Revealing New Biology from Multiplexed, Metal Isotope-tagged, Single-cell Readouts

Abstract

Mass cytometry (MC) is a recent technology that pairs plasma-based ionization of cells in suspension with time-of-flight mass spectrometry to sensitively quantify single cell abundance of metal isotope-tagged affinity reagents to key proteins, RNA and peptides. Given the ability to multiplex readouts (~50 per cell) and capture millions of cells per experiment, MC offers a robust way to assay rare, transitional cell states that are pertinent to human development and disease. Here, we review MC approaches that let us probe dynamics of cellular regulation across multiple conditions and sample types in a single experiment. Additionally, we discuss current limitations and future extensions of MC as well as computational tools commonly used to extract biological insight from single-cell proteomic datasets.

Mass cytometry: Expanding on traditional proteomic tools

Various breakthrough findings in the past 3 decades have been the result of accurately quantifying specific biological molecules such as protein or RNA in a group of cells or in more recent years, in single cells[1]. The ability to capture the change in abundance of a molecule-of-interest across experimental conditions and samples has spearheaded the deduction of cell types, signalling pathways, cellular response to perturbation such as drugs and disease-driven dysregulation of cellular function. Antibodies are the most widely used affinity reagent for proteins; and their first application was in providing a measurable surrogate for protein-of-interest in bulk assays (e.g., western blots) [2]. However, sufficient cell lysate (from 0.1–2 million cells) is necessary for robust detection of protein-of-interest [2]. This bulk method was not amenable to the study of human tissues such as bone marrow given the inherent cell-to-cell variation and low cell numbers present. Given the complexity of human biology, researchers need tools that can capture a more expansive picture at a single cell resolution to derive fundamental understandings of cellular behaviour that can be leveraged for new biotechnologies, diagnostics, and therapeutics.

With the application of micro-fluidics to cell suspensions, single cell detection became possible [3,4]. Combined with the development of more sensitive reagents for antibody labelling such as lanthanide metal chelate complexes and fluorescent molecules, single cell readouts of a few proteins-of-interest (up to 8) was possible [5–7]. However, these technologies have a limited ability to measure multiple targets simultaneously due to spectral bleeding between fluorophores and additionally, can suffer from signal artifacts such as autofluorescence (Fig 1a). In contrast, metal lanthanide isotopes as molecular labels that are measured by Inductively Coupled Plasma Mass Spectrometry (ICP-MS) introduced a new approach which directly benefited from the high sensitivity and specificity of lanthanide isotopes, as well as expanded the number of simultaneous readouts captured (Fig 1b) [5]. Given the mass-charge ratio is relatively consistent per metal isotope, the lanthanide signals have comparatively minimal signal overlap compared to spectral overlap in fluorescent readouts. Additionally, there is no endogenous signal interference given absence of heavy metal isotopes in biological tissues (unlike autofluorescence in flow cytometry) resulting in higher signal-to-noise ratios in MC data (Fig 1).

Figure 1.

Comparison of signal detection of fluorophore-conjugated affinity reagents versus metal isotope-conjugated affinity reagents. a) Fluorescent flow cytometry uses affinity reagents linked with fluorophore molecules to measure abundance of target molecules in individual cells using light-based detection. b) Single-cell mass cytometry uses affinity reagents conjugated to stable, elementally pure lanthanide metal isotopes to measure multiplex abundance of target molecules in individual cells using mass-spectrometry based elemental analysis.

These advantages, paired with advancements in chemical conjugation of stable lanthanide metal-isotope polymer chelates to affinity reagents (i.e., purified antibody or nucleic acid probes) allowed for the development of high-throughput and highly multiplexed single-cell proteomic assay with mass cytometry [8]. Mass cytometry (MC) combines high cellular throughput with the sensitivity and parameterization of Time-of-Flight (TOF) mass spectrometry to capture relative protein expression in single cells (in comparison to bulk assays like western blot, ELISA or other approaches utilizing spectrally confined fluorescent reporter molecules to measure biomolecule abundance) [8,9]. With MC, researchers can routinely profile the targeted single-cell expression of >40 protein, RNA, post-translational modifications, or other molecular targets, in addition to general cellular parameters (i.e. DNA content, viability, cell cycle phase) at a throughput of up to 400 cells per second, and millions of cells per day (Fig 2a) [10–13]. The multiplexed parameter space and cellular throughput of MC enables exploration of entire cell-systems in one round of data acquisition, as has been demonstrated in the human immune system (Fig 2a).

Figure 2.

Applications of mass cytometry to investigate cellular composition and behaviour of biological systems. Single-cell mass cytometry uses affinity reagents tagged with purified metal isotope reporters to measure multiplex expression of target molecules. a) Cellular features which are commonly measured using mass cytometry and a general workflow of a MC experiment. b) Mass cytometry analysis is best suited for biological systems which can be isolated as single cells, such as in vitro maintained cell lines or collected primary tissue. Demonstrated applications of MC can be used individually or in combination to investigate cellular behaviours underlying biological systems. Computational analysis of collected single-cell data enables identification of underlying cellular features, such as cell clusters or expression dynamics during cellular progression.

The advent of technologies like MC allows us to design experiments to better understand human biology and particularly the complex, coordinated processes in human development and homeostasis with subsequent dysregulation in disease (Fig 2b). We can accurately define multiplexed cell state and function simultaneously and in rare cell populations, which are pertinent in human systems (Fig 2b). This review explores the different ways MC accelerates the discovery of new biology and enables systems-minded interrogation of key biological processes, and more importantly how researchers can apply new experimental and computational techniques to better answer their biological questions. Furthermore, within the current context of multiple competing single cell technologies, we discuss the place of MC in this evolving landscape and the ways MC technology can complement and augment other widely used assays.

Mass cytometry enables novel cell type discovery through deep and robust cellular profiling

The earliest developed and most widely used application of MC is profiling immune cell frequency in blood or tissue (i.e., immunophenotyping) [14,15]. This approach captures abundance of lineage-defining (e.g., CD3, CD57) and functional molecules in millions of immune cells found in a blood or tissue sample. In general, the experimental design constitutes the collection of primary tissue specimens of immune cells (typically peripheral blood or bone marrow) from a pre-determined cohort followed by analysis with MC to examine the relative abundance and activity of multiple immune cell types across different experimental groups. As major immune subsets are readily identified using only a small fraction of the MC panel of markers, researchers can in parallel assess other surface and/or intracellular factors that inform immune cell function such as exhaustion markers (e.g., PD1, LAG3) or intracellular cytokine levels (e.g., Interferon-gamma, Interleukin-15) [14,15].

The immunophenotyping capacity of MC technology makes it well-suited to clinical trial applications, (e.g. immunotherapy) where we do not yet understand the complex effects of and responses to chimeric antigen receptor-based cell therapy or checkpoint blockade biologics [16–18]. We need to better chart changes in immune cell state and function across patients to improve immunotherapy targets and/or design combination therapies. To this effect, large patient cohorts in clinical trials are run on MC and use sample and live cell barcoding, and antibody panel lyophilization to control for experimental variation (e.g. antibody staining) and doublet contamination (Table 1) [19–21]. To ensure consistent measurement, data from 5-metal beads run with sample is used to normalize for time-of-flight acquisition inconsistency across a single barcoded or multiple runs. Lastly, signal spillover in MC data from adjacent (+1 mass) and oxidised (+16 mass) metal isotopes disproportionately affects low abundance markers (e.g. T cell exhaustion proteins) and reduces accuracy of measurement. We highlight recent tools to compensate for these linear spillover effects and ensure accurate relative protein quantification in single cells [22].

Table 1.

Sample barcode modalities for mass-cytometry

Barcode Type	Specific Application	Detection Approach	Barcoding	Capacity	Ref
Genetic barcode	Examine targeted gene function in individual cells	Antibody panel dedicated to peptide-barcodes expressed with gRNA (uses panel space)	Genetically encoded surface-expressed peptide barcode	100+ distinct	20
Antigen barcode	Antigen-specific T-cell identification	Lanthanide isotope conjugated to tetramer (uses panel space)	Peptide-MHC with unique mass barcode	10–20 Specific pMHC	43,44
Panel/cell population barcode	Screen surface expressed molecules in specific cell population	Antibody panel to identify major populations across all screen panels	Conserved phenotype sub-panel unifies all panels	380 molecules across 12 panels (ability to add more panels)	45
Mass-tag barcode	Compare experimental treatments (small molecules, differentiation, etc.,) across samples	Non-antibody conjugation channels used (do not need to sacrifice antibody space)	Covalently linked to cell. Can be performed live (demonstrated for organoid) or on fixed cells	20 conditions – limited by mass-tag barcode reagents	8,12
Antibody barcode/Live Cell barcode	Compare immunophenotyping across patient samples	Antibody against commonly expressed surface epitope	Antibodies against broadly expressed surface epitope	20 samples/conditions	25

While immune phenotyping experiments are predominately employed on peripheral blood and bone marrow, similar efforts are demonstrated in other cell systems and tissues, such as the central nervous system and muscle [23–25]. Additionally, profiling multiple interacting immune cell types in these complex systems can quantify cooperation between distinct cell populations and/or markers and reveal features obscured in bulk assays. Collectively, these investigations highlight a broad utility of MC to capture immune cell frequencies and function. We highlight its application to other complex and immune-infiltrated tissue such as lymphoid organs or gut to better understand differences between circulating and tissue-resident immune cells and cell state changes as they move between these compartments.

Owing to the large number (1–10 million cells) of high dimensional (~50 features) cell states captured per sample and number of samples (20–50 samples per barcoded run) collected in 1 experiment, the resulting sizable and complex MC datasets require well-designed computational pipelines (Fig 3a) (Table 2) [26]. Often both the quantification of known cell type frequencies as well as discovery of novel cell states are a key focus of such studies [27]. To this effect, computational approaches like cell clustering identifies conserved states held by a population of measured cells in immunophenotyping and more broadly in cellular profiling studies (Fig 3b) [28]. Current clustering tools allow researchers to identify conserved cell states and discover rare cell populations (i.e. less abundant cell state) which are then validated through functional readouts and assays (Fig 3b) [29–34]. Extending this into a clinical setting, cell clustering can be paired with regression analysis using independent clinical features such as treatment response or disease stage to identify diagnostically or therapeutically relevant cell populations and markers [35,36].

Figure 3.

Computational techniques for MC data analysis to find novel cell subsets and transitions in complex biological systems. a) MC data typically is represented in a data frame with features (measured target molecules) as columns and individual cells as rows. b) Single-cell data points embedded in a high dimensional space represented by selected features are grouped together or clustered by calculating the similarity of their target molecule expression c) Cells are projected onto a 2-dimensional space from the chosen multi-dimensional feature space, where spatial layout in 2 dimensions captures the global and local structure of data in the multi-dimensional feature space. Uniform Manifold Approximation and Projection (UMAP) is one commonly used algorithm to carry out dimensionality reduction. d) Cells in asynchronous, dynamic processes such as differentiation occupy different transitionary states that can be modelled as cellular trajectories through differentiation time.

Table 2.

Computational approaches to analyze mass cytometry data

Biological question	Computational Approach	Applications	Algorithm Developed/Used	Biological applications	Ref
Cell population identification	Clustering	Identify, organize cell types, compare abundances across conditions	SPADE, PhenoGraph, flowSOM, X-Shift	Immune cell subset identification, novel signaling/metabolic state identification	46–50,52,53
Global data relationships	Dimensionality Reduction	Visualize global and local relationships in single-cell data	PCA, LDA, tSNE, UMAP, Diffusion maps, PHATE	Visualize global structure of dataset, identify cells/features for further analysis	55–62
Developmental pseudo-time	Trajectory Inference	Trace cell trajectories, calculate pseudo-time	Wanderlust, wishbone, Scorpius, monocole, PAGA	Mapping developmental progression	38,63–65
Novel biomarker discovery	Supervised analysis	Identify cell populations corresponding with condition of interest	Citrus, Cydar, diffcyt, MetaCyto, FAUST, cellCNN	Identifying biomarkers for disease diagnosis/progression or for drug resistance	66–71
Relationship between different regulatory landscapes	Multi-omic analysis	Integrate single-cell datasets	Modified, nested elastic net models	Identifying features of disease progression	72

Traditionally, cellular profiling is largely limited to cell type information from cell surface markers. MC can also capture diverse cellular regulation information such as signalling, metabolism, global translational activity, chromatin content and lineage priming. The following sections discuss key open questions in various biological fields that are being answered by diverse applications of the MC technology. We also highlight MC extensions and integration with orthogonal techniques which further extends MC capabilities.

Multiplexed signalling states of cells reveal transient and coordinated signalling network activity

The activity of intracellular signalling pathways underlies cellular response to changes in the surrounding environment. Signalling pathway activation is a multifaceted process, which typically starts with ligand binding receptor leading to Post-Translational Modification (PTM) of downstream targets in a tightly regulated fashion to solicit a specific cellular response such as change in gene expression (e.g. T cell receptor activation) [37]. Monitoring this multifaceted process is difficult due to inherent stochasticity of cell signalling response and degeneracy in pathways [38]. Therefore, approaches such as MC which can capture multiple single cell signalling proteins are useful in measuring fractional responses, dynamics and cooperation/antagonism of multiple pathways.

MC allows capture of affinity reagents which recognize specific PTMs (e.g., phosphorylation, ubiquitin, methylation) implicated in various signalling pathways to interrogate and profile responses to cellular perturbation. A well-studied and clinically relevant signalling phenomenon is primary T-cell signalling dynamics in response to T-cell receptor engagement [37]. MC was effectively applied to quantify the fast and coordinated signalling dynamics upon TCR binding [39]. Signalling activity of primary mouse T-cells revealed consistent pathway activity in naïve and memory subsets, but stronger signalling response in naive T-cells [39]. Researchers can similarly uncover pathway dynamics and marker relationships in MC data using their mutual information metric [39].

The stochastic nature of cell signalling responses of individual cells within larger populations can be readily captured with MC, as seen in a study examining the diversity of human B cell phenotype and function [40]. Investigations of intracellular signalling in additional immune settings, co-culture organoid models of oncogenesis, and even during cellular reprogramming demonstrate a common approach to examine signal pathway activity of individual cells in diverse biological systems using MC [41–44]. Finally, these investigations which capture receptor signalling cascades alongside cell profiles illustrate the ability to scale MC applications down to investigate signal pathway cascades in individual cell populations or up to examine interacting cell systems (e.g. immune cells in peripheral blood versus in tissue).

Probing novel cell states and manifold learning tools to understand cell regulation

Apart from quantifying signal pathway activity, MC can also be used to robustly approximate cellular biosynthesis activity and metabolic and epigenetic states. Global macromolecule biosynthesis and metabolic pathway activity can be tracked in cellular processes and transitions as delineated by other phenotypic markers in the same experiment. Given cellular proliferation and DNA synthesis is highly regulated to ensure homeostasis of tissue progenitor pools while preventing dysplasia, approaches that assess phenotype alongside cell cycle will be able to identify cell types or developmental periods with altered cellular division [13]. Pyrimidine analog, 5-Iodo-deoxyuridine, is routinely used to label cells in S phase undergoing DNA replication actively for measurement on MC [12]. Additionally, small molecule reagents can be used to approximate global RNA and protein synthesis, which were used to capture the fine-scale global repression of these activities during cell cycle progression into mitosis, and their dynamic activities in human immune cell subsets during ex vivo activation [45]. Finally, a newly developed amino-acid analog containing a heavy metal atom of tellurium was demonstrated as an additional approach to approximate protein synthesis activity in cell lines and in mice [46]. Collectively, these MC methods enable parallel approximation of cellular biosynthesis activity and importantly can provide unmatchable interrogation of cell identity and underlying biosynthesis activities in complex biological settings when used in tandem with cell phenotype analysis.

In addition to cellular biosynthesis, the metabolic status of single cells informs their current function as well as ability to respond to environmental signals. However, few approaches can capture cellular metabolism characteristics in individual cells. A new MC approach assays single-cell metabolic states, specifically in primary human T-cells under various conditions [47]. To do this, a MC panel with antibodies to rate-limiting enzymes and major metabolic regulator function was developed. Markers were organized into metabolic modules, such as glycolysis and amino acid transport [47]. To benchmark this, a single-cell metabolic score from the cumulative expression of metabolic targets was compared to gold standard metabolic consumption assays [47]. This novel metabolic MC assay has been used to delineate new metabolic states of T cell exhaustion in colon cancer and we anticipate its application to other contexts of metabolic dysregulation [47]. Similarly, recent work uses histone markers in conjunction with lineage-defining markers in peripheral blood to track epigenetic states and changes to their heterogeneity in aging [48]. It is important to note that with the expansion of interrogated markers from cell type-defining surface proteins, the validation and optimisation of antibodies becomes more involved. As with any antibody-based assay, negative and positive controls are integral to ascertain marker specificity. However, in cases where post-translationally modified proteins such as phosphorylated kinases or histone modifications are being assayed, drug treatments and/or knockdowns need to be used to validate antibody. Further titrations need to be carried out to get optimal staining concentrations for best signal-to-noise ratios.

Given the ability to measure numerous features simultaneously in a single cell, MC enables researchers to capture key cell regulatory information alongside phenotype and other functional markers. However, the high dimensionality of the data presents unique challenges where traditional cytometry methods of data visualisation such as 2D scatter plots are insufficient to represent data distributions [49]. In order to project high dimensional single cell data such as multiplexed metabolic or epigenetic states captured by MC into lower dimensional representations, manifold learning methods are applied onto the data [50]. The resulting embedding ensures positions of cells in low dimensional space recapitulates cell positions in high dimensional space. Tools such as t-distributed stochastic neighbour embedding (tSNE) and the newer Uniform Manifold Approximation and Projection (UMAP) and potential of heat diffusion for affinity-based transition embedding (PHATE) allow researchers to examine underlying relationships between cell states, the heterogeneity of cell states in system and therefore populations in an unsupervised, data-driven manner (Fig 3c) (Table 2) [51–54]. However, researchers need to bear in mind that these representations are for visualisation of high dimensional data and do not constitute deep analysis of the data. Often, clustering results are projected onto embeddings or embeddings are used to track cellular transitions between phenotypes across a calculated trajectory.

Single-cell trajectories can estimate transitions through cell phenotypes-of-interest

Dynamic cellular processes such as TRAIL (TNF-related apoptosis inducing ligand) drug resistance, Epithelial-to-Mesenchymal Transition (EMT) and B-cell maturation in bone marrow, inherently contain a spectrum of cell states, many of which are temporally transient and occur in low frequency [55–58]. MC is well-equipped to capture these cell states with its ability to assay millions of cells in a short amount of time and delineate key cell state transitions and quantify implicated regulatory factors. To better understand EMT, a well-known process that is highly active during embryonic development and commonly re-activated during disease/cancer progression, cells were profiled through experimentally-driven EMT in human lung cancer cell lines [58]. A comprehensive map of EMT charted the cellular paths traversed during this process, and subsequently this map was used to identify the EMT status of primary human lung cancer samples to then predict its clinical outcome [58]. Importantly, this study demonstrates the ability to take observations from model systems and translate them onto primary clinical samples; a practice that can be expanded to other disease models.

Beyond tracking cellular transitions in drug resistance and disease context, MC can also be used to map protein expression during developmental transitions or developmental reprogramming, such as B-cell maturation or induced pluripotent reprogramming [44,56]. Researchers can use MC approaches described above to identify disease-induced cellular states and their relationships with other cell types. Furthermore, MC is commonly used to track cellular progression through asynchronous biological processes such as cancer progression, stem cell differentiation and immune cell development with the aid of trajectory inference algorithms. In these applications, cellular features which display dynamic expression patterns are used to derive a pseudo-time trajectory (Fig 3d) (Table 2) [59,60]. Using Wanderlust trajectory inference, B-cell maturation was mapped and intracellular signalling and biosynthesis activity coordination was identified at major transitions [45,56]. Not long after, additional trajectory finding algorithms that modelled branching trajectories such as Wishbone and Monocle have been used to map more complex cell trajectories and branching points in differentiation [61,62]. Newer algorithmic approaches, such as Partition-based graph abstraction (PAGA), attempt to capture both discrete cell states through clustering and continuous cell transitions by uncovering links between these states using graph abstraction of single-cell clustered data [63].

In general, trajectory algorithms benchmarked for single cell RNA sequencing (scRNA seq) can be applied to MC datasets [59]. However, given significant differences in data structure between the 2 data types (higher dimensionality and data sparsity in scRNA seq), standard processing steps such as PCA for scRNA seq can be omitted for MC data. The ability of MC to measure numerous multiplexed cell states in a system or across conditions paired with computational tools that order cells by their high-dimensional similarities allows researchers to ask specific, pertinent questions in different fields of human biology. We anticipate future studies in other organ development/dysregulation and viral disease onset contexts such as early neural development and neurodegeneration, other autoimmune diseases such as rheumatoid arthritis and tracking immune remodelling in COVID and other acute viral infections.

Current limitations

A key limitation of lanthanide metal-based labelling of affinity reagents is the inability to further multiplex measured features. Approaches that label multiple metals at a time have been developed for the application of peptide major histocompatibility (MHC) tetramer staining on MC, but it is difficult to adapt this to antibody labelling [73]. Due to this constraint, other labelling approaches that allow for greater than 1-plex barcoding such as DNA oligonucleotide-tagging of affinity reagents can measure more than 50 features per cell [74,75]. Further advancements in multi-omic techniques that leverage oligo-barcoding enable information from multiple regulatory layers to be captured at once [76,77]. However, despite their higher multiplexing and multi-omic capacities, these techniques do not make MC technology obsolete. As these techniques use tags that are measured using single cell sequencing, there is a lower sensitivity and therefore limit of detection (low abundance protein such as transcription factors cannot be detected easily), a lower number of cells can be captured per experiment (10,000s vs millions of cells) and there is a prohibitively high cost and complexity that is not amenable to large clinical studies. Single cell sequencing readouts also have higher data sparsity and therefore lower signal-to-noise ratios and accuracy of relative protein abundances. Other considerations include more rigorous optimization of oligo-tagged antibody staining due to higher non-specific and unbound signal than in MC [78]. Taken together, though oligo-tagged affinity reagent measurements are useful for large, ‘fishing expedition’ type of studies where the magnitude of features inform the patterns in the biological system tested; many other hypothesis-driven or exploratory studies with a priori knowledge on major players in biological process benefit from more ‘supervised’ and accessible experimental technologies like MC and flow cytometry.

A further limitation of MC is that samples cannot be recovered since cells need to be ionized to readout antibody-conjugated heavy isotope abundance per cell unlike in flow cytometry and fluorescence-activated cell sorting assays where downstream assays can be carried out on sorted cells. Additionally, current high parameter flow cytometry machines (eg. BD FACSymphony) also have similar single cell multiplexing capability (up to 50 readouts per cell) and higher cellular throughput (~40,000 in high speed cell analyser) than MC. However, there are significant differences between dynamic range and sensitivity of readouts in the 2 platforms. Certain fluorophores have a higher sensitivity and therefore dynamic range than metal isotopes, and can capture lower abundance proteins with greater accuracy. However sensitivity varies much more across fluorophores than metal isotopes (12–15 times less variable), making MC measurements more consistently robust across markers [8].

Concluding remarks

The behaviour and function of individual cells underlie organism-wide homeostasis and are dysregulated in disease. Single-cell MC offers a platform to interrogate the identity and behaviours of cells functioning in larger cellular systems, which would previously require purification (i.e., sorting), perturbation (i.e., chemical synchronization), or generalizations (i.e., bulk assays) in order to investigate underlying biology. Importantly, this advantage is afforded by the ability to identify cell states with affinity reagents (e.g., purified antibody and RNA probes), used to routinely measure expression of cellular biomarkers [11]. As the search continues to identify new biomarkers of human health and disease, their integration into MC approaches can shed light on their contributions to underlying biology across cellular transitions or in rare cell populations [64,65]. Additionally, MC enables biomarker discovery as well through screening experiments routinely conducted on sample such as peripheral blood and dissociated tissue [66]. This can be further extended to in vitro cultured organoids used to model human disease, which is in many cases more recapitulative of human biology than mouse models [67].

Though this review discusses MC applications in isolation, it must be emphasized that MC data insights can be followed up with orthogonal, complementary experiments to generate addition de novo insights including, InTAC-seq, RNA-seq and BCR/TCR-seq [40,56,68,69]. Significant cell states and populations identified by MC can also be prospectively isolated using fluorescence-activated cell sorting and serve as input into downstream functional assays. This extends the utility of MC from a systems biology experimentation tool to a method that integrates with other techniques to provide mechanistic insight into key biological processes. Furthermore, antibody and probe panels developed for use on MC platforms with metal isotope labelling can be extended to imaging technologies that use laser ablation and time-of-flight mass spectrometry to reveal spatial patterns of key markers in human tissue. Namely, imaging mass cytometry and multiplexed ion beam imaging (MIBI-TOF) have been used to reveal spatial architectures of breast cancer and spatial co-locations of metabolic states in colon cancer in extension to MC [70–72].

There are key aspects of MC technology and application that can be further developed for more robust and accessible single cell measurements (see Outstanding Questions). While there are over 100 non-biological elemental isotopes, we are currently only using around 50 for multiplexing on MC. We believe new chemistries can be designed to utilise more elemental isotopes for antibody conjugations. In parallel, massively parallel targeted and de novo untargeted proteomic methods are improving in resolution (i.e. few 100 cells/sample) and throughput, and may surpass MC in research assay utility in the near future. However, on the clinical side, low parameter flow cytometry is still routinely used for hematopoietic diagnostics. With improved ease-of-use and standardization of reagents, we believe MC can be adopted into the clinic and provide in-depth information for more accurate diagnostics [79].

Outstanding Questions.

There exist over 100 non-biological elemental isotopes that could be used for targeted mass spectrometry-based single cell proteomics. Current efforts demonstrate the ability to target ~50 simultaneous features per cell.

• Will new chemistries and assays be created to unlock the full potential of multiplexing in these single-cell assays?

• What improvements in technology or sample handling can improve MC assay throughput, sensitivity, and robustness?

• Can mass reporters be combined (i.e. barcoded) to unlock higher levels of multiplexing?

• Will de novo, untargeted proteomic methods catch up (and overtake) single cell MC in terms of throughput, economics, and assay utility?

Low parameter fluorescence flow cytometry remains the mainstay of hematopoietic immune clinical diagnostics for liquid biopsies.

• Will MC-like approaches offer advantages for next-generation immune monitoring and clinical decision making?

• If so, how will complexities in sample preparation, instrumentation, and data analysis evolve to accommodate higher-volume, lower skill diagnostic laboratory environments?

Next generation embodiments of MC go beyond cell enumerating cell immunophenotypes by capturing regulatory cell states (i.e. cellular metabolism), and use mass spectrometry imaging techniques to visualize histological context, identifying cell–cell interaction information.

• How will MC-based approaches be adapted or combined in the future to create new cell state approximations and analytical modalities?

In summary, this review highlights the utility of MC as a powerful tool that interrogates cellular states at various levels of regulation (i.e., signalling, metabolism, biosynthesis, transcription factor expression, etc.) across experimental settings and human tissues (i.e. blood, colon, muscle, bone). Partnered with single-cell analysis tools, MC provides a framework to understand underlying biology of complex biological systems and processes. Researchers can effectively use this technology platform in conjunction with various complementary techniques to answer key biological questions across multiple fields.

Highlights.

• Mass Cytometry (MC) has catalyzed a new generation of single cell proteomic discovery in immunology and beyond.

• Quantification of single cell states (i.e. cell cycle, apoptosis, metabolism, signaling, etc.) offer regulatory insights into cell behavior.

• New algorithms can organize MC data to discover new cell populations (i.e regulatory states and immunophenotypes) and organize cells into relationships based on differentiation status.

• Opportunities exist to use MC datasets to inform orthogonal experiments for complementary functional and molecular profiling studies (i.e. gene expression and epigenetic analysis).

• New assays for different single cell states, new analysis modalities (i.e. imaging), as well as additional elemental reporter isotopes offer numerous opportunities for future development.