NEW YORK – Researchers at Pacific Northwest National Laboratory have developed a single-cell proteomics workflow capable of reproducibly quantifying 1,000-plus proteins per cell.
Detailed in a study published this month in Molecular & Cellular Proteomics, the work represents one of the most in-depth characterizations of single-cell proteomes thus far and highlights how key parts of the mass spec workflow can be optimized for single-cell work.
The effort builds on previous research by the PNNL team, which last year published a study in Analytical Chemistry detailing what they call their Boosting to Amplify Signal with Isobaric Labeling (BASIL) strategy.
The method uses isobaric labeling to enable quantification of extremely low abundance targets. It builds off of the TMT Calibrator method originally developed by Proteome Sciences, inventor of the TMT isobaric labeling reagents. In that method, TMT-labeled peptides from the sample of interest are mixed with TMT-labeled peptides from another sample source where the target proteins are produced in higher abundance. By including the second supplementary sample source at a higher concentration than the sample of interest, researchers are able to ensure that even analytes present in low abundance in the sample of interest are present at high abundance in the overall sample, making them more likely to be fragmented and detected by the mass spec.
In 2017, the technique was adopted by researchers at Northeastern and Harvard Universities and applied to single-cell proteomics.
PNNL's BASIL method likewise uses TMT to label a carrier sample to allow for quantification of very low abundance proteins.
In the MCP study the researchers present a refined version of the approach that they termed iBASIL. The method differs from the original BASIL technique in that it optimizes the amount of carrier versus sample as well as the mass spectrometer's automatic gain control settings. In an analysis of 104 acute myeloid leukemia cells, the researchers were able to identify 2,500 proteins per cell. For roughly 1,500 of those proteins, they were able to quantify them in 70 percent of 104 cells the looked at. They were able to collect quantitative data in 100 percent of the cells for 900 of those 2,500 proteins.
Using that data they were able to separate the three AML cell lines comprising their 104-cell sample set into distinct clusters and observe differences in protein expression that correlated with the lines' known driver mutations.
The work demonstrates that relatively in-depth proteomic analysis can be done at the single-cell level with high reproducibility, said Tao Liu, a senior scientist at PNNL and senior author on the study.
He added that while single-cell proteomics is a fast growing area, much of the early data produced by such efforts has been low quality.
"In the early stages, you are just happy that you can identify some proteins," he said. "But now we are trying to use it for real biological interpretation, and so you have to have a higher bar in terms of quantitation quality."
Liu added that missing values have been a common issue for single-cell data, as they are for quantitative proteomics experiments generally.
"Basically, you don't get a signal from a lot of your sample channels," he said. "You have signal here and there, but it isn't consistent. And what causes this is basically you are not injecting enough ions into the mass spec to get high-quality measurements. So you have basically noise-level signal from the sample channels and then you try to make sense of it."
To address this problem, the researchers looked at two things — the ratio of carrier cells to sample cells used in an experiment and the mass spec's automatic gain control and ion injection settings, which control the flow of ions into the analyzer.
In the first case, the goal is to use enough carrier cells to trigger measurement of as many proteins as possible but not so much that the carrier cells overwhelm the cells from the sample actually being analyzed.
"We view that as a common problem in the single-cell field," Liu said. "We felt like it was an urgent task to lay it out there so people would know it was an issue."
In the case of the automatic gain control and ion injection settings, the researchers found that increasing those parameters to allow more ions to enter improved proteome coverage as well as the quality of their quantitative data.
The downside of these adjustments is an increase in cycle time, which means fewer measurement cycles in a given analysis, which can reduce the depth of analysis.
"That's why you have to balance [the factors," Liu said, though he added that in the case of single-cell studies "most of the time you are limited by signal, not the complexity of the sample. That is why it is very effective to increase the [automatic gain control settings] in this case."
The PNNL researchers looked at two widely used instruments, Thermo Fisher Scientific's Orbitrap Lumos and Q Exactive. Liu said, though, that researchers should optimize their settings for their particular instrument and sample.
They also explored whether quantifying the cell proteomes at the MS3 level as opposed to the MS2 level improved their data. Quantification at the MS3 level adds another level of ion isolation and fragmentation, which eliminates precursor interference challenges that reduce the accuracy of quantitative data from isobaric labeling experiments. It also, however, increases the instrument's duty cycle, which reduces depth of coverage. In line with this, the researchers found that while MS3-level quantification increased the dynamic range of their analysis, it reduced the number of proteins quantified.
Liu said he expected that the RTS-MS3 method, which uses real-time database searching to prioritize ions for MS3 fragmentation, would improve this situation. Developed by Harvard University researcher Steven Gygi, the approach has enabled his lab to reduce precursor interference without any trade-off in depth of coverage.
The technique was developed for use on Thermo Fisher's new Orbitrap Eclipse Tribrid, which Liu said his lab does not currently have. However, he said he and his colleagues are working with Thermo Fisher demoing the approach on the new instrument and he believes it will further improve their single-cell analyses.