Quantitative Label-Free Single-Cell Proteomics on the Orbitrap Astral MS

Abstract

Single-cell proteomics by mass spectrometry (scp-MS) holds the potential to provide unprecedented insights into molecular features directly linked to the cellular phenotype while deconvoluting complex organisms into their basic building blocks. Tailored sample preparation that maximizes the extracted amount of material that is introduced into the mass spectrometer has rapidly propelled the field forward. However, the measured signal is still at the lower edge of detection, approaching the sensitivity boundary of current instrumentation. Here, we investigate the capacity of the enhanced sensitivity of the Orbitrap Astral mass spectrometer to facilitate deeper proteome profiles from low-input to single-cell samples. We carry out a comprehensive data acquisition method survey to pinpoint which parameters provide the most sensitivity. Furthermore, we explore the quantitative accuracy of the obtained measurements to ensure that the obtained abundances are in line with expected ground truth values. We culminate our technical exploration by generating small datasets from two cultured cell lines and a primary bone marrow sample, to showcase obtainable proteome coverage differences from different source materials. Finally, as a proof of concept, we explore protein covariation to showcase how information on known protein complexes is captured inherently in our scp-MS data.

Keywords: ASTRAL; Quantitative accuracy; single-cell proteomics.

Graphical abstract

Fig. 1

Evaluating the performance of Orbitrap Astral for limited input proteomics.A, identified peptide amount (n = 3) at 65 samples per day throughput. Number notes the average number of peptides found. B, Violin plots of the CVs on non-normalized peptide abundances on MS1 level. t test was used to assess the significance between the CV distributions. ∗∗∗∗ notes a p-value ≤0.0001. C, Histogram showing the identified peptide number relative to the log2 transformed peptide abundances. A 16-min gradient (75 samples per day) was used.

Fig. 2

Assessing the utility of FAIMS for Orbitrap Astral with wide-window DIA.A, Scatter plot showing the relative changes of proteome coverage with and without FAIMS with different DIA acquisitions methods where the injection time on MS2 level is increased and the scan cycle time is compensated by doubling the isolation window. The shortest injection time (3 ms) is set as the control value in all cases. B, barplot showing the absolute proteome coverage related to (A). Bar height represents mean proteome coverage, and the points represent individual measurements. C, Violin plots of CVs on MS1 and MS2 levels for the 250-pg sample. A 16-min gradient (75 samples per day) was used.

Fig. 3

Optimizing Astral parameters for quantitative limited-input proteomics.A, Barplots showing the protein groups and peptides identified for 250pg input with different MS2 (Astral) injections times. B, similar to (A) but with different injection times on the MS1 (Orbitrap) level. C, method optimization for single-cell input. Barplots show identified protein groups and peptides with different injection times on MS2 (40 and 60 ms) and MS1 (100 and 200 ms). All runs carried out at a throughput of 65 samples per day (∼22 min run-to-run).

Fig. 5

Evaluating quantitative accuracy at single-cell input level.A, Violin plot of peptide level CV distributions with different input levels. Only the peptides that are found in all the loads with no missing values are used (n = 1064). B, Violin plot of peptide log2 transformed abundances. Median value is indicated by the dot and the lines represent the 0.05 and 0.95 quantile boundaries. Numbers note the numeric median value. For both (C and D), 100 pg is set as a reference for the other inputs (50, 40, 30, 20, and 10 pg) as relative targets. C, CDF plots of the absolute error. D, histograms showing the relative error distribution based on MS1. The numbers denote the peaked full-width at half maximum (2.634σ) with the uncertainty calculated from three replicates. Replicate 2 data is shown. E, error distribution is the same as in (D) but for 20:40 pg and 10:20 pg comparisons. Spectronaut was used to obtain peptide abundances. All runs were carried out at a throughput of 65 samples per day (∼22 min run-to-run).

Fig. 6

Evaluating the peptide ID propagation quantitative accuracy.A, schematic depicting the ID propagation to the 50pg reference samples from libraries from different inputs with respect to the reference. B, Bar plot of the total number of peptides identified without and without libraries. Arrows and numbers denote the fractional increase in peptide identifications. C, a CDF plot of the absolute error for the 1:2 comparison (50: 100 pg). D and E, tile plots showing the fraction of peptide identifications below 25% error (D) and above 50%. E, numbers represent the fraction from the total number of peptides added and the color represents the value size. The same data is used as for Figure 4. Spectronaut was used to obtain the peptide abundances.

Fig. 7

Profiling protein covariation at single-cell resolution.A, Histograms showing the distribution of protein groups obtained from different types of cells HEK293 culture embryonic kidney cells, U937 culture monocytes, and primary BM CD34+ cells. B, scatter plot of single-cell sample intensity (HEK293) dependency on the isolated cell size. Forward side scatter area (FSC-A) is shown on the x-axis, and the y-axis reflects the median log2 transformed sample intensity. The color denotes the number of protein groups identified in each cell, and the line is a linear fit of the data with the confidence interval in grey shade. C, upper diagonal correlation of the map of proteins from the HEK293 cells. The map is ordered by hierarchical clustering and stratified into specific groups based on the dendrogram (not shown). Color represents Spearman correlation value. Red squares and arrows note the section used to create (D and E). D and E, cutouts of the correlation map showing the capture covariation of the MCM complex and the proteasome subunits. All runs carried out at a throughput of 65 samples per day (∼22 mins run-to-run).

Fig. 4

Accuracy comparison of MS1 and MS2 level based quantification.A, histograms showing the relative error distribution on MS2 (top) and MS1 (bottom) levels. 50 pg is set as reference, and the ratios correspond to comparison with 100, 150, 200, 250, and 500pg. The numbers note the mean standard deviation of the error distribution from three replicates. B and C, cumulative density function plot showing the absolute error trend for different ratios. The embedded bar plot shows the fraction of peptides below and the absolute 25% error. All runs were carried out at a throughput of 65 samples per day (∼22 min run-to-run).