Click here to close
Hello! We notice that you are using Internet Explorer, which is not supported by Xenbase and may cause the site to display incorrectly.
We suggest using a current version of Chrome,
FireFox, or Safari.
Angew Chem Int Ed Engl
2025 Aug 03;6445:e202510692. doi: 10.1002/anie.202510692.
Show Gene links
Show Anatomy links
Real-Time Eco-AI, Electrophoresis-Correlative Data-Dependent Acquisition with AI-Based Data Processing Broadens Access to Single-Cell Mass Spectrometry Proteomics.
???displayArticle.abstract???
Single-cell mass spectrometry (MS) offers unprecedented sensitivity for profiling cellular proteomes, yet widespread adoption is hindered by the cost of advanced instrumentation. Here, we broaden access to single-cell proteomics by combining capillary electrophoresis (CE), data-dependent acquisition (DDA) with electrophoresis-correlative (Eco) ion sorting, and artificial intelligence (AI)-assisted spectral deconvolution via CHIMERYS (Eco-AI). This "Real-Time Eco-AI" workflow was implemented on a custom-built CE platform coupled to a legacy hybrid quadrupole-orbitrap mass spectrometer (Q Exactive Plus). Despite slower scan speed, lower resolution, and inferior ion transmission efficiency, real-time Eco-DDA sampling and CHIMERYS processing enabled identification of up to ∼15 peptides per spectrum-performance on par with modern Orbitrap Fusion Lumos tribrid systems. From 1 ng of HeLa digest, 2142 proteins were identified, surpassing the 969 proteins detected on a contemporary nanoLC Orbitrap Fusion Lumos. Even from ∼250 pg (a single-cell equivalent), 1799 proteins were identified in <15 min of effective separation, raising a theoretical throughput of 48 samples per day. As proof of principle, Real-Time Eco-AI profiled 1524 proteins from single precursor cells (50-75 µm diameter) in Xenopus laevis blastulae, revealing proteome asymmetry during neural versus epidermal fate specification. These results establish Real-Time Eco-AI as a budget-conscious yet powerful strategy for single-cell proteomics using CE-MS.
Figure 1. Strategy for affordable single‐cell proteomics using capillary electrophoresis (CE), electrospray ionization (ESI), and mass spectrometry (MS). Neural (D11) and epidermal (V11) precursor blastomeres of 16‐cell Xenopus laevis embryos (St. 5) were microinjected with red/green fluorescent dyes, tracked to the blastula stage (St. 8), and isolated for proteome analysis. CE was coupled to a legacy Orbitrap (Q Exactive Plus, Thermo) with electrophoresis‐correlative (Eco) ion sorting and real‐time data‐dependent acquisition (DDA). Artificial intelligence (CHIMERYS) resolved the resulting highly chimeric tandem mass spectra, achieving sensitivity comparable to modern tribrid instruments. D11 and V11 descendants were profiled to reveal proteome reorganization during early differentiation. (Created with BioRender.com).
Figure 2. Advancing deeper proteomics through Real‐Time Electrophoresis‐Correlative (Eco) data acquisition with AI‐aided data processing on a legacy Orbitrap mass spectrometer. A 10 ng HeLa digest was analyzed by CE– and nanoLC–MS. a) Eco sorting grouped the peptide ions into temporally correlated data clusters based on mass‐to‐charge (m/z) and separation time (2+ charge shown, Pearson ρ ≥ 0.93), whereas b) nanoLC spread them broadly over the dimensions (ρ = 0.29). c) Eco‐data independent acquisition (DIA), using 5 segmented m/z–migration time windows, identified ∼38% more proteins than the control while using <50% of the analytical bandwidth. d) Data‐dependent acquisition (DDA) with standard isolation width (1.6 Th) tracked the m/z–migration correlation with up to 99% success. e), f) MS/MS events aligned with the density of the identified peptide spectral matches (PSMs), reflecting high real‐time sequencing efficiency (ρ = 0.93). g), h) AI‐assisted data processing software (CHIMERYS, Thermo) improved deconvolution of highly chimeric spectra, identifying over 3× more molecular features (MFs). Together, these results demonstrate that Real‐Time Eco–AI enables deep proteome coverage even on earlier‐generation Orbitrap instruments.
Figure 3. Configuration of empirical proteome depth using electrophoresis‐correlative data acquisition with AI‐aided data processing (Eco–AI). HeLa digest (1 ng and 250 pg) was analyzed under fixed MS cycle duration while varying key parameters: quadrupole isolation window (w), number of targeted precursor ions (Top 10 versus Top 20), and Orbitrap resolution (35 000 versus 70 000 FWHM). a) Eco–AI identified up to 1799 proteins in 15 min from ∼250 pg of digest—approximating the protein content of a single HeLa cell. b) The MS/MS spectra showed an increased peptide group depth (examples in Supporting Information), while c) wider quadrupole precursor ion isolation window widths exacerbated spectral interference. d) Label‐free quantification (LFQ intensity) confirmed superior sensitivity of Real‐Time Eco–AI over the recent (scheduled) Eco CE–DIA method. ***p < 0.001, Mann–Whitney U test.
Figure 4. Benchmarking Real‐Time electrophoresis‐correlative (Eco) acquisition with AI‐aided processing against a modern nanoLC–tribrid–WWA workflow. A 1 ng HeLa digest was analyzed by capillary electrophoresis (CE) on a legacy Orbitrap mass spectrometer (Q Exactive Plus, QE+), and compared to nanoLC–MS on a tribrid Orbitrap (Fusion Lumos, Thermo) with higher ion transmission, faster scan rate (maximum 15 versus 13 Hz), and greater resolution (tested: QE+ at 70 000 FWHM and Fusion Lumos at 120 000 FWHM). a) Despite lower hardware specifications, Real‐Time Eco–AI yielded a marked sensitivity gain, validated also against independently obtained data (see Supporting Information). b) Peptide spectral match (PSM) rates reached levels comparable to the modern tribrid system. c) This gain resulted from richer tandem MS spectra acquired with Real‐Time Eco–AI compared to nanoLC‐WWA.
Figure 5. Proteome profiling of single cells undergoing differentiation in Xenopus laevis embryos. a) Dorsal‐animal (D11) and ventral‐animal (V11) blastomeres (n = 8) were microinjected with red and green fluorescent dyes at the 16‐cell stage; these lineages give rise to neural (central somites, CSs) and epidermal (Epi.) tissues in the larva, respectively. b) Descendant cells (∼50–75 µm) were isolated at the blastula stage using a micropipette under fluorescence guidance. ∼500 pg of proteome digest (∼2% of total cellular protein) was analyzed using the Real‐Time electrophoresis‐correlative AI workflow. c) Principal component analysis (PCA) of the cell proteomes revealed systematic differences among sample types (scores and loadings plots), driven by varying protein expression levels. Representative proteins are labeled. d) Hierarchical cluster analysis (HCA, z‐score scale) of the top 200 proteins grouped them into three major abundance profiles (#1–3), supporting cell‐type‐specific proteome remodeling. Scale bars: 500 µm (black), 2 mm (gray), 100 µm (white).
