Single-Cell Proteomics: A Different Story from Bulk Analysis

Where bulk cell analysis provides average values for proteins, genes, or metabolites, single-cell analyses uncover differences among cells within a tissue.

“The point is to determine if certain cell types or subsets of cells from bulk populations are responsible for creating the response that caused a disease, for example cancer,” says Matt Salem, Marketing Director at Namocell, which specializes in single-cell dispensing.

That knowledge provides insights into cellular and molecular cause-and-effect, and ultimately helps explain how complex disease states arise and persist.

“The goal is to get the most data granularity from a given study that allows us to explain the biology,” says Daniel Lopez-Ferrer, Ph.D., Director of Proteomics and Translational Research at Thermo Fisher Scientific. For example, studying the tumor microenvironment, which consists of multiple cell types, offers the possibilities of enhancing early tumor detection as well as deciphering therapeutic drug targets.

Single-cell analysis allows profiling of these individual cellular components, thereby uncovering huge chunks of the biological activity hidden by the tumor cell heterogeneity. “Understanding the different cell types and their functions is impossible with bulk analysis, which provides only the signal that something is happening and a readout representing the average cellular response.” Bulk analysis provides mean values, not just of characteristics of interest (e.g., a cell surface marker or other phenotype), but of artifacts and cellular “mistakes,” thereby obfuscating the contributions to these events by specific cells.

The inherent limitations of single-cell analysis arise from the microscopic sample size (a single cell) and the vanishingly small quantity of protein they hold: around 150 million copies for yeast and about 10 billion protein molecules in mammalian cells.

“Getting cells isn’t the problem,” says Lopez-Ferrer. “The issue is that to reach statistical power you have to analyze hundreds or thousands of cells. At about one hour per LC/MS analysis, a simple study can take ten days, and that’s assuming everything goes according to plan.” Moreover, human cells produce between 80,000 and 400,000 different proteins. Not all are expressed at once, and some are isoforms of the same basic molecule, but the “needle in a haystack” metaphor is appropriate. The best single-cell proteomic methods can quantitatively profile around 2500 unique proteins per cell at a throughput of about 300 cells per day.

Single-cell workflows

Single-cell workflows begin with the sample source. Solid samples are treated to detach and suspend cells in buffer, followed by separation through fluorescence-activated cell sorting (FACS). Single cells are dispensed into wells on a microtiter plate or into a chip-based analysis system, where they undergo lysis. Lysing agents that interfere with MS are removed, but since cleanup results in sample losses, researchers try to avoid them.

FACS, the method of choice for cell sorting, suffers from (sometimes very) high losses, requires trained personnel, and is expensive to own and operate, says Salem.

But cell sorting is not the only source of losses. “Sticky proteins are hard to separate by LC,” Salem adds, “so there are losses associated with that step as well. This creates further sensitivity issues or analysis bias based on which proteins get through. Sometimes you just can’t create enough ions from the proteins from a single cell.”

To address these sensitivity issues some protocols add unlabeled carrier proteins to the mix to mitigate losses associated with the small sample sizes. While these approaches help, advances in MS, particularly in ionization efficiency, will be needed to handle the small sample and working volumes.

“These workflows would not be possible without automated liquid-handling systems, which provide accurate, nanoliter-scale fluid manipulations, plus consistency and robustness,” Lopez-Ferrer says.

Early attempts at single-cell proteomics used matrix-assisted laser desorption-ionization (MALDI), a “soft” ionization technique that preserves labile molecular characteristics, and minimizes both sample losses and required sample preparation.

MALDI-based methods were first used in the 1990s and provided, through time-of-flight (TOF) detection, a functional single-cell assay, but these techniques suffer from three main drawbacks: MALDI does not separate peptides in the time domain or allow sequencing of more than a handful of proteins, and the variability of MALDI ionizations undermine the accuracy of quantitation.

The other MS method for ionizing proteins, electrospray ionization (ESI), more readily allows sequencing since it is easily integrated with separation methods (e.g., capillary electrophoresis). ESI requires extensive sample prep, however, which results in unacceptable losses.

SCoPE and nanoPOTS

Most single-cell methods these days use ESI and are based on some variant of single-cell proteomics by mass spectrometry (SCoPE-MS), a method developed by Prof. Nikolai Slavov at Northeastern University. SCoPE-MS uses either nano-LC or electrophoresis as the separation front-end. Labeled carrier proteins are added before the separation step to reduce losses due to interactions with the nano-LC column surface. The first MS scan provides identity through the molecular ion peak. Proteins of interest undergo fragmentation and further MS analysis. Normally only high-abundance peaks or those derived from readily cleaved peptide linkages were available for sequencing. SCoPE-MS overcomes this limitation by pooling identical (and identically tagged) proteins from multiple cells.

A more efficient implementation of the SCoPE method is nanoPOTS (Nanodroplet processing in one pot for trace samples), which was designed for analyzing small cell populations. NanoPOTS enhances sample/target efficiency and recovery by working in volumes below 200 nL. When combined with ultrasensitive LC/MS, nanoPOTS allows identification of between 1500 and 3000 proteins from as few as ten cells. The original paper  announcing nanoPOTS described a method for quantifying around 2400 proteins from single human pancreatic islets.

The separation step for single-cell MS is unremarkable for proteomic workflows. Lopez-Ferrer notes that small-bore columns provide the best sensitivity, “but that comes at the cost of robustness. We are working with down to 20 um ID columns, which they clog if you look at them the wrong way. Some users are okay with the pain, as long as they get the proteome depth they want, others will go with more routinely used nano-columns and they get very good results (1500 proteins quantified) while making sure the study gets done within less than a week.”

Addressing challenges

Single-cell proteomics is not for the faint-hearted. Much work is required for the technique to reach cost, sensitivity, robustness, reliability, and ease-of use requirements demanded by everyday research, biomanufacturing, and diagnostics.

“But before that happens the industry must address challenges with single-cell isolation and the validation of MS techniques and workflows. Plus, technicians in clinical settings will require more turnkey workflows. It currently takes two to three hours to run sixteen samples, and that needs to improve as well.”

Lopez-Ferrer says he hasn’t seen direct applications of either method in biomanufacturing, where single-cell analysis could assist in clone selection for monoclonal antibody production, “but we will definitely see this in the future as biopharma companies are very interested in these approaches.”