Novel Methods For Mass Spectrometry-based Single-cell Proteomics

Proteomic information is directly related to cell function and state,making single-cell proteomic profiling an invaluable technique that can directly determine protein composition and abundance in individual cells.However,single-cell proteomic methods lag far behind single-cell transcriptomic methods in terms of protein coverage and analysis throughput,and these limitations make it difficult to apply single-cell proteomics to large populations.Therefore,this work aims to improve the sensitivity and throughput of mass spectrometry-based single-cell proteome analysis through innovative research on liquid chromatography-mass spectrometry(LC-MS)method,data analysis and sample preparation approaches.To overcome the shortcomings of conventional data-dependent acquisition(DDA)and long-gradient separation,we developed an approach,SCP-MS1,that combines rapid low-flow chromatographic separation,MS1-only acquisition,deep learning-based retention time(RT)prediction and reference library-based 4D peptide feature matching to achieve high-throughput and highly sensitive single-cell proteomics.The chromatographic separation was performed on a 5 cm×25μm nanocolumn at a flow-rate of 100 n L/min to improve both sensitivity and throughput.MS1 only acquisition,which eliminated the acquisition of MS/MS spectra,was particularly compatible with the rapid separation and could avoid signal loss due to fragmentation.Auto RT,a transfer learning-based RT prediction software that requires a small number of peptides to train the model,improved the accuracy of RT predictions for single cells.The use of 4D feature matching enhanced the specificity of identification and thus reduced the false discovery rate(FDR).Combining the above four tools,SCP-MS1 was able to identify more than 3000 proteins and 10000 peptides from 0.2 ng of 293T cell lysed peptides using an active gradient of 14 minutes.As a result of filtering the identification with at least two unique peptides per protein,the number of identified proteins and peptides remained at 2000 and 7500,respectively,and the FDRs were reduced from 2.5%to 0.3%at the peptide level and from 7.6%to 0.8%at the protein level.When SCP-MS1 was applied to FACS sorted 293T and He La single-cells,1715±204 and 1604±224 proteins were identified.Further application of this method to investigate the mechanisms of induced sorafenib resistance in Hep G2 cells revealed that the resistant cell population was heterogeneous and that enhanced protein translation and stability could be exploited by the Hep G2 cells to improve their resistance.Using SCP-MS1,single-cell proteome analysis could be performed with increased sensitivity and throughput.However,this approach had the drawback of requiring DDA results from a large number of cells to establish a reference peptide feature library,making it unsuitable for rare cell studies.Therefore,given the low utilization rate of the spectra in DDA analysis of single-cell and low-input samples,we developed a three-stage search strategy combining database reduction and retention time(RT)filtering to improve the sensitivity of protein and peptide identification.This strategy was simple and easy to implement without the need for additional data collection.The first step of this strategy was to merge the data to be analyzed and search them against the Uni Prot database with a loose FDR of 40%at both the PSM and protein levels to obtain as many proteins as possible that could be detected in the experiment.These identified proteins were used as a new reduced database,and then a second search was performed with 10%FDR to gain more reliable identifications.Finally,ΔRT filtering was used to retain correct identifications and filter out false identifications.Using this strategy,the number of identified peptides and proteins in 0.2 ng samples was increased by 143.1%and 87.0%on average with a final protein FDR of≤1%.Application of this strategy to the analysis of single neuron cell proteome data reported in the literature increased the number of differential proteins found by more than 50%.A number of nano-volume reaction systems have been developed to reduce protein loss due to surface adsorption and improve the digestion efficiency in the preparation of single-cell proteomic samples.However,these methods significantly increased the processing time and compromised convenience.Here,we proposed a temperature and ultrasonic assisted fast digestion method for the preparation of single-cell and low-input samples.In addition to reducing sample loss and improving the identification numbers,this method also greatly reduced sample preparation time.Cells were first lysed on ice for 5 min,denatured at 90℃for 10 min,and finally digested at 49℃for 20 min in an ultrasonic cleaner(single-cells were digested at 42℃for 30 min).All operations were performed in a 0.2 ml PCR tube without sample transfer.This method allowed the preparation to be completed within 35/45 min.It was then applied to the proteome profiling of different regions of the mouse brain.Approximately 2800 proteins were obtained from 0.04 mm~2 of tissue and heterogeneous spatial distribution of the identified proteins was discovered.Further analysis revealed that these differentially expressed proteins were related to metabolism and nervous system function.In short,this study improved the sensitivity and throughput of single-cell proteomics by innovating the LC-MS method,data analysis method and sample preparation method,which provided technical support for the large-scale application of single-cell proteomics.And the application potential of the methods developed in this study was demonstrated in the single-cell proteome profiling of sorafenib-resistant Hep G2 cells and the spatial proteomic study of mouse brain.