Protocol: Palladium-based mass tag cell barcoding

Palladium-based mass tag cell barcoding with a doublet-filtering scheme and single-cell deconvolution algorithm

Nature Protocols
Published online


Mass-tag cell barcoding (MCB) labels individual cell samples with unique combinatorial barcodes, after which they are pooled for processing and measurement as a single multiplexed sample. The MCB method eliminates variability between samples in antibody staining and instrument sensitivity, reduces antibody consumption and shortens instrument measurement time. Here we present an optimized MCB protocol. The use of palladium-based labeling reagents expands the number of measurement channels available for mass cytometry and reduces interference with lanthanide-based antibody measurement. An error-detecting combinatorial barcoding scheme allows cell doublets to be identified and removed from the analysis. A debarcoding algorithm that is single cell–based rather than population-based improves the accuracy and efficiency of sample deconvolution. This debarcoding algorithm has been packaged into software that allows rapid and unbiased sample deconvolution. The MCB procedure takes 3–4 h, not including sample acquisition time of ∼1 h per million cells.

At a glance


View all figures


  1. Isothiocyanobenzyl-EDTA(palladium) MCB cell labeling reagent.
    Figure 1
  2. MCB cell labeling by the isothiocyanobenzyl-EDTA(palladium) chelate.
    Figure 2
  3. Doublet-filtering MCB scheme.
    Figure 3
  4. Single-cell barcode deconvolution.
    Figure 4
  5. Single-cell debarcoding software.
    Figure 5
  6. Doublet removal with the 6-choose-3 MCB scheme and single-cell debarcoding.
    Figure 6
  7. Single-cell deconvolution versus Boolean gating for MCB samples with and without large differences in MCB labeling intensity.
    Figure 7
  8. Palladium barcode staining intensity across cell lines of different cell sizes.
    Supplementary Fig. 1
  9. Time-of-flight traces of palladium barcodes of singlets and doublets with overlapping event lengths and Ir-intercalator intensities.
    Supplementary Fig. 2
  10. Gates used for Figure 6b.
    Supplementary Fig. 3
  11. 96-well plate layout for MCB reagent titration in triplicate.
    Supplementary Fig. 4
  12. Plate layout for 6-choose-3 MCB combinatorial doublet-filtering scheme.
    Supplementary Fig. 5
  13. Pooled sample groups for 20-sample MCB combinatorial plate testing and validation.
    Supplementary Fig. 6



Barcode multiplexing

As a general approach, pooled sample analysis has been used to improve the efficiency and comparability of a diverse range of biological assays, including microsphere-based ELISA1 and high-throughput DNA sequencing2, 3. For these applications, assay-specific identifiers such as fluorochrome combinations or oligonucleotide sequences are used as barcodes to uniquely label each sample, and the barcoded samples are pooled together for processing and measurement. Multiplexing in this manner eliminates sample-to-sample assay variability, increases assay throughput and reduces reagent consumption. After pooled measurement, the uniquely identifiable barcodes are used to recover the individual samples for further analysis.

This multiplexing strategy was adapted to flow cytometry by use of the fluorescent cell barcoding (FCB) technique, which uses unique combinations of cell-reactive fluorophores to covalently label cell samples before pooled antibody staining and flow cytometry analysis4. Mass cytometry, a recently developed variation of flow cytometry, uses rare earth metal isotopes instead of fluorophores as detection reagents, allowing over 40 simultaneous antibody-based measurements at the single-cell level5. The principles of FCB were extended to mass cytometry by the MCB technique, which uses cell-reactive metal chelators to covalently label cell samples with combinatorial barcodes6.

Advantages and disadvantages of MCB

Both FCB and MCB use a single antibody cocktail to stain all samples simultaneously within a single tube, ensuring that all samples are exposed to the same antibody concentration at the same cell density. This uniform antibody exposure removes tube-to-tube variability from the assay, and it is especially important when antibodies are used at nonsaturating concentrations, as is often the case with mass cytometry because antibody concentrations must be titrated low enough to prevent ion detector saturation.

Analysis of multiplexed samples offers additional benefits that are specific to mass cytometry. The ion detection sensitivity of a mass cytometer will drift during instrument use and vary after each maintenance, and although this effect can be mitigated by normalization using bead standards7, measuring samples after pooling further reduces intersample variability. In addition, the sample introduction loop of a mass cytometer is a potential source of carryover between samples, but the possibility of sample cross-contamination is bypassed by MCB because the samples are individually labeled with a unique barcode. Further improvements to MCB described here include (i) palladium-based cell labeling reagents, (ii) a combinatorial doublet-filtering scheme and (iii) an improved barcode deconvolution algorithm implemented as a software application, all of which markedly improve the quality of mass cytometry data.

One drawback to the previously described MCB method is that paraformaldehyde (PFA) fixation and ​methanol permeabilization must be performed before the antibody staining step, which can adversely affect the quality of antibody staining for some epitopes. In our experience, ∼50% of cell surface epitopes are adversely affected by ​methanol treatment, and <5% are adversely affected by PFA treatment. To address this problem, we have recently modified the MCB protocol to permit barcode staining before ​methanol permeabilization, allowing ​methanol-sensitive surface markers to be assessed in combination with MCB multiplexing8. This modified protocol relies on transient permeabilization with saponin, and it is included here as an alternate procedure. If PFA-sensitive epitopes must be used for an experiment, there are two options that still allow for MCB multiplexing. One option is to identify another antibody clone that recognizes a different, PFA-compatible epitope on the same marker of interest. A second option is to perform an initial stain with antibodies against all PFA-incompatible epitopes before PFA fixation, followed by the remaining MCB protocol as described here, therefore still gaining the advantages of sample multiplexing for all PFA-compatible antibodies.

Palladium-based MCB reagents

Lanthanide-based MCB reagents perform well as cell labeling reagents6, but their utility for mass cytometry analysis is limited in two ways. First, lanthanides are used as antibody tags for mass cytometry, and therefore their use as MCB reagents reduces the number of antibody-based measurement parameters available. Second, MCB lanthanide reagents may interfere with other measurement channels owing to isotopic impurity of enriched lanthanide isotopes and owing to mass spectrometry effects. These effects include the ‘+1 effect,’ which is due to the time-of-flight (TOF) mass trace distribution being skewed toward larger mass and overlapping with the next mass unit integration window, and the ‘+16 effect’ of oxidation during ionization in the plasma ion source. These contaminating effects are typically minor, but they become more pronounced when a high-intensity MCB signal spills over into a low-intensity antibody signal.

In order to avoid both measurement channel cross talk and the reduction of antibody measurement channels, palladium was identified as a potential mass tag for MCB. As palladium is not compatible with the ​diethylene triamine pentaacetic acid (​DTPA)-based polymer that is used for antibody labeling, it is not used for mass cytometry antibody measurements9. Palladium has six stable isotopes with masses of 102, 104, 105, 106, 108 and 110 amu, which are commercially available with enriched purities of 91%, 96%, 98%, 99%, 99% and 99%, respectively. These isotopes fall well below the 139–176 amu mass range of the lanthanides, and therefore their use in MCB reagents does not impinge upon lanthanide-based antibody detection. However, Pd110 does overlap with the Cd110 present in quantum-dot–conjugated antibodies, which are therefore not recommended for use in combination with the palladium-based MCB protocol. All six palladium isotopes were obtained in their 2+ charge state as nitrate salts, and dissolved in 5 N ​HCl (Fig. 1a). The addition of the bifunctional molecule isothiocyanobenzyl-EDTA yielded a palladium chelate (Fig. 1b), which may be subsequently used to covalently label cells (Fig. 1c).

Figure 1: Isothiocyanobenzyl-EDTA(palladium) MCB cell labeling reagent.
Isothiocyanobenzyl-EDTA(palladium) MCB cell labeling reagent.

(a) ​Palladium nitrate is converted to ​palladium chloride after dissolving in 5 N ​HCl. (b) Palladium chelation by isothiocyanobenzyl-EDTA. (c) Cell labeling by the isothiocyanobenzyl-EDTA(palladium) chelate.

For MCB multiplexing, a binary labeling strategy is used in which each sample is either positive or negative for each of the six palladium isotopes (Fig. 2a), and unique combinations of these positive and negative labels are used as sample-identifying barcodes. The isothiocyanobenzyl-EDTA(palladium) chelates label cells rapidly in PBS, reaching completion between 0.5 and 1 min at 4 °C (Fig. 2b). Care must be taken to wash excess FBS or ​BSA away from the samples before MCB labeling, because nucleophile-containing proteins will compete for isothiocyanate reactivity and decrease cell labeling intensity (Fig. 2c). In addition, care must be taken to quantify the number of cells in each sample to be labeled, because the isothiocyanobenzyl-EDTA(palladium) labeling reaction reaches completion rapidly in a stoichiometric manner, and therefore it is highly sensitive to differences in cell number (Fig. 2d). Although the EDTA-palladium dissociation rate is permissible for metal exchange between chelating groups on the timescale of days, in practice this effect and any resulting cross-contamination between labeled cell samples is observed to be negligible.

Figure 2: MCB cell labeling by the isothiocyanobenzyl-EDTA(palladium) chelate.
MCB cell labeling by the isothiocyanobenzyl-EDTA(palladium) chelate.

(a) Binary MCB labeling of PFA-fixed, ​methanol-permeabilized cells. (b) One million PFA-fixed, ​methanol-permeabilized cells were incubated with 100 nM isothiocyanobenzyl-EDTA(palladium) at 4 °C for the indicated times. Median counts are shown as connected blue circles, and they are overlaid on individual contour plots for each sample with Ir-intercalator along their hidden x axes. (c) One million PFA-fixed, ​methanol-permeabilized cells were mixed with the indicated concentrations of ​BSA before incubation with 300 nM isothiocyanobenzyl-EDTA(palladium) at 4 °C for 30 min. Median counts are shown as connected blue circles, and they are overlaid on individual contour plots for each sample with Ir-intercalator along their hidden x axes. (d) PFA-fixed, ​methanol-permeabilized cells were incubated at the indicated cell densities with the indicated isothiocyanobenzyl-EDTA(palladium) concentrations at 4 °C for 30 min.

Doublet-filtering barcode scheme

In mass cytometry, physical or coincident doublets will result in the false interpretation of two or more cells as a single cell. Physical doublets occur when multiple cells are physically attached owing to some form of cell-cell interaction or owing to incomplete enzymatic separation of cells derived from tissue or adherent culture. Coincident doublets occur when physically separate cells pass through the instrument in too quick a succession to be identified as separate events by the mass cytometer’s cell detection software. Cell doublets will confound any single-cell analysis, including algorithms such as SPADE10, visNE11 and Citrus12, and they are especially problematic for the investigation of rare or uncharacterized cell types.

To improve doublet removal from mass cytometry data sets, a doublet-identifying barcode scheme was developed. Error-correcting codes such as Hamming codes have been used in barcode design for error detection and correction in multiplexed high-throughput sequencing13. This strategy relies on redundancy in the sample barcode: if one measurement is incorrect, the error is detected and it may even be corrected depending on the level of barcode redundancy. Here, an n-choose-kbarcoding scheme was chosen as the minimally redundant code that allows doublet identification and removal while maximizing the number of unique combinatorial identifiers, where n-choose-kequals n!/(k!(nk)!).

With six palladium isotopes available for MCB labeling, a 6-choose-3 barcoding scheme was used, in which each of the 20 individual barcodes are positive for exactly 3 of the 6 possible palladium MCB reagents (Fig. 3a). In this scheme, the combination of any two barcodes resulting from a cell doublet yields an ‘illegal’ barcode with at least four barcode channels that are positive—i.e., one that cannot belong to a single-cell event (Fig. 3b). This scheme cannot detect coincident doublets between two cells from the same sample, but every other combination can be detected and removed. By this reasoning, 95% of the coincident doublets can be removed when using the 6-palladium 20-sample scheme, because for any single cell within a coincident doublet the other cell within the doublet will have a different barcode 19 out of 20 times. However, in practice, the actual doublet removal rate may be <95% depending on the number of physical doublets present in the samples.

Figure 3: Doublet-filtering MCB scheme.
Doublet-filtering MCB scheme.

(a) The 6-choose-3 doublet-filtering barcode scheme. Each well is positive (gray) and negative (white) for exactly three out of the six MCB reagents. (b) Examples of a barcode singlet (three positive barcode channels) and a barcode doublet (>3 positive barcode channels) as seen in the time-of-flight spectra used to visualize cells while acquiring data at the instrument. (c) Maximum number of available barcodes as a function of the number of barcode channels for both doublet-filtering n-choose-k schemes and for the nonredundant 2n binary scheme.

One tradeoff with this doublet-filtering scheme is a reduced number of cell samples that can be multiplexed by MCB. For every n metals used for barcoding, only n-choose-k combinations are available, which is maximized when k = floor(n/2), rather than the 2n that are available with a nonredundant binary scheme (Fig. 3c). With six palladium MCB reagents, the n-choose-k scheme results in 20 possible samples rather than 64. An appropriate barcode scheme must be chosen for each experiment on the basis of (i) the number of measurement channels available for barcoding, (ii) the desired number of samples to be multiplexed, and (iii) the importance of doublet removal for sample analysis.

Single-cell deconvolution algorithm

Traditionally, individual samples have been recovered from FCB and MCB data sets using Boolean combinations of manually drawn gates, but this method is not ideal because cell events that fall outside these gates must be discarded. This problem is exacerbated, and cell yield is made even lower, when there is variability in barcode staining intensity between pooled samples. This variability can be caused either by differences in cell number between samples without appropriate adjustment of MCB reagent (Fig. 2d) or by differences in cell size within samples (Supplementary Fig. 1), and this variability may lead to systematic depletion of certain cell types or samples from the data set during deconvolution. To address this problem, an alternative deconvolution algorithm was developed that treats each cell individually instead of using gates to demarcate populations of cells. This deconvolution strategy is termed single-cell debarcoding (SCD).

The SCD algorithm first rescales and then sorts the barcode intensities for each cell, in order to identify the largest barcode separation between adjacent barcode levels. This barcode separation is used as a boundary to define which barcode channels are ‘positive’ and ‘negative’ for each individual cell (Fig. 4). Cell events are assigned to a sample when they contain (i) a barcode separation that is larger than a threshold value and (ii) positive barcode channels corresponding to a combination used in the barcode scheme. Cell events remain unassigned if they contain (i) a barcode separation that is lower than the defined threshold, such as low signal debris, or (ii) positive barcode channels that do not correspond to a barcoded population, such as doublets. When an n-choose-k doublet-filtering barcode scheme is used, the SCD algorithm uses the khighest and nk lowest barcode channels instead of the largest barcode separation to assign positive and negative barcode values, and the barcode separation threshold is applied to the difference between the kth and (k–1)th highest normalized barcode intensities. Once preliminary barcodes have been assigned, outliers are filtered out by applying a Mahalanobis distance threshold to each barcode population, which takes into account the covariance of the barcode populations. Finally, each barcode population is output to a corresponding Flow Cytometry Standard (FCS) file, and the cells that were discarded by the algorithm are output to a separate FCS file.

Figure 4: Single-cell barcode deconvolution.
Single-cell barcode deconvolution.

Five events from a 6-choose-3 MCB-multiplexed FCS file are shown in single-cell format displayed on a vertical dashed line. Events 1–3 correspond to barcode singlets as indicated by the barcode key, Event 4 is a barcode doublet and Event 5 is classified as debris. The red line segments indicate ‘barcode separation’, assuming the 6-choose-3 scheme, which is always set as the distance between the third- and fourth-highest barcode intensities. Without this assumption, the last two events would have larger barcode separations but would still be discarded because their barcodes would not match any in the 20-sample scheme.

To facilitate deconvolution of barcoded data sets, the SCD algorithm was implemented as a standalone MATLAB application. By using the MATLAB Compiler Runtime (MCR), this application does not require a MATLAB installation or license. A flowchart describing the debarcoding workflow is shown in Figure 5a. The inputs are an FCS file that contains a barcoded data set and a spreadsheet in CSV format that defines the barcoding scheme, referred to as the ‘Barcode Key’. After selection of the input FCS file and Barcode Key in the control panel of the GUI (Fig. 5b), a preliminary round of barcode assignment is performed for a range of barcode separation thresholds. A histogram of cells binned by barcode separation and a plot of the number of total events yielded for each barcoded sample as a function of the separation threshold are then displayed in the top right panel of the GUI (Fig. 5c). This view of yield versus separation threshold, as well as a single-cell view for each resulting barcode population (Fig. 5d) and biaxial scatter plots (Fig. 5e), aid the choice of deconvolution parameters that can favor either barcode stringency or cell yield.

Figure 5: Single-cell debarcoding software.
Single-cell debarcoding software.

Mouse splenocytes were collected from 20 individual mice and treated with benzonase to minimize cell aggregates. 1.5 × 106 cells from each sample were MCB-labeled and then pooled for mass cytometry processing. The pooled sample was blocked with anti-​CD16/32, and then stained with an antibody cocktail including anti-​CD4 and anti-CD8, followed by mass cytometry measurement. (a) A flowchart of the single-cell debarcoding process. (b) The menu of the single-cell debarcoder. The lower plot portion dynamically changes depending on the plot type selected. (c) The analysis window that is used to guide selection of the barcode separation threshold parameter. The distribution of barcode separations is shown by green bars, and the resulting cell yields for each of the 20 unique populations after debarcoding are displayed as a function of this barcode separation threshold. The dashed line indicates a separation threshold value that best balances barcode assignment stringency and cell yield. (d) The ‘Event’ plot shows all cell events assigned to barcode 100101, with each cell event represented as a vertical line on which the 6 MCB reagent intensities are plotted, as in Figure 4. (e) The ‘All BC Biaxials’ plot type colored by Mahalanobis distance for barcode 100101 with the chosen parameters. All animal studies were performed in accordance with the investigators’ protocols approved by the Stanford University institutional animal care and use committee.

In most cases of MCB debarcoding, there is no tradeoff necessary between stringency and yield, and high stringency settings may be used without substantially lowering the cell yield. In certain exceptional cases with large differences in cell number, cell size or the amount of debris between samples, the MCB staining may not be uniform and the researcher may then adjust the parameters according to his or her desire for barcode stringency versus cell yield. A valuable internal control for the MCB protocol that provides an estimation of the deconvolution error rate is to leave a single well empty—for example, only use 19 cell samples with the 20-sample 6-choose-3 MCB scheme. This allows for an estimation of incorrect sample assignment if any cells are assigned to the empty well after sample deconvolution, and it is useful whether Boolean gating or the single-cell deconvolution algorithm is used.

Doublet identification and removal

Unlike fluorescence-based flow cytometry, mass cytometry does not have scatter measurements to identify cell doublets for removal. Previous attempts to eliminate cell doublets from mass cytometry data sets used gating to remove events that are high for both Ir-intercalator (iridium-intercalator) staining intensity and the number of TOF scans per cell (event length). This gating strategy is imperfect for doublet removal because the doublet and singlet populations substantially overlap on the Ir-intercalator × event-length scatter plot, as revealed by n-choose-k barcoding (Fig. 6a andSupplementary Fig. 2). In addition, this gating strategy may result in the systematic removal of specific cell types, because event length is dependent on the total amount of metal labeling each cell, which in turn is cell type– and staining panel–dependent. When the stringency of the Ir-intercalator × event length gate matches the stringency of the SCD separation threshold—i.e., the two methods produced the same cell yield—the SCD algorithm consistently filtered out more doublets than the ‘singlet’ gate (Fig. 6b). The ‘singlet’ gates used for this analysis are shown inSupplementary Figure 3; to achieve a similar range of yields, the SCD separation threshold was varied from 0 to 0.7, with the Mahalanobis distance threshold set to 30. In addition to its application for mass cytometry as presented here, this doublet-identifying barcode scheme may be applied to other single-cell analysis methods, and it may be especially useful when experimental analysis requires high-confidence discrimination between unique cells and removal of cell doublets.

Figure 6: Doublet removal with the 6-choose-3 MCB scheme and single-cell debarcoding.
Doublet removal with the 6-choose-3 MCB scheme and single-cell debarcoding.

(a) Biaxial plot of event length × Ir-intercalator of events that were assigned a barcode, and of events that were left unassigned. (b) Percentage of cells assigned by gating (green squares) or debarcoding (purple circles) versus percentage of assigned cells that are ​CD4+CD8+ doublets. The different yields were acquired by variable event length × Ir-intercalator gates (green squares) or debarcoding threshold stringency (purple circles). The arrow indicates the debarcoding parameters used in Figures 5 and 6a.

Comparison of SCD with Boolean gating

Although Boolean gating has been used successfully for MCB deconvolution6, it relies on distinct populations with consistent positive and negative MCB labeling intensities across all samples and cell types. If high-intensity, low-background staining is achieved consistently in all samples, the performance of Boolean gating and SCD is essentially the same (Fig. 7a–d). However, consistent MCB labeling can be difficult to achieve for several reasons, such as if cell number varies unexpectedly, if ​BSA washout is incomplete, or if one or more samples contain substantial amounts of debris. An important attribute of the SCD algorithm is that it does not depend on uniform MCB labeling across samples or even within samples, and it performs well on suboptimally MCB-multiplexed samples (Fig. 7e–h) in which deconvolution by Boolean gating fails (Fig. 7i,j). Because the SCD algorithm assigns samples by identifying the highest MCB measurements from each cell individually rather than on a population basis, it works for multiplexed samples with a continuous distribution of MCB labeling intensity, as well as for multiplexed samples with well-separated bimodal distributions.

Figure 7: Single-cell deconvolution versus Boolean gating for MCB samples with and without large differences in MCB labeling intensity.
Single-cell deconvolution versus Boolean gating for MCB samples with and without large differences in MCB labeling intensity.

(a) PFA-fixed, ​methanol-permeabilized mouse embryonic fibroblast (MEF) and mESC cells were aliquotted into a 2-ml 96-well plate in a 20-sample checkerboard pattern at 0.2 × 106 and 0.5 × 106 cells per well, respectively. The cells were incubated with MCB reagents in a 6-choose-3 combinatorial scheme at 300 nM isothiocyanobenzyl-EDTA(palladium). (b) After MCB labeling and pooling of the checkerboard-arranged samples, the MEF-specific antibody against ​CD44 and the mESC-specific antibody against ​Oct4 were used to differentiate between the two cell types. (c,d) Single-cell debarcoding (c) and Boolean gate debarcoding (d) produce similar cell yields and accuracies. (e) PFA-fixed, ​methanol-permeabilized U937 and OVCAR-3 cells were aliquotted into a 2-ml 96-well plate in a 20-sample checkerboard pattern at 30,000 and 100,000 cells per well, respectively. A large percentage of the OVCAR-3 cells were lost during the PBS wash steps before MCB labeling, which resulted in unusually high MCB staining intensity for these samples. The cells were incubated with MCB reagents in a 6-choose-3 combinatorial scheme at 30 nM isothiocyanobenzyl-EDTA(palladium). (f) After MCB labeling and pooling of the checkerboard-arranged samples, U937-specific antibodies against ​CD33 and ​CD45 and OVCAR-3-specific antibodies against ​CD24 and ​E-cadherin were used to differentiate between the two cell types. (g) Gating based on ​CD33, ​CD45, ​CD24 and ​E-cadherin reveals the difference in MCB-labeling intensity between the U937 and OVCAR-3 cells. (h) Single-cell debarcoding successfully recovers both the U937 and OVCAR-3 populations. (i) Boolean gates bisecting the populations at a low MCB intensity primarily recover U937 cells. The low percentage of recovered OVCAR-3 cells is highlighted in red. (j) Boolean gates bisecting the populations at a high MCB intensity primarily recover OVCAR-3 cells. The low percentage of recovered U937 cells is highlighted in red.

It may not be possible to achieve equal levels of barcode staining across all wells when comparing multiple patient samples or different tissue types, yet it is precisely these situations in which the benefits of barcoding, in particular uniform antibody staining, will have the greatest benefit on the quality of the data. Performing barcode deconvolution with the single-cell, rather than population-based, method overcomes the challenges associated with variable barcode staining levels. In addition, in contrast to the variability of manually chosen gate boundaries, the results of the single-cell method are solely determined by the chosen distance parameters, and therefore the deconvolution is reproducible.

Experimental design

The main PROCEDURE describes how to prepare cell samples, label with MCB reagents and analyze the obtained results. Box 1 describes how to prepare MCB labeling reagents. Box 2describes how to titrate MCB labeling reagents. Box 3 describes how to make combinatorial plates.Box 4 describes how to test the combinatorial plates. All the procedures in the boxes must be performed before performing an MCB labeling experiment, as they are all important components of the required setup (see also Supplementary Figs. 4,5,6).

Box 1: Preparation of palladium MCB cell labeling reagents • TIMING overnight

Box 2: Titration of palladium MCB reagents • TIMING 4–6 h

Box 3: Preparation of combinatorial MCB plates • TIMING ∼2 h

Box 4: MCB combinatorial plate validation • TIMING ∼2 h



  • Cells or tissue sample. We have successfully used this protocol for suspension and adherent cell lines; mouse tissues including blood, spleen, bone marrow, spinal cord and brain; and human tissues including blood, bone marrow, leukemia and ovarian tumor biopsies

    Caution:All experiments should adhere to relevant institutional and governmental ethics guidelines and regulations. Informed consent should be obtained from donors of human blood or tissue. All of our animal studies were performed in accordance with the investigators’ protocols approved by the Stanford University institutional animal care and use committee.

  • Palladium MCB cell labeling reagents, prepared as described in Box 1
  • Cell culture medium (cell type appropriate)
  • Cell dissociation reagent (cell type appropriate, such as: PBS-EDTA/Versene (Life Technologies, cat. no. 15040-066), ​trypsin (Life Technologies, cat. no. 25200-056), TrypLE (Life Technologies, cat. no. 12605-010), Accutase (Innovative Cell Technologies, cat. no. AT 104) and collagenase (Life Technologies, cat. no. 17100-017) (only required for adherent cell types and solid tissue samples)
  • Nylon mesh cell strainer, 40 μm (Corning, cat. no. 352340), 70 μm (Corning, cat. no. 352350) or 100 μm (Corning, cat. no. 352360) (only required for adherent cell types and solid tissue samples)
  • DMSO (Sigma-Aldrich, cat. no. D2650)
  • Methanol (Fisher Scientific, cat. no. A412-4)

    Caution:Methanol is flammable. Keep it away from heat. Avoid contact with skin and eyes. Avoid inhalation.

  • Sodium hydroxide (Sigma-Aldrich, cat. no. S-8045)

    Caution:Sodium hydroxide is corrosive. Avoid contact with skin and eyes.

  • Sodium phosphate dibasic heptahydrate (Sigma-Aldrich, cat. no. S9390)
  • Potassium phosphate monobasic anhydrous (Sigma-Aldrich, cat. no. P0662)
  • Potassium chloride (Fisher Scientific, cat. no. P330)
  • Sodium chloride (Fisher Scientific, cat. no. S271)
  • Ammonium acetate (Sigma-Aldrich, cat. no. A7330)
  • Saponin (Sigma-Aldrich, cat. no. S-7900)
  • Tween-20 (Sigma-Aldrich, cat. no. P-9416)
  • BSA fraction V (Sigma-Aldrich, cat. no. A-2153)
  • Sodium azide (​NaN3; Sigma-Aldrich, cat. no. S-8032)

    Caution:NaN3 is highly toxic. Avoid contact with skin, eyes and clothing. It is fatal if swallowed.

  • 16% Paraformaldehyde (PFA), electron microscopy (EM) grade (Electron Microscopy Sciences, cat. no. 15710)

    Caution:PFA is an irritant. Avoid contact with skin and eyes. Avoid inhalation.

  • Cluster tubes, 1.2 ml polypropylene (Fisher Scientific, cat. nos. 07-200-319 and 07-200-317)
  • Deep-well microplates, 96 well, polypropylene, 2.0 ml nonsterile (VWR, cat. no. 40002-012)
  • Aluminum block, 96 well (BioExpress, cat. no. R-2027-S)
  • Combinatorial MCB plate prepared as described in Box 3 and validated as described in Box 4
  • Hydrochloric acid (​HCl; Fisher Scientific, cat. no. A466)


  • Multichannel pipettes, P20, P200 and P1200 (Rainin, cat. nos. L8-20, L8-200 and L8-1200)
  • Vortex mixer (GeneMate, cat. no. S-3200-1)
  • Tabletop refrigerated swinging-bucket centrifuge (Allegra 6R, Beckman Coulter)
  • Hemocytometer (Hausser Scientific, cat. no. 3200)
  • Tabletop microcentrifuge (5424, Eppendorf)
  • Tabletop 96-well format refrigerated centrifuge (Allegra X-22R, Beckman Coulter)
  • Aspirator, 96 well (VP Scientific, cat. no. VP 177A-1)
  • Lyophilizer (FreeZone 4.5, Labconco)
  • Mass cytometer (CyTOF, Fluidigm)
  • Aluminum 96-well block (GeneMate, cat. no. R-2027-S)
  • Debarcoding software, available at


  • HCl, 5 N

    • Dilute concentrated ​HCl solution to 5 N in ddH2O and store it indefinitely at room temperature (20–25 °C).
  • Sodium hydroxide, 5 N

    • Dissolve ​sodium hydroxide to 5 N in ddH2O and store it indefinitely at room temperature.
  • PBS, 10×

    • Dissolve 320 g of ​sodium chloride (1.37 M), 8 g of ​potassium chloride (27 mM), 46 g of ​sodium phosphate dibasic heptahydrate (43 mM) and 8 g of ​potassium phosphate monobasicanhydrous (15 mM) in 3 liters of ddH2O; adjust the pH to 7.4 with 5 N ​sodium hydroxide, and then bring the final volume to 4 liters and filter the solution using a 0.8-μm filter. Store it for up to 5 years at room temperature.
  • PBS

    • Dilute 10× PBS at a 1:10 ratio in ddH2O and store it for up to 5 years at room temperature.
  • PBS-S

    • Dissolve solid saponin powder in PBS to 0.02% (wt/vol), and store it for up to 6 months at room temperature.
  • Fixation solution, 2×

    • Dilute 16% (wt/vol) PFA at a 1:5 ratio (to 3.2%) into PBS and store it at 4 °C shielded from light for up to 2 weeks.
  • Cell staining medium

    • Dissolve ​BSA to 0.5% (wt/vol) and ​NaN3 to 0.02% (wt/vol) in PBS. Store it for up to 6 months at 4 °C.
  • Ammonium acetate buffer

    • Dissolve ​ammonium acetate to 20 mM in ddH2O, and then confirm that the solution is at pH 6.0. Store it for up to 1 year at 4 °C.
  • Intercalator solution

    • Dilute 16% PFA at a 1:10 ratio (to 1.6%) into PBS, and then add Ir-intercalator diluted at a 1:10,000 ratio. Store the solution at 4 °C shielded from light for up to 2 weeks.


Troubleshooting advice can be found in Table 1.

Table 1: Troubleshooting table.


  • Steps 1–8, preparation of fixed, permeabilized cells: ∼1 h
  • Steps 9–20, MCB labeling and antibody staining: ∼3 h
  • Step 21, mass cytometry measurement: ∼1 h for every million cells
  • Steps 22–28, sample deconvolution by single-cell debarcoder: ∼30 min
  • Box 1, preparation of palladium MCB cell labeling reagents: overnight
  • Box 2, titration of palladium MCB reagents: 4–6 h
  • Box 3, preparation of 1,000× and 100× combinatorial MCB plates: ∼2 h
  • Box 4, MCB combinatorial plate validation: ∼2 h

Anticipated results

The MCB protocol allows multiple samples to be pooled for antibody staining and mass cytometry analysis. The use of the six palladium MCB reagents and the doublet-filtering n-choose-k barcoding scheme allows 20-sample multiplexing. Although cell yield depends on several factors, including cell type, container type, starting cell number and procedures for liquid transfer and mixing, the typical yield for MCB multiplexing as described is ∼50%; if the procedure is begun with 500,000 cells per well at Step 10 (1 × 107 total from 20 individual samples), then ∼5 × 106 cells will remain in the pooled, antibody-stained sample at Step 21 for mass cytometry analysis. We observe cell loss to be especially pronounced when the initial cell number is low (Fig. 7e–g), and therefore we have found the MCB protocol to improve cell yield greatly for low-abundance samples owing to the early sample pooling step, which results in a single high-abundance sample rather than multiple low-abundance samples.

After pooled sample processing and measurement on the mass cytometer, a single FCS file is produced for each multiplexed cell sample. The 20 individual samples are recovered from this pooled FCS file by the debarcoding software, which produces 20 FCS files plus an additional FCS file that contains the cell events that were unassigned by the deconvolution algorithm. The MCB protocol and software are flexible enough to allow for additional MCB reagents and barcode schemes. If a researcher desires to multiplex >20 cell samples, a nonredundant barcoding scheme may be used at the expense of doublet filtering, or with additional MCB reagents using lanthanide chelates6 but at the expense of additional antibody measurement channels. The procedure described here is designed to enable a researcher to barcode and to antibody-stain four groups of 20 samples in 3–4 h. The attributes of MCB described in this report will enhance the quality of mass cytometry data.