Multiplexed detection and isolation of viable low-frequency cytokine-secreting human B cells using cytokine secretion assay and flow cytometry (CSA-Flow)

The ability to functionally characterize cytokine-secreting immune cells has broad implications in both health and a range of immune-mediated and auto-immune diseases. Low-frequency cytokine-defined immune-cell subsets can play key immune-regulatory roles, yet their detailed study is often hampered by limited clinical sample availability. Commonly used techniques including intracellular cytokine staining require cell fixation, precluding subsequent functional interrogation. The cytokine-secretion assay (CSA) can overcome this limitation, though has mostly been used for detection of relatively high-frequency, single-cytokine secreting cells. We examined how adaptation of the CSA in combination with multiparametric flow-cytometry (CSA-Flow) may enable simultaneous isolation of multiple, low-frequency, cytokine-secreting cells. Focusing on human B cells (traditionally recognized as harder to assay than T cells), we show that single-capture CSA-Flow allows for isolation of highly-purified populations of both low-frequency (IL-10+; GM-CSF+) and high-frequency (TNF+) cytokine-defined B cells. Simultaneous detection and isolation of up to three viable and highly-purified cytokine-secreting B-cell subpopulations is feasible, albeit with some signal loss, with fractions subsequently amenable to gene expression analysis and in vitro cell culture. This multiplexing CSA-Flow approach will be of interest in many human cellular immunology contexts aiming to functionally characterize cytokine-secreting immune cells, especially when sample volumes and cell numbers are limited.

). However, the isolation of viable and purified cytokine-expressing immune cells has presented a technical challenge, as commonly used intracellular cytokine-staining (ICS) or flow-fluorescent in-situ hybridization approaches involve cell fixation and permeabilization [19][20][21][22][23][24][25] . The resultant loss of cell viability and integrity seriously limits the use of cells for comprehensive functional profiling and downstream applications [19][20][21][22][23][24][25] . While viable cells can be sorted based on certain cell surface-markers that have been reported to enrich for particular cytokine-expressing immune cells, such markers are inevitably neither fully sensitive nor specific [26][27][28][29][30][31][32][33] . The cytokine-secretion assay (herein referred to as CSA) was designed to overcome this limitation and to allow for the isolation of viable cytokine-defined immune cells [34][35][36] . CSA relies on bispecific antibodies (bsAbs) bound to the cell surface that capture secreted cytokines on viable cells, thereby preserving cell viability and integrity, in stark contrast to ICS. In essence, CSA typically involves four steps applied to the mixed population of cells: (1) activation (e.g. polyclonal or antigen-specific); (2) labelling with bsAbs that bind on one hand to a cell surface antigen (e.g. CD45 for immune cells) and on the other hand to the particular cytokine of interest; (3) secretion and capture of the cytokine and (4) detection of the cytokine-secreting cells with fluorochrome-conjugated anticytokine antibody for flow cytometry analysis (Fig. 1). However, cytokine secretion assays have more typically been applied to capture high-frequency cells [37][38][39] . In particular, the assay has not been used for simultaneous isolation of multiple low-frequency cytokine-secreting cells, which would be beneficial in many human cellular and immunology contexts, especially when sample volumes and cell numbers are limited.
In this study, we demonstrate the utility of the CSA-Flow for the detection and isolation of viable human cytokine-secreting cells including simultaneous sorting of both low-and high-frequency populations. To illustrate this approach, we focus our attention on human B cells as a growing body of work implicates cytokinesecretion by B cells as relevant in orchestrating immune responses in both health (e.g. host-protection from infection) and different autoimmune diseases (recently reviewed in 12,40 ).

Results
Detection of low-frequency cytokine-secreting human B cells in vitro. We first isolated highly purified (> 98%) B cells from peripheral blood mononuclear cells (PBMC) and then assessed, in parallel, the ability to detect individual, low-frequency cytokine-producing human B cells, using either intracellular cytokine staining (ICS) or the cytokine secretion assay (CSA). IL-10 + and GM-CSF + B cells, known to be expressed by relatively low frequencies of human peripheral blood B cells 41 , can be detected following brief (4 h) stimulation with PMA and ionomycin, using either ICS (Fig. 2a) or CSA (Fig. 2b). The CSA was on par with, if not better than, ICS for the detection of either of these low frequency IL-10 + (CSA: 2.22 ± 0.59% vs. ICS: 0.81 ± 0.28%, p = 0.0319) or GM-CSF + (CSA: 5.49 ± 1.17% vs. ICS: 2.65 ± 0.60%, p = 0.0649) B cells (Fig. 2c).
Sorting purified and functional low-frequency cytokine-secreting B cells using CSA. We next assessed whether the CSA could be used to sort and purify viable and functional low frequency cytokine-defined human B cell subpopulations. Following brief stimulation of purified human B cells, and parallel application of the CSA for either IL-10 + or GM-CSF + B cells, fluorescence-activated cell sorting (FACS) was used to isolate the cytokine-producing B cells. Flow cytometry following sorting confirmed enrichment of > 95% (Fig. 3a). qPCR analysis demonstrated that IL10 and CSF2 expression was strikingly and reciprocally enriched in the IL-10 + and GM-CSF + CSA-sorted B cells respectively (Fig. 3b). This data is in keeping with the near reciprocal expression of these cytokines observed when total B cells are co-stained using intracellular cytokine staining (Fig. 3c)    41 . We next demonstrated that the CSA sorted cells were able to respond to subsequent re-stimulation, with secretion of either IL-10 or GM-CSF confirmed by ELISA in culture supernatants of the respectively sorted sub-populations (Fig. 3d). These experiments demonstrated that functional, highly-enriched low-frequency cytokine-producing B cells can be isolated using CSA and flow cytometry, and integrated into downstream applications such as gene expression analysis or in vitro cell culture.

Simultaneous isolation of two low frequency cytokine-secreting human B cells. Concurrent
detection and isolation of multiple low frequency-cytokine-secreting cell sub-populations would allow for more efficient use of limited cell numbers often obtained from rare and/or limited clinical samples. We therefore next sought to adapt the CSA-Flow approach by multiplexing the capture reagents for concurrent detection and isolation of two low-frequency cytokine-expressing (IL-10 + and GM-CSF + ) B cell subpopulations. We found that the www.nature.com/scientificreports/ multiplexing successfully yielded highly purified cytokine-defined cell-subpopulations (Fig. 4a). Compared with single cytokine-captured B cells, however, the presence of a second capture reagent resulted in a reduction in the frequencies of IL-10 + (single-capture: 3.83 ± 0.84% vs. double-capture: 2.88 ± 0.65%, p < 0.0001) and GM-CSF + (single-capture: 5.25 ± 1.26% vs. double-capture: 4.28 ± 1.08%, p = 0.0139) B cells (Fig. 4b,c). Broad titration of the capture reagents did not overcome this limitation (data not shown). We conclude that detection and isolation of two, highly purified, viable low-frequency cytokine-secreting B cells are feasible using multiplexed CSA-Flow, though can result in reduced frequencies of the individual subpopulations. www.nature.com/scientificreports/ concurrent detection of three cytokine-secreting B cell subpopulations. We considered that concurrent isolation of low-and high-frequency cytokine-secreting cells may also be of interest. TNF is expressed by a high proportion of circulating human B cells as compared to GM-CSF or IL-10 27, 30, 32, 41 . After validating the detection and isolation of single-captured TNF + B cells using CSA (Fig. S1), we integrated TNF capture into our multiparametric CSA to test the feasibility of simultaneous detection and isolation of IL-10 + , GM-CSF + and TNF + B cells (t-SNE plot; Fig. 5a). To better understand assay performance, we evaluated the impact of simultaneous co-capture of two or three cytokines on the detection of individual cytokine-secreting B cells. We found that combined double-detection of the high-frequency (TNF + ) subpopulation with either of the low frequency (IL-10 + or GM-CSF + ) subpopulations, generally resulted in lesser decreases in yields of the low-frequency sub-populations than the decreased yields observed with combined detection of the two low-frequency cytokine secreting sub-populations (Fig. 5b). While triple-capture tended to result in somewhat greater decreases in yields of each of the three cytokine-defined populations (compared to single-or double-capture; Fig. 5b), the IL-10 + , GM-CSF + and TNF + B cell fractions isolated by flow cytometry exhibited excellent purities www.nature.com/scientificreports/ (Fig. 5c) and viabilities (Fig. S2). Further, the isolated IL-10 + and GM-CSF + B cells again exhibited the expected reciprocal enrichment for their corresponding cytokines by both qPCR (Fig. 5d) and as secreted cytokines measured by ELISA following further activation (Fig. 5e). In keeping with our prior work assessing co-expression of multiple B cell cytokines by intracellular cytokine staining 41 , we observed that TNF was expressed (both at the RNA level using qPCR and as secreted protein measured by ELISA) by a broad range of cytokine-sorted B cells, including those expressing IL-10 + or GM-CSF + (Fig. 5d,e). Overall, we describe a successful approach for simultaneous detection, and high-purity isolation, of both low-and high-frequency cytokine-defined B cell subsets using a combined CSA and flow cytometry approach (CSA-Flow) and provide useful insights on the impact of multiplexing different cytokine-capture reagents on signal detection, which will guide users in their particular applications.

Discussion
In this study, we demonstrate the utility of the cytokine secretion assay combined with flow cytometry (CSA-Flow) in isolating viable, low-and high-frequency cytokine secreting B cells, both individually or concurrently. Isolated cell fractions were highly purified, exhibited excellent viability, and could be integrated into downstream applications such as analysis of gene expression or further in vitro activation/cell culture. The assay relies on sequential antibody-based cell surface labelling, thereby preserving cell integrity and minimizing cell loss 34,35 . These features, along with the ability to isolate viable and functional low frequency cytokine-defined cells, encapsulate the main competitive advantage of this assay over traditional intracellular cytokine-staining or flow fluorescent in situ hybridization [19][20][21][22][23][24][25] . Assay multiplexing is attractive for maximizing the output of potentially limited clinical sample volumes, and has been previously used with CSA to both isolate and distinguish distinct cytokine-secreting immune cells 42,43 . Our data suggests that cell purities and viabilities are excellent upon multiplexing but the downside can be variable loss of signal. We speculate that signal loss during assay multiplexing was due to saturation of CD45, the cell surface anchor for the bispecific capture reagents 35 , which may impact on assay sensitivity. Reagent titration is important in optimizing signal to noise for each of the different multiplexed cytokines (general guidelines on flow-cytometry antibody titration have been discussed in recent publications 21,44,45 ), though this does not fully mitigate partial loss of signal. While this can impact yields of the isolated subpopulations, their purities (including of the concurrently isolated low-frequency populations) following the multiplexed CSA was nonetheless excellent, making them amenable to highly sensitive interrogation. We note relatively broad variation in the magnitude of cytokine expression by the isolated subpopulations, both at the RNA-expression and secreted-protein levels which, in addition to the expected inter-individual variability, may also reflect cellular-level heterogeneity 46 which could be of interest to elucidate.
CSA-Flow can be seamlessly integrated into multi-parametric FACS panels as part of complex cell-sorting strategies 44,47 , as the anti-cytokine antibodies can be pre-mixed with other cell surface antibodies and simultaneously applied to cells. Prior work with CSA has included pre-enrichment of cytokine-secreting immune cells using magnetic-based isolation 34,48,49 which can facilitate the subsequent study of particularly rare populations, and additional applications have included studies of cytokine detection kinetics 42 , cellular plasticity of cytokine expression 43,50 and interrogation of antigen-specific cytokine-expressing cells 34,49 . We envision that CSA-Flow can also be integrated in the workflow for next-generation and single-cell technologies. This would enable analysis of molecular pathways and transcriptional networks of cells secreting different cytokines, as well as analyses of cells secreting the same cytokine, in different contexts. As an example of the latter, distinct subsets of IL-17-expressing T cells have been shown to harbor homeostatic versus pathologic properties 51,52 . Further cellular state diversity could be gleaned from combining CSA-Flow, antigen receptor sequencing and single-cell RNA-seq, as it would allow to draw relationships between antigenic specificity and cell transcriptome among seemingly homogeneous functional cellular profiles 3,4 , extending prior antigen-specific CSA applications.
In summary, we present a method for the simultaneous isolation of viable low-and high-frequency cytokinesecreting human B cells using CSA-Flow, and highlight strengths and limitations of this approach. Application of this approach should provide users with an opportunity to simultaneously and more fully elucidate the biology of distinct cytokine-defined cell subpopulations, with in-depth analysis and further functional characterization predicated on isolation of highly purified and viable cells from limited clinical samples.

Materials and methods
cell isolation and culture. Fresh blood was obtained from healthy individuals recruited from the Montreal Neurological Institute and Hospital (MNI/H), McGill University and University of Pennsylvania. All subjects provided an informed consent as approved by the corresponding institutional ethics review boards. The study was approved by and carried out in accordance with the guidelines and regulations of the Institutional Review Board of the Montreal Neurological Institute (McGill University) and University of Pennsylvania and in accordance with the Declaration of Helsinki. Peripheral blood mononuclear cells (PBMC) were isolated from whole blood using Ficoll-Paque density gradient centrifugation (GE Healthcare) and a strict standardized protocol 53 . B cells were selected from PBMC using CD19 microbeads (Miltenyi Biotec) according to manufacturer's recommendations. Typical purities routinely assessed by flow cytometry were > 98%. B cells were cultured in serum-free X-Vivo medium (Lonza) and plated in flat-bottom 24 well plate at 2 × 10 6 /well in a total volume of 1500 μl of medium. Cells were activated with phorbol 12-myristate 13-acetate (PMA; 20 ng/ml, Sigma-Aldrich) and Ionomycin (500 ng/ml, Sigma-Aldrich) for 4 h. In the case of intracellular cytokine staining (ICS), Golgi stop (Monensin, BD Biosciences) was added to cells at the start of stimulation.  41 . Briefly, cells were stained with LIVE/DEAD fixable Aqua dead cell stain (Thermo Fisher Scientific) for 20 min on ice following which cell-surface marker staining was performed using mouse antihuman CD20 (BD Biosciences; clone: 2H7) and mouse anti-human CD3 (BD Biosciences; clone: UCHT1). Cells were then fixed and permeabilized using fixation/permeabilization buffer (BD Biosciences). Rat anti-human GM-CSF (clone: BVD2-21C11), rat anti-human IL-10 (clone: JES3-19F1) and mouse anti-human TNF (clone: MAb11) antibodies (BD Biosciences) or matching isotype controls were added and cells were incubated for 30 min on ice. Cells were washed and resuspended in FACS buffer (PBS/1%FCS) until analysis on a FACS LSR Fortessa (BD Biosciences).
Cytokine secretion assay (CSA). For the cytokine secretion assay, cells were harvested and washed once in ice-cold MACS buffer (PBS/2 mM EDTA/5%BSA). B cells were then resuspended in serum-free X-Vivo medium and labelled with capture antibodies for GM-CSF (final dilution 1:20), IL-10 (final dilution 1:10) or TNF-alpha (final dilution 1:10) (Miltenyi Biotec) for 10 min on ice. Cells were further diluted in pre-warmed medium at 1 × 10 5 cells/ml and incubated while rotating (using a MACS rotor) at 37° C/5% CO 2 for 45 min. B cells were then placed on ice for 10 min to terminate the secretion phase. Cells were subsequently washed twice in ice-cold MACS buffer before labelling with detection antibodies for GM-CSF (final dilution 1:20), IL-10 (final dilution 1:10) or TNF (final dilution 1:40) along with mouse anti-human CD20 (clone: 2H7) and mouse anti-human CD3 (clone: UCHT1) antibodies. We employed the anti-human TNF antibody (clone: MAb11; BD Biosciences) as a detection antibody. B cells were washed once in ice-cold MACS buffer and analyzed on BD LSRFortessa (BD Biosciences). Alternatively, cytokine-secreting B cells were sorted on BD FACSAria (BD Biosciences). Data analysis was performed with FlowJo Software (TreeStar).

Enzyme-linked immunosorbent assay (ELISA).
Cytokine levels in cultured supernatants were measured by ELISA kits (BD Biosciences or Thermo Fisher Scientific) following manufacturer's protocols.
Quantification of mRNA expression by quantitative, real-time polymerase chain reaction. Total RNA extraction was performed using RNeasy Plus Micro kit (Qiagen, Valencia, CA) following manufacturer's protocols. The RNA was stored -80C and used to generate single-stranded cDNA in a standard reverse transcription (RT) reaction using high-capacity cDNA reverse transcription kit with RNase inhibitor (Thermo Fisher Scientific). Analysis of gene expression was performed using the following TaqMan probes: CSF2 (Hs00929873_m1), IL10 (Hs00961622_m1) and TNF (Hs01113624_g1). 18 s (Hs03003631_g1) was used as a housekeeping gene. Fold change calculations were performed using the -ΔΔCT method.
Statistics. All values are expressed as mean ± SD. Student's paired t-test or two-way ANOVA with a Tukey post-hoc test was used for statistical comparisons between two groups. GraphPad Prism 8 was used to perform all the statistical analyses. P values of 0.05 or less were considered statistically significant.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.