A German-Danish team of researchers has developed a new imaging technology that is able to quantify the number of expressed proteins in a given cell, map tissue and cell-type specific proteomes, and identify drug targets.
Using the technique, called Deep Visual Proteomics (DVP), it was possible to identify spatially resolved proteome changes as normal melanocytes transition to fully invasive melanoma, in archived primary melanoma tissue.
That revealed pathways which changed in a spatial manner as the cancer progresses. For example, mRNA splicing dysregulation during metastatic vertical growth – when a melanoma grows deeper into the tissue – coincided with reduced interferon signaling and antigen presentation.
The ability to combine visual features of a tumor with deep profiling, and to visualize the protein signature in aberrant cells that are next to surrounding normal cells, promises to give unprecedented insight into disease processes, the researchers say.
"DVP could be a game changer for molecular pathology," said Andreas Mund, associate professor at the Novo Nordisk Foundation for Protein Research in Copenhagen, and co-lead author of a paper describing the technique in Nature Biotechnology, May 19, 2022.
"With this method, we take a tissue sample with tumor cells and can identify and determine 1,000s of proteins in a minute of time and effort," Mund said. "These proteomics signatures reveal the mechanisms that drive tumor development and directly expose new therapeutic targets from a single tissue slice of a cancer patient biopsy. It exposes a cosmos of molecules inside these cancer cells."
Despite the availability of imaging-based mass spectrometry approaches to spatial proteomics for the visualization of proteins in their native cellular environment, a gap has remained in connecting images with single cell resolution measurements of protein abundance.
DVP overcomes this by combining artificial intelligence-based image analyses of cellular phenotypes with automated single cell laser microdissection and ultra-high sensitivity mass spectrometry.
That makes it possible to link protein abundance to cellular or subcellular phenotypes, while preserving the spatial context. "Working with the proteomic signatures, we reflect the direct manifestation of genetic and epigenetic alterations, and are thus investigating diseases in closest relation to the actual phenotype," say the researchers.
"By merging the power of microscopy, AI and ultra-sensitive mass spec-based proteomics, we have developed a method that is very powerful for elucidating the molecular wiring of healthy versus diseased cells," said research leader Matthias Mann, head of the department of proteomics and signal transduction at the Max Planck Institute of Biochemistry in Munich.
To apply DVP, any formalin fixed and paraffin embedded stored tissue can be used. Tissue sections are imaged in high resolution using commercially available microscopes.
Working from the images, cells present in the tissue of interest are classified by machine learning and artificial intelligence. Single cells of interest are then excised by laser capture microdissection.
Following this, the 1,000s of proteins present in excised cells are detected simultaneously using mass spec, with subsequent bioinformatics analyses generating protein maps across the different cell populations.
The ability to map protein profiles of different cell types onto the tissue architecture means properties such as the capacity of a tumor to invade surrounding tissue can be traced back to the cells that are responsible for this.
The researchers first tested DVP on salivary gland acidic cell carcinoma, a rare and understudied cancer of the epithelial secretory cells. Bioinformatics analyses of tumor cells and their neighboring normal cells revealed significant biological differences, with the tumor cells showing upregulation of interferon response proteins and of the proto-oncogene Src. As the researchers note, both of these are actionable therapeutic targets.
They next turned attention to decoding molecular alterations in the development and progression of melanoma. The pathogenic mutations in melanoma are largely catalogued, supporting the direct study of spatially resolved proteomes of distinct cellular phenotypes of melanoma progression.
Precancerous melanoma in situ and primary melanoma showed differences in proteins involved in immune cell signaling and cell metabolism, and also coincided with reduced production of melanin. The advanced stages of radial and vertical growth showed well-defined metabolic activation along with disease progression. Expression of proteins involved in oxidative phosphorylation and mitochondria function gradually increased from melanocytes, to melanoma in situ, and on to invasive melanoma, indicating a dependency on mitochondrial respiration in advanced tumor stages.
Conversely, proteins involved in antigen presentation and interferon response were downregulated as the tumor progressed, in line with known immune evasion responses in melanoma.
Overexpression of CD146 (cluster of differentiation 146), is implicated in melanoma progression, and the direct comparison of spatially defined regions of cells enriched in CD146-high regions further highlighted critical features of cancer metastasis. One example is extracellular matrix remodeling through collagen degradation and upregulation of platelet-derived growth factor.
These tumor-driven changes support growth, increase migration of tumor cells and remodel the extracellular matrix to facilitate metastasis via adjacent blood vessels.
The examples of salivary gland cancer and melanoma were used to validate DVP. Co-author Fabian Coscia, head of the spatial proteomics research group at the Max Delbruck Center for Molecular Biology in Berlin, said it can now be more widely applied. "The technique can be used to characterize all other tumor types in similar detail," he said.
The intention is to tap into patient biobanks and apply DVP to find new targets for personalized cancer therapies, including for treatment-resistant tumors.
It is not only tumor cells that can be imaged with DVP, Coscia said. "You can for example, analyze the protein in nerve cells to find out exactly what happens in cells in neurodegenerative diseases, such as Alzheimer's or Parkinson's."