Individual cells’ proteins vary. Single-cell proteomics can now show how

Scientists have long had the tools to identify the genes being transcribed in a single cell. Those measurements can uncover cell-to-cell differences that reveal clues about the underpinnings of disease and lead to better treatments. Now, researchers to want to analyze a single cell’s proteins too. They hope the added information will reveal facets of biology that can’t be seen with nucleic acids alone.

But proteins are a tougher challenge than nucleic acids. To detect low levels of nucleic acids, researchers can use long-standing methods for boosting even the faintest of genetic signals by making copies. There’s no analogous method for copying proteins.

People have been able to detect a handful of proteins in individual cells using fluorescence- or antibody-based methods. Now, the goal is a full readout of all the proteins being made in a cell. No technology can do that yet, but mass spectrometry comes closest. And as the sensitivity of mass spectrometers has improved, researchers have realized that they can finally achieve single-cell proteomics.

Nikolai Slavov, a systems biologist at Northeastern University, is a pioneer in single-cell proteomics. As a graduate student and postdoc, he analyzed DNA and RNA in single cells. He wanted to do similar measurements of proteins, but “it went against the grain to think that mass spectrometry would have enough sensitivity to analyze single cells at high scale,” Slavov says.

When he started his own lab at Northeastern University in 2015, Slavov finally got the chance to try single-cell proteomics. He built on protocols developed for analyzing proteins in large samples to study much smaller ones. “From the very beginning, it gave encouraging results,” he says. “When I look back on those encouraging results, they were very modest. But we saw there was signal there. We also knew that what we were doing was primitive, and we could improve it tremendously.” In 2017, he published the results of that first attempt on a preprint server—and later in the peer-reviewed journal Genome Biology—and it immediately got attention, Slavov recalls.

In the years since that paper, Slavov and others have honed older methods—and developed new ones—to make it easier for researchers across many disciplines to probe the proteins in cells. Advances in sample preparation and improvements in the sensitivity of mass analyzers, including Orbitraps and high-resolution time-of-flight instruments, are driving rapid growth in the field. Already, scientists are using single-cell proteomics to answer long-standing questions in developmental biology, neuroscience, cancer biology, and other areas.

Matthias Mann, a proteomics expert at the Max Planck Institute of Biochemistry and the Novo Nordisk Foundation Center for Protein Research at the University of Copenhagen, was initially skeptical that single-cell proteomics would be possible. “About 5 years ago, nobody really thought single-cell proteomics would work,” Mann says. But as the sensitivity of instruments improved and his group developed more sensitive workflows, he was persuaded that single-cell proteomics was indeed feasible, he says.

Researchers are interested in single-cell analysis of proteins because analyzing tissues as a whole—what scientists call a bulk measurement—often hides biological differences. Most tissues, whether healthy or diseased, are a hodgepodge of cell types, making results from those bulk measurements difficult to interpret.

Bulk tumor samples, for example, “contain normal cells, tumor cells, matrix cells, and a lot of immune cells, and they’re all mixed together,” Slavov says. “If we measure a high abundance of interferon γ, this may be because that particular sample had more immune cells or because the immune cells expressed more interferon γ.”

Bulk measurements make sense when studying cell lines that have been grown in a lab and feature mostly identical cells. But even in those cases, the cells can be in different states. “There is always heterogeneity,” Slavov says. “And in some cases, that’s biologically interesting and meaningful.”

Scaling down

When Slavov embarked upon his single-cell proteomics experiments, he purposely restricted himself to using commercially available equipment. “There have been points when we had to decide between going for the highest performance possible or going for something that others would be able to reproduce,” he says. “I always chose to go with something that others can reproduce.”

Slavov’s method, which he calls SCoPE-MS, uses 100–200 cells as a carrier proteome for the single cells he’s interested in (Genome Biol. 2018, DOI: 10.1186/s13059-018-1547-5). The proteins from each cell and the carrier proteome are digested into peptides and labeled with different tags. All the proteins in a single cell have the same mass tag, and all the cells providing the carrier proteome have the same mass tag. During mass spectrometry, the tags fragment and release reporter ions that reveal which cell a peptide came from. One function of the carrier proteome is to increase the likelihood that any losses caused by sticking to chromatographic columns or other surfaces come from the carrier proteome. The carrier proteome also increases the signals so that low-abundance peptides can be sequenced and identified. The reporter ions enable the proteins to be quantified and the cells they came from to be identified, while the carrier proteome allows Slavov to identify more proteins than would be possible with just the individual cells.

Other researchers are building on Slavov’s work to help establish best practices for single-cell protein analysis. A team led by Genentech’s Christopher M. Rose and Bernhard Küster of the Technical University of Munich performed controlled experiments to determine how the amount of the carrier proteome affects the accuracy of the protein readout (Nat. Methods 2020, DOI: 10.1038/s41592-020-01002-5). It found that as more cells are used for the carrier proteome, more peptide ions need to be sampled to ensure that the proteins from the single cells can be detected and accurately quantified. The researchers recommend using about 20 cells for the carrier proteome. If more cells are needed, the number shouldn’t exceed 100.

“People were operating single-cell proteomics experiments the same way they had been doing regular proteomics experiments,” Rose says. “They had been not letting the instrument sample enough to get that quantitative signal higher when they had a big carrier proteome. The goal was to give people guidance on how to get better data with single-cell proteomics.”

Max Planck’s Mann has worked with technology developers at Bruker Daltonics and Evosep to develop a workflow that enables the analysis of proteins in single cells without the need for a carrier proteome. They achieved this with a combination of low-volume sample processing and improved MS sensitivity. Mann and his collaborators were able to identify an average of 843 proteins from individual HeLa cells (bioRxiv 2021, DOI: 10.1101/2020.12.22.423933). The work has been published as a preprint, so it has not yet been peer-reviewed. Since then, they say, they’ve improved their methods to be able to identify about 2,000 proteins in a single cell.

Other researchers have been focused on ways to streamline the sample preparation process. In conventional bulk proteomics experiments, the proteins must be released from the cells and then broken up by enzymes into peptides, which in turn are labeled with tags that let researchers quantify them. Finally, those peptides are analyzed by liquid chromatography/mass spectrometry. At each step, some of the sample is lost.

“If you start with a bunch of sample, you don’t care that you lose a few nanograms as you take it through the workflow,” says Ryan T. Kelly, who’s working on single-cell proteomics at Brigham Young University. “But that’s your whole sample at the single-cell level.”

Researchers have found multiple ways to overcome the sample-loss challenge. Most rely on fewer processing steps and eliminating some of the surfaces to which proteins adhere during processing. And multiple groups have developed methods for processing cells in nanodroplets instead of in conventional vials or wells on microtiter plates.

Nanodroplets’ small processing volumes mean the sample’s concentration is much higher than in conventional proteomics methods, enabling scientists to get lots of information from a tiny cell. “In our first study, we were identifying more proteins from 10 cells than were previously identified from 5,000 cells,” says Kelly, who with Ying Zhu and others at Pacific Northwest National Laboratory developed a method they call nanoPOTS, short for “nanodroplet processing in one pot for trace samples” (Nat. Commun. 2018, 10.1038/s41467-018-03367-w).

Most workflows for single-cell proteomics, much like those for bulk proteomics, incorporate some form of liquid chromatography to separate the digested, labeled peptides. Peter Nemes, who develops instruments and methods for single-cell analysis at the University of Maryland, College Park, relies on capillary electrophoresis (CE) instead. That method, in which electric fields are used to separate molecules in electrolytic solutions in a narrow tube, is ideal for the small volumes involved in single-cell measurements, Nemes says. It also has other advantages, such as sharp peaks and the ability to easily switch between types of analytes, such as small molecules and proteins.

Jonathan V. Sweedler, an analytical chemist and neurochemist at the University of Illinois Urbana-Champaign, uses methods that allow him to achieve higher throughput and even analyze an individual cell by multiple methods. He uses matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) to analyze peptides directly from cells without going through the typical proteomics digestion. MALDI-MS sacrifices the outstanding performance of the more commonly used electrospray MS but means he can study more cells in a single experiment. “One of the things we gain is the ability to look for rare cells,” Sweedler says. For example, by using MALDI-MS to study cells from human islets, clusters of pancreatic cells involved in producing insulin, his team has been able to identify cells that are processing proglucagon into oxyntomodulin, a peptide that has been associated more with the gut than with the pancreas. He’s also found that there’s enough material left in cells after his analysis to study them with other techniques, including electrospray CE/MS or even single-cell transcriptomics.

Getting at biology

Researchers are using single-cell proteomics to answer questions in a variety of biological fields, including developmental biology, cancer biology, basic cell biology, and cardiology.

Nemes is using single-cell proteomics to study developmental biology in Xenopus laevis, a species of frog. His students target individual cells in 16-cell frog embryos by inserting a sharp capillary into a cell to remove material, which they then use for a proteomic or metabolomic analysis. “This platform works so well and so easily that we can actually teach high school students to conduct the measurement,” Nemes says.

They collect such a small fraction of the cell that they can take multiple samples without harming the cell or the embryo. In fact, embryos from which his students extract material grow into larvae and tadpoles that behave normally. “This is exciting because that means that now you can not only link early chemistry to developing cells and trajectories, but now you can also start talking about functional insights,” Nemes says. “To get functional insight into an organism, you have to have a functioning organism.”

Nemes and his colleagues can also do multiple analyses on the same cells. They have done proteomic and metabolomic measurements of individual X. laevis cells (Angew. Chem., Int. Ed. 2021, DOI: 10.1002/anie.202100923). They detected about 150 metabolites, 57 of which they were able to identify, and 738 proteins.

They found differences in both proteins and metabolites between different cells in the embryo. “We discovered that there are small molecules that can reprogram the normal developmental trajectory of embryonic stem cells,” Nemes says. “And we actually found evidence that at the very early stages of development, the cells already have imprinted on them chemical differences.”

Alex Kentsis, a cancer biologist and pediatric oncologist at Memorial Sloan Kettering Cancer Center, is interested in using single-cell proteomics to study cancers that affect children and young adults. These cancers, such as leukemia, often have their origins in developmental processes that have gone awry. “Single-cell measurements are so exciting because they allow us to define precisely the cellular state that these tumor cells adapt,” he says. Such information can help identify how some cancers don’t respond to therapies.

Kentsis is using single-cell proteomic measurements to create a panel of targeted proteomics assays for measuring biological activity and cell states related to regulation and signaling in cancer cells. He calls the collection the Quantitative Cancer Proteomic Atlas.

Erwin Schoof, a proteomics expert at the Technical University of Denmark, is using single-cell proteomics to study cell differentiation in the blood system. He’s particularly interested in hematopoietic stem cells and cells called multipotent progenitors (MPPs). Both types of cells can differentiate to form any type of cell in the blood. But MPPs can divide only a limited number of times, whereas stem cells can continue to divide indefinitely. “We are really keen on figuring out what is different between the hematopoietic stem cell and the MPP,” Schoof says.

He already knows from previous work that not many genes or proteins are expressed at different levels in the two cell types. He thinks that previous studies weren’t working with pure enough cell populations to be able to pinpoint the differences.

For Schoof’s work, the ability to look at single cells is imperative. The stem cell and MPP cell populations are so small that “unless we can do single-cell proteomics, we are never going to be able to characterize the cell-specific proteome landscapes,” he says. He and his team first tried to do bulk proteomic measurements of mixtures of about 50,000 purified cells. They were able to identify between 5,000 and 8,000 proteins, but they couldn’t find which proteins were specific to the MPPs and stem cells. Schoof suspects that the key players are going to be proteins that control gene expression and are turned on in some cells but not in others.

Other researchers are using single-cell proteomics to understand how to tailor therapies. Jennifer Van Eyk of Cedars-Sinai Medical Center is gearing up to use single-cell proteomics to study how cardiac muscle cells differ across various forms of heart failure. Previous studies have shown that cardiac muscle cells experience different forces and are regulated differently according to where they are in the heart. But researchers don’t know what that means in terms of the proteins they make.

Cedars-Sinai is a large heart transplant center, so Van Eyk will have access to hearts removed from people undergoing transplantation. She wants to understand how the ratios of proteins that control muscle contraction change in different diseases. Diseased cardiac muscle cells often become larger, and they can grow in either length or width. “We don’t know at the protein level how that is different,” Van Eyk says.

Van Eyk wants to make sure that single-cell proteomic measurements she does are accurately and reproducibly quantitative. The contractile proteins that she’s interested in are highly regulated. “It’s hard to imagine a change of 10–20% of any of those proteins,” Van Eyk says. “If we have to be measuring an accuracy of 10–20% difference between the components, reproducibility becomes key.”

Eventually, Van Eyk wants to use these single-cell measurements in studies for developing personalized treatments for heart failure.

UIUC’s Sweedler, meanwhile, is pushing proteomics beyond single cells to single organelles. “Cells are not a bag of cytoplasm,” he says. “They have organelles.” In many cases, differences in the location, number, and contents of those organelles control cell function, he says.

Sweedler is using mass spectrometry to analyze individual dense-core vesicles isolated from the atrial gland in the California sea hare, Aplysia californica (Nat. Methods 2021, DOI: 10.1038/s41592-021-01277-2). The vesicles all look the same under a microscope, but his students found three distinct types characterized by differing ratios of peptides and lipids. He and his team now want to push the method further to work with smaller organelles, such as mitochondria.

Slavov is amazed to see the rapid growth in the number of researchers adopting approaches that many thought were out of reach not long ago. One sign of the surge in interest can be found in the single-cell proteomics conference Slavov has organized at Northeastern every year since 2018. About 60 people attended the first meeting. The most recent meeting, which had participants both in person and online, had more than 1,200 registrants. The attendees were a mix of technology developers and people who want to harness technology for biological applications.

“At this point, the field is so interdisciplinary that oftentimes a single lab will not have the expertise to drive both the biological investigations and the technological developments,” Slavov says. He hopes that meetings like the ones he organizes will help bring together people with biological questions and others with technical know-how.