SPPUSM: An MS/MS spectra merging strategy for improved low-input and single-cell proteome identification.

Single and rare cell analysis provides unique insights into the investigation of biological processes and disease progress by resolving the cellular heterogeneity that is masked by bulk measurements. Although many efforts have been made, the techniques used to measure the proteome in trace amounts of samples or in single cells still lag behind those for DNA and RNA due to the inherent non-amplifiable nature of proteins and the sensitivity limitation of current mass spectrometry. Here, we report an MS/MS spectra merging strategy termed SPPUSM (same precursor-produced unidentified spectra merging) for improved low-input and single-cell proteome data analysis. In this method, all the unidentified MS/MS spectra from multiple test files are first extracted. Then, the corresponding MS/MS spectra produced by the same precursor ion from different files are matched according to their precursor mass and retention time (RT) and are merged into one new spectrum. The newly merged spectra with more fragment ions are next searched against the database to increase the MS/MS spectra identification and proteome coverage. Further improvement can be achieved by increasing the number of test files and spectra to be merged. Up to 18.2% improvement in protein identification was achieved for 1 ng HeLa peptides by SPPUSM. Reliability evaluation by the "entrapment database" strategy using merged spectra from human and E. coli revealed a marginal error rate for the proposed method. For application in single cell proteome (SCP) study, identification enhancement of 28%-61% was achieved for proteins for different SCP data. Furthermore, a lower abundance was found for the SPPUSM-identified peptides, indicating its potential for more sensitive low sample input and SCP studies.

Sensitive N-Glycopeptide Profiling of Single and Rare Cells Using an Isobaric Labeling Strategy without Enrichment.

Single-cell omics is critical in revealing population heterogeneity, discovering unique features of individual cells, and identifying minority subpopulations of interest. As one of the major post-translational modifications, protein N-glycosylation plays crucial roles in various important biological processes. Elucidation of the variation in N-glycosylation patterns at single-cell resolution may largely facilitate the understanding of their key roles in the tumor microenvironment and immune therapy. However, comprehensive N-glycoproteome profiling for single cells has not been achieved due to the extremely limited sample amount and incompatibility with the available enrichment strategies. Here, we have developed an isobaric labeling-based carrier strategy for highly sensitive intact N-glycopeptide profiling for single cells or a small number of rare cells without enrichment. Isobaric labeling has unique multiplexing properties, by which the "total" signal from all channels triggers MS/MS fragmentation for N-glycopeptide identification, while the reporter ions provide quantitative information. In our strategy, a carrier channel using N-glycopeptides obtained from bulk-cell samples significantly improved the "total" signal of N-glycopeptides and, therefore, promoted the first quantitative analysis of averagely 260 N-glycopeptides from single HeLa cells. We further applied this strategy to study the regional heterogeneity of N-glycosylation of microglia in mouse brain and discovered region-specific N-glycoproteome patterns and cell subtypes. In conclusion, the glycocarrier strategy provides an attractive solution for sensitive and quantitative N-glycopeptide profiling of single/rare cells that cannot be enriched by traditional workflows.

A three-stage search strategy combining database reduction and retention time filtering to improve the sensitivity of low-input and single-cell proteomic analysis.

When performing proteome profiling of low-input and single-cell samples, achieving deep protein coverage is very challenging due to the sensitivity limitation of current proteomic methods. Herein, we introduce a three-stage search strategy that combines the advantages of database reduction and Δ retention time (ΔRT) filtering. The strategy improves peptide/protein identification and reproducibility by retaining more correct identifications and filtering out incorrect identifications. The raw data were first merged and searched against a Uniprot database with a relaxed false discovery rate (FDR) of 40% to identify the possible detectable proteins. The identified proteins were then used as a new database to search the raw data against with a tighter FDR of 10%. After this, the results were filtered using ΔRT (the difference between the measured and predicted RT) to reduce the incorrect identifications and maintain the FDR below 1%. This strategy resulted in over 30% improvement in proteome coverage for single-cells and samples of similar size. The reproducibility of identification and quantification was also enhanced for the low-input samples. Moreover, the 50% higher number of differential proteins found in the two types of single neurons indicates the application potential of this strategy.

A rapid and sensitive single-cell proteomic method based on fast liquid-chromatography separation, retention time prediction and MS1-only acquisition.

Single-cell analysis has received much attention in recent years for elucidating the widely existing cellular heterogeneity in biological systems. However, the ability to measure the proteome in single cells is still far behind that of transcriptomics due to the lack of sensitive and high-throughput mass spectrometry methods. Herein, we report an integrated strategy termed "SCP-MS1" that combines fast liquid chromatography (LC) separation, deep learning-based retention time (RT) prediction and MS1-only acquisition for rapid and sensitive single-cell proteome analysis. In SCP-MS1, the peptides were identified via four-dimensional MS1 feature (m/z, RT, charge and FAIMS CV) matching, therefore relieving MS acquisition from the time consuming and information losing MS2 step and making this method particularly compatible with fast LC separation. By completely omitting the MS2 step, all the MS analysis time was utilized for MS1 acquisition in SCP-MS1 and therefore led to 65%-138% increased MS1 feature collection. Unlike "match between run" methods that still needed MS2 information for RT alignment, SCP-MS1 used deep learning-based RT prediction to transfer the measured RTs in long gradient bulk analyses to short gradient single cell analyses, which was the key step to enhance both identification scale and matching accuracy. Using this strategy, more than 2000 proteins were obtained from 0.2 ng of peptides with a 14-min active gradient at a false discovery rate (FDR) of 0.8%. Comparing with the DDA method, improved quantitative performance was also observed for SCP-MS1 with approximately 50% decreased median coefficient of variation of quantified proteins. For single-cell analysis, 1715 ± 204 and 1604 ± 224 proteins were quantified in single 293T and HeLa cells, respectively. Finally, SCP-MS1 was applied to single-cell proteome analysis of sorafenib resistant and non-resistant HepG2 cells and revealed clear cellular heterogeneity in the resistant population that may be masked in bulk studies.

A new integrated method for tissue extracellular vesicle enrichment and proteome profiling.

Extracellular vesicles (EVs) are membranous vesicles released by cells that carry a number of biologically important components such as lipids, proteins, and mRNAs. EVs can mediate cancer cell migration, invasion, angiogenesis, and cell survival, greatly contributing to cell-to-cell communication in the tumor microenvironment. Additionally, EVs have been found to have diagnostic and prognostic significance in various cancers. However, the direct isolation of pure EVs remains challenging, especially from tissue samples. Currently available EV isolation approaches, e.g., ultracentrifugation, are time-consuming, instrumental dependent, and have a low recovery rate with limited purity. It is urgent to develop rapid and efficient methods for enriching tissue EVs for biological and clinical studies. Here, we developed a novel isolation approach for tissue EVs using an extraction kit combined with TiO2 microspheres (kit-TiO2). The EVs were first precipitated from the tissue fluid using a precipitation agent and then further enriched using microspheres based on the specific interaction between TiO2 and the phosphate groups on the lipid bilayer of the EVs. Kit-TiO2 approach led to improved purity and enrichment efficiency of the isolated EVs, as demonstrated by western blot and proteomic analysis, compared with previously reported methods. A total of 1966 protein groups were identified from the tissue EVs. We compared the proteomic profiles of the liver tissue EVs from healthy and hepatocellular carcinoma (HCC) bearing-mice. Twenty-five significantly upregulated and 75 downregulated protein groups were found in the HCC EVs. Among the differentially expressed proteins, Atic, Copa, Cont3, Me1, Anxa3, Fth1, Anxa5, Phb1, Acaa2, ATPD, and Glud1 were reported to be highly relevant to HCC. This novel isolation strategy has provided a powerful tool for enriching EVs directly from tissues, and may be applied in biomarker discovery and drug screening of HCC.

Synthesis of Surface-Functionalized Molybdenum Disulfide Nanomaterials for Efficient Adsorption and Deep Profiling of the Human Plasma Proteome by Data-Independent Acquisition.

Blood is one of the most important clinical samples for protein biomarker discovery, as it provides rich physiological and pathological information and is easy to obtain with low invasiveness. However, the discovery of protein biomarkers in the blood by mass spectrometry (MS)-based proteomic strategies has been shown to be highly challenging due to the particularly large concentration range of proteins and the strong interference by the high-abundant proteins in the blood. Therefore, developing sensitive methods for low-abundant biomarker protein identification is a key issue that has received great attention. Here, we report the synthesis and characterization of surface-functionalized magnetic molybdenum disulfide (MoS2) for the large-scale adsorption of low-abundant plasma proteins and deep profiling by MS. MoS2 nanomaterials resulted in the coverage of more than 3400 proteins (including a single-peptide hit) in a single LC-MS analysis without peptide prefractionation using pooled plasma samples, which were five times more than those obtained by the direct analysis of the plasma proteome. A detection limit in the low ng L-1 range was obtained, which is rare compared with previous reports.

A New Strategy for High-Efficient Tandem Enrichment and Simultaneous Profiling of N-Glycopeptides and Phosphopeptides in Lung Cancer Tissue.

N-glycosylation and phosphorylation, two common posttranslational modifications, play important roles in various biological processes and are extensively studied for biomarker and drug target screening. Because of their low abundance, enrichment of N-glycopeptides and phosphopeptides prior to LC-MS/MS analysis is essential. However, simultaneous characterization of these two types of posttranslational modifications in complex biological samples is still challenging, especially for tiny amount of samples obtained in tissue biopsy. Here, we introduced a new strategy for the highly efficient tandem enrichment of N-glycopeptides and phosphopeptides using HILIC and TiO2 microparticles. The N-glycopeptides and phosphosites obtained by tandem enrichment were 21%-377% and 22%-263% higher than those obtained by enriching the two PTM peptides separately, respectively, using 160-20 µg tryptic digested peptides as the starting material. Under the optimized conditions, 2798 N-glycopeptides from 434 N-glycoproteins and 5130 phosphosites from 1986 phosphoproteins were confidently identified from three technical replicates of HeLa cells by mass spectrometry analysis. Application of this tandem enrichment strategy in a lung cancer study led to simultaneous characterization of the two PTM peptides and discovery of hundreds of differentially expressed N-glycosylated and phosphorylated proteins between cancer and normal tissues, demonstrating the high sensitivity of this strategy for investigation of dysregulated PTMs using very limited clinical samples.

[Research progress and application of retention time prediction method based on deep learning].

In "shotgun" proteomics strategy, the proteome is explained by analyzing tryptic digested peptides using liquid chromatography-mass spectrometry. In this strategy, the retention time of peptides in liquid chromatography separation can be predicted based on the peptide sequence. This is a useful feature for peptide identification. Therefore, the prediction of the retention time has attracted much research attention. Traditional methods calculate the physical and chemical properties of the peptides based on their amino acid sequence to obtain the retention time under certain chromatography conditions; however, these methods cannot be directly adopted for other chromatography conditions, nor can they be used across laboratories or instrument platforms. To solve this problem, in recent years, deep learning was introduced to proteomics research for retention time prediction. Deep learning is an advanced machine-learning method that has extraordinary capability to learn complex relationships from large-scale data. By stacking multiple hidden neural networks, deep learning can ingest raw data without manually designed features. Transfer learning is an important method in deep learning. It improves the learning process a new task through the transfer of knowledge from an already-learned related task. Transfer learning allows models trained using large datasets to be utilized across conditions by fine-tuning on smaller datasets, instead of retraining the whole model. Many retention time prediction methods have been developed. In the process of training the model, the sequences of peptides are encoded to represent peptide information. Deep learning considers the relationship between the characteristics of the peptides and their corresponding retention times without the need for manual input of the physical and chemical properties of the peptides. Compared with traditional methods, deep learning methods have higher accuracy and can be easily used under different chromatography conditions by transfer learning. If there are not enough datasets to train a new model, a trained model from other datasets can be used as a replacement after calibration with small datasets obtained from these chromatography conditions. While the retention times of modified peptides can also be predicted, the predictions are inadequate for complex modifications such as glycosylation, and this is one of the main problems to be solved. The predicted retention times were used to control the quality of peptide identification. With high accuracy, the predicted retention times can be considered as actual retention times. Therefore, the difference between predicted and observed retention times can serve as an effective and unbiased quantitative metric for evaluating the quality of peptide-spectrum matches (PSMs) reported using different peptide identification methods. Combined with fragment ion intensity prediction, retention time prediction is used to generate spectral libraries for data-independent acquisition (DIA)-based mass spectrometry analysis. Generally, DIA methods identify peptides using specific spectrum libraries obtained from data-dependent acquisition (DDA) experiments. As a result, only peptides detected in the DDA experiments can be present in the libraries and detected in DIA. Furthermore, it takes a lot of time and effort to build libraries from DDA experiments, and typically, they cannot be adopted across different laboratories or instrument platforms. In contrast, the pseudo spectral libraries generated by retention times and fragment ion intensity prediction can overcome these shortcomings. The pseudo spectral libraries generate theoretical spectra of all possible peptides without the need for DDA experiments. This paper reviews the research progress of deep learning methods in the prediction of retention time and in related applications in order to provide references for retention time prediction and protein identification. At the same time, the development direction and application trend of retention time prediction methods based on deep learning are discussed.