NEW YORK – According to a recent ChemRxiv preprint by researchers at Johns Hopkins University School of Medicine and Harvard Medical School, since the introduction of single-cell proteomics workflows seven years ago, the throughput of the average experiment has declined while the average cost has steadily risen.
This trend, the authors noted, runs counter to other single-cell technologies, which have seen dramatic increases in throughput along with declining costs, and raises questions about whether the field has overprioritized depth of proteomic coverage at the expense of other metrics.
Benjamin Orsburn, a researcher at JHU and senior author on the preprint, said that the trajectory of single-cell RNA sequencing experiments, which over the last decade have gone from analyzing thousands to millions of cells, led him to question why single-cell proteomic experiments still typically analyze on the order of hundreds of cells.
Such limited sample sizes, he said, make it difficult in many cases to generate meaningful biological insights from single-cell experiments.
Orsburn cited as an example his lab's interest in single-cell analysis of brain cells, noting that in its experience dissociating mouse brains for mass spec experiments, typically between one in 10 and one in 100 cells will be neurons.
"If you're interested in neurons, if you do 100 cells and statistically one to 10 of those are neurons, it makes no sense," he said. "Unless you're doing thousands of cells, you can't get to some of the cells you're interested in. I think for real biology, we need these high numbers."
The field, however, has largely focused on lower-throughput single-cell experiments aimed at producing the deepest possible proteomics coverage, Orsburn said. Because these experiments are commonly run on the newest, most expensive mass spec instruments available, the average cost of a single-cell proteomic analysis has also risen over time, he added.
Orsburn warned that this combination of high costs and low throughput could limit single-cell proteomics' uptake among biologists outside the world of cutting-edge mass spec and, consequently, the field's broader impact.
One of the major drivers of this trend toward lower-throughput, higher-cost experiments has been the increased uptake of label-free approaches as opposed to the multiplexed, isobaric-labeling-based methods that kickstarted the field almost a decade ago.
Early single-cell techniques like the SCoPE-MS method developed by Nikolai Slavov, professor of bioengineering at Northeastern University and director of the Parallel Squared Technology Institute (PTI), used Thermo Fisher Scientific's TMT isobaric labeling reagents to boost mass spec sensitivity enough to measure a large number of proteins at the single-cell level. This label-based approach also meant these experiments were multiplexed, measuring a dozen or more cells per run.
In the years following the introduction of SCOPE-MS and similar methods, advances in mass spec instrumentation and bioinformatics enabled label-free single-cell proteomic workflows. Ryan Kelly, associate professor of chemistry and biochemistry at Brigham Young University, said that these label-free workflows proved popular due to both their relative simplicity and their superior depth of coverage compared to label-based experiments.
Kelly said that he is a fan of both label-based and label-free approaches and that his lab uses both, but added that "if you are relatively new to the field, it is a no-brainer that you should start with label-free."
"With label-free, the prep is now super easy," he said. By contrast, "there are a million ways to screw up a TMT experiment at the single-cell level. It's really difficult to get it to all work well."
Karl Mechtler, head of the proteomics tech hub at Vienna's Research Institute of Molecular Pathology, echoed Kelly's comments regarding the challenges of label-based single-cell experiments, noting that label-free approaches produce deeper coverage and better quantitative results.
Slavov said, however, that he believes the challenges involved in label-based experiments can be overcome. He said that wider adoption of sample preparation approaches like the nPOP (nano-proteomic sample preparation) method developed by his lab could ease the difficulty of label-based methods while better optimization of the TMT process can improve the quality of data generated in these experiments.
Slavov's lab has also developed a method called plexDIA that combines labeling with mTRAQ mass tags with the data-independent acquisition mass spec approach commonly used in label-free experiments.
To date, the highest level of multiplexing published using the approach is five-plex, but Slavov said he and his colleagues have managed up to nine-plex in unpublished work. He added that they have higher plexes "in the works" and expect in the near future to be able to analyze as many as 1,000 cells per day with the method.
Slavov said his lab also has a paper in press at Nature Protocols using multiplexing with 32-plex TMT reagents to analyze more than 1,000 single cells per day.
Kelly said that despite the challenges of multiplexed single-cell workflows, he believes they are "the only path forward to 1,000 cells per day," which means greater focus is needed on optimizing and streamlining these approaches to make running them feasible for a wider range of labs.
He said that on the label-free side, his lab is working to optimize a workflow capable of running around 290 cells per day. On the TMT-based side, he said he expects his lab will "in short order" be at 1,000 cells per day.
Kelly said his lab is taking two approaches in its TMT-based work, one using 32-plex reagents on the Thermo Fisher Orbitrap Exploris and another using 16-plex reagents on the Thermo Fisher Orbitrap Astral. The first uses the higher resolution of the Exploris to enable the higher level of multiplexing, while the other leverages the higher speed of the Astral to compensate for the lower level of multiplexing. Both approaches allow for analysis of around 1,000 cells per day, Kelly said.
He noted, however, that throughput on the order of millions of cells as seen in single-cell seq experiments will likely remain out of reach for single-cell mass spec.
"The nature of [mass spec] analysis just does not scale the way sequencing-based approaches can scale," he said.
Biology could help compensate for this limitation, however. Kelly noted that any given cell has around 30,000 corresponding proteins for every RNA transcript. The low number of transcripts means sequencing experiments require large numbers of samples to account for the noise inherent in measuring such small numbers of analytes.
"It may be the case that we get more information from 10,000-scale single-cell proteomics than from million-scale single-cell transcriptomics because of that inherent noise," Kelly said.