate if clinically meaningful, high-quality data can be obtained from archived protein-rich flow-through (here named the ‘protein-fraction’) that is usually discarded during this kit protocol. In addition, they investigate if these data can be acquired at scale and speed, enabling DNA, RNA and protein analyses all from the same patient samples in a clinical setting.

Schematic summary of preprint aim

Key findings

• The authors suggest utilizing proteomic data to improve patient survival in cases of metastatic cancers with exhausted treatment options. They demonstrated that the information generated from leftover samples obtained during sequencing stages can provide a valuable layer of information. Additionally, the information generated comes from the same tumor material processed for sequencing, which facilitates correlation between different omics.
• First, the authors optimized a proteomic processing workflow for the leftover material from the QIAGEN AllPrep kit following DNA/RNA extraction from precision oncology samples. Their results demonstrate that the proteome integrity is preserved even after 5 years of storage, which opens the door to retrospective studies.
• The authors expanded the study to investigate if the information generated could provide meaningful biological insights. The proteomic analysis of primary and metastatic samples recapitulates biological tissue-specific markers and cancer-associated proteins. Additionally, the proteome retains consistent cell-type-specific information at different storage times (one or five years), enabling retrospective identification of the origin site.
• Next, the authors demonstrated the clinical utility of the information generated by proteomics in a case of colon cancer (Her2-Neu) that metastasized to the liver. Upon treatment failure, genomic analyses were unable to explain the mechanism of resistance. A comparison of protein fractions in two metastatic lesions (baseline and progression) identified the focal adhesion protein kinase 1 (PTK2/FAK1) as a possible explanation for resistance.
• To improve the proteome coverage of their workflow, the authors successfully implemented data-independent acquisition methods, doubling the number of identifications when compared to the original data-dependent acquisition method (9,173 protein groups in DIA vs 4,351 groups by DDA). The information generated accurately reflects the biology of the analyzed samples and covers a good percentage of targetable proteins (as reported in the curated gene lists from FoundationOne CDx – Roche).
• In a final optimization step, to create a method that can be easily translated into clinical settings, the paper shows a 2.5-fold increase in sample throughput (30 samples per day) by reducing the gradient length from 120 to 44 minutes, a great correlation between proteins identified in the two gradients with minimal impact on proteome depth.
• Additionally, the authors tested the potential of the Orbitrap Astral MS instrument to enhance proteome coverage in an automated version of this workflow (SP3 on a KingFisher Flex). Their optimized LC gradient (21 min) doubled the sample throughput, further increasing it from 30 to 60 samples per day, allowing for the identification of an even higher set of proteins with low abundance variability (10,197 protein groups).
• Finally, the authors incorporated phosphoproteomic analysis as an additional layer of complexity to enhance our understanding of intracellular signaling during cancer progression. Leftover peptides from proteomic analysis were used for phosphopeptide enrichment with low input material (as low as 29 µg). Analysis of a cancer cohort led to the identification of 20,000 phosphopeptides in 4,839 proteins, which covers half of the FoundationOne targetable markers. Phosphorylation sites in well-known cancer markers were detected (i.e., EGFR, AKT1, BRAF).

What we like about this preprint

The ability to integrate proteomics into existing workflows in a clinical setting is an ongoing challenge. The authors present an interesting perspective to use an otherwise discarded by-product of routine DNA and RNA extraction. This could be a step forward in creating accessible proteomics analyses by reducing barriers in sample acquisition. As the authors mention, this method also allows a direct comparison of DNA, RNA and protein analyses from the same sample. From a biological perspective, this will enable scientists to determine that any differences observed are not due to differences in sample location within the tumour. The authors also show that clinically relevant information can be obtained from the flow-through of samples stored for as long as five years, circumventing the need for immediate processing.

Future directions and questions for the authors

• Multiple sections in this preprint show a similar number of identifications for samples stored for one or five years. Have the authors considered comparing these results with freshly processed samples to determine if there are significant losses for markers with low stability, such as transcription factors or proteins with short half-lives? Additionally, to determine if the Qiagen kit has any bias for a specific group of proteins, have the authors thought about comparing with the proteome obtained by conventional digestion methods (in-solution, on-column, or SP3) from fresh tumor material, rather than leftover flow-through fractions?
• Have the authors considered correlating RNA and protein expression for each sample?
• Do the authors plan to follow up on some of the proteomic findings? For example, PTK2, the protein suggested as a potential marker of treatment resistance, or the phosphorylation sites identified, if commercial antibodies are available?
• The authors established an excellent workflow for the different stages of metastatic cancer, including baseline and progression upon treatment. The workflow incorporates the Clinical Knowledge Graph; however, for that analysis, the authors considered the upregulated proteins. It would be interesting to see the role that downregulated proteins play in the ERBB2 (HER2) pathway. Have the authors considered this?

Tags: precision oncology, proteomics

doi: https://doi.org/10.1242/prelights.41416

(No Ratings Yet)