Proteomics moves toward the clinic
Broad researchers are showing how protein analysis is not just a scientific tool but could also help improve patient care.
In September 1844, Thomas Alexander McBean of London was vacationing in the Scottish countryside when he collapsed, overcome by violent chest pain. Over the next year, McBean’s pain worsened as he became thin and weak. After a slew of treatments failed, his doctor took a sample of his urine, which was opaque and dense but when heated, formed a clear yellow solid.
The doctor sent a sample of the urine to Henry Bence Jones, a London-based physician and chemist specializing in urine composition, who concluded that the substance that made the urine dense was an indicator of McBean’s disease — multiple myeloma, a type of blood cancer that develops in bone marrow.
The substance is recognized today as the first protein marker of cancer and is named after Bence Jones. Doctors still test for the protein in urine to diagnose multiple myeloma. Broadly, McBean’s story also shows what happens when scientists pay attention to proteins, and how they can be early signatures of disease and can provide mechanistic insights into biological processes.
More than 150 years after McBean’s death, researchers at the Broad Institute of MIT and Harvard and elsewhere are continuing to find ways to profile, measure, and learn from proteins, using more high-tech methods to achieve the same goal: improving the diagnosis and treatment of disease. Scientists at Broad’s Proteomics Platform are increasingly showing the promise of proteomics — the global study of proteins together with their myriad modifications in health and disease — to impact patients. They are developing better analytical methods that work on patient samples and using those and other approaches to, for example, find early markers of severe disease, design personalized medicines, and track how patients respond to treatment. And with each new improvement to technologies that identify and quantify proteins, scientists rapidly gain more insight into the role of key proteins in disease, moving proteomics closer to the clinic.
“Clinical proteomics is still a work in progress, and to improve approaches and throughput, we’ll need a lot of incremental steps,” said Steve Carr, senior director of the Proteomics Platform. “The potential these methods have to impact patients — to lead to new therapeutic interventions or better use of existing therapies — is really exciting. And that’s already started happening.”
People and tools
The Proteomics team has been a part of the Broad since its founding in 2004. It now consists of about 35 scientists including research associates, postdoctoral fellows, and staff scientists with decades of experience. The team pursues their own research projects and also closely collaborates with other Broad scientists from the earliest stages of their experiments. These projects range from development of new technologies to hypothesis-generating exploratory or hypothesis-driven translational studies. This work has led to more than 150 papers in the last five years including uncovering a protein involved in blood cell development, tracking how proteins travel between organs, and finding protein biomarkers for tuberculosis, type 2 diabetes, bacterial pneumonia, and various cancers.
Carr is tirelessly enthusiastic about proteins. He and other proteomic experts argue that while the genome encodes the cell’s instructions, the proteome provides a valuable snapshot of what is actually happening in the cell. Proteins carry out most biological processes, from shuttling nutrients around the cell to controlling the rate of nearly every biochemical reaction.
The main protein-analysis tool is mass spectrometry, often called “mass spec” for short. It allows scientists to identify proteins by first vaporizing samples under the influence of high voltage, creating charged ions. The mass spec instrument measures these ions as well as ion fragments induced to form inside the mass spec. Scientists analyze the resulting mass spectra to deduce the proteins’ identity.
Signs of disease
Of all the biological fluids scientists could study with mass spec to learn about disease, blood is among the most complex. It contains more than 10,000 different proteins that differ dramatically in concentration, as well as several abundant proteins that make detecting scarcer proteins, sometimes one trillion times less abundant, challenging. But in 2017, Broad Proteomics researchers developed a method that identifies more proteins in blood than other comparable methods, a major step toward making proteomics useful in the clinic. Mining the full range of blood proteins is critical to finding biomarkers that can give clinicians what McBean lacked — a diagnosis before serious disease develops.
The method, developed by platform research scientist Hasmik Keshishian and her colleagues, focuses on the colorless, cell-free part of blood called plasma, where most of the blood’s proteins reside. The team designed their method to detect trace levels of proteins, which other methods have struggled to achieve.
With their approach, published in Nature Protocols, the team detected about 4,500 proteins in each of 16 patient samples, and identified more than 300 potential biomarkers of heart attacks. Keshishian said the method can find about half of the available proteins in plasma — more than any other mass spec-based technique.
Although they honed their protocol for plasma, the team says others could adapt it to measure proteins in urine or cerebrospinal fluid. Keshishian also hopes to see other improvements in sample processing and analysis technologies that could increase the throughput for plasma proteomics without reducing the total number of proteins detected.
“Plasma is still the biofluid of choice for disease biomarkers,” she said. “Any technology that helps us detect them will greatly impact the discovery of biomarkers for any disease.”
A window into the cancer cell
Studying proteins can also illuminate the biological pathways that cause a disease or lead to drug resistance — pathways that scientists could then precisely target with new drugs.
The Broad Proteomics team is part of the Clinical Proteomic Tumor Analysis Consortium, a nationwide group of researchers formed by the National Cancer Institute to use proteogenomics — the integrated study of proteins, DNA, and RNA — to learn about the molecular basis of cancer.
Through this effort, Broad researchers recently discovered several potential drug targets for lung adenocarcinoma, the most common lung cancer. They have also identified subtypes of breast cancer that could potentially be more easily treated than others, and subsets of lung squamous cell carcinoma tumors that are particularly adept at evading the immune system.
Shankha Satpathy, a senior group leader in the Proteomics Platform who helped lead these projects together with Michael Gillette, also a senior group leader, hopes this work will lead to a future in which a patient receives genomic and proteomic data as readily as they do standard imaging results at their oncologist’s office. “You could then tell the patient the mutations they have, but we’d also be able to tell them if the mutations are having an impact on signaling in their tumor,” Satpathy said. “Using all these different levels of information in real time, we could come up with better treatment options.”
Personalized vaccines
Other scientists in the platform share Satpathy’s vision for proteomics as a key part of precision medicine, specifically for the design of cancer vaccines. Cancer vaccines are a new way of stimulating a cancer patient’s immune system to attack tumors and have shown promise in clinical trials. These vaccines contain key proteins, or peptides, that are found on the surface of a patient’s own tumor. The vaccines then present those peptides to the immune system to trigger specific immune responses.
Determining which peptides from a patient’s tumor to use in a cancer vaccine is painstaking, but proteomics promises to make this process more efficient, says Susan Klaeger, a mass spectrometrist who spent several years as a postdoctoral fellow and staff scientist with the platform before joining Genentech in 2022.
Ultimately, Klaeger says, cancer patients could provide labs with tumor tissue. Scientists would use mass spectrometry to determine which peptides are on the tumor’s surface, and figure out which of those peptides would elicit an immune response — and thus be useful components of a cancer vaccine.
“Cancer immunotherapy has been revolutionary, and proteomics can make a huge contribution because we look at proteins and peptides directly,” Klaeger said. “These are the biomolecules that actively function in the body, so getting a better understanding of what they do can really help us design therapies.”
Klaeger and colleagues have made progress towards this goal. In 2019, for example, they showed that a machine learning model could help generate rules to predict which peptides are presented on tumor surfaces. Other work continues in the proteomics group with collaborators across the Broad community, led by research scientist Jenn Abelin and team members.
But there’s still a long way to go before a proteomics approach can be used to design and optimize cancer vaccines for clinical use. It can take months for researchers to go from a patient’s tumor to peptides to that patient’s vaccine.
“For high throughput application in the clinic, we really do need robust, sensitive instruments that are easy to run and to set up before this can be more like an FDA-approved diagnostic test,” Klaeger said. “But developments are going in that direction — it's really an exciting time.”
Real-time response
Satpathy, Gillette, and their team recently showed how proteomics can prove useful in a clinical setting, by developing and deploying a method that can analyze proteins from tumor biopsies. Typical proteomic methods require five times more tissue than an average biopsy provides, so in 2020 the researchers came up with an approach to extract protein data from a single biopsy taken with a needle only a millimeter and a half wide.
The method could make it possible for doctors to learn more from a biopsy sample, such as whether a tumor will respond to certain drugs, which could help them make better treatment decisions sooner.
Using this approach, the Broad researchers collaborated with oncologists at the Baylor College of Medicine to study tumors from breast cancer patients undergoing chemotherapy that targeted the ERBB2 protein. They found that just 48 hours after treatment, ERBB2 was less active and less abundant in patients whose tumors ultimately disappeared completely — signs that the treatment was working.
“We now know that within 48 hours of therapy, rather than two or three months, we can actually get a readout of a drug’s impact on a tumor,” said Gillette, who is also a pulmonary and critical care medicine attending and associate physician at Massachusetts General Hospital.
Some patients in the study did not respond to treatment. To understand why, the team used proteogenomics to develop mouse models of cancer that had a similar genetic makeup to the tumors of those patients. They used the lab animals to test hypotheses about how the tumors resisted chemotherapy, and found that some showed higher levels of gel-forming proteins called mucins, a response they could target with other therapeutics.
“I’m hopeful that proteomics can help us make important strides in guiding precision therapy, identifying evidence of resistance and redirecting treatment strategies much earlier than is possible now,” Gillette said.
Gillette and his colleagues have started applying their method to larger clinical trials; a recent study published in Cancer Discovery, for instance, revealed markers of resistance to chemotherapy. Satpathy hopes that insights gained from proteomic analysis of patient samples will compel researchers to design clinical trials more deliberately.
“Many clinical trials are being done to study new cancer treatments, but investigators often don’t think about collecting biopsy samples to power these analyses,” Satpathy said. “If we can be thoughtful about subsequent studies, that would open up all kinds of possibilities.”
Future clinics
To mine the full potential of clinical proteomics, the researchers say they are working on new approaches to extract even more data from patient samples. The Proteomics Platform is continuing to develop methods such as MONTE (Multi-Omic Native Tissue Enrichment), which can analyze a larger fraction of the proteome including proteins that have been enzymatically tagged with phosphate, acetyl, and ubiquityl groups. These proteins include some that are more active in the cell and targetable by drugs.
“As we gather more data we may learn more and more about the ways in which these post-translational modifications are important to cellular signaling,” Gillette said. “The ability to look at the interplay between different modifications, while a nascent field, is potentially really powerful.”
The team is also partnering with flow cytometry and imaging teams at the Broad to couple highly sensitive mass spectrometry with advanced microscopes to study and map proteins to specific areas of tissue, and even analyze proteins from individual cells. They also hope to find a way to process samples that are chemically fixed, embedded in wax, and stored in banks, sometimes for decades.
“There are many tens of thousands — if not hundreds of thousands — of these samples from patients sitting in freezers around the world, with good clinical information and data about disease outcomes,” Carr said. “These are a huge, largely untapped resource for clinical proteomics.”
Gillette is excited about the field’s potential to impact patient care, but emphasizes that scientists need to integrate proteomics with the study of other molecules to capture the complex interactions between DNA, RNA, and proteins.
“If you’re trying to impact disease in patients and you leave out any level of biology, you’re going to miss really important things,” Gillette said. “Proteomics won’t replace genomics. There’s room for everyone to be working together — that’s how we arrive at comprehensive molecular portraits that are much more likely to inform our understanding of disease.”
Carr adds that proteomics is playing an increasingly important role in disease-related research. “Proteomics is now on the critical path for many, if not most, biological and clinical studies aimed at understanding disease and improving human health,” he said. “The future is very bright for our rapidly evolving field.”
The research discussed in this story was supported in part by the National Cancer Institute Clinical Proteomic Tumor Analysis Consortium, the National Institutes of Health Molecular Transducers of Physical Activity Consortium, and Dr. Miriam and Sheldon G. Adelson Medical Research Foundation.