Challenges and Opportunities for Single-cell Computational Proteomics.

Single-cell proteomics is growing rapidly and has made several technological advancements. As most research has been focused on improving instrumentation and sample preparation methods, very little attention has been given to algorithms responsible for identifying and quantifying proteins. Given the inherent difference between bulk data and single-cell data, it is necessary to realize that current algorithms being employed on single-cell data were designed for bulk data and have underlying assumptions that may not hold true for single-cell data. In order to develop and optimize algorithms for single-cell data, we need to characterize the differences between single-cell data and bulk data and assess how current algorithms perform on single-cell data. Here, we present a review of algorithms responsible for identifying and quantifying peptides and proteins. We will give a review of how each type of algorithm works, assumptions it relies on, how it performs on single-cell data, and possible optimizations and solutions that could be used to address the differences in single-cell data.

HyperSCP: Combining Isotopic and Isobaric Labeling for Higher Throughput Single-Cell Proteomics.

Recent developments in mass spectrometry-based single-cell proteomics (SCP) have resulted in dramatically improved sensitivity, yet the relatively low measurement throughput remains a limitation. Isobaric and isotopic labeling methods have been separately applied to SCP to increase throughput through multiplexing. Here we combined both forms of labeling to achieve multiplicative scaling for higher throughput. Two-plex stable isotope labeling of amino acids in cell culture (SILAC) and isobaric tandem mass tag (TMT) labeling enabled up to 28 single cells to be analyzed in a single liquid chromatography-mass spectrometry (LC-MS) analysis, in addition to carrier, reference, and negative control channels. A custom nested nanowell chip was used for nanoliter sample processing to minimize sample losses. Using a 145-min total LC-MS cycle time, ∼280 single cells were analyzed per day. This measurement throughput could be increased to ∼700 samples per day with a high-duty-cycle multicolumn LC system producing the same active gradient. The labeling efficiency and achievable proteome coverage were characterized for multiple analysis conditions.

Calculating Sample Size Requirements for Temporal Dynamics in Single-Cell Proteomics.

Single-cell measurements are uniquely capable of characterizing cell-to-cell heterogeneity and have been used to explore the large diversity of cell types and physiological functions present in tissues and other complex cell assemblies. An intriguing application of single-cell proteomics is the characterization of proteome dynamics during biological transitions, like cellular differentiation or disease progression. Time-course experiments, which regularly take measurements during state transitions, rely on the ability to detect dynamic trajectories in a data series. However, in a single-cell proteomics experiment, cell-to-cell heterogeneity complicates the confident identification of proteome dynamics as measurement variability may be higher than expected. Therefore, a critical question for these experiments is how many data points need to be acquired during the time course to enable robust statistical analysis. We present here an analysis of the most important variables that affect statistical confidence in the detection of proteome dynamics: fold change, measurement variability, and the number of cells measured during the time course. Importantly, we show that datasets with less than 16 measurements across the time domain suffer from low accuracy and also have a high false-positive rate. We also demonstrate how to balance competing demands in experimental design to achieve a desired result.

Data-Dependent Acquisition with Precursor Coisolation Improves Proteome Coverage and Measurement Throughput for Label-Free Single-Cell Proteomics.

We combined efficient sample preparation and ultra-low-flow liquid chromatography with a newly developed data acquisition and analysis scheme termed wide window acquisition (WWA) to quantify >3,000 proteins from single cells in rapid label-free analyses. WWA employs large isolation windows to intentionally co-isolate and co-fragment adjacent precursors along with the selected precursor. Optimized WWA increased the number of MS2-identified proteins by ≈40 % relative to standard data-dependent acquisition. For a 40-min LC gradient operated at ≈15 nL/min, we identified an average of 3,524 proteins per single-cell-sized aliquot of protein digest. Reducing the active gradient to 20 min resulted in a modest 10 % decrease in proteome coverage. Using this platform, we compared protein expression between single HeLa cells having an essential autophagy gene, atg9a, knocked out, with their isogenic WT parental line. Similar proteome coverage was observed, and 268 proteins were significantly up- or downregulated. Protein upregulation primarily related to innate immunity, vesicle trafficking and protein degradation.

Features of Peptide Fragmentation Spectra in Single-Cell Proteomics.

The goal of proteomics is to identify and quantify the complete set of proteins in a biological sample. Single-cell proteomics specializes in the identification and quantitation of proteins for individual cells, often used to elucidate cellular heterogeneity. The significant reduction in ions introduced into the mass spectrometer for single-cell samples could impact the features of MS2 fragmentation spectra. As all peptide identification software tools have been developed on spectra from bulk samples and the associated ion-rich spectra, the potential for spectral features to change is of great interest. We characterize the differences between single-cell spectra and bulk spectra by examining three fundamental spectral features that are likely to affect peptide identification performance. All features show significant changes in single-cell spectra, including the loss of annotated fragment ions, blurring signal and background peaks due to diminishing ion intensity, and distinct fragmentation pattern, compared to bulk spectra. As each of these features is a foundational part of peptide identification algorithms, it is critical to adjust algorithms to compensate for these losses.

Simplified and Unified Access to Cancer Proteogenomic Data.

Comprehensive cancer data sets recently generated by the Clinical Proteomic Tumor Analysis Consortium (CPTAC) offer great potential for advancing our understanding of how to combat cancer. These data sets include DNA, RNA, protein, and clinical characterization for tumor and normal samples from large cohorts of many different cancer types. The raw data are publicly available at various Cancer Research Data Commons. However, widespread reuse of these data sets is also facilitated by easy access to the processed quantitative data tables. We have created a data application programming interface (API) to distribute these processed tables, implemented as a Python package called cptac. We implement it such that users who prefer to work in R can easily use our package for data access and then transfer the data into R for analysis. Our package distributes the finalized processed CPTAC data sets in a consistent, up-to-date format. This consistency makes it easy to integrate the data with common graphing, statistical, and machine-learning packages for advanced analysis. Additionally, consistent formatting across all cancer types promotes the investigation of pan-cancer trends. The data API structure of directly streaming data within a programming environment enhances the reproducibility. Finally, with the accompanying tutorials, this package provides a novel resource for cancer research education. View the software documentation at https://paynelab.github.io/cptac/. View the GitHub repository at https://github.com/PayneLab/cptac.

Fully Automated Sample Processing and Analysis Workflow for Low-Input Proteome Profiling.

Recent advances in sample preparation and analysis have enabled direct profiling of protein expression in single mammalian cells and other trace samples. Several techniques to prepare and analyze low-input samples employ custom fluidics for nanoliter sample processing and manual sample injection onto a specialized separation column. While being effective, these highly specialized systems require significant expertise to fabricate and operate, which has greatly limited implementation in most proteomic laboratories. Here, we report a fully automated platform termed autoPOTS (automated preparation in one pot for trace samples) that uses only commercially available instrumentation for sample processing and analysis. An unmodified, low-cost commercial robotic pipetting platform was utilized for one-pot sample preparation. We used low-volume 384-well plates and periodically added water or buffer to the microwells to compensate for limited evaporation during sample incubation. Prepared samples were analyzed directly from the well plate with a commercial autosampler that was modified with a 10-port valve for compatibility with 30 µm i.d. nanoLC columns. We used autoPOTS to analyze 1-500 HeLa cells and observed only a moderate reduction in peptide coverage for 150 cells and a 24% reduction in coverage for single cells compared to our previously developed nanoPOTS platform. To evaluate clinical feasibility, we identified an average of 1095 protein groups from ∼130 sorted B or T lymphocytes. We anticipate that the straightforward implementation of autoPOTS will make it an attractive option for low-input and single-cell proteomics in many laboratories.

Simple and Efficient Data Analysis Dissemination for Individual Laboratories.

Scientific progress comes as we build upon the work of others. Implicit in this advance is that we have access to and can thoroughly examine the work of others. It is important to recognize that our scholarly work as scientists encompasses not only experimental design and data collection but also our analytical methods. Thus when communicating biology experiments, especially those that utilize molecular omics data, the analysis methods that connect raw data to scientific conclusions must be presented with sufficient clarity that others can reproduce our exact work. Although there are many resources for sharing raw data files, there is currently not a widely utilized method for sharing analysis methods. We present a semistructured pattern for sharing analysis methods that is simple and efficient and can be implemented by individual laboratories using existing software. This pattern requires three types of files in a publicly accessible repository, such as GitHub: (1) data files, (2) a universal I/O script that parses all data files, and (3) analysis scripts creating figures and metrics reported in the manuscript. We suggest additional conventions to improve the readability and provide a template repository for the pattern. Sharing our exact analysis methods as software, in addition to their narrative description in a manuscript, will ensure reproducibility and transparency. Importantly, the pattern we present does not require new infrastructure and can be achieved without advanced computing skills.

Pushing the Isotopic Envelope: When carrier channels pollute their neighbors' signals.

Individual cells are the foundational unit of biology, and understanding their functions and interactions is critical to advancing our understanding of health and disease. Single cell proteomics has seen intense interest from mass spectrometrists, with a goal of quantifying the proteome of single cells by adapting current techniques used in bulk samples. To date, most method optimizations research has worked towards increasing the proteome coverage of single cells. One prominent technique multiplexes many individual cells into a single data acquisition event using isobaric labels. Accompanying the single cells, one label is typically used for a mixed set of many cells, called a carrier or boost channel. Although this improves peptide identification rates, several groups have examined the impact on quantitative accuracy as more cells are included in the carrier channel, e.g. 100x or 500x. This manuscript explores how impurities in the multiplexing reagent can lead to inaccurate quantification observed as a measurable signal in the wrong channel. We discover that the severe abundance differential between carrier and single cell, combined with the reagent impurities, can overshadow several channels typically used for single cells. For carrier amounts 100x and above, this contamination can be as abundant as true signal from a single cell. Therefore, we suggest limiting the carrier channel to a minimal amount and balance the goals of identification and quantification.

TDP-43-stratified single-cell proteomics of postmortem human spinal motor neurons reveals protein dynamics in amyotrophic lateral sclerosis.

A limitation of conventional bulk-tissue proteome studies in amyotrophic lateral sclerosis (ALS) is the confounding of motor neuron (MN) signals by admixed non-MN proteins. Here, we leverage laser capture microdissection and nanoPOTS single-cell mass spectrometry-based proteomics to query changes in protein expression in single MNs from postmortem ALS and control tissues. In a follow-up analysis, we examine the impact of stratification of MNs based on cytoplasmic transactive response DNA-binding protein 43 (TDP-43)+ inclusion pathology on the profiles of 2,238 proteins. We report extensive overlap in differentially abundant proteins identified in ALS MNs with or without overt TDP-43 pathology, suggesting early and sustained dysregulation of cellular respiration, mRNA splicing, translation, and vesicular transport in ALS. Together, these data provide insights into proteome-level changes associated with TDP-43 proteinopathy and begin to demonstrate the utility of pathology-stratified trace sample proteomics for understanding single-cell protein dynamics in human neurologic diseases.

TDP-43-stratified single-cell proteomic profiling of postmortem human spinal motor neurons reveals protein dynamics in amyotrophic lateral sclerosis.

Unbiased proteomics has been employed to interrogate central nervous system (CNS) tissues (brain, spinal cord) and fluid matrices (CSF, plasma) from amyotrophic lateral sclerosis (ALS) patients; yet, a limitation of conventional bulk tissue studies is that motor neuron (MN) proteome signals may be confounded by admixed non-MN proteins. Recent advances in trace sample proteomics have enabled quantitative protein abundance datasets from single human MNs (Cong et al., 2020b). In this study, we leveraged laser capture microdissection (LCM) and nanoPOTS (Zhu et al., 2018c) single-cell mass spectrometry (MS)-based proteomics to query changes in protein expression in single MNs from postmortem ALS and control donor spinal cord tissues, leading to the identification of 2515 proteins across MNs samples (>900 per single MN) and quantitative comparison of 1870 proteins between disease groups. Furthermore, we studied the impact of enriching/stratifying MN proteome samples based on the presence and extent of immunoreactive, cytoplasmic TDP-43 inclusions, allowing identification of 3368 proteins across MNs samples and profiling of 2238 proteins across TDP-43 strata. We found extensive overlap in differential protein abundance profiles between MNs with or without obvious TDP-43 cytoplasmic inclusions that together point to early and sustained dysregulation of oxidative phosphorylation, mRNA splicing and translation, and retromer-mediated vesicular transport in ALS. Our data are the first unbiased quantification of single MN protein abundance changes associated with TDP-43 proteinopathy and begin to demonstrate the utility of pathology-stratified trace sample proteomics for understanding single-cell protein abundance changes in human neurologic diseases.

Proteogenomic and metabolomic characterization of human glioblastoma.

Glioblastoma (GBM) is the most aggressive nervous system cancer. Understanding its molecular pathogenesis is crucial to improving diagnosis and treatment. Integrated analysis of genomic, proteomic, post-translational modification and metabolomic data on 99 treatment-naive GBMs provides insights to GBM biology. We identify key phosphorylation events (e.g., phosphorylated PTPN11 and PLCG1) as potential switches mediating oncogenic pathway activation, as well as potential targets for EGFR-, TP53-, and RB1-altered tumors. Immune subtypes with distinct immune cell types are discovered using bulk omics methodologies, validated by snRNA-seq, and correlated with specific expression and histone acetylation patterns. Histone H2B acetylation in classical-like and immune-low GBM is driven largely by BRDs, CREBBP, and EP300. Integrated metabolomic and proteomic data identify specific lipid distributions across subtypes and distinct global metabolic changes in IDH-mutated tumors. This work highlights biological relationships that could contribute to stratification of GBM patients for more effective treatment.