NEW YORK – Recent developments in single-cell proteomics are pushing the field toward higher levels of throughput and coverage while also bringing together two formerly distinct modes of analysis.
Improvements in informatics and instrumentation along with workflow innovations that allow for multiplexing of data-independent acquisition mass spectrometry experiments have made it possible to quantify around 1,000 proteins per cell with run times of 10 minutes.
In the March issue of Molecular Systems Biology, a team led by researchers at the Max Planck Institute of Biochemistry detailed the performance of Bruker's timsTOF SCP, a version of the company's timsTOF system optimized for high-sensitivity analyses of very small sample sizes, down to single cells. Bruker launched the system in June 2021.
The lab of Matthias Mann, head of the department of proteomics and signal transduction at Max Planck, initially published a bioRxiv preprint covering the development and use of the system in January 2021. Since then, the researchers have continued to advance their single-cell work, with key improvements on the informatics side of the equation, said Andreas-David Brunner, first author on the preprint and MSB paper and formerly a graduate student in Mann's lab. (Brunner has since taken a position as a scientist at drug company Boehringer Ingelheim.)
One of the most significant improvements came from the integration of the DIA-NN analysis software into the lab's single-cell proteomics workflow, Brunner said. Developed in 2019 by researchers at the Francis Crick Institute and the University of Cambridge, DIA-NN uses a combination of signal correction strategies and deep learning algorithms to better interpret the complex spectra generated by DIA mass spec methods, reducing interferences and improving the confidence of peak identifications. The approach was originally developed to enable analysis of DIA experiments using very short liquid chromatography runs, but it has since become broadly used for DIA experiments generally, including single-cell proteomics experiments based on DIA mass spec.
DIA-NN "uses a lot of machine learning that really helps with signal deconvolutions in the context of DIA data and also some algorithms that correct for imperfections in LC-MS/MS data, and it turned out to work immensely well," Brunner said, noting that using DIA-NN he and his colleagues were able to quantify around 1,800 proteins in single cells.
The MSB authors noted that the performance of DIA-NN was also boosted by the fact that the timsTOF SCP instrument generates ion mobility data for the ions being analyzed, which provides another dimension of data the software can use to make peptide identifications. Bruker offers the DIA-NN software package for use on the timsTOF SCP as part of the PASER 2022 proteomics software it released at the end of February.
Shortly after release of the MSB paper, a team of researchers including Markus Ralser, the developer of the DIA-NN software, published a bioRxiv preprint presenting a workflow for multiplexing DIA called plexDIA and demonstrated its use for both bulk and single-cell proteomic analyses.
Brunner, who was not involved in the study, called plexDIA "a very interesting development" for single-cell proteomics.
To date, researchers have typically taken one of two approaches to single-cell proteomics. One uses isobaric labeling to boost mass spec sensitivity enough to measure large number of proteins at the single-cell level. In these experiments, researchers label peptides from both the single-cell sample of interest and another, larger sample source (such as dozens or hundreds of cells of equivalent type) termed the "carrier proteome." By including the carrier proteome sample, they are able to ensure that even analytes present only at low abundance in the single-cell samples are present in relatively high abundance in the overall sample, making them more likely to be fragmented and detected by the mass spec.
The other common approach is label-free DIA-based analysis like that presented by the Mann lab in the MSB paper.
Both have their advantages and downsides. With isobaric labeling, for instance, researchers must be careful not to use too much carrier proteome in their experiments or else quantitative accuracy will suffer. Additionally, isobaric labeling can be expensive and integrating data generated across large sample sets can be challenging.
On the other hand, use of isobaric labeling increases throughput as researchers are able to measure a dozen or more individual cells per experiment. Throughput has been a limitation for label-free DIA approaches, Brunner noted.
"We had throughput [initially] of like 40 cells per day, but you want to have thousands of cells being measured quickly," he said. "We thought DIA multiplexing might be a thing, but we were not there yet."
The plexDIA method combines elements of these two flavors of single-cell proteomics, using a labeling approach to enable multiplexed DIA measurements. In the preprint, the researchers used the technique to run single-cell experiments in triplicate and with LC-MS/MS run times of around 10 minutes, which would, in theory, enable analysis of more than 400 cells per day.
"For sure this could be the way to go for single cell, multiplexing them and running them in short [LC] gradients," Brunner said.
PlexDIA uses three-plex mTRAQ mass tags from Sciex to label samples prior to DIA analysis, allowing researchers to keep track of which cells are producing particular peptide spectra. Multiplexing cells adds to the complexity of the spectra that must be deconvoluted to make peptide identifications, but by using a DIA-NN module designed to use expected patterns in the data such as the known mass-shifts created by the mass tags, the researchers were able to confidently identify and quantify upwards of 1,000 proteins per cell.
"One of the beautiful things about labeling is you are building known, regular structure into your data," said Nikolai Slavov, director of the single-cell proteomics center at Northeastern University and senior author on the plexDIA preprint. Slavov is one of the inventors of the SCOPE-MS method for single-cell proteomics, a commonly used carrier proteome approach. "And that regular structure can actually enhance your data interpretation."
In fact, he said, he and his colleagues observed improved quantitative accuracy and precision using plexDIA compared to label-free DIA.
Like Brunner, Slavov stressed the importance of improving the throughput of single-cell proteomics experiments.
"It's not about analyzing one cell or a handful of cells," he said. "I think that for single-cell proteomics to really drive biology it has to get into the realm of many thousands of cells analyzed."
Slavov said the researchers' experience using plexDIA for bulk analysis made him confident that they can significantly boost single-cell multiplexing beyond the three-plex demonstrated in the preprint.
"From the bulk samples we know that we can quantify 250,000 precursors in a single experiment," he said, adding that with single cells they typically generate between 10,000 and 20,000 precursors per cell.
"That means that we can do at least 10-plex before we get to the density of precursors we have already measured in the bulk analysis," he said. "So I have very high confidence that we can substantially increase the plex for single-cell proteomics. For bulk [assays] I also have confidence, but I don't have proof in the same way."
Slavov said that he and his colleagues are currently working on designs of additional labels for use in higher plex experiments and that he plans to commercialize them in some fashion, though he did not specify a particular path to market.
"We would certainly like to commercialize them," he said. "They have to be accessible to everybody who would like to buy them. If the demand for them increases, I want to run a lab, not a manufacturing facility for mass tags."