Quantitation and Identification of Thousands of Human Proteoforms below 30 kDa

Abstract

Top-down proteomics is capable of identifying and quantitating unique proteoforms through the analysis of intact proteins. We extended the coverage of the label-free technique, achieving differential analysis of whole proteins <30 kDa from the proteomes of growing and senescent human fibroblasts. By integrating improved control software with more instrument time allocated for quantitation of intact ions, we were able to collect protein data between the two cell states, confidently comparing 1577 proteoform levels. To then identify and characterize proteoforms, our advanced acquisition software, named Autopilot, employed enhanced identification efficiency in identifying 1180 unique Swiss-Prot accession numbers at 1% false-discovery rate. This coverage of the low mass proteome is equivalent to the largest previously reported but was accomplished in 23% of the total acquisition time. By maximizing both the number of quantified proteoforms and their identification rate in an integrated software environment, this work significantly advances proteoform-resolved analyses of complex systems.

Keywords: Fourier transform mass spectrometry; GELFrEE; cellular senescence; differential mass spectrometry; label-free; proteoform; quantitative proteomics; top-down proteomics.

Figure 1

The process of TDQ and intelligent identification of differentially modified proteoforms from senescent versus control cells. Senescence was induced in growing (Gr) IMR90 fibroblasts by oncogenic Ras expression and verified by positive SA-β-gal staining (top left). Error bars are standard deviation. After positive senescent cells were obtained, growing and Ras-induced senescent (R or RIS) cells were separated by molecular weight using GELFrEE (visualized using a silver-stained slab gel, bottom left). Fractions containing <30 kDa proteins were analyzed by Fourier transform mass spectrometry. (Top) First, MS¹-only spectra were acquired and processed through a linear statistical model. Fold-changes in proteoform abundance between senescent and growing cells were displayed in a volcano plot (upper right). (Bottom) Autopilot acquisition generates confident identifications of previously quantitated proteoforms by tandem MS. Identification-centric runs were performed only after all proteoform quantitation was complete.

Figure 2

Top-down quantitation yields changes in abundance of 1038 cytoplasmic proteoforms. (A) Differences in proteoform abundance are mapped according to their log₂ fold change (x-axis) and the −log₁₀ of a confidence metric (y-axis) best described as an instantaneous FDR (also called a q-value). The dotted line indicates the arbitrary 5% FDR threshold (i.e., q-values of 0.05 and below). Negative values on the x-axis signify a decrease in the level of the proteoform in senescence, while positive values indicate an increase. Each QMT/proteoform is indicated with a dot, with the identified QMTs denoted by a yellow color. The three large circles are selected examples highlighted in panels B–D. (B) A proteoform of the enhancer of rudimentary homologue (P84090) was downregulated by 3.6-fold in OIS compared to growing cells; the q-value associated with this observation was 0.00005, which converts via −log₁₀ into 4.3. (C) One of the largest decreases in senescence was nine-fold, observed with a q-value of 0.0016 for Isoform 1 of FAM107B (large green circle; Q9H098). (D) Guanine nucleotide-binding protein subunit gamma-5 (GBG5; P63218) was upregulated 6.1-fold in senescence with a q-value of 5.0 × 10⁻¹⁰ (or converted via −log₁₀, 9.3). In the box and whisker plots, the first and third quartiles are the ends of the boxes with the median included. The whisker demarcate the minimum and maximum data points for the proteoform.

Figure 3

Advanced Autopilot acquisition confidently identified 1599 unique Uniprot accession numbers from one subcellular fractionation experiment. (A) Multitiered searching was utilized to maximize total proteome coverage. Absolute mass searches (windowed search type based around intact mass) yielded 1555 identifications <30 kDa at a 5% FDR threshold. Through additional GRBM searches aimed at improving proteoform characterization, 44 new proteins were identified, while 653 were shared. The main goal of GRMB searching was to improve proteoform characterization. (B) A comparison of the number of confident identifications found under 30 kDa between this study and a previous one is shown. Additionally, a comparison of the number of LC–MS runs needed to achieve each set of results reveals the same number of protein identifications were obtained in 1/4 of the time.