Osmium-Labeled Microspheres for Bead-Based Assays in Mass Cytometry

This article is featured in In This Issue, p.2821

Flow and mass cytometry are employed to characterize complex cellular systems with single-cell resolution. In conventional, fluorescence-based flow cytometry, small nanoscale-sized microspheres (beads) are routinely used as a reference to assist in instrument setup and performance tracking. Beads equipped with Abs on their surface, (i.e., functionalized beads) are used in a variety of tools and assays for data curation and gaining additional information from the acquired cell samples. For example, compensation beads with Ab capture functionality are used to assess fluorescence spillover for subsequent correction of raw data (commonly termed compensation) (1). Likewise, quantification beads, such as Quantum Simply Cellular (QSC) beads, equipped with known, gradually increasing amounts of Ab-binding sites capturing a particular, receptor-specific, fluorescence-tagged Ab of interest, serve as reference for approximating the absolute number of Ab targets on the cell surface (Ab-binding capacity; ABC) (2). In other assays, beads serve as a spike-in reference for normalizing cell counts (e.g., in the BD Trucount assay) (3), as the solid phase for multiplexed ELISAs implemented in the Luminex platform or the cytometric bead assay (4) or in phagocytosis assays (5).

In conventional flow cytometry, functionalized beads are detected based on their distinctive light scatter properties and/or inherent fluorescence, whereas the fluorescence of the captured Ab conjugates is usually determined in an independent cytometric channel. Although the application of such bead-based assays in mass cytometry is highly desired, the adoption of respective applications from flow cytometry is challenged by the fundamentally different means of signal detection. Mass cytometry has no analog to light scatter, and all particles to be detected by mass cytometry have to be associated to a minimum amount of at least one element or isotope within the mass cytometer’s detection range for their recognition. Particles not or insufficiently labeled with such elements are not reliably detected. Consequently, beads designed for flow cytometry assays are per se not detectable by mass cytometry, and functionalized beads equipped with an inherent metal signal are commercially not available. In contrast, metal-containing europium or normalization beads, regularly used in mass cytometry, are not functionalized and not suitable for immunoassays (6−8). Nevertheless, functionalized beads intended for fluorescence-based flow cytometry have been employed in mass cytometry (i.e., for quality control of Ab conjugates and compensation) (8−11), in which, however, the detection of the beads relied on the signal elicited by the captured metal-conjugated Ab. In this study, we exemplify that this approach can lead to selective detection or ignorance of beads by the mass cytometer, which may importantly confound the readout signal. To overcome this limitation, we introduce labeling of functionalized beads with the tetroxide of osmium, a polyisotopic element with an average mass of 190 Da, enabling their detection in mass cytometry. We exemplify the application of osmium-labeled beads for compensation of signal spillover and for combining the power of high-dimensional immune profiling by mass cytometry with the quantification of the cell surface receptors, providing new insight into immune aberrations in patients with the autoimmune disease systemic lupus erythematosus (SLE).

Millipore-filtered, deionized water was used as CyTOF sample carrier and to prepare 1× PBS from 10× PBS (Rockland Immunochemicals, Gilbertsville, PA; pH was adjusted to 7.4). CyTOF staining medium (CSM) was prepared from 1× PBS supplemented with 0.5% (w/v) BSA (PAN-Biotech, Aidenbach, Germany) and 0.02% sodium azide (Sigma-Aldrich, St. Louis, MO). Buffers were filtered through 0.22-μm membranes (EMD Millipore, Billerica, MA) and stored in Stericup single-use bottles (EMD Millipore) at 4°C. Four percent (weight per volume) osmium tetroxide (OsO₄) solution was prepared from crystalline OsO₄ (no. 19100; Science Services, Munich, Germany) and distilled water. The stock solution was diluted in 1× PBS to generate OsO₄ solutions ranging from 0.01 to 0.0001%. Stocks were stored in brown glass vials at −80°C, working solutions were stored in brown glass vials at −20°C. Native Abs and Ab conjugated to lanthanides or fluorochromes (Supplemental Table I) were purchased from Fluidigm (South San Francisco, CA), BioLegend (San Diego, CA), eBioscience (San Diego, CA), Miltenyi Biotec (Bergisch-Gladbach, Germany), and Novus Scientific AB (Uppsala, Sweden) or were produced in house (German Rheumatism Research Centre Berlin, Berlin, Germany). Rituximab was kindly provided by Dr. B. Hoyer, Charité Berlin. In-house Ab conjugations with lanthanide and indium isotopes were carried out using Maxpar X8 labeling kits (Fluidigm) according to the manufacturer’s instructions. Platinum or palladium labeling was performed as described previously (10, 12). Highly isotopically enriched metal salts not available through Fluidigm were purchased from Trace Sciences (Richmond Hill, ON, Canada). Cisplatins carrying isotopically enriched Pt were purchased from or kindly provided by Fluidigm. Cell events were detected using an iridium-containing DNA intercalator (Fluidigm). mDOTA-103Rh, used for discrimination of dead cells, was prepared from DOTA-maleimide (Macrocyclics, Dallas, TX) and rhodium chloride (Sigma-Aldrich) as described previously (13) and stored at 4°C. Polyisotopic cisplatin (Enzo Life Sciences, Lörrach, Germany), alternatively used for discrimination of dead cells, was dissolved at 25 mM in DMSO and stored at −20°C.

Polystyrene beads were aliquoted dropwise into new 1.5-ml reaction vials (Safe Seal; Sarstedt, Nümbrecht, Germany) from homogenously dispersed bead stocks. In experiments comparing numbers of acquired beads, equal numbers of drops were used per aliquot. Prior to OsO₄ staining, polystyrene beads were pelleted by centrifugation (according to the individual manufacturers’ recommendations, 1500 – 3000 × g) and storage buffer was removed by aspiration. For osmium staining, beads were resuspended in 20 μl of 0.01–0.0001% OsO₄ solution per 1 × 10⁶ beads and incubated for 30 min at room temperature (RT) in 1.5-ml reaction tubes (Safe Seal). After incubation, 1 ml of CSM was added, beads were pelleted, and supernatant was discarded (“washing”). Beads were washed at least three times with 1 ml of CSM to remove unbound OsO₄. The beads were then available for further protocols or resuspended in Millipore water for direct acquisition at the CyTOF instrument.

OsO₄ is a hazardous chemical that requires specific safety measures to avoid skin and eye contact and aspiration, such as working in a fume hood. We suggest discussing the osmium-labeling procedure and waste disposal with the laboratory safety officer to implement safety precautions meeting the individual laboratories’ safety requirements. All work involving fresh/unbound OsO₄ was conducted in a chemical hood. The resulting OsO₄-labeled beads were further processed in a normal laboratory environment.

Peripheral blood samples were drawn by venipuncture from healthy individuals and patients diagnosed with SLE or inclusion body myositis (Table I) under approval by the local ethic committees in accordance with the Helsinki II Declaration (approval identifier EA1/132/116). All donors gave written informed consent prior to blood donation. For some experiments, PBMC were isolated from buffy coats obtained from the Institute of Transfusion Medicine, Charité University Medicine, Berlin, Germany.

PBMCs were prepared from anticoagulated whole blood samples by density gradient centrifugation using Ficoll (800 × g, 20 min; GE Healthcare; Chicago, IL). PBMC were subsequently washed twice in 45 ml of CSM, counted by volumetric flow cytometry (MACSQuant; Miltenyi Biotec), cryopreserved in freezing medium (10% DMSO [Sigma-Aldrich]/90% FCS [Biowest, Nuaillé, France]) and stored in liquid nitrogen gas phase until use. For each experiment, PBMC were gently thawed at RT, transferred to 50 ml of prewarmed (37°C) RPMI 1640 (Thermo Fisher Scientific, Waltham, MA) supplemented with 10% FCS and 5 U/ml benzonase (Sigma-Aldrich). To counteract metabolic activity, PBMC were immediately pelleted for 10 min at 500 × g and 4°C and washed once in 15 ml of cool PBS (500 × g, 4°C), followed by volumetric cell counting by flow cytometry. Finally, PBMC were resuspended in CSM and kept on ice for further use.

Ab–metal conjugates were used at concentrations optimized to separate cell population expressing or lacking a particular Ag. Ab conjugates applied in ABC assays were used at saturating conditions, which were determined by escalating the Ab concentration in the staining reaction. The Ab concentration at which the maximum separation between stained and unstained cell populations was observed, while not increasing background staining, was considered as optimal [QSC Manual, Bangs Laboratories (14)]. Ab mixtures were prepared fresh and kept on ice until use or were cryopreserved at −80°C. For experiments involving multiple samples of different donors, live-cell barcoding was performed before staining of cell surface Ags using a set of palladium and platinum isotope-labeled β2-microglobulin (B2M) Abs (10, 12). For live-cell barcoding, PBMC were incubated for 30 min at 4°C with combinations of B2M conjugates in a 5-choose-2 barcode scheme. Next, the barcoded cells were washed four times with 1.2 ml CSM, pooled and further processed together. For staining cell surface Ags by isotope-tagged Abs, up to 2 × 10⁶ PBMC were resuspended in 50 μl Ab mixture diluted in CSM and incubated for 30 min at 4°C in 1.5 ml reaction vials. The cell pellet (500 × g, 4°C, 5 min), was resuspended and incubated in 1 ml 10 μM mDOTA-¹⁰³Rh solution in PBS for 5 min at RT, to label dead cells for their later exclusion. The sample volume was filled up to 1500 μl with CSM followed by cell pelleting and supernatant aspiration. Next, the cells were washed once with 1 ml CSM, once with 1 ml PBS, and finally resuspended in 2% paraformaldehyde (PFA) solution (diluted from 16% stock with PBS; Electron Microscopy Sciences, Hatfield, PA) and incubated at 4°C overnight. On the next day, 500 μl CSM was added, cells were pelleted (700 × g, 5 min, 4°C) and subsequently washed once with 1 ml CSM. Next, the samples were incubated for 25 min at RT in 500 μl 1 × permeabilization buffer (diluted with Millipore water from 10 × permeabilization buffer; Thermo Fisher Scientific) supplemented 1:500 v/v with 0.125 μM iridium-based DNA intercalator (Fluidigm). Cells were then washed with 1 ml CSM. Thereafter, samples were resuspended in 1 ml PBS and counted by flow cytometry. Prior to acquisition on the CyTOF instrument, samples were washed twice with 1 ml Millipore water and pelleted by centrifugation at 800 × g, 5 min, 4°C. Cells were then resuspended in an appropriate volume of water to a maximum of 5 – 7.5 × 10⁵ cells/ml supplemented with EQ Four Element beads (1:10, v/v) (Fluidigm) and were optionally filtered through 35 μm cell strainer cap tubes (BD, San Jose, CA) prior to acquisition.

For flow cytometry, up to 2 × 10⁶ PBMC were resuspended in 50 μl of fluorescent Ab mixture diluted in CSM. Assay concentrations of Ab conjugates intended for ABC assays were optimized to saturate their respective Ag without increasing the nonspecific background staining. Cell staining was performed in 1.5-ml reaction vials (30 min, 4°C), washed once with 1 ml of CSM, and pelleted (500 × g, 5 min). Directly before acquisition on the flow cytometer, DAPI (Sigma-Aldrich) was added to label dead cells for their latter exclusion by gating. The data were acquired on a MACSQuant flow cytometer (Miltenyi Biotec).

Mass cytometry was performed on CyTOF version 1 [operated as described before (9, 15)] and Helios instruments (Fluidigm). Instruments were daily prepared for acquisition by tuning and cleaning according to the manufacturer’s advice, using tuning and cleaning solutions (Fluidigm). Data of cells or bead suspensions were acquired in cell acquisition mode at event densities of <5 × 10⁵ events/ml for CyTOF v1 and <7.5 × 10⁵ events/ml for the Helios instrument.

The sample supply was set to 45 μl/min for CyTOF version 1, or to 30 μl/min for the Helios instrument. Both instruments were run in dual calibration mode, with noise reduction turned on and event length thresholds set to 10 and 75. The lower event length limit was set to three in some experiments where indicated. The lower convolution threshold was kept at the default value of 300.

OsO₄-labeled beads were detectable at atomic mass channels 186, 187, 188, 189, 190, and 192, and were typically evaluated or gated according to signal in the channel with the highest isotopic abundance (41%, ¹⁹²Os). After acquisition of more than 50 osmium-labeled bead samples, osmium signals typically returned to baseline levels after routine washing with Fluidigm washing solution, HNO₃, and water. At a single instance, additional injections of 3% HNO₃ were required to wash out remaining osmium.

For Ab capturing assays, aliquots of up to 2 × 10⁶ Ab capture beads from different vendors labeled or not with OsO₄ were stained in 50 μl of CSM containing a single Ab conjugate (1:50, v/v) for 30 min at 4°C. Next, beads were washed three times with 1 ml CSM and finally resuspended and incubated in 500 μl 2% PFA solution in PBS overnight at 4°C. Fixation was stopped by adding 1 ml of CSM followed by centrifugation and aspiration of supernatant. Finally, beads were washed twice in 1 ml of CSM, once in 1 ml of PBS, and twice in 1 ml of Millipore water before acquisition. Unless otherwise indicated, BD CompBeads (anti-mouse Ig, anti-rat Ig) were applied for compensation (Supplemental Table I). For the experiment shown in Fig. 3, a large batch of beads was labeled with OsO₄, aliquoted into cavities of a 96-well Deep Well Plate (Corning, Corning, NY), and incubated with individual Ab conjugates (Supplemental Table I). After loading of beads with Ab conjugates and washing, bead aliquots were pooled and either stored at −80°C (Supplemental Fig. 1B) or directly combined with cell suspensions previously stained with Intercalator-Ir and the same set of isotope-tagged Abs (1:4, v/v) for simultaneous acquisition.

For absolute quantification experiments (ABC assays), QSC quantification beads (Bangs Laboratories, Fisher, IN) were implemented according to the manufacturer’s instructions after labeling with OsO₄. To distinguish different bead population in mass cytometry data, OsO₄ labeling was performed with the following different concentrations of OsO₄ to achieve distinct osmium signal intensities: bead 0, 0.01% OsO₄; bead 1, 0.0075% OsO₄; bead 2, 0.005% OsO₄; bead 3, 0.0025% OsO₄; and bead 4, 0.001% OsO₄. After OsO₄ labeling, beads (except bead 0) were incubated with the Ab–metal conjugate targeting the cellular receptor of interest for 30 min at 4°C. The amount of applied Ab conjugate was determined by escalating the Ab conjugate concentration to achieve target saturation. For this, the concentration at which the maximum cytometric separation of stained bead 4 from unstained beads in the channel of interest, while not increasing background, was considered optimal and applied to beads 1–4 in the ABC assay. After incubation, the supernatant was removed by centrifugation (2500 × g, 5 min). Different from Ab capture beads used for compensation, QSC beads were not exposed to PFA to mirror conditions applied to the beads in the flow cytometry–based ABC assays. QSC beads were always prepared freshly at the day of the experiment. For acquisition, the different bead populations were pooled and in some cases combined with PBMC and prepared for mass cytometric ABC assay (1:4, v/v).

Beads for quantification experiments performed by flow cytometry were not labeled with OsO₄. The amount of Ab-fluorochrome conjugates necessary for saturation was determined in a similar manner as for mass cytometric quantification beads. The beads were processed according to the manufacturer’s instructions and acquired on a MACSQuant flow cytometer (Miltenyi Biotec).

Raw mass cytometry data were converted to Flow Cytometry Standard 3.0 files during acquisition. Data were normalized based on EQ Calibration Bead signals using CyTOF 6.7 software (Fluidigm). Flow Cytometry Standard files were analyzed using FlowJo (version 10.4; TreeStar, Ashland, OR), Cytobank Premium (http://www.cytobank.org) and CytoSplore (16). Data from beads and cells were distinguished according to their osmium (¹⁹²Os at CyTOF 1, ¹⁸⁸Os at Helios) and iridium (^191/193Ir) signal. The retrieval of single Ab conjugate–positive Os⁺ beads was achieved by manual gating or by gating in t-distributed stochastic neighbor embedding (t-SNE) plots (17, 18). The Catalyst algorithm was used for calculating spillover matrices; channels occupied by osmium isotopes were not considered in this analysis, and bead data were downsampled to <15,000 events (11). Manual gating of PBMC subsets is represented in Supplemental Fig. 2. Statistical analyses were performed with GraphPad Prism (version 5.4; GraphPad, San Diego, CA). Receptor quantifications were based on median signal intensities of manually gated cell and bead populations or FlowSOM clusters.

For FlowSOM clustering, each Flow Cytometry Standard file containing data of live, single CD45⁺ cells was loaded into R (R version 3.5.1) in RStudio (version 1.1.463) using the flowCore Bioconductor package, and, in particular, the flowSet data structure, as it was done before in the CyTOF workflow Bioconductor package (19). Raw expression data were first arcsinh-transformed with a scale argument of five, as is typical for mass cytometry data preprocessing (20). Combined preprocessed data were clustered using FlowSOM (21), based on all markers listed in Panel C1 (Supplemental Table I) except CD45, B2M, CCR6, CD4, Siglec-1, and HLA-DR and MetaClustered as described before (22). We used the default 10 × 10 self-organizing map to generate 100 clusters, followed by consensus hierarchical MetaClustering down to 60 clusters. For both these functions, we used the FlowSOM Bioconductor package. The clustered data were converted into a cluster frequency table, in which the rows are clusters 1–60, the columns represent individual samples, and each element is the sample-specific percentage of CD45⁺ cells residing within that cluster. From this frequency table, Mann–Whitney U tests were performed between sample groups of interest. The resulting p values were adjusted for multiple comparisons using the Benjamini­–Hochberg procedure within the R Stats package. The frequency table and the median marker expressions per cluster were plotted as heatmaps using the pheatmap function in R.

First, the feasibility of labeling beads with natural abundance OsO₄ solution was tested to create a general bead identification signal suitable for mass cytometry that is applicable to commercially available polystyrene beads. Osmium labeling was achieved by incubating Ab capture beads with OsO₄ diluted in PBS for 30 min at RT, followed by the removal of unbound OsO₄ by washing with PBS. Osmium-labeled beads were readily detectable by the mass cytometer, based on their uniform osmium labeling and independent of the captured Ab conjugate, exhibited no background in non-osmium channels and, at most, minimal background in osmium isotope channels in spectra recorded between the passing of beads (Fig. 1A). The distinct osmium signal enabled the clear discrimination of the OsO₄-labeled beads from 4-element normalization beads or Ir–­DNA intercalator–labeled cells recorded simultaneously (Fig. 1B, Supplemental Fig. 2). Osmium staining solutions with concentrations of up to 0.01% OsO₄ were evaluated for the labeling of beads. Only the incubation of beads with solutions containing at least 0.001% OsO₄ was sufficient to generate a bead-associated osmium signal with event length values >10, securing their reliable detection with default acquisition settings on the mass cytometer (Figs. 1C, 2B). Hence, for all the following experiments, 0.005–0.01% OsO₄ staining solutions were used unless indicated differently. Osmium-labeled beads and their osmium signal were stable for at least 4 wk when stored in CSM at −80°C or at 4°C (Fig. 1D).

In summary, OsO₄ labeling converts polystyrene beads, originally intended for use in conventional flow cytometry, into beads readily applicable in mass cytometry.

Next, we confirmed that labeling beads with OsO₄ had only minor, if any, impact on the Ab-capturing functionality of the beads, as evidenced by the successful capture of CD3–¹⁷⁰Er after osmium labeling (Fig. 2A). Moreover, osmium-labeled beads not only successfully captured lanthanide Ab conjugates but also palladium Ab conjugates and platinum Ab conjugates, confirming the compatibility of osmium labeling with different types of metal Ab conjugates (Fig. 2B). Labeling beads with OsO₄ after Ab capturing also preserved the signal of the captured Ab conjugate. However, as OsO₄ prelabeling of Ab capture beads did not impede the captured Ab signal, provided a higher degree of flexibility for application in various assays and minimized handling of toxic OsO₄, we choose to prelabel beads with OsO₄ throughout all the following experiments. That strategy proved successful for polystyrene beads from multiple vendors with different capturing specificities (Supplemental Fig. 1A).

We then compared bead event length values, signals of the captured Ab conjugates, and numbers of beads detected by the CyTOF software between beads labeled with OsO₄ or not. Omitting the OsO₄-labeling lead to critically low event length values of the beads, equal to or only slightly higher than the event length cut-off value, even when that cut-off value was lowered to three, indicating that ion clouds of these beads did not contain sufficient amounts of ions of the mass cytometers measurement range in a sufficient number of consecutive mass spectra to trigger their detection as “events” by the CyTOF software. Consistently, bead samples were only incompletely detected in the absence of the osmium label. Less than 28% of bead events were detected in the suspension as compared with beads that were labeled with osmium (n = 2). When labeled with osmium, more than 99% of the beads exceeded an event length value of 10, typically used as a lower cut-off in routine mass cytometry sample acquisitions. Consistent with a selective, and nonrepresentative detection of beads in the absence of the osmium, median signal intensities of the captured identical Ab conjugates differed depending on whether the beads were labeled with OsO₄ (reductions of 18 and 66%, gain of 36%, Fig. 2B). Moreover, histograms of Ab conjugate signal intensities showed Gaussian distributions when captured by OsO₄-labeled beads, but not when captured by beads lacking osmium (Fig. 2B).

Taken together, labeling of beads with an identifier tag, such as osmium, is required for obtaining accurate results from bead-based assays. Osmium labeling was proven successful for a variety of beads and did not interfere with capturing Ab conjugates.

Recently, the concept of compensation for signal spillover between channels resulting from metal oxide formation and isotopic impurities of metal labels has been introduced to mass cytometry by Chevrier et al. (11). To generate a spillover matrix, the authors stained aliquots of Ab capture beads (compensation beads) individually with a specific Ab conjugate from a mass cytometric Ab panel, pooled them, and acquired the bead convolute separately from cells. In this study, we extend this approach using osmium-labeled Ab capture beads, which can be combined with the cell sample for joint acquisition (Fig. 3A). The osmium label discriminated the bead fraction from DNA intercalator–labeled cells, and all bead populations carrying individual Ab conjugates were detectable within the bead fraction (Fig. 3B). Catalyst (11) was applied to bead data to calculate a spillover matrix, which was used to compensate spillover effects in bead and cell data. For example, a representative Ab panel used CD20–¹¹³In with an ¹¹³In purity of 93.1% beside CD3–¹¹⁵In for detection of B and T cells, respectively. As expected, the majority of impurity in the ¹¹³In was explained by the presence of ¹¹⁵In, which resulted in an artifactual signal by B cells in the CD3–¹¹⁵In detection channel (calculated spillover, 6.5%). Compensation using osmium-labeled beads and Catalyst properly corrected the spillover artifact, thereby improving data accuracy and facilitating proper interpretation (Fig. 3C, 3D). Importantly, once prepared, compensation bead pools were storable at −80°C for at least 2 wk without affecting the signal intensity of the captured Ab conjugates (Supplemental Fig. 1B), simplifying its use as a routine spike-in compensation control in mass cytometry.

Finally, we exemplify the utility of osmium-labeled beads for combining cell surface receptor quantification assays with high-dimensional immune cell phenotyping by mass cytometry. We analyzed five SLE-active patients and four control donors using 42-dimensional mass cytometry (Table I). After analyzing the abundance of FlowSOM clusters in SLE versus control PBMC, revealing multiple aberrations, including reduced frequencies of CD8⁺ T cell, CD4⁺ T cell and B cell clusters, and increased frequencies of a monocyte cluster (Supplemental Fig. 3G, 3H) consistent with known lymphopenia in SLE (23, 24), we used the same data to analyze the expression of four different receptors (CD4, HLA-DR, Siglec-1, and CCR6) at the level of absolute ABC quantification across all clusters in high-dimensional phenotypic space. For each target, one set of QSC beads equipped with known, gradually increasing numbers of Ab-binding sites were labeled with OsO₄, incubated with the Ab conjugate, pooled, and acquired directly before acquisition of the cell sample by the mass cytometer. Bead data served as reference for the determination of ABC (Fig. 4A, 4B). Notably, calibration curves exhibited very similar slopes for the CD4 and HLA-DR assays that were performed by both flow and mass cytometry (flow cytometry: 0.9168; mass cytometry: 0.9817, Fig. 4C, Supplemental Fig. 3C). This shows that despite employing biophysically fundamentally different detection methods, the assay calibration worked comparably well in flow and mass cytometry. In addition, mass cytometric ABC assays were reliably reproducible (Supplemental Fig. 3D), confirming the robustness of the approach.

Because CD4 ABC have been quantified in various studies before (3, 25−28), we first analyzed CD4 ABC in select PBMC subsets of our data to enable a direct comparison with previous data. We determined a median of 140,677 CD4 Abs per cell on Th cells (minimum: 126,861; maximum: 160,378 Abs per cell; 10 donors, Supplemental Fig. 3A). As expected, the CD4 ABC of monocyte subsets and plasmacytoid dendritic cells was much lower than that of Th cells (median: 20,367–44,886 Abs per cell, Fig. 4A; gating in Supplemental Fig. 2). Because those data were largely consistent with other studies (Table II) (25, 26, 28, 29), providing confidence in the approach, we then performed a systematic comparison of CD4, HLA-DR, Siglec-1, and CCR6 ABC across all PBMC subsets derived from FlowSOM clustering between five SLE patients and four healthy controls. Despite the few individuals analyzed, this analysis revealed several distinct clusters with significantly reduced CD4 or HLA-DR ABC or enhanced Siglec-1 ABC in patients with SLE. CCR6 ABC showed high interindividual variation and were not found significantly dysregulated in SLE (Fig. 4D). HLA-DR ABC were found diminished in cluster 7, showing the phenotype of myeloid dendritic cells (mDC) (high expression of CD36, CD11c, and CD1c; low or no expression of CD3, CD19, CD56, CD16, and CD14, Supplemental Fig. 3G), and the increase of Siglec-1 ABC was documented in seven clusters with dominant monocyte or mDC features.

Because all clusters with reduced CD4 ABC were Th cells, we exemplarily performed manual gating and CD4 ABC analysis of CD4⁺ T cells and their major subsets, confirming the result of the systematic analysis (reduction of CD4 ABC by 14%, p = 0.0156, Supplemental Fig. 3B; data of manual subsetting not shown). Interestingly, low HLA-DR ABC on mDC were strongly associated with low CD4 ABC of certain Th cell clusters (Supplemental Fig. 3E).

Combining unsupervised clustering of high-dimensional mass cytometry data with receptor expression level analysis sensitively identified known and novel immune cell aberrations in active SLE, illustrating the utility of bead-based assays in biomarker identification studies. In summary, osmium-labeled beads facilitated the integration of ABC assays in high-dimensional immune phenotyping studies by mass cytometry, extending its application range and providing a means to report standardized receptor expression data in mass cytometry.

In this study, we present a simple and stable bead-labeling approach for mass cytometry using OsO₄, a highly reactive compound used before for cell membrane and interior staining in electron microscopy and mass cytometry (30−33). The approach enables osmium labeling of common, commercially available functionalized polystyrene beads and facilitates their unambiguous detection by mass cytometry by introducing a stable and high-intensity signal independent of the readout of the bead functionality. Corresponding to the labeling stability observed in this study, OsO₄ has been shown to covalently bind to unsaturated bonds, such as present in benzyl groups, which are abundant in polystyrene (32−34). Beads were likewise successfully labeled with ruthenium tetroxide (data not shown). Notwithstanding the potential use of OsO₄ containing isotopically enriched Os for a higher degree multiplexing of bead-based assays, ruthenium labeling of beads directly offers a second, independent detection channel for functionalized beads.

The acquisition of osmium-labeled beads did not interfere with the detection of lanthanides, palladium, and platinum (i.e., metals used in analytical reagents, sample barcoding, etc.), and osmium-labeled beads successfully captured metal conjugates irrespective of which conjugation strategy was employed to generate these conjugates. Mixtures of osmium-labeled beads with cells and normalization beads were routinely acquired on the mass cytometer, and their data were deconvoluted according to osmium, iridium, and cerium signals. Although not formally tested, osmium-based cell size measurements or sample barcoding (30, 31) are likely to be compatible with the use of beads if the beads are prepared to elicit an osmium signal intensity distinct from that of cells or labeled with ruthenium tetroxide being detectable in atomic mass channels distinct from osmium or by the potential use of different osmium tetroxides holding isotopically purified Os isotopes, all of which should allow for a clear discrimination of beads from cells. The osmium label also permitted exclusion of bead doublets according to high versus average osmium signal intensity, and bead labeling with OsO₄ did not interfere with Ab capture. Taken together with their storability, either with or without captured Ab conjugates, alone or pooled with other beads, osmium-labeled beads proved a highly versatile and compatible tool for various bead-based assays in mass cytometry.

Consistently, we successfully employed osmium-labeled beads for compensation of signal spillover using Catalyst. In this study and, likewise, in Ab quality control experiments, osmium labeling secures that Ab conjugate signal intensity readouts are based on representative bead populations and that beads with no or low signal intensity of the captured Ab are reliably detected. Notably, for the purpose of compensation, osmium-labeled beads can be spiked into the cell sample for spillover correction on a per sample basis.

The new possibility to assess low bead-associated Ab conjugate signal intensities also facilitated bead-assisted quantification of cellular receptors in ABC assays modified for mass cytometry, requiring a standard based on beads that capture many, few, or no Ab conjugates. We obtained ABC values for CD4 on Th cells largely agreeing with published data obtained by flow cytometry and mass spectrometry (3, 25−29) (Table II). Remaining variance in CD4 ABC could be because of deviations in the cell preparation protocol, in which cell fixation is mandatory for mass but not flow cytometry, differences in Ab conjugation for mass versus flow cytometry, or different production lots of QSC beads. We recommend preparing and running the matched cell and bead samples in pairs to achieve ideal conditions for the quantification of cellular ABC.

By pairing the power of high-dimensional immune phenotyping by mass cytometry to quantify large numbers of immune cells subsets from a single sample with the ability of ABC assays to quantify the expression levels for individual receptors, we analyzed the ABC of four receptors in 60 cell clusters in a pilot study of five SLE patients and controls. Although the accuracy of this analysis largely benefits from sample barcoding, the conversion of marker expression signal intensities to ABC values enabled by osmium-labeled QSC beads establishes comparability of cellular receptor expression data across different mass cytometry platforms. Our data confirmed the increase of Siglec-1 on monocytes and mDC clusters in SLE consistent with an IFN-α signature in these patients (9, 35, 36) and, in accordance with data of in vitro–generated dendritic cells in SLE, downregulated HLA-DR on ex vivo mDC (37). We further identified reductions of CD4 ABC in SLE Th cells, affecting memory rather than naive CD4⁺ T cell clusters. All Th clusters with significantly reduced CD4 ABC expressed CD25 or CCR4, increased levels of CD95, and CCR7, CXCR3, CD39, and CD127 to various extents, indicating that Th1, Th2, and regulatory T cell clusters were affected. By contrast, CD4 ABC of activated CD38⁺HLA-DR^high CD4⁺ T cells (cluster 30) were comparable between patients and controls. Notably, both CD4 expressed by Th cells as a coreceptor of the TCR, and HLA-DR expressed by APCs, such as mDC, directly interact in the immunological synapse (38). Consistently our results may indicate defective cross-talk between Th/regulatory T cells and mDC and related limitations in cellular immunity and immune regulation in SLE.

By pairing the power of high-dimensional immune phenotyping by mass cytometry with the ability of ABC assays to quantify individual receptors, we circumvent the high efforts of previous approaches (25) and introduce the reporting of receptor expression data that are comparable across different CyTOF instruments and other platforms, facilitating standardization in biomedical research. By converting conventional flow cytometry beads (e.g., those used for spillover correction and quantification) into osmium-labeled beads suitable for respective assays in mass cytometry, our approach paves the way to a broad application of bead-based assays in mass cytometry.

We are grateful to Vera Bockhorn for revising the implementation of the FlowSOM script.

L.B., A.R.S., and H.E.M. are listed inventors on a related patent application that was submitted by their employer, the German Rheumatism Research Centre Berlin (EP 18169878.8). The other authors have no financial conflicts of interest.