Researchers have used proteomics to identify and validate a druggable metabolic vulnerability in a child’s rare, treatment-resistant cancer.
Led by teams at the University of British Columbia (UBC) and BC Children’s Hospital Research Institute, Canada, the study combined deep proteomic profiling of preserved biopsy tissue with a fast, in vivo testing model. Using LC-MS/MS, the researchers analyzed formalin-fixed samples and identified elevated levels of SHMT2 – a key enzyme in one-carbon metabolism. They then grafted the patient’s tumor onto the membrane of a chicken egg (CAM model) to assess drug response in a living system.
The antidepressant sertraline, known to inhibit SHMT2, slowed tumor growth in this model and was recommended for off-label use. This treatment reduced the patient’s tumor growth rate, although the disease continued to progress.
We spoke with co-senior author Dr. Philipp Lange – Canada Research Chair in Translational Proteomics of Childhood Malignancies, and Associate Professor of Pathology and Laboratory Medicine at UBC – about the rationale behind this approach and the implications for real-time precision oncology.
What was your main inspiration for this work and the approach you took?
Precision oncology has provided new treatment opportunities for children with cancer after standard of care therapies have been exhausted. Largely driven by genome sequencing and transcriptome analysis, this strategy now identifies actionable genome alterations and mRNA changes in the majority of patients. Unfortunately, however, only a subset of patients experiences clinical benefit, which is often short-lived.
As most drugs act on proteins and substantial regulation occurs between the genome, transcriptome and targeted proteome, the logical next step was to directly measure proteome changes to complement genomic information. Inspired by the opportunity to provide potentially life-changing information to patients and their families, we sought to adapt our proteomics workflows to integrate seamlessly with clinical practice. This led us to a LC-MS/MS proteomics approach on formalin-fixed paraffin-embedded (FFPE) tissues that used deep data-independent acquisition (DIA) for relative quantification of thousands of proteins in tumor versus normal tissue. Importantly, since this was a research study, we could rapidly integrate non-validated functional precision oncology approaches following identification of a drug target; for example using stable isotope tracing to confirm the drug mechanism of action and patient-derived xenograft models to confirm sensitivity to the identified drug through the BRAvE initiative.
Were there any difficult analytical challenges you had to overcome during this study?
There were a number of challenges to overcome and it really took an interdisciplinary team effort of experts and emerging scientists from across the country to make this study a success. We had to build on a number of recent technical and analytical advances by us and others.
Firstly, the ability to process FFPE tissues was key to integration with routine sample handling in the clinical laboratory. This was optimized by collaboration between my laboratory and clinical pathologists.
The biggest analytical challenge was arguably establishing the optimal comparator for differential analysis. Quantitative comparison across proteomic studies, in particular when acquired on different instruments or sites, remains an unsolved challenge. Comparison of a patient’s tumor against large normal tissue or cancer-type compendia, as commonly done for transcriptome analyses, lacks precision. To overcome this, we opted for personalized comparators using earlier disease timepoints and several tumor adjacent non-cancerous tissues. To orthogonally validate our findings, we then used immunohistochemistry to confirm the elevated abundance of the identified drug target relative to other tumors with demonstrated drug-sensitivity.
We expect that the increased throughput on newer generation mass spectrometers and advances in cross-study normalization and quantification will soon make comparison against large tissue and disease proteome atlases possible.
What implications do your findings have for the future of proteomics-based precision medicine?
Demonstrating the ability of proteomics to inform real-time treatment decisions, even for a single case, has fundamentally changed the conversation with clinicians and patient advocates. Generally seen as a complex research exercise only a few years ago, most now see the value of proteomics as a complementary layer.
I believe if we keep advancing analytical technologies at the current pace, we will soon reach truly personalized healthcare. Longitudinal point-of-care or at-home sampling will allow us to establish individual baselines, eliminating population-based reference ranges with their inherent biases and limitations. Innovations in AI will enable high-dimensional multi-parametric assessments that have the potential to not only identify single treatment opportunities but also detect disease heterogeneity and predict personalized strategies to limit the development of drug resistance.
Could the approach taken in your study be scaled up? If so, what hurdles do you envision?
We are currently establishing this research approach at multiple centres across Canada; indeed, this study was a huge team effort of PROFYLE (PRecision Oncology for Young peopLE), a key initiative of the pan-Canadian pediatric cancer network ACCESS (Advancing Childhood Cancer Experience, Science and Survivorship). A real hurdle will be translation into clinical tests that are recognized by the relevant certification bodies. We are still a few steps away from validated whole proteome-scale assays that could be run routinely in a clinical laboratory. In the meantime, we are planning a feasible intermediate step using targeted mass spectrometry to analyze panels of known cancer-associated proteins. These will be particularly powerful in pediatric precision oncology where tumors have a lower mutational burden and fewer treatment options.
In the future, would it be possible to include additional “omes;” for example, metabolomics-based approaches?
Our study has shown the value of a multiomic approach, incorporating genomic, transcriptomic and proteomic analyses. While not incorporating a full metabolomics characterization of the patient’s tumor, we did validate our findings at the metabolite level using stable isotope tracing. This shows that incorporating metabolomics upfront can be very valuable. Seamless integration with clinical workflows, in particular the use of FFPE biopsies for solid tumors, may pose challenges but I am optimistic that the field can overcome these.
What's next for this work?
The next milestone is to establish proteome-based treatment recommendations for 50 or 100 patients across the country and then re-evaluate the value proposition. Together with lessons learned in similar studies in, for example, the US, Europe and Australia, this will give us a good idea if, and where, this is best applied and where further improvement is required. Ultimately, the key goal will be to move from the research environment into clinical routine, importantly at an affordable price to reach for all patients in public and private healthcare settings.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
