Optimized Approaches for High-Throughput Protein Quantitation

Goal To develop and assess qualitative and quantitative performance of label-free quantitation (LFQ) with an optimized data-independent acquisition (DIA) method on a Thermo Scientific™ Orbitrap™ Exploris 480 mass spectrometer using a long (60 min active gradient), short (30 min active gradient), and high throughput (9 min active gradient) method for large-scale proteomics analysis. Introduction Quantitative proteomics is an essential tool for understanding global protein expression and the mechanisms of biological processes and disease states. Accurately quantifying the abundances of proteins of interest in complex samples is a prerequisite for developing suitable statistical models to gain biological insights from experimental data sets. Statistical significance is improved by decreasing variability in measurements and/or increasing the sample set. However, increasing throughput means decreasing acquisition time, which often comes at a cost to measurement quality. Therefore, acquisition methods must be extensively optimized and validated to ensure that the data will produce meaningful biological insights. Traditional data-dependent analysis (DDA) approaches have been widely employed for LFQ experiments, but they suffer from runto-run inconsistencies due to intensity-based stochastic triggering of precursors, often leading to under sampling especially of low-abundant proteins. Missing values become more likely as sample size increases, DIA has emerged as a popular technique for large scale quantitative analyses. High-throughput high-resolution data-independent acquisition workflow on an Orbitrap Exploris 480 mass spectrometer for accurate label-free quantitation Authors Kevin Yang1 , Julia Kraegenbring2 , Julian Saba3 , Sally Webb1 , Maciej Bromirski4 , and Amirmansoor Hakimi1 1 Thermo Fisher Scientific, San Jose, California, USA 2 Thermo Fisher Scientific, Bremen, Germany 3 Thermo Fisher Scientific, Mississauga, ON, Canada 4 Thermo Fisher Scientific, Warsaw, Poland Keywords Data-independent acquisition, DIA, Velocity DIA, Orbitrap Exploris 480 mass spectrometer, μPAC Neo HPLC column, Vanquish Neo UHPLC, precision, quantitation, accuracy, label-free quantitation, LFQ, HeLa, CHIMERYS, Spectronaut software, Proteome Discoverer software Technical note | 002684 Omics In contrast to DDA, DIA addresses missing value concerns by equally cycling through defined m/z windows along the survey scan range. The inherent tradeoff between measurement selectivity (isolation window size), frequency (cycle time), scope (mass range), and sensitivity (ion accumulation time) necessitates careful method optimization, and a suboptimal combination of parameters can have disastrous consequences. Additionally, the resulting spectral complexity of the mixed precursor fragmentation and mixed product ions is often addressed by employing large spectral libraries. However, recent developments in data analysis software (e.g., using machinelearning approaches for in silico prediction of high-quality spectral libraries) have made library-free approaches a valid time- and cost-effective alternative. The need for analyzing large numbers of samples, especially in clinical and biomarker discovery studies, makes LFQ DIA-based workflows an obvious choice for ensuring highthroughput and accurate quantitative analyses. A suitable analytical workflow addresses the need for reproducible sample preparation, robust separations, high-quality quantitative measurements, and reliable data analysis. To this end, we present the Velocity LFQ DIA workflow, an endto-end solution for quantitative proteomics (Figure 1 and Table 1). Briefly, sample preparation can be automated to increase throughput and decrease technical variability with the Thermo Scientific™ AccelerOme™ sample preparation platform. Robust LC-MS/MS analysis is performed on the Thermo Scientific™ Vanquish™ Neo UHPLC system with a Thermo Scientific™ μPAC™ Neo HPLC column coupled to an Orbitrap Exploris 480 mass spectrometer. The Vanquish Neo UHPLC system combined with the μPAC Neo HPLC column has been shown to increase sensitivity and retention time stability.1 Additionally, the Orbitrap Exploris 480 MS uses Orbitrap technology to deliver high resolution measurements with low background ions and high sensitivity. We compare multiple different analysis methods, including Thermo Scientific™ Proteome Discoverer™ software (v3.1.0.638) with CHIMERYS™ intelligent search algorithm by MSAID. Overall, we demonstrate that the Velocity LFQ DIA workflow on the Orbitrap Exploris 480 mass spectrometer enables a high level of quantitative performance across short, medium, and long gradients to meet a variety of experimental needs. In addition, in large cohort studies, a robust setup (separation technology, column, and mass spectrometer) that can run stably for an extended period is a necessity. The Vanquish Neo UHPLC system delivers maximum performance for reproducible and versatile LC-MS experiments. New technologies in chromatographic separations also help achieve robustness. Micropillar-array LC columns like the μPAC Neo HPLC column have been shown to deliver increased sensitivity and higher retention time stability,1 making them ideal candidates for setting up a robust and reproducible workflow. Aside from robustness and reproducibility, confidence in identification and quantitation is an imperative for impactful proteomics research. Confidence in quantitative results is driven not only by ensuring accurate and precise measurements, but also by careful data analysis and rigorous validation methods like controlled strict false discovery rates (FDR). Thermo Scientific™ Orbitrap™ technology fulfills the prerequisites for confident measurements by delivering both highly accurate mass as well as high resolution, providing sensitivity and specificity in an unparalleled manner. These two factors are key to reliable identifications and the ability to accurately and precisely detect and resolve ion species in complex DIA scans. The challenge for applications in large-scale clinical studies is to achieve high levels of performance for all these factors— sensitivity, mass resolution, mass accuracy, quantitative accuracy, and precision—while increasing sample throughput to advance biological insights. In many cases, these performance factors are influenced by counteracting method parameters, and a fine balance must be found in an optimized setup to best fulfill all these criteria. Here, we present a robust and reproducible workflow on the Orbitrap Exploris 480 mass spectrometer for accurately quantifying and identifying hundreds to thousands of proteins from single cell-line to complex sample mixtures with a high background of human peptides (Figure 1 and Table 1). The Velocity LFQ DIA workflow could be coupled with the optional AccelerOme automated sample preparation platform to improve throughput and minimize variations caused by manual sample handling (Figure 1). The mass spectrometric method was adopted to account for long gradient lengths (60 min active gradient), short gradient lengths (30 min active gradient), and high throughput (9 min active gradient) without compromising identification rates at great quantitation accuracy and precision. 2 Experimental Consumables • Fisher Scientific™ LC-MS grade water with 0.1% formic acid (P/N LS118-500) • Fisher Scientific™ Optima™ LC-MS grade 80% acetonitrile with 0.1% formic acid (P/N LS122-500) • Fisher Scientific™ Optima™ LC-MS grade 100% acetonitrile with 0.1% formic acid (P/N LS120-212) • Thermo Scientific™ Pierce™ HeLa Protein Digest Standard (P/N 88329) • Waters™ MassPREP™ E. coli Digest Standard (P/N 186003196) • Promega™ Mass Spec-Compatible Yeast Protein Extracts (P/N V7461) • Evosep Thermo Scientific™ EASY-Spray™ Adapter (P/N EV1072) • Evosep fused silica emitters 10 µm (P/N EV1111) • Fluidics and consumables used to set up the Vanquish Neo UHPLC system for direct injection are given in Table 1. Workflow components Description Liquid chromatography Vanquish Neo UHPLC system: Binary Pump N, Split Sampler NT, Solvent Rack, Vanquish System Controller, System Base with drawer, Vanquish Display (P/N 6036.1180), Vanquish Split Sampler Sample Loop, 100 µL (P/N 6252.1950), Vanquish Column Compartment N (P/N VN-C10-A-01) Column µPAC Neo HPLC column, 50 cm, 180 µm bed width, 16 µm pillar length (P/N COL-NANO050NEOB) Emitter • Thermo Scientific™ EASY-Spray™ Nano & Capillary adapter (P/N ES993) • EvoSep Thermo Scientific™ EASY-Spray™ adapter (P/N EV1072) • EvoSep Fused Silica Emitter (P/N EV1111) Source Thermo Scientific™ EASY-Spray™ ion source (P/N ES082) Mass spectrometer Orbitrap Exploris 480 mass spectrometer Data analysis software • Spectronaut™ 18 software (Biognosys) • DIA-NN software (v1.8.1) • Proteome Discoverer software using CHIMERYS (v3.1.0.638) Figure 1. Graphical schematic of the Velocity DIA workflow for label-free quantitation on the Orbitrap Exploris 480 MS together with the optional AccelerOme automated sample preparation platform Table 1. List of workflow components with part numbers Sample preparation Pierce HeLa Protein Digest Standard, Waters E. coli MassPREP Standard and Promega Mass Spec-Compatible Yeast Protein Digest were dissolved in 0.1% formic acid (FA) with 30 seconds of vortexing. For the three-proteome mix, E. coli peptide digest and yeast peptide digest were added to a fixed amount of HeLa digest (325 ng) at amounts of 100 ng to 25 ng, and 75 ng to 150 ng, respectively, yielding an E. coli peptide ratio of 1:4 and a yeast peptide ratio of 0.5:1 (Figure 2A and 2B). LC-MS method HeLa digest and three-proteome mixtures were loaded onto a 50 cm μPAC Neo HPLC column and separated at a 350 nL/min flow rate in direct injection mode using a Vanquish Neo UHPLC system over 9 min, 30 min, and 60 min active LC gradients, respectively, before being transferred into the Orbitrap Exploris 480 mass spectrometer (Figure 2C). Source parameters, including spray voltage and ion transfer tube temperature, are tunable parameters and must be optimized for the individual setup. The details of the LC gradient, LC parameters, and MS method are reported in Table 2. Vanquish Neo UHPLC system µPAC Neo 50 cm HPLC column EASY-Spray nano source Orbitrap Exploris 480 MS with FAIMS Pro interface Velocity LFQ DIA Software of choice AccelerOme automated sample preparation platform 3 Figure 2. Experimental sample and active gradient design of the Velocity LFQ DIA workflow for label-free quantitation. Two different sample sets were used for assessing the identification and quantitative performance of the Velocity LFQ DIA workflow on the Orbitrap Exploris 480 MS. (A) 200 ng HeLa digest to access quantitation precision and proteome coverage of human samples. (B) The three-proteome mix contains a medium human background of 325 ng HeLa peptides together with yeast and E. coli peptides digested in ratios of 0.5:1 and 1:4, respectively. The mixtures have been chosen because they closely mimic biological samples with larger or smaller protein expression changes. (C) Three different active gradient lengths, including 60 min, 30 min, and 9 min, were selected to develop the Velocity LFQ DIA workflow for different throughput needs. For the ultrahigh throughput 9 min active gradient setup, two methods were developed to obtain deep proteome coverage (Max ID) or excellent quantitation performance (Max Quan). 200 ng HeLa digest E. coli HeLa Sample loading (ng) Yeast 100 25 150 75 325 325 A B Mixture 500 ng Proteome mix 0 10 20 30 40 50 60 70 80 90 100 0 5 10 15 20 25 30 35 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 0 10 20 30 40 50 60 70 80 90 100 0 2 4 6 8 10 12 0 10 20 30 40 50 60 70 80 90 100 0 2 4 6 8 10 12 Relative abundance 60 min 30 min 9 min (Max ID) 9 min (Max Quan) Active gradient (C) (A) (B) Data analysis and post-processing Acquired data was processed by Spectronaut 18 software using a directDIA approach, DIA-NN (v1.8.1) or Proteome Discoverer software (v3.1.0.638) using CHIMERYS intelligent search algorithm by MSAID. For Spectronaut software, default settings were used except that Cross-Run Normalization > Normalization Filter Type was set to “FASTA name filter” and the “FASTA name” was defined to be the human protein database. Peptide and protein identifications were filtered for 1% FDR, and a Q-value cutoff of 1% was used for the DIA analysis. FASTA files for human, yeast, and E. coli were downloaded from UniProt™. For analysis of the human protein group abundance ratios in the three-proteome mix, the default filter “Absolute AVG Log2 Ratio” in the candidate table was disabled. For CHIMERYS in Proteome Discoverer software, default settings were used for both the processing and consensus workflow. All the PSM, peptides, and proteins were filtered at 1% FDR. For the three-proteome mix, species map and species names were set as True in the “Protein Marker” node. For DIA-NN software, default settings were used for either direct DIA or library search. The resulting candidate tables and report files for data searched with either Spectronaut software and CHIMERYS in Proteome Discoverer software were exported to .csv or .tsv files. The ensuing tables were imported to Python™ or a spreadsheet for downstream data analysis and visualization. Results and discussion The Velocity LFQ DIA workflow was initially developed on the Thermo Scientific™ Orbitrap Exploris™ 240 Mass Spectrometer.2 Here we expand this workflow to Orbitrap Exploris 480 mass spectrometer (Figure 1). In addition, we also evaluated the benefits of Thermo Scientific™ FAIMS Pro™ interface in protein identification. Our results showed that the FAIMS Pro interface allowed for an additional 5% increase in protein group identifications (Figure 3). Consequently, the following experiments were all coupled with the FAIMS Pro interface. 4 Separation column specifications (set in the Vanquish Neo system) Inner diameter 75 µm Length 50 cm Maximum pressure 450 bar Maximum flow 0.7 µL/min Maximum temperature 60 °C LC method (60 min method duration) Gradient Time (min) %B 0 4 45 30 60 45 60.0–64.9 97.5 64.9–65 4 LC parameters Flow rate 350 nL/min Column temperature 50 °C Fast loading/equilibration PressureControl Pressure loading/equilibration 350 bar Equilibration factor 2.0 Sampler temperature 7 °C Table 2. Summary of all LC and MS method parameters. Parameters not mentioned in the table are set to default values. LC method (30 min method duration) Gradient Time (min) %B 0 4 22.5 30 30 45 30.1–33.0 97.5 33.0–33.1 4 LC parameters Flow rate 350 nL/min Column temperature 50 °C Fast loading/equilibration PressureControl Pressure loading/equilibration 350 bar Equilibration factor 2.0 Sampler temperature 7 °C LC method (9 min method duration) Gradient Time (min) %B 0 4 0–7.5 30 7.5–9.0 45 9.0–12 97.5 LC parameters Flow rate 350 nL/min Column temperature 50 °C Fast loading/equilibration PressureControl Pressure loading/equilibration 350 bar Equilibration factor 2.0 Sampler temperature 7 °C MS method for 30 and 60 min (Application mode “peptide”) Global parameters Use ion source settings from Tune True Expected peak width 10 Advanced peak determination True Default charge state 2 MS parameters Resolution MS1 /MS2 60,000/15,000 Scan range (m/z) MS1 400–900 Scan range (m/z) MS2 145–1,450 Normalized AGC target (%) MS1 /MS2 300/800 Maximum injection time mode MS1 /MS2 Auto Isolation window (m/z) 12 Window overlap (m/z) 1 Window placement optimization On Normalized HCD Collision Energy (%) 30 MS method for 9 min MaxID (Application mode “peptide”) Global parameters Use ion source settings from Tune True Expected peak width 10 Advanced peak determination True Default charge state 2 MS parameters Resolution MS1 /MS2 60,000/15,000 Scan range (m/z) MS1 400–800 Scan range (m/z) MS2 145–1,450 Normalized AGC target (%) MS1 /MS2 300/800 Maximum injection time mode MS1 /MS2 Auto Isolation window (m/z) 8 Window overlap (m/z) 1 Window placement optimization On Normalized HCD Collision Energy (%) 30 MS method for 9 min MaxQuan (Application mode “peptide”) Global parameters Use ion source settings from Tune True Expected peak width 10 Advanced peak determination True Default charge state 2 MS parameters Resolution MS1 /MS2 60,000/30,000 Scan range (m/z) MS1 650–770 Scan range (m/z) MS2 145–1450 Normalized AGC target (%) MS1 /MS2 300/800 Maximum injection time mode MS1 /MS2 Auto Isolation window (m/z) 12 Window overlap (m/z) 1 Window placement optimization On Normalized HCD Collision Energy (%) 30 5 Figure 4. Determination of proteome coverage and quantitation precision of the human proteome. (A-C) Bar charts showing the number of proteins and peptides identified from 200 ng HeLa digests with different active gradients evaluated in the present study. (D-F) Violin plots of all four tested methods reveal high precision of protein quantities in technical replicates. Data was analyzed with Spectronaut software (left panel), CHIMERYS on Proteome Discoverer software (middle panel), or DIA-NN (right panel). The initial set of experiments consisted of maximizing identification and quantitative performance for various active gradients of 9, 30, and 60 minutes for tryptically digested HeLa standards (Figure 2C). With the 30 min active gradient, 6,300+ proteins and 40,000+ peptides were identified, along with a protein group CV of approximately 5% (Figure 4), suggesting that a 30 min active gradient method provides the perfect amount of time to achieve throughput while maximizing identification and quantitative performance. We extended this workflow to a 60 min active gradient and successfully identified close to 7,200 proteins and >60,000 peptides, highlighting that deeper proteome coverage can be achieved in the Velocity LFQ DIA workflow by using a longer gradient. We also performed a 9 min active gradient to develop an ultra-high throughput method. For the 9 min active gradient experiments, we focused on developing two methods. One for maximizing identifications (Max ID) and the other for maximizing quantitative (Max Quan) performance. In the Max ID method, we were able to identify 5,500+ proteins in a 9 min active gradient, highlighting the sensitivity and scanning speed of the Orbitrap mass spectrometer in a high throughput setup. In contrast, the Max Quan method, although it compromised proteome coverage, provided better quantitation performance, as evidenced by ~6% protein group CV in such an ultra-high throughput setup (Figure 4). To understand if a higher load benefits the Velocity LFQ DIA workflow in maximizing identification, 500 ng of HeLa digest was analyzed with the 60 min active gradient experiments. The results indicated 7,400+ protein groups were identified, with CV being <5% (Figure 5). Thus, we recommend injecting 500 ng of digest, if possible, to maximize the performance of the Velocity LFQ DIA experiment and thus obtain higher quality data. (A) (B) (C) (D) (E) (F) Spectronaut software Proteome Discoverer software using CHIMERYS DIA-NN Figure 3. Evaluation of the FAIMS Pro interface in the Velocity LFQ DIA workflow. Box plots depicting the number of proteins and peptides identified from 200 ng of HeLa digest in the 30 min active gradient method with the FAIMS Pro interface at a CV of -45 V. Data were analyzed with (A) Spectronaut software, (B) Proteome Discoverer software using CHIMERYS, or (C) DIA-NN. (A) (B) (C) No FAIMS FAIMS No FAIMS FAIMS No FAIMS FAIMS No FAIMS FAIMS No FAIMS FAIMS No FAIMS FAIMS 6 Figure 5. A long gradient coupled with a higher sample load affords the superior performance in the Velocity LFQ DIA workflow. (A) Bar chart of protein groups identified from 500 ng of HeLa digest in the 60 min active gradient method and analyzed with different software demonstrate even deeper proteome coverage. (B) Violin plots reveal high precision of protein quantities in technical replicates. Data was analyzed with Spectronaut software (left), Proteome Discoverer software using CHIMERYS (middle), or DIA-NN (right). High accuracy and precision of quantitation In addition to protein identification, quantitative data is necessary to study biomarkers and get biological insights. The quantitative data must be highly precise and accurate to reflect subtle changes in biological systems. Distorted quantitation will mislead the direction of biomedical research and cause significant waste of resources down the road. Our results from the HeLa digest indicated a protein CV of approximately 4–5%, suggesting excellent quantitation precision. To test the reliability of quantitation accuracy, in this data set, we used two samples with different amounts of spiked microbial proteins to mimic biological samples where proteins might be up- or downregulated under different conditions. We tested the quantitative performance for different active gradients (9, 30, and 60 min). The Velocity LFQ DIA workflow used for relative quantitation of E. coli and yeast proteomes in a high amount of human peptides as background yields excellent quantitation accuracy across all ratios with median values extremely close to the theoretical ratios, as well as a narrow distribution of all data points around the median values, indicating high quantitative accuracy and precision of the workflow (Figure 6). Additionally, the three-proteome mix experiment further highlights the proteome depth that can be achieved with an Orbitrap mass spectrometer. In the 60 min active gradient, >10,000 protein groups were identified. The numbers of quantified proteins differ by species, with nearly 7,200 human protein groups, approximately 2,800 yeast protein groups, and close to 870 E. coli protein groups (Figure 6). The data demonstrated deep proteome coverage and excellent quantitation accuracy afforded by the Orbitrap Exploris 480 mass spectrometer. Evaluating data processing software in DIA analysis Being able to accurately quantify protein abundances and achieve high proteome coverage are important prerequisites for investigating the underlying mechanisms and proteins of interest in biological processes. However, protein grouping and identification must be confident to avoid false positives as much as possible. The data analysis and post-processing of the analysis results are therefore essential for meaningful LFQ results. To assess data analysis software, we used the commercially available Spectronaut 18 software as a benchmark to evaluate Proteome Discoverer software with CHIMERYS (v3.1.0.638) and DIA-NN (v1.8.1),3 an academic software package (Figures 4-6). For the three-proteome mix, 7,000+ to 9,700+ protein groups were identified in the 30 min active gradient experiment across different software. For the 60 min active gradient, the number of protein groups spans from ~7,000 to >10,800 (Figure 6). Thus, software should be taken into account when accessing the performance of mass spectrometers. Library-based search in DIA proteomics enhances proteome coverage and improves peptide identification. To evaluate if Velocity LFQ DIA data can be benefited from library search, we generated a general library using Pulsar on Spectronaut 18 software from single shots of 12 cell lines acquired on a Thermo Scientific™ Orbitrap™ Astral™ mass spectrometer. The raw files were reprocessed with library-based search on DIA-NN. In line with the benefits of library-based search, we observed a 3–10% increase in protein identification from results searched by using DIA-NN (v1.8.1) depending on the LC gradient (Figure 7). (A) (B) 7 Figure 7. Library search improved proteome coverage. Bar charts showing the number of protein groups gained through library search by using DIA-NN Figure 6. Determination of microbial and human protein abundance ratios in a three-proteome mixture. Bar plots showing the total protein groups identified from three-proteome mix in different active gradients and whisker box plots of protein abundance ratios of all three species demonstrate excellent quantitation accuracy by being consistent with the theoretical ratios (gray dotted line). Data was analyzed with Spectronaut software (A), Proteome Discoverer software using CHIMERYS (B), and DIA-NN (C). (A) (B) (C) Spectronaut software Proteome Discoverer software using CHIMERYS DIA-NN 8 General Laboratory Equipment – Not For Diagnostic Procedures. © 2024 Thermo Fisher Scientific Inc. All rights reserved. All trademarks are the property of Thermo Fisher Scientific and its subsidiaries unless otherwise specified. CHIMERYS is a trademark of MSAID GmbH. Waters and MassPREP are trademarks of Waters Corp. Promega is a trademark of Promega Corp. Spectronaut is a trademark of Biognosys. Python is a trademark of the Python Software Foundation. UniProt is a trademark of the European Molecular Biology Laboratory. This information is presented as an example of the capabilities of Thermo Fisher Scientific products. It is not intended to encourage use of these products in any manner that might infringe the intellectual property rights of others. Specifications, terms and pricing are subject to change. Not all products are available in all countries. Please consult your local sales representative for details. TN002684-EN 0124S Learn more at thermofisher.com/orbitrapvelocitydia Conclusion The high-resolution DIA workflow for LFQ setup on an Orbitrap Exploris 480 mass spectrometer coupled to a Vanquish Neo HPLC system running with a 50 cm μPAC Neo UHPLC column was shown to fulfill the following performance criteria: • Excellent quantitation accuracy and precision for small amounts of bacterial and fungal proteomes from challenging sample matrices • Sample throughput and quality of the obtained data while achieving high proteome coverage References 1. Stadlmann, J.; Hudecz, O.; Krššáková, G. et al. Improved Sensitivity in Low-Input Proteomics Using Micropillar Array-Based Chromatography. Anal. Chem. 2019, 91(22), 14203–14207. https://pubs.acs.org/doi/10.1021/acs.analchem.9b02899 2. Thermo Scientific Technical Note 1251 (TN1251): High-throughput high-resolution data-independent acquisition workflow for accurate label-free quantitation. https:// assets.thermofisher.com/TFS-Assets/CMD/Technical-Notes/tn-001251-ms-lfq-diaorbitrap-tn001251-en.pdf 3. Demichev, V.; Messner, C.B.; Vernardis, S.I. et al. DIA-NN: neural networks and interference correction enable deep proteome coverage in high throughput. Nat. Methods 2020, 1, 41–44. https://doi.org/10.1038/s41592-019-0638-x