What’s New in Single-Cell Proteomics

Abstract

In recent years, single-cell proteomics has undergone significant evolution, enabling the analysis of thousands of proteins within single cells. This progress is driven by advances in experimental design, with maturing label-free and multiplexed methods, optimized sample preparation, and innovations in separation techniques, including ultra-low-flow nanoLC. These factors collectively contribute to improved sensitivity, throughput and reproducibility. Cutting-edge mass spectrometry platforms and data acquisition approaches continue to play a critical role in enhancing data quality. Furthermore, the exploration of spatial proteomics with single-cell resolution offers significant promise for understanding cellular interactions giving rise to various phenotypes. Single-cell proteomics has far-reaching applications in cancer research, biomarker discovery and developmental biology. Here we provide a critical review recent advances in the field of single-cell proteomics.

Graphical Abstract:

Introduction

Single-cell protein analysis using mass spectrometry has a rich, decades-long history dating back to the quantification of a small number of high-abundance proteins in human erythrocytes [1,2] and in-depth profiling of proteins in large cells such as oocytes [3,4]. The modern era of single-cell proteomics began ~6 years ago with reports of hundreds of proteins being profiled from typically sized mammalian cells [5,6]. These analyses were respectively enabled by the novel experimental design of Single Cell ProtEomics by Mass Spectrometry (SCoPE-MS) that incorporated a high-input isobaric carrier channel into a tandem mass tag (TMT) set [6], and the near-lossless sample preparation provided by Nanodroplet Processing in One pot for Trace Samples (nanoPOTS) [7]. Very rapid progress has been made since these early publications, and it is now possible to achieve a measurement throughput of hundreds of single cells per day and a profiling depth of thousands of proteins per cell (although these metrics cannot yet be obtained simultaneously). Accessibility, reproducibility and quantitative accuracy have also rapidly improved. We most recently reviewed the field of single-cell proteomics (SCP) in 2020 [8]. Here we briefly summarize salient developments that have taken place since that time.

Sample Preparation and Experimental Design

Label-free and multiplexed approaches to SCP are both being actively developed and improved. Clear advantages are emerging for each strategy, which may dictate one approach over the other for specific use cases.

Label-free techniques eliminate the need for reacting peptides with labelling compounds, relying instead on the direct quantification of proteins based on the mass spectral intensities of their constituent peptides. This approach is favored for its simplicity in sample preparation, broad dynamic range, and the avoidance of biases introduced by labels. Earlier workflows for label-free sample preparation [7] required custom robotics, extensive user intervention and overnight digestions, but these have been gradually replaced by fully or semi-automated workflows [9,10], a reduction in the number of reagents and digestion time [11], as well as lessening the need for expensive or custom equipment. For example, we recently isolated and prepared single cells using the HP D100 Single Cell Dispenser [12] (now Tecan Uno) as illustrated in Figure 1. Cells were dispensed into standard 384-well PCR plates and a 1-step sample preparation reagent mixture containing high-temperature stabilized trypsin/lys-C and the MS-compatible surfactant n-dodecyl-β-maltoside was added to the wells. Samples were ready to analyze after a single 1-h incubation. The 0.5 μL reagent volumes can also be manually pipetted if necessary, making the workflow accessible to virtually any proteomics laboratory that can isolate single cells into a well plate.

Figure 1.

A streamlined workflow using the Tecan Uno Single Cell Dispenser, providing cost-effective and accurate sample preparation for SCP. Adapted with permission from [13]. Copyright © 2023, American Chemical Society.

While label-free SCP has many advantages over multiplexed approaches, the substantial throughput advantages of multiplexing make it highly compelling for many applications. To date, multiplexed SCP has primarily utilized isobaric labels such as TMT, which are currently available in up to 18-plex sets. Single cells can be prepared individually and then pooled for analysis in an LC-MS run using, e.g., nested nanoPOTS [13], nPOP [14] or the proteoChip [15]. However, these approaches all rely on Cellenion’s cellenONE platform, which is not accessible to all proteomics labs. Low-loss sample preparation and pooling of multiplexed sets that can be accomplished using low-cost instrumentation will thus be beneficial.

Non-isobaric approaches to SCP multiplexing have recently been reported in the form of plexDIA [16] and mDIA [17], as have combined isobaric/non-isobaric approaches for higher-order multiplexing [18,19]. For example, our group developed hyperSCP based on a combination of TMT and stable isotope labeling by amino acids in cell culture (SILAC) using nested nanoPOTS chips for up to 28-plex SCP [18]. Budayeva et al. [19] developed improved software for selecting both the heavy and light SILAC pairs to increase data completeness. While all multiplexing approaches can potentially improve throughput, great care must be taken to minimize the challenges of added complexity in the mass spectra, distorted protein quantification due to precursor co-isolation, isotopic contamination, incomplete labeling, overlabeling, and reduced signal-to-noise ratios caused by a carrier proteome [20,21].

Both label-free and multiplexed workflows generally employ low-volume, one-pot preparations in microwells or nanowells. However, alternative approaches are being explored. For example, Gebreyesus et al. [22] developed microfluidic devices with integrated valves based on multilayer soft lithography to trap single cells and sequentially add reagents for sample preparation. Matsumoto et al. [23] developed an automated cell processing platform for TMT-based SCP, utilizing acoustic levitation to handle cell samples in an airborne environment, thus completely avoiding any sample losses associated with adsorption to container walls. Such approaches offer promise, provided they add functionality without significantly increasing the cost or complexity of the workflow.

Separations

High-performance separations are crucial for in-depth protein identification and quantification in single-cell proteomics. Ultra-low-flow liquid chromatography (nanoLC) and capillary electrophoresis separations operated below ~50 nL/min have proven effective at increasing ionization efficiency at the electrospray source, and thus providing greater MS signal per analyte molecule relative to standard proteomics separations performed at ~300 nL/min [24–28]. Such reduced flow rates are best achieved for nanoLC by reducing the bore of the separation column rather than operating a standard column at reduced pressures, as this allows an optimal linear velocity to be maintained. However, it is challenging to prepare narrow-bore columns with void-free stationary phase, and low-flow LC can lead to long sample loading times and low overall duty cycles. Maintaining high throughput separations while preserving superior ionization efficiency at low flow rates may be achieved using multiple trapping [29] and/or analytical columns [30]. Higher duty cycles can also be achieved by operating the separation at lower flow rates/pressures and then increasing the flow rate during the ‘overhead’ steps for faster sample loading, column washing and regeneration. This approach is implemented in both the Evosep and Thermo Vanquish Neo systems. While effective at increasing duty cycle, the separation takes place at a lower pressure than what the system is capable of, thus compromising on separation efficiency/peak capacity and leading to a tradeoff between duty cycle and separation performance.

An alternative to packed bed columns is microfabricated pillar array (μPAC) columns [31]. These promise to provide high peak capacity separations at moderate LC pump pressures due to well-ordered column beds achieved through photolithographic patterning rather than high-pressure particle packing [32,33]. In our experience, these provide similar performance to standard packed columns, which may indicate that some of the efficiency gains from the ordered separation beds are lost during transfer from the chip to the long transfer capillary and electrospray emitter. A μPAC column with an integrated emitter may achieve the best of all worlds if it were developed.

Given the costs and challenges associated with alternatives to standard 75 μm bore commercial packed capillaries, it is certainly fair to question whether such alternatives are worthwhile to implement. To partially address this, we recently compared home-packed 30 μm i.d. columns operated at ~40 nL/min to commercial 50 μm i.d. columns operated at 100 nL/min and found that the narrow-bore column provided a >50% increase in proteome coverage for single HeLa cells using the Orbitrap Exploris 480 platform [12]. This and prior studies [34,35] show there are clear benefits derived from miniaturizing LC separations, but such gains should be balanced against potential robustness challenges, the need for custom instrumentation, etc.

MS Instrumentation & Data Acquisition

For many years, deciding which mass spectrometer to utilize for SCP, and indeed all proteomics analyses, was a simple one, as the Thermo Orbitrap line of mass spectrometers was essentially peerless in the global proteomics space. Similarly, data-dependent acquisition (DDA) has historically been the dominant data acquisition mode for all global proteomics studies. However, the rise of single-cell proteomics over the past few years has coincided with the introduction and rapid improvement of the trapped ion mobility spectrometry time-of-flight (timsTOF) line of instrument from Bruker Corp., as well as the maturing of data-independent acquisition (DIA) software and more efficient data acquisition schemes [36–38]. The timsTOF provides lower resolution mass spectra than what an orbitrap is capable of, but the up-front TIMS ion mobility separation compensates by adding overall resolving power. In addition, the dual-region TIMS tunnel enables one ion packet to be separated while the subsequent ion packet is accumulated, providing high ion utilization efficiency. This ‘parallel acquisition-serial fragmentation’ (PASEF) strategy is especially beneficial for limited samples such as in SCP. As such, researchers must now choose between DIA and DDA (and the hybrid wide-window data-dependent acquisition (WWA) [26,39], as well as which MS platform to select. Further, the Orbitrap Astral platform from Thermo Scientific has recently been released, which enables high-resolution MS2 spectra to be rapidly acquired and provides deep proteome coverage in a fraction of the time that was previously required [40–42]. The SCP researcher now faces multiple decisions, and given the very recent releases of both the timsTOF Ultra and Orbitrap Astral platforms, it will take time to match specific use cases with the optimal MS platform.

Still, some decisions are clear. For isobaric labeling workflows, the Orbitrap line of instruments remains the best choice, as these instruments alone can resolve the isotopologous reporter ions in e.g. TMT and TMTpro sets. Similarly, despite some exploratory studies with alternative approaches, isobaric labeling should generally use DDA with narrow isolation windows to minimize precursor co-isolation. For label-free studies, there is a clear trend towards DIA, particularly for the timsTOF platform [43], and the timsTOF SCP/Ultra and the Orbitrap Astral represent tremendous advances for the proteomics field in general and SCP field more specifically. However, access to the latest equipment should not be a barrier to entry into SCP, and prior generation mass spectrometers can still be used to great effect. For example, the Slavov lab recently utilized an older Q-Exactive instrument with a novel acquisition technique termed prioritized SCoPE (pSCoPE) [44] to maintain high proteome coverage and data completeness, and we profiled >3,000 proteins per cell in a label-free workflow utilizing a previous-generation Orbitrap Exploris 480 instrument in combination with WWA [26].

Data Analysis

As MS instrumentation continues to advance in speed and sensitivity, the resulting datasets become increasingly complex. In single-cell proteomics, a key challenge is the substantial reduction of ions introduced into the mass spectrometer, impacting the characteristics of collected MS2 fragmentation data compared to bulk samples [45]. Hence, there is a pressing need for data analysis tools that are tailored to the unique challenges of SCP.

Comparative studies between various software packages such as DIA-NN/Spectronaut etc. [46] for DIA, or between DDA data processing tools like Proteome Discoverer/MaxQuant etc. [47], have been undertaken, although the findings may quickly become dated due to the rapid software development cycles. Efforts are also being directed towards improving existing platforms [48,49] and refining algorithms for preprocessing, feature extraction, normalization, quantification and noise reduction [50]. Some approaches leverage additional features for matching, such as ion mobilities or FAIMS compensation voltages along with retention times and accurate masses [51]. Furthermore, advances in peptide-spectrum matching through deep learning are being explored [52,53]. Ultimately, to discern whether variations between cells result from their inherent heterogeneity or are merely due to technical variation, it is crucial to employ adequate statistical power and a well-defined experimental design [54–56]. In addition, as SCP datasets inevitably grow to comprise thousands of cells, there will be an urgent need for fully automated data management, processing, statistical treatment, etc. Thermo’s recent Ardia platform aspires to provide such an end-to-end solution, and open access, vendor-neutral options will be invaluable for managing such large datasets.

Beyond Global Proteome Profiling of Dissociated Cells

Although single-cell proteomics has seen notable progress in recent years, the addition of spatial distribution of proteins within tissues at single-cell resolution remains at an early stage. Spatial SCP generally relies on laser capture microdissection (LCM), enabling the isolation of specific cells or regions of interest from frozen or formalin-fixed, paraffin-embedded (FFPE) tissues [57]. When combined with high-content imaging, artificial intelligence, and ultrasensitive multiplexed mass spectrometry, this approach can provide tissue phenotyping at single-cell resolution [58,59]. However, LCM remains a tedious, low-throughput process, and the need for special membrane-coated slides poses challenges for tissue adhesion and optical imaging. To address these limitations, Zhu et al. developed spatial proteomics based on deep ultraviolet laser ablation directly from tissues affixed to standard glass slides. This approach provides greater spatial resolution and improved protein extraction due to the formation of small tissue particles during the laser ablation process [60].

Post-translational modifications (PTMs) play a crucial role in regulating protein function and cellular processes. By harnessing TIMS in conjunction with the parallel accumulation serial fragmentation for reporter ion quantification (pasefRiQ), multiple classes of protein PTMs have been quantified at the single-cell level [61]. We expect much more development and application of PTM analysis within single cells, but clear challenges remain such as the low stoichiometry of such modifications, etc.

Conclusion

Single-cell proteomics has advanced rapidly, enabling the analysis of thousands of proteins within single cells. This progress has been driven by improvements in experimental design, sample preparation, separation techniques, MS instrumentation and data analysis.

Label-free and multiplexed methods are both being actively developed, and each approach possesses unique advantages. Advances in sample preparation have resulted in simplified workflows and optimized reagents, making the process more affordable and efficient. Within the domain of separation techniques, innovations such as ultra-low-flow nanoLC, microfabricated pillar array columns, and capillary electrophoresis have significantly improved sensitivity and resolution.

Although in its nascent stages, the exploration of protein distribution across tissues with single-cell spatial resolution holds promise, with techniques such as LCM and laser ablation combined with ultrasensitive analytical workflows promising high-resolution tissue phenotyping. Applications of single-cell proteomics are expected to be highly impactful, particularly in cancer research, where it is expected to aid in biomarker identification, drug response analysis, and the study of circulating tumor cells [62].

Looking to the future, single-cell proteomics holds great potential for shaping personalized medicine, diagnostics, and therapeutic advancements. Challenges remain in terms of limited depth of proteome coverage per cell and the few cells (hundreds) that can be measured per day, but there are clear paths forward to dramatically improve on both fronts. As such, SCP stands at the threshold of uncovering novel frontiers in life sciences and effectively addressing urgent challenges within medicine and biology.

Figure 2.

Relative ion utilization efficiency for model peptides in a direct infusion experiment. Ion utilization (MS peak intensity divided by amount of introduced analyte) increases by a factor of ~32 as flow rate decreases from 1 μL/min to 10 nL/min.

Highlights:

• Recent advances in single-cell proteomics

• Improvements in sample preparation, separation techniques, mass spectrometry instrumentation, and data acquisition/analysis