Mass cytometry and type 1 diabetes research in the age of single-cell data science

Abstract

Purpose of Review:

New single-cell technologies developed over the past decade have considerably reshaped the biomedical research landscape, and more recently have found their way into studies probing the pathogenesis of type 1 diabetes (T1D). In this context, the emergence of mass cytometry (MC) in 2009 revolutionized immunological research in two fundamental ways that also affect the T1D world: 1., its ready embrace by the community and rapid dissemination across academic and private science centers alike established a new standard of analytical complexity for the high-dimensional proteomic stratification of single-cell populations; and 2., the somewhat “unexpected” arrival of MC awoke the flow cytometry field from its seeming sleeping beauty stupor and precipitated substantial technological advances that by now approach a degree of analytical dimensionality comparable to MC.

Recent Findings:

Here, we summarize in detail how mass cytometry has thus far been harnessed for the pursuit of discovery studies in T1D science; we provide a succinct overview of other single-cell analysis platforms that already have been or soon will be integrated into various T1D investigations; and we briefly consider how effective adoption of these technologies requires an adjusted model for expense allocation, prioritization of experimental questions, division of labor, and recognition of scientific contributions.

Summary:

The introduction of contemporary single-cell technologies in general, and of MC in particular, provides important new opportunities for current and future T1D research; the necessary reconfiguration of research strategies to accommodate implementation of these technologies, however, may both broaden research endeavors by fostering genuine team science, and constrain their actual practice due to the need for considerable investments into infrastructure and technical expertise.

Introduction

With a prevalence exceeding 0.3% in the United States and an incidence that until recently increased worldwide at an annual pace of 3–4%, type 1 diabetes (T1D) poses a considerable challenge to afflicted individuals, to the development of effective prevention and treatment regimens, to the allocation of research funds, and to public health initiatives at large [1–4]. While the precise interactions between inherited susceptibilities and environmental factors remain to be elucidated in detail [5, 6], T1D development is mediated by a complex interplay of innate and adaptive autoimmune processes that damage and destroy insulin-secreting pancreatic beta cells and eventually lead to elevated blood glucose levels as well as serious disturbances of protein, fat and carbohydrate metabolism. To date, no cure or effective prevention is available and despite insulin treatment, long-term complications such as kidney failure, heart attack and stroke remain a serious challenge to health and well-being [1, 2, 7]. Progress in T1D research is slated to greatly benefit from the rapid expansion of new single-cell technologies, in particular if applied to broad collections of T1D-focused tissue repositories [8–13]. Here, the high-dimensional proteomic single-cell profiling by mass cytometry (MC) [14], also called cytometry by time of flight (CyTOF), is uniquely situated at the juncture between traditional flow cytometry (FC) and emerging sequencing-based approaches. Furthermore, adaptation of MC technology to tissue imaging introduces a similar degree of analytical dimensionality to the spatial reconstruction of cellular protein expression [15].

Mass cytometry: the coming-of-age of a transformative single-cell technology

High-dimensional cytometry is a powerful technique to elucidate complex phenotypic and functional characteristics of heterogeneous cell populations, which can identify cellular biomarkers and provide valuable insights into disease mechanisms in the context of many experimental and clinical conditions. MC is conceptually analogous to conventional FC; however, antibodies and other affinity reagents are conjugated to rare-earth metal isotopes instead of fluorophores. After staining with these reagents, metal-labeled single cell suspensions are nebulized and passed through an argon plasma that atomizes and ionizes the cells into individual ion clouds, which then pass through a series of mass optics and are analyzed by time-of-flight spectrometry. The multi-atomic signals associated with the labeled antibodies are integrated to derive single cell protein expression data much like that generated by conventional flow cytometers. By overcoming the limitations of spectral overlap and cellular autofluorescence, MC allows for a greater number of simultaneous single cell measurements than conventional FC with less signal spillover between signals, allowing for better-resolved high-dimensional cytometry data. While the current 3^rd generation Fluidigm Helios MC instrument has 135 detection channels, the number of parameters that can be measured in a single experiment is limited by the available isotopic reagents that can be effectively used for cell labeling. Current reagents allow for the simultaneous measurement of up to 60 cell-associated metal signals, including over 40 antibodies against surface and intracellular targets, nucleic acid labels to identify cells, assess their viability and evaluate cell cycle, and additional channels for sample barcoding, and QC.

MC offers the capacity for higher-dimensional measurements than traditional FC yet this advantage comes at the expense of lower sample throughput (<500 cells/second relative to 1000s of cells/second) and lower sensitivity, which can present challenges when applying the technology to the study of extremely rare cell populations or to the quantification of low abundance protein expression. In addition, the recent revolution in massively parallel single cell sequencing technologies and the development of multi-modal single cell proteomic and transcriptional approaches now allow for the simultaneous measurement of a much higher number of parameters, though at much higher cost and even lower throughput. Thus, as a relative mature single-cell profiling technology, MC currently straddles the divide between conventional FC and sequencing-based approaches, and in a setting of suitable infrastructure access and service provision can offer a reasonable balance of dimensionality, throughput and cost that is particularly well suited to conducting translational research studies.

Harnessing mass cytometry for T1D research

Of the more than 900 articles published to date on MC, less than 1% constitute primary T1D-related work. Given the mature status of the MC technology platform, its relatively wide-spread dissemination and unique utility as a powerful discovery tool, this is a somewhat surprising state of affairs, perhaps arising from a combination of the need for considerable upfront research investments and limited access to suitable tissue samples. The latter constraint, however, should not apply to murine studies yet the only MC work conducted in this area pertains to an elegant delineation of T cell receptor signaling events. Published in 2014, this study demonstrates that in non-obese diabetic (NOD) mice, CD4⁺T cells at all stages of development/differentiation compound small impairments of proximal CD3ζphosphorylation into larger downstream signaling defects such as reduced ERK activation [16]. In human T1D, three studies have thus far characterized blood-borne beta cell-specific CD8⁺T cell populations (visualized with individual and/or pooled HLA-A2 tetramers) and consistently emphasized their overall heterogenous phenotypic composition [17–19]. Of these, the first two studies interrogated only small cohorts of T1D patients (3 T1D or 6 T1D vs. 6 healthy control [HC] individuals, respectively) [17, 18], but the observation of increased CD57⁺CD45RO⁺ ZnT8-reactive CD8⁺T cells in T1D vs. HC subjects [18] is noteworthy in light of a larger longitudinal study of new onset T1D patients that employed FC to correlate a relative expansion of CD57⁺ beta cell-specific “effector memory” (CCR7^–CD45RA^–CD27^–) CD8⁺T cells with prolonged C-peptide preservation [20]. Although CD57 is typically considered a marker of T cell exhaustion or senescence, the above population exhibited features of enhanced cytotoxicity leading the authors to speculate that this phenotype is calibrated in direct response to the functional insulin reserve [20].

The third study by Wiedeman et al. evaluated 20 HC and 46 T1D subjects the majority of which were divided into rapid progressors (<5 years disease duration, undetectable C-peptide) and slow progressors (≥5 years disease duration, C-peptide >0.1ng/mL) [19]. Using a 30-parameter phenotypic MC staining panel in conjunction with combinatorial tetramer staining [21] to improve identification of beta cell-specific CD8⁺T cells (and which remains a particular challenge for low avidity auto-reactive CD8⁺T cells), the authors established a direct correlation between specific CD8⁺T cell phenotypes and the rate of disease progression (rapid progressors harbored greater proportions of HELIOS⁺ [a marker for T cell activation/proliferation, [22]] CD8⁺T cells with a “transitional memory” phenotype; slow progressors presented with an enrichment of “exhausted” beta cell-specific CD8⁺T cells regardless of disease duration and age). Interestingly, the association between “exhaustion phenotype” and slow disease progression also pertained to other (i.e., tetramer⁻) CD8⁺T cell populations, supporting the notion that more generalized T cell exhaustion in autoimmune disease may serve as a predictive biomarker as well as a correlate for therapeutic efficacy in the wake of immunomodulatory interventions [23, 24]. A correlation with CD57⁺ CD8⁺T cell subsets, however, was not observed, likely a consequence of the fact that dynamic regulation of CD57 expression preferentially occurs early after T1D onset and in children younger than 12 years of age [20].

Other T1D-related MC studies delineated complex phenotypes of in vitro generated T_REGs [25], or, in complementary work conducted by the same group, defined alterations of the PBMC compartment in children at high-risk (9 multiple autoantibody⁺ and 9 HC subjects) as well as individuals with new-onset T1D and long-standing T1D (7 and 9 T1D cases, respectively) [26, 27]. The latter studies employed the same 32-parameter MC staining panel, incorporated functional immune cell profiling (PMA/ionomycin-induced IFNγ and IL-4/10/17A production), and even included samples from donors with other autoimmune endocrine disorders (Hashimoto’s thyroiditis, Grave’s disease, Addison’s disease [27]) to effectively harness the discovery potential of MC analyses. Yet the relevance of the reported results, not least on account of the small study cohort size, is difficult to ascertain because of challenges to relate particular immune subset alterations to other and at times seemingly contrasting observations (what to make, for example, of the increased preponderance of different NK cell subsets in at-risk children and new onset T1D cases [26, 27] vs. a reduction of total and various NK cell subsets previously reported for new onset T1D cases in ref.[28] and several other studies; or the elevated vs. reduced frequencies of different CD4⁺T_REG subsets in at-risk individuals that only share a CD127⁻CCR4⁺CCR7⁻ backbone [26, 29]; or the decrease of a CD8⁺T_EMRA subpopulation [CD45RA⁺CD27⁻CXCR3⁺CCR7⁻Tbet⁺IFNγ⁺] in new-onset individuals [27] vs. elevated numbers of CD8⁺T_SLEC [short-lived effector-like cells] defined by a CD57⁺CD27⁻CD28⁻CD127⁻ phenotype in seroconverted subjects [29]?). The preceding discussion, including its cumbersome references to complex cellular phenotypes, also highlights a terminological conundrum exacerbated by the very potential of high-dimensional immune cell sub-setting, namely the continued use of descriptive short-hand definitions that equate limited phenotypic properties with cellular differentiation states and functional properties (e.g., CD45RA is not only a marker for naïve T cells and a terminally differentiated memory T cell subset [“T_EMRA”] but in fact is expressed by genuine memory T cells with substantial proliferative reserves [30, 31]). Rather than carrying forward these sobriquets and forcing the astonishing breadth of phenotypic heterogeneity as revealed by MC and related technologies into a Procrustean bed of historical contingency, we would be well advised to leverage the insights of deep immune profiling efforts for a development of improved classification and labeling schemes that facilitate coherent communication.

Moving forward, it is therefore imperative to conduct comprehensive longitudinal deep immune profiling studies that will establish consensus population structures and thus will serve as a general reference frame for better deconvolution of new immune cell subsets across different study cohorts, and eventually for a more effective elucidation of their potential contribution to T1D pathogenesis. To this extent, we are currently leading a Juvenile Diabetes research Foundation (JDRF) and TrialNet sponsored nested case-control study in which MC is applied to the interrogation of samples collected for the Pathway to Prevention study [9]. This study employs three complementary 39 parameter MC panels to comprehensively characterize immune cell composition and phenotype in longitudinal PBMC samples collected from 89 multiple autoantibody-positive subjects as they progress to T1D onset, together with samples collected from age and sex matched controls over the same approximate timeframe; additional examination of the same samples with a broad emphasis on functional single-cell parameters will further expand the analytical scope of this study. Our experimental results will ultimately represent one of the largest and best-controlled datasets of longitudinal changes in immune cell phenotype during T1D progression that, beyond their primary discovery value, may also serve as a coordinate system for the improved contextualization of many other T1D studies. Lastly, among related and other ongoing investigations that employ MC technology for the high-dimensional stratification of immune cell subsets in T1D research are several projects conducted under the umbrellas of the Human Pancreas Analysis program (HPAP) [12, 32], the Human Islet Research Network (HIRN) [33], and the nPOD Autoimmunity Working Group [34].

Mass cytometry of the human pancreas

As a technology platform for single-cell analyses, MC interrogations are not limited to immune cell population. In a pioneering study published in 2016, the Kaestner group stratified dissociated human islets from 17 non-diabetic (ranging in age from 18 days to 59 years) and 3 T2D donors according to 25 MC parameters [35]. Their study demonstrated the principal feasibility of “islet cell MC”, provided further evidence for the consolidating concept of beta cell heterogeneity [36], profiled islet endocrine cell proliferation by Ki67 expression over 6 life-time decades, and showed that alpha cells exhibit both higher baseline replication and greater responsiveness to the mitogen and DYRK1A protein kinase inhibitor harmine [37, 38] than beta, gamma, delta or epsilon cells [35]. Despite the considerable promise, however, no follow-up report has been issued but additional islet samples are continuously processed for MC analysis and the data are made readily available through the HPAP PANC-DB portal [12, 32]. In related ongoing work, we are building and applying complementary MC staining panels that aim to reveal the breadth of phenotypic properties among major constituent cell types in the healthy and diabetic human pancreas (endocrine, acinar, ductal, endothelial and immune cells; see Fig.1 for representative transcription factor expression profiles in beta cells); we are seeking to broaden access to these analysis tools [39]; and we are developing “islet cell MC” as a screening platform to monitor the consequences of mitogenic activation by combinations of harmine and inhibitors of SMAD signaling or GLP-1 receptor agonists [40, 41] on endocrine cell proliferation and differentiation/preservation of respective cellular identities.

Figure 1. Transcription factor profiles of human beta cells.

Islets from a non-diabetic adult donor were dissociated, stained with a 38-parameter MC antibody panel and acquired on a Helios mass cytometer. The plots (gated on live CD45⁻ singlets) demonstrate expression of C-peptide (CPEP) together with proinsulin (ProINS), CD9 or selected transcription factors (respective metals used for antibody conjugation are indicated). Note that robust NKX6.1 expression is largely restricted to beta cells. In contrast, PDX1 is not only expressed by beta cells but also by SST⁺ delta cells and at lower levels by a PPY⁺ gamma cell subset; a very similar expression pattern by delta and gamma cells is also observed for CD9, a recently proposed beta cell subset marker (not shown).

Imaging mass cytometry

A particularly promising avenue for MC-based T1D research is the adaptation of cell suspension MC technology to the high-dimensional spatial interrogation of both fresh and archival tissue sections [15, 42]. As with suspension MC, imaging mass cytometry (IMC) uses metal-labeled antibodies to simultaneously label tissue sections with up to 40 antibodies. The tissue is then ablated using a laser with a ~1 micron spot diameter, which rasters over the imaging field of view. The ablated tissue from each spot is aerosolized, atomized and ionized, and subsequently analyzed by time-of-flight mass cytometry. It is important to note that IMC of metal-labeled tissue sections can also be performed using Multiplexed Ion Beam Imaging (MIBI) [43]; instead of a laser, MIBI uses a tuneable primary ion beam to raster the tissue, and the resulting secondary ions are analyzed by time-of-flight mass cytometry. Despite the considerable specific technical differences between these two technologies, in both cases the resulting isotopic abundance signals from the metal-labeled antibodies are integrated at a pixel level to generate a high-dimensional spatial reconstruction of tissue protein expression at subcellular resolution. These approaches both offer the same relative advantages of high-dimensional measurement as suspension MC, but with the added advantages of resolving the spatial relationships between cells and their surrounding microenvironment and being readily applicable to archival formalin fixed paraffin embedded (FFPE) tissue specimens.

Thus far, two simultaneous reports have described the application of IMC technology to the study of T1D pathogenesis. Damond et al. evaluated the expression of 35 biomarkers in pancreatic tissue sections of non-diabetic, new-onset and long-standing T1D cases (4 donors each) and reconstructed the progression of T1D development from these cross-sectional data by organizing 1,581 individual islets along a pseudotimeline [44]. Accordingly, the transition from pseudostage 1 to 2 is characterized by a focused alteration of beta cell phenotypes (reduced insulin, proinsulin, IAPP and PTPRN expression but maintenance of signature [NKX6.1, PDX1] and pan-endocrine [synaptophysin, CD99] markers) before pronounced beta cell loss in pseudostage 3 and residual beta cells resembling those at pseudostage 2. These observations reinforce the notion that substantial beta cell-intrinsic changes precede their eventual T cell-mediated destruction, emphasize a likely essential role for other yet to be elucidated early pathogenetic and compensatory processes [45], and demarcate a potential window for therapeutic interventions [44]. In fact, while recruitment of CD8⁺ and to a lesser extent also CD4⁺T cells to individual islets appears to occur simultaneously at T1D onset and before beta cell destruction, T and other immune cells are rarely observed in direct contact with islet cells [44] consistent with the prolonged and histopathologically heterogenous disease course [46]. Similarly, Wang et al. employed a 33 parameter IMC staining panel to interrogate pancreatic tissue sections of 6 non-diabetic and 12 T1D cases ranging from new onset to 23 years of duration and reported a series of broad T1D-associated alterations pertaining to islet architecture (reduced islet perfusion and structural integrity), endocrine cell composition and identity (including evidence for potential transdifferentiation of alpha toward beta cell fates as also reported in ref.[47]), and the numbers, spatial distributions and basic functional properties (GzmB: killing capacity; Ki67: proliferation) of major immune cell subsets (CD8⁺ and CD4⁺T cells, B cells, macrophages, NK cells) [48]. Collectively, both studies establish a uniquely granular perspective onto the pronounced heterogeneity of T1D pathology and progression, provide more detail about potential pathogenetic mechanisms, and tentatively point towards adaptive mechanisms that may be targeted with novel therapeutic approaches. In the meantime, newly emerging pancreatic IMC data sets are made available to interested investigators through the HPAP [12, 32] thus providing an important opportunity for collective data mining and applied team science.

T1D research in the age of single-cell data science

In addition to MC, recent and remarkable technological advances have provided access to different analytical modalities at the single-cell level, and multiple current efforts in fact aim to promote methodological convergences towards highly integrative “single-cell omics” [49]. Most prominently, genome-wide transcriptomic landscaping at single-cell resolution as achieved by single-cell RNA-sequencing (scRNA-seq) permits the unbiased distinction of cell clusters, the demarcation of transitional and other cell types, and the organization of cell populations according to novel hierarchies [50]. The power, potential and promise of scRNA-seq, however, has to contend with a number of important challenges that range from the relative sparsity of scRNA-seq data (owing to the stochastic and inefficient capture of mRNA species) to the need for integration of single-cell data across samples, experiments, and types of measurement as well as a requirement for stringent validation and benchmarking of analysis tools [51]. Yet to date, relatively few studies have applied scRNA-seq technology to the study of T1D. Among these are the characterization of human and murine beta cell-specific CD4⁺T cell populations [52–55]; the identification of T and NK cell-expressed IL32 as a potential biomarker preceding seroconversion in a small cohort of young at-risk children [56]; murine CD4⁺T cells recovered from human islets transplanted into non-obese diabetic (NOD) mice under conditions of graft rejection or tolerance [57]; and immune, endothelial and mesenchymal cells isolated from NOD islets in the early course of T1D development [58].

Since 2016, considerable strides have also been made in deconvoluting the transcriptomic single-cell signatures of mouse as well as human islets in health and T2D disease (reviewed in [59–61]) but the corresponding T1D space remains little explored [62]. In fact, until this year only 1 alpha, 6 beta and 66 ductal cells from a single T1D donor had been captured in scRNA-seq analyses [63], but in very recent work by P. MacDonald’s group, 348 alpha, beta, gamma, delta and ductal cells from 3 T1D and 3 control donors were interrogated by combined scRNA-seq and electrophysiological measurements of exocytosis and channel activity (patch-seq) [64]. This study, also notable for the successful use of cryopreserved islets, indicates that transcriptomic and functional signatures of surviving beta cells are largely preserved while the corresponding properties of alpha cells are compromised consistent with similar findings reported in other recent studies employing different methodologies [47, 65].

A particularly promising avenue for future investigations into single-cell biology consists of transcriptomic profiling in conjunction with a high-dimensional survey of cell surface proteomes [66–69]; by documenting a limited correlation between RNA and protein expression at the single-cell level, these studies also illustrate a fundamental limitation of scRNA-seq technology, namely that transcript levels by themselves are not sufficient to predict protein levels in many scenarios [70]. While these technology platforms have not yet been harnessed for T1D research, a most recent publication by B. Youngblood and associates employed scATAC-seq (single-cell assay for transposase-accessible chromatin using sequencing) in combination with the generation of a DNA methylation-based T cell “multipotency index” to profile the epigenetic landscape of individual human beta cell-specific CD8⁺T cells [71]. Remarkably, these cells present with a hybrid of naïve and effector-associated programs resembling that of stem cell memory T cells, a signature that also applies to beta cell-specific CD8⁺T cells recovered from spleen and pancreatic draining lymph node in the NOD mouse and that contrasts with a terminal differentiation status of corresponding CD8⁺T cells in the murine pancreas. These observations offer a unified perspective on states of cellular differentiation that may help reconcile the seemingly contradictory naïve-, stem-, effector- and exhaustion-like phenotypes of autoreactive CD8⁺T cells described in earlier studies [20, 72, 73].

In parallel to the introduction of these new technology platforms, the “original single-cell technologies” of FC and immunohistochemistry/immunofluorescence have been considerably expanded and enhanced over the past decade, in part undoubtedly accelerated by the successful introduction and dissemination of the higher-dimensional MC platform. Cytometry configurations with increasing numbers of spatially separated lasers and detectors as well as improved optics now allow the distinction of at least 28 fluorescent parameters [74, 75]. Furthermore, spectral flow cytometry approaches permit the simultaneous use of fluorophores that cannot be effectively distinguished by conventional cytometry, but can be resolved by spectral unmixing. Widespread community adoption of commercial spectral cytometry systems, such as the Cytek Aurora, have further increased the number of parameters that are now routinely measured in flow cytometry experiments [76]. These advances continue to narrow the current dimensionality gap between FC- and MC-based approaches.

While the greater complexity of higher-dimensional FC analyses provides both unique opportunities and challenges, adherence to fundamental quality control practices remains an integral part of experimental design, execution and analysis. For example, a recent report postulated the existence of individual lymphocytes expressing both T and B cell receptors and in T1D, these “dual expressers” were described to be predominantly of a single clonotype encoding a potent CD4⁺T cell autoantigen in its antigen binding site [77]. Even without resorting to the maxim that “extraordinary claims require extraordinary evidence” [78], more prosaic explanations for observed phenomena need to be excluded and according to independent follow-up work, it appears that the “dual expressers” may not be a new lymphocyte subset but rather doublets of conventional T and B cells [79].

Finally, an extent of multiplexing similar to contemporary FC has also been achieved for tissue imaging of protein and RNA expression at single-cell resolution [15, 80, 81], a further expansion of analytical dimensionality is ongoing [82], and genome-wide in situ RNA coverage is emerging as a new frontier in spatial transcriptomics [83, 84]. The integration of these technologies in T1D research is only beginning to emerge, and in addition to the aforementioned IMC studies [44, 48], co-detection by indexing (CODEX) technology [85] has gained considerable traction and features prominently in current work of HPAP-affiliated [12] and other T1D investigators.

Conclusion: big cytometry data & the new economics of T1D science

The increasing adoption of MC and other high-dimensional cytometry approaches in biomedical research has dramatically accelerated the pace and volume of single-cell data generation, and these efforts have only begun to exert their transformative effect on T1D research. The principal utility of single-cell data science pertains to its discovery potential as achieved by an unbiased and comprehensive description of individual cells across larger populations and anatomical contexts. While certain technological constraints (e.g., the a priori choice of particular antibody staining panels for MC experiments) are associated with some residual bias, single-cell data science will establish a new perspective onto T1D immunology, islet cell biology and beyond that in turn will launch a plethora of investigations the specific nature and extent of which has to remain speculative but that undoubtedly will substantially refine our understanding of T1D pathophysiology and contribute to the development of novel diagnostic, preventive and therapeutic modalities. In addition to these ample opportunities for discovery and collaborative team science, however, many new single-cell technology platforms also present with a number of challenges including the high cost of initial equipment purchase, maintenance and upgrade; the expenses, complexities and stringent QC measures associated with the execution of experiments; and the considerable demands for data management, visualization and analysis. Generation of robust high-dimensional cytometry data is therefore typically contingent on access to both advanced institutional infrastructure and deep technical expertise to fully appreciate and appropriately control for potential platform-specific issues and artifacts. High-dimensional cytometry datasets are also generally not well-suited for conventional analysis strategies (e.g., manual gating approaches) and are instead processed, visualized and interpreted using a range of clustering and dimensionality approaches, the implementation of which needs to rely on extensive computational and statistical expertise. Even when applied appropriately to facilitate unbiased discovery from complex data, these approaches do not fully address the range of specific hypotheses that may be tested using a given dataset and that typically require considerable biological and domain-specific knowledge. Accordingly, expertise and resources needed to effectively address these challenges are often going to be distributed across multiple research labs and institutions. In addition, the richness of these high-dimensional datasets offers unique opportunities for re-analysis, integration with other data, and new discoveries beyond those reported as part of their initial publication. In this era of big cytometry data, collaborative team science is increasingly important to ensure that technologies are used effectively to generate the most valuable data possible, and to guarantee that costly resources are not wasted on unnecessary duplication of experiments. The emphasis on data generation and initial publication must also be coupled with an equal emphasis on effective public sharing of both raw and annotated cytometry data to maximize their value to the broader scientific community. The extent to which such a reconfiguration of research practices will affect the allocation and distribution of research funding, the attribution and recognition of scientific discoveries, and the opportunities and challenges for individual career choices and trajectories will have to be discussed and resolved sooner rather than later.

Bullet points.

• Mass cytometry (MC) permits the deconvolution and organization of single-cell populations according to ~5 dozen phenotypic, functional, barcoding and QC parameters; imaging mass cytometry (IMC) allows for additional microanatomic reconstruction of cellular protein expression.

• MC and IMC are powerful discovery tools particularly suited for translational research studies.

• T1D research is slated to benefit from the incorporation of MC, IMC and other single-cell technologies through a more granular description of pathological states and spaces, the reconstruction of dynamic pathogenetic processes, and the identification of novel pathways that may be targeted for diagnostic, prophylactic and/or therapeutic interventions.

• The considerable infrastructure and resource investments necessary to effectively harness single-cell technology platforms for T1D research will require a broad reconfiguration of scientific practices ranging from the sourcing and distribution of research funds, study design and execution, and division of labor to strategies for data analysis, communication and the recognition of scientific contributions.