In vitro-Induced Human IL-10+ B Cells Do Not Show a Subset-Defining Marker Signature and Plastically Co-express IL-10 With Pro-Inflammatory Cytokines

Regulatory B cells (Breg) have been described as a specific immunological subsets in several mouse models. Identification of a human counterpart has remained troublesome, because unique plasma membrane markers or a defining transcription factor have not been identified. Consequently, human Bregs are still primarily defined by production of IL-10. In this study, we sought to elucidate if in vitro-induced human IL-10 producing B cells are a dedicated immunological subset. Using deep immune profiling by multicolor flow cytometry and t-SNE analysis, we show that the majority of cells induced to produce IL-10 co-express pro-inflammatory cytokines IL-6 and/or TNFα. No combination of markers can be identified to define human IL-10+TNFα−IL-6− B cells and rather point to a general activated B cell phenotype. Strikingly, upon culture and restimulation, a large proportion of formerly IL-10 producing B cells lose IL-10 expression, showing that induced IL-10 production is not a stable trait. The combined features of an activated B cell phenotype, transient IL-10 expression and lack of subset-defining markers suggests that in vitro-induced IL-10 producing B cells are not a dedicated subset of regulatory B cells.


INTRODUCTION
A large number of studies demonstrated the existence of a regulatory subset of B cells (Bregs) in mice (1,2). Still, it is difficult to pinpoint a unique Breg phenotype, as the unspoken consensus that murine Bregs are CD1d Hi CD5 + CD19 + overlap in surface markers of B-1a cells (CD11b low CD5 + IgD + ), marginal zone B cells (CD1d Hi CD23 − IgM Hi ) and T2 marginal zone precursor B cells (CD1d Hi CD23 + IgM + ) (3). Indeed, IL-10 production seems to be the only distinguishing feature of Bregs , explaining their alias of "B10" cells (4). Once activated, IL-10-producing B cells can inhibit pro-inflammatory immune responses. The mechanisms of action of this process are largely unknown. Some of these regulatory effects act through IL-10, others can be IL-10 independent [reviewed by (5)]. It was shown that IL-10-producing B cells inhibit CD4 + T cell proliferation, increase FoxP3 expression in these cells and reduce T H 1 skewing. These effects were not only dependent on secreted IL-10, but also on CD80/86 costimulatory signals (6)(7)(8).
In human, specific B cells were also shown to exhibit regulatory functions (9)(10)(11)(12)(13)(14)(15)(16). The human equivalent of murine IL-10-producing Bregs however, remains matter of intense debate, both in terms of its actual existence and its potential defining phenotype as a specific B cell subset. Initially, human IL-10-producing B cells were described as being CD19 + CD24 Hi CD38 Hi . In one report the majority of the IL-10 + B cells were also described as being IgM Hi IgD Hi CD1d Hi CD5 Hi CD10 + CD20 + CD27 − (11). Another report showed that different subsets of transitional IL-10 + CD24 Hi CD38 Hi B cells have distinct regulatory properties (17). Other studies describe entirely different phenotypes for IL-10 + B cells, such as CD19 + CD25 + CD71 + CD73 − CD274 − , CD24 Hi CD27 + CD48 Hi CD148 Hi and CD20 + CD27 + CD43 + CD11b + (1,(18)(19)(20). Because the reported human IL-10-producing B cells show resemblance to immature, transitional, marginal zone, activated and memory B cells and even plasma cells, it is conceptually difficult to attribute a single regulatory function to these cells or coin them a Breg subset with confidence (1,(18)(19)(20)(21)(22)(23)(24). The marginal overlap between murine Bregs and human IL-10-producing B cells makes translation from animal studies to potential human Breg function all the more difficult.
Another challenging aspect of investigating in vivo relevance of IL-10-producing B cells is that IL-10 production by most B cells may not be an innate trait of a specific subset. Spontaneous IL-10 production in mice and men is only rarely reported, making the existence of 'natural Bregs' controversial (19,25). To properly assess the capacity of B cells to produce IL-10, ex vivo stimulation is routinely performed. Non-cognate triggering of B cells seems particularly potent in inducing IL-10, for example via TLR4 and TLR9 or via CD40L (11,(25)(26)(27)(28)(29)(30)(31)(32). Other IL-10inducing stimulations, such as IL-21, autoantigens, vitamin D3 and human chorionic gonadotropin (hCG) have been reported, but these have not gained broad recognition (33)(34)(35). Besides this non-cognate triggering, IL-10 can also be induced by B cell receptor (BCR) triggering (30), although data concerning simultaneous stimulation of BCR and TLR9 show conflicting results. In one study, simultaneous BCR ligation augmented CpG-induced IL-10 production (29). The opposite was found in another study with BCR ligation reducing the efficacy of CpG in inducing IL-10 in B cells, making it unclear what the effects of combined stimulations are on IL-10 production by B cells (3). In all of these cases, it is unclear whether Bregs develop from a specific pre-Breg lineage, or whether many B cells can adopt a regulatory phenotype after receiving the appropriate Abbreviations: Breg, regulatory B cell; hCG, human chorionic gonadotropin; BCR, B cell receptor; GCs, germinal centers; PBMCs, Peripheral blood mononucleated cells; FCS, fetal calf serum; PMA, phorbol 12-myristate 13-acetate; BFA, Brefeldin A; BSA, bovine serum albumin; t-SNE, t-distributed stochastic neighborhood embedding; DAPI, 4 ′ ,6-diamidino-2-phenylindole. signals (36). The production of IL-10 by subsets resembling different B cell subtypes supports the latter theory. Finally, it has been shown that IL-10 + B cells can also produce the proinflammatory cytokine TNFα (37). Co-expression of different cytokines suggests IL-10 can be produced by a range of B cells and is not a trait of a dedicated IL-10 producing regulatory B cell.
It is important to realize, although often underappreciated, that in lymph nodes IL-10 can have decidedly immunoactivatory effects, especially on B cell differentiation and humoral immune responses. IL-10 reduces B cell apoptosis in germinal centers (GCs) and induces plasma cell differentiation (23,(38)(39)(40), antibody production and promotes Ig isotype switching (13,40). Thus, in contrast to the proposed predominant regulatory role of Bregs on immunity, autocrine secretion of IL-10 by B cells is important in supporting humoral immune responses. Therefore, IL-10 may on the one hand be secreted by B cells at specific stages of B cell activation and function to direct immunity against specific antigens toward humoral immunity, while simultaneously acting as immune regulator for other arms of the immune system. The label "Breg subset" for IL-10 producing B cells would in that case be unfortunate and may give rise to undesired conclusions about identification of these cells in settings of human health or disease.
A true IL-10 + "Breg subset" would be expected to express some subset-defining, unique markers, transcription factors or other co-expressed regulatory molecules.
We therefore investigated the potential of B cells to stably produce IL-10 after stimulation with different agents, and investigated if they exhibit a unique and stable phenotype.

Isolation of Human B Cells
Buffycoats of healthy human donors were obtained from Sanquin Blood Supply upon informed consent and approval by local ethical committee (Sanquin Amsterdam) and in line with the Declaration of Helsinki. Peripheral blood mononucleated cells (PBMCs) were isolated from buffycoats using a Lymphoprep (Axis-Shield PoC AS) density gradient. CD19 + cells were separated using magnetic Dynabeads (Invitrogen) following manufacturer's instructions; resulting in >98% purity.

Flow Cytometry and Detection of IL-10
CpG-stimulated B cells were restimulated with PMA (100 ng/mL, Sigma-Aldrich), ionomycin (1 µg/mL, Sigma-Aldrich) and Brefeldin A (BFA; 10 µg/mL, Sigma-Aldrich) for 4 h on day of experiment. B cells were then washed with PBS and stained with LIVE/DEAD Fixable Near-IR (Dead cell stain kit, Invitrogen) for 30 min at room temperature in the dark. Subsequently, B cells were washed with PBS/1% bovine serum albumin (BSA; Sigma Aldrich). An extracellular staining was performed by incubating the B cells for 30 min at room temperature in the dark with fluorescently labeled antibodies (Supplemental Table 1). Cells were then fixed in 4% paraformaldehyde (PFA; Sigma-Aldrich), permeabilized with 0.5% saponin (Calbiochem) in PBS/1% BSA and stained with anti-IL-10, anti-TNFα and anti-IL-6 (Supplemental Table 1) antibodies for 30 min at room temperature in the dark. B cells were washed and resuspended in PBS/1% BSA and measured on an LSRII or LSRII Fortessa flow cytometer (BD Biosciences). Analysis was performed using FACS Diva (v6.1.2; BD Biosciences) and FlowJo (v7.6.5; Treestar) software. Live CD19 + cells were selected by setting a lymphocyte gate in FSC-A and SSC-A, followed by single cell gates using FSC-W/FSC-H and SSC-W/SSC-H and subsequent live CD19 + gating using the LIVE/DEAD Fixable Near-IR channel (APC-Cy7) and CD19 expression. Dead cells were excluded in Figures 1C,D in the absence of LIVE/DEAD Fixable Near-IR staining via strict gating on CD19 + cells. In all other figures, LIVE/DEAD Fixable Near-IR staining was used to exclude dead cells.

Multimarker Analysis Using t-SNE
B cells were stimulated with CpG for 2 and 4 days, restimulated with PMA, ionomycin and BFA for 4 h on the day of experiment and stained with two different antibody mixtures (Supplemental Table 1). B cells were measured on LSRII Fortessa flow cytometer. Live B cells were gated in Flowjo v10.3 analysis software and exported as separate fcs-files for seven different donors. Populations were randomly down sampled to 13,000 events and subsequently concatenated into 91,000 events using plugins in Flowjo v10.3 to normalize contribution between donors. Next, concatenated samples were analyzed using the t-distributed stochastic neighborhood embedding (t-SNE) plugin in Flowjo v10.3. t-SNE is a non-linear dimension reduction technique creating a low-dimensional map of highdimensional data that preserves distances between pairs of points (42). Basically, it "performs dimensionality reduction, allowing visualization of complex multi-dimensional data in fewer dimensions while still maintaining the structure of the data" (Flowjo v10 Documentation). This method has one main parameter, the perplexity value, which "sets the number of effective nearest neighbors" (Flowjo v10 Documentation). In our analysis, perplexity was set to 70, theta to 0.5, eta to 200 and iterations to 1,000. IL-10, TNFα, and IL-6 were selected as t-SNE parameters in Figures 3, 4 and Supplemental Figures 1A,B and the analysis was performed within live CD19 + cells. All markers including IL-10, TNFα, and IL-6 were selected as t-SNE parameters in Supplemental Figures 1C,D and the analysis was performed within IL-10 + B cells. Next, cytokinepositive cells (IL-10, TNFα, and IL-6) were manually gated in the concatenated sample file and used to distinguish the cytokine-positive subpopulations in the generated t-SNE data space. This allows to manually gate several IL-10 + TNFα − IL-6 − populations and IL-10 + cell populations co-expressing TNFα and/or IL-6 in the t-SNE data space. Finally, expression of the cell surface markers, contained in the two antibody mixtures, on these cytokine-positive cell populations were visualized as halfoffset overlay histograms. Quantification of cell surface marker expression was performed in the original non-down sampled, non-concatenated fcs data file of each individual donor.

Statistical Analysis
Statistical analysis was performed using GraphPad Prism (version 6; GraphPad Software). Data were analyzed using a Wilcoxon test, or Friedman's test where appropriate. Results were considered significant at p < 0.05. Significance was depicted as * (p < 0.05) or * * (p < 0.01).

RESULTS
Most in vitro-Induced Human IL-10 + B Cells Also Secrete TNFα and/or IL-6 Human CD19 + B cells isolated from peripheral blood were stimulated with the known IL-10 inducing TLR9 stimulus CpG (16,29) and analyzed at different time points (Figures 1A,B).
Within the first 4 days of stimulation the frequency of IL-10 + B cells steadily increased ( Figure 1B). No further increase was observed at later time points (data not shown). Comparison of the efficacy of IL-10 induction by other reported antigenindependent TLR-stimuli (Figures 1C,D) demonstrated that IL-10 can indeed be induced by various stimuli, but that CpG stimulation was a potent and consistent in vitro inducer of IL-10 production by B cells in this culture system, both by itself or in combination with other TLR-stimuli (Figures 1C,D). In all conditions, the potency for IL-10 induction varied between donors. Antigen-dependent stimuli activating B cells via the B cell receptor (BCR) or CD40, showed similar IL-10 induction by CD40 and CpG, while BCR ligation did not significantly induce IL-10 ( Figure 1D). Simultaneous addition of both CD40 and CpG did not yield a higher percentage of B cells expressing IL-10 ( Figure 1D). To define the overall pro-or anti-inflammatory potential of the IL-10-producing B cells, co-expression of the cytokines TNFα and IL-6 was assessed (Figures 1E,F). A large proportion of IL-10 expressing B cells co-express TNFα and/or IL-6 ( Figure 1E) that mostly does not reside in the same intracellular vesicles (Figure 1G). The percentage of IL-10 + TNFα − IL-6 − single-positive B cells and IL-10 + TNFα + IL-6 + triple-positive B cells increased over time (Figure 1H).
In contrast, IL-10 − TNFα + IL-6 + double-positive B cells were induced within 1 day of B cell stimulation, but diminished in frequency after that time point (Figure 1H). Co-expression of TNFα and/or IL-6 within the IL-10 + B cell population remained relatively stable over time ( Figure 1I) without a significant decrease or increase in IL-10 + TNFα + IL-6 − and IL-10 + TNFα − IL-6 + double-positive populations ( Figure 1J). These data show that most CpG-stimulated peripheral blood human B cells express IL-10 together with TNFα and/or IL-6.
These data indicate that in vitro stimulation of B cells by CpG induces IL-10 to a certain extent in transitional B cells, but not in plasmablasts.

IL-10 Production by in vitro-Stimulated Human B Cells Is a Transient Feature
To investigate if IL-10 production is a constitutive trait of a particular (regulatory) subset of B cells, or a transient feature of activated B cells, IL-10 + and IL-10 − B cells induced by CpG were sorted after 3 days, cultured for 3 additional days with B cell medium only and then re-stimulated. Most of the cells that were originally IL-10 + had lost their capacity to produce IL-10, while part of the formerly IL-10 − cells did produce IL-10 after this second round of stimulation ( Figure 6A). Comparison of IL-10 production in the re-culture of IL-10 + and IL-10 − sorted cells showed similar percentages of cells expressing IL-10 ( Figure 6B). This suggests that IL-10 production upon TLR9 stimulation represents a transient property of activated B cells.

DISCUSSION
To date, conflicting data on the phenotype and stability of human regulatory B cells lead to an ongoing debate on the functional regulatory importance of human Bregs. A few unique populations of human B cells have been identified that have all the characteristics of true regulatory cells, like CD24 Hi CD27 + , CD24 Hi CD38 Hi , and CD25 Hi CD71 + B cells (36). This is in part confirmed by our data showing that CD24 Hi CD38 Hi transitional B cells contain a large proportion of IL-10 + TNFα − IL-6 − B cells. These IL-10 + TNFα − IL-6 − transitional B cells only constitute 0.14-0.22% of the CD19 + cells, whereas the total population of IL-10 + TNFα − IL-6 − cells constitute 2-5% of the CD19 + cells, indicating that the majority of IL-10 + TNFα − IL-6 − cells consist of other B cell subsets. In general however, the main identifier of regulatory B cells is still considered to be IL-10 and in mice, indeed IL-10 seems to be expressed by B cell populations that are accepted to be Bregs (4). In human, B cells that produce IL-10 upon in vitro stimulation are therefore also often coined Bregs. Defined markers are however lacking and the stability of IL-10 production in these cells is unclear. There is continuous debate about the question if human IL-10-producing B cells define a true regulatory B cell subset.
In this paper, we show that induced IL-10 expressing B cells often co-express pro-inflammatory cytokines, which argues against a fully anti-inflammatory capacity of these cells. Although part of the cells produce only IL-10, these cells do not represent a single population as shown by variation of IL-10 expression level and do not express a unique surface marker signature. Finally, induced IL-10 expression is not stable: IL-10 producing B cells lose their capacity to produce IL-10 after re-stimulation. Taken together, these data suggest that in vitro-induced IL-10 + B cells do not represent a dedicated Breg subset but rather represents an activated B cell phenotype.
CpG stimulation resulted in the development of several populations of IL-10 producing cells, some of which display a distinctly pro-inflammatory phenotype, co-expressing TNFα and/or IL-6. Roughly half of the IL-10 + B cells express TNFα, while IL-6 expression was present in approximately 60% of IL-10 + B cells. Furthermore, 27-37% of IL-10 producing B cells co-expressed both cytokines. The population that expressed only IL-10 comprised approximately 25% of all IL-10 producing B cells. Co-expression of TNFα by IL-10 + B cells has been demonstrated before (37), but, since TNFα can also be associated with tolerance (43), this trait does not argue against an overall regulatory phenotype of these B cells. Our finding, however, of co-expression of the pro-inflammatory cytokine IL-6 does show that these cells also have pro-inflammatory properties, which may compromise the regulatory function of B cell-produced IL-10 on other cell types. Further research is needed.
Several subsets of B cells have been described to possess regulatory capacity by means of IL-10 production, especially transitional B cells (11) and plasmablasts (15). Within the transitional B cell population, the amount of IL-10 + TNFα − IL-6 − B cells is enriched, with approximately 65% of IL-10 + B cells not producing TNFα or IL-6. However, IL-10 producing transitional B cells only constitute 0.19-0.36% of the CD19 + cells in vitro, whereas around 2% IL-10 + B cells are still present after IL-10 + B cells were sorted and re-cultured for 3 more days (Figure 6). This indicates that even though IL-10 + transitional B cells may be Bregs they are not the only possible Breg subset. In contrast, within the plasmablast pool, TNFα was the predominant cytokine, with only very few B cells expressing IL-10.
Since cytokine analysis of CpG stimulated cells indicates several distinct populations of IL-10 + B cells, unbiased clustering based on IL-10, TNFα, and IL-6 was performed using t-SNE analysis. This demonstrated not only that there are several populations of IL-10 producing cells based on cytokine production, but furthermore, that IL-10 + TNFα − IL-6 − B cells are further divided into several distinct populations. However, no clear differences in surface marker expression were observed between these IL-10 + TNFα − IL-6 − B cell populations.
Arguably, cells that produce IL-10 alone and not TNFα and/or IL-6 are more likely to have a regulatory function similar to dedicated Bregs. We therefore compared IL-10 + TNFα − IL-6 − B cells to IL-10 + TNFα + IL-6 + B cells using flow cytometry. Indeed, IL-10 + TNFα + IL-6 + B cells display increased expression of CD80, indicating they might be more activated (44) and suggesting that IL-10 produced by these cells is a trait of activation. Interestingly, expression of CD5 was increased on IL-10 + TNFα − IL-6 − B cells. CD5 + B cells in mice negatively regulate inflammatory responses (10). Although expression of CD5 was increased on IL-10 single-positive cells, compared to TNFα single-positive cells as well, expression of these markers was not significantly higher than on IL-6 single-positive cells arguing against this human CD5 + B cell subset possessing regulatory functions to dampen inflammatory responses. Furthermore, t-SNE clustering on all surface markers including IL-10, TNFα and IL-6 within IL-10 + B cells revealed a large overlap of IL-10 + TNFα − IL-6 − with IL-10 + B cells co-expressing the pro-inflammatory cytokine IL-6 (IL-10 + TNFα − IL-6 + ), indicating that within the markers we tested, all of which have been linked to regulatory B cells in literature, none of the markers, either alone or combined, can define a unique subset of IL-10 expressing B cells.
Dedicated Bregs should at least retain the capacity to produce IL-10 when (re-)stimulated. Murine B10 cells however, were shown to secrete IL-10 transiently (45). Therefore, IL-10 expression upon restimulation was investigated in our assay. Cells that produce IL-10 after induction with CpG lose IL-10 expression after 3 days of in vitro culture. When restimulated, only a small fraction of these cells produced IL-10 again. This fraction was comparable to the fraction of formerly IL-10 − cells that produced IL-10 after a second round of stimulation. Together with the co-expression of pro-inflammatory cytokines and the lack of defining Breg markers, this suggests that IL-10 production is a transient trait of activated B cells at specific time points during differentiation and may not be a true Breg subset identifier. To unequivocally assess regulatory capacity of the IL-10 B cells, suppression assays using the various IL-10expressing B cells would be required. Currently, this is however, prohibited by the lack of surface markers distinguishing the IL-10 + TNFα − IL-6 − B cells from B cells co-expressing TNFα and/or IL-6.
The finding that transitional B cells are relatively enriched within the IL-10-expressing population [ Figure 2C and (11)] indicates that IL-10 expression is an early trait during B cell differentiation. Since IL-10 plays an active role in B cell differentiation, stimulating antibody production, expression of this cytokine after B cell stimulation is not surprising. The balance between dynamic production and consumption of IL-10 and the pro-inflammatory cytokines co-expressed by activated B cells, rather than the absolute amount of cytokine production, is likely to influence the outcome of local immunity. In this respect, it is of interest to mention the possibility for an autocrine loop, with IL-10 either suppressing or stimulating B cell activity (13); and may represent a general state of activation of the B cell.
The realization that IL-10-induction in B cells in vitro might not define a regulatory B cell subset per se but is part of an activated B cell phenotype co-expressing pro-inflammatory cytokines is important for the field of immunomonitoring and immunotherapy. Enrichment of IL-10 producing B cells in vitro in certain patient populations does not imply in itself that these patients exhibit overall immune regulation. Functional analysis of the regulatory capacities of IL-10 + B cells on the immunological effector function of various immune cells and the insight in stability of the IL-10 + B cells is warranted to get a better insight in the meaning of potential accumulation of IL-10 producing B cells in disease. The consideration of using induced IL-10 + B cells for tolerance-inducing therapies, either by itself or in combination with other tolerogenic cell types depends on the outcomes of these studies.

AUTHOR CONTRIBUTIONS
LL and P-PU participated in acquisition of data, analyses and interpretation of data, drafting and writing of the article and approved the final version. DWV, DV, EV, and AT participated in acquisition of data, critically reviewed the article and approved the final version. AtB, TR, and SvH participated in design of research, interpretation of data and contributed to writing, reviewing and finalizing the article.