Non-redundant roles of IL-21 and IL-4 in the phased initiation of germinal center B cells and subsequent self-renewal transitions

Abstract

We examined the unique contributions of the cytokines IL-21 and IL-4 on germinal center (GC) B cell initiation and subsequent maturation in a murine model system. Similar to other reports, we found T follicular helper cell (Tfh) expression of IL-21 begins prior to Tfh migration into the B cell follicle and precedes that of IL-4. Consistent with this timing, IL-21 signaling has a greater influence on the peri-follicular pre-GC B cell transition to the intra-follicular stage. Notably, Bcl6^hi B cells can form in the combined absence of IL21R and STAT6 derived signals, however these nascent GC B cells cease to proliferate and are more prone to apoptosis. When B cells lack either IL21R or STAT6, aberrant GCs form atypical centroblast and centrocytes that differ in their phenotypic maturation and co-stimulatory molecule expression. Thus IL-4 and IL-21 play non-redundant roles in the phased progression of GC B cell development that can initiate in the combined absence of these cytokine signals.

Introduction

Effective immune responses to pathogens and vaccines critically depend on germinal centers (GC) to generate long-lived, high affinity memory B cells and plasma cells (1,2). The current model of GC development envisions that activated B cells, after interacting with cognate T cells, commit to one of several potential fates; they either re-enter the follicle and commit to the formation of a new GC, or they migrate away from the follicle and differentiate into short-lived antibody forming cells or early memory B cells. Similarly, a proportion of activated CD4⁺ T cells differentiate into a helper cell subset, known as T follicular helper cells (Tfh), that have rewired follicular chemotactic propensities and supply the cytokines IL-4 and IL-21 to GC B cells (3, 4).

Our prior studies examining these early events leading to GC initiation indicate that B cells dwell at the follicular boundary and between follicles for several days, while in close and recurring contact with differentiating T cells. Within this milieu, the first events of GC B cell differentiation and Tfh cell maturation occur, including elevated expression of Bcl6, a transcriptional repressor required for their formation (5–10). Bcl6 controls GC B cell differentiation by regulating cell cycle genes, repression of terminal differentiation factors and suppression of some signaling pathways, including B cell receptor signaling (8, 11). Repressed target genes in mice include CD38, a member of an ecto-enzyme family (12) and CD23. However, the extent of repression of potential Bcl6 target genes is influenced by the composition of Bcl6 corepressor complexes that can differ functionally (13).

Following interactions with antigen-specific T cells at the follicle periphery, Bcl6-expressing B cells committed to the GC pathway re-enter the follicle, divide at an extremely rapid rate, and form very large aggregates that organize into mature GCs with microanatomically distinct compartments. The light zone (LZ) is distinguished by the presence of follicular dendritic cells (FDC), a stromal cell subset, the fine processes of which form a reticular network with high levels of the complement receptor CD35 (14). The LZ also harbors a high density of Tfh cells. Much of the terminology used to describe murine GC B cells is based on the appearance of tonsillar GCs that harbor “centrocyte” LZ B cells that are typically smaller and “centroblast” DZ B cells that are larger and more often mitotic (15). Whereas centrocytes are predominantly CXCR4lo CD86 hi and express higher levels of CD40 and CD86, centroblasts are characterized as typically CXCR4hi and CD86lo (16). Intravital imaging studies of Tfh contacts with LZ B cells (17, 18), coupled with prior observations of inter-zonal migrations (19–21) support a model of GC B cell dynamics in which Tfh cell engagement of centrocytes within the LZ instructs a rewired transcriptional program that propels them to migrate to the DZ as centoblasts. During extensive clonal expansion, GC B cells with higher affinity BCR variants are enriched, eventually differentiating further into long-lived plasma cells and memory cells, a process that is only observed when effective cognate T follicular helper (Tfh) cell responses are also evoked (2, 22).

The formation of small GCs that fail to mature in the absence of a T cell adaptive immune response calls into question the absolute requirement of T helper cell-derived cytokines for the onset of GC B cell differentiation (23). T cell support for GCs is known to depend on the cytokines IL-21 and IL-4. In the absence of signaling through either IL-4 or IL-21 receptors in B cells, GC formation occurs in vivo, but the resulting GCs are diminished in size (24–27). In the case of IL-21, smaller GCs are thought to occur from a direct effect on B cells that results in reduced Bcl6 levels within GC B cells (24, 25). In vitro, both IL-21 and IL-4 can enhance Bcl6 transcription and/or translation in B cells (28, 29). Both cytokines are also known to influence the formation of antibody secreting cells (ASC) (30–35). Interestingly, Tfh cells can differ in their cytokine expression profiles at different stages in their differentiation and at distinct locations within lymphoid tissue (35–37). Whereas Tfh cells that emerge early in an immune response predominantly express IL-21 but substantially less IL-4, the reverse is true of distal LZ Tfh within fully mature GCs (37). IL-4R and IL-21R complexes activate distinct signaling pathways; IL-21 predominantly activates STAT1 and STAT3, whereas IL-4 and IL-13 signal uniquely via STAT6 (38, 39).

Here we examined the earliest stages of GC B cell formation, its reliance on IL-21 and IL-4 signaling, and its temporal relationship to the expression of these cytokines by Tfh cells. Our results support a multi-phased model of GC B cell development. Notably, we found that Bcl6^hi nascent GC B cells still develop in the combined absence of both IL-21 and IL-4-derived signaling, but are unable to progress further to form a fully mature GC. Similar to prior reports (36, 37, 40), we observed IL-4 and IL-21 to be differentially expressed by emerging Tfh cells; IL-21 expression is evident as early as 2–3 days post immunization, while increased IL-4 expression becomes apparent when Tfh cells reside within follicles. Consistent with this timing, we found that B cell intrinsic IL-21 signaling played a stronger role than IL-4 during the transition from the peri-follicular pre-GC B cell phase to the intra-follicular phase.

Our results further indicate that IL-21 and IL-4 play non-redundant roles in subsequent GC B cell maturation and self-renewal in vivo. In the absence of either of these signaling pathways, GC B cells form aberrantly and are compromised in their self-renewal in that environment. IL21R and STAT6 deficient GC B cells have distinct pathologies, differing in their phenotypic maturation and co-stimulatory molecule expression. While IL-21 signaling is requisite to properly instruct the centocyte to centroblast transition, IL-4 dependent signaling has a greater influence on the acquisition of the centrocyte state. Therefore, in the absence of either of these signaling pathways, Bcl6+ B cells can form, however their failure to establish and maintain a properly functioning GC has revealed non-redundant roles for IL-21 and IL-4 signaling in the phased progression of GC B cell development.

Materials and Methods

Mice and Genotyping

B1–8 mice (41)were crossed to a homozygous deletion of the Jκ locus (42). The B1–8 K/I gene carries the Vh 186.2 Ig heavy chain derived from the B1–8 hybridoma which generates a BCR with moderate affinity for the hapten (4-hydroxy-3-nitrophenyl) acetyl (NP) when paired with λ1 or λ3 light chains (43). B1–8 cells obtained from mice with a homozygous deletion of the Jκ locus are thus highly enriched for the anti-NP specificity due to the absence of kappa expressing B cells. Hapten-specific mice were crossed with strains of mice either harboring a homozygous deletion of the transcription factor Stat6 (#5977; B6.129S2(C)-Stat6^tm1Gru/J; Jackson Labs), or a homozygous deletion of the receptor for the cytokine IL-21 (44), to be used as sources of hapten-specific B cells lacking the ability to signal through Stat6, or IL-21 respectively. IL21R−/− mice were generated at Lexicon Genetics, provided by ZymoGenetics (a Bristol-Myers Squibb Company) and backcrossed 12 generations onto the C57BL/6 background prior to crossing with hapten specific strains. The lack of IL-21R expression in IL-21R^−/− mice was confirmed by PCR analysis and flow cytometry.

Ovalbumin-specific TCR-transgenic (OT-II) mice (#4194; Tg(TcraTcrb)425Cbn/J; Jackson Labs) were used as a source of carrier-specific T cells. Hapten-or carrier-specific mice were further crossed with strains of mice expressing fluorescent proteins within all nucleated cells that produced either dsRed under control of the β-Actin promoter (#6051; Tg(CAG-DsRed*MST)1Nagy/J); Jackson Labs) or eGFP via the ubiquitin promoter (#4353; Tg(UBC-GFP)30Scha/J); Jackson Labs). Mice carrying a transgene encoding a TCR specific for a LCMV epitope (SMARTA) were used as recipients for cell transfers (45). Some experiments used wild type C57BL/6 mice as recipients (#664; Jackson Labs). All mice had been fully backcrossed onto the C57BL/6 background and were maintained under specific-pathogen-free conditions. All experiments were approved by the Yale University Institutional Animal Care and Use Committee (IACUC).

Homozygosity of the Igh^VNP gene and the Jκ knockout allele was detected by PCR as described (7). Homozygosity of the Stat6 knockout allele was detected by the presence of a ~380 bp PCR product (mutant) and the absence of a ~275 bp PCR product (wild type) with the following primer sequences: Stat6 wild type 5’-AAGTGGGTCCCCTTCACTCT-3’, Stat6 common 5’-ACTCCGGAAAGCCTCATCTT-3’, and Stat6 mutant 5’-AATCCATCTTGTTCAATGGCCGATC-3’. Homozygosity of the IL-21R knockout allele was detected by the presence of a ~380bp PCR product (mutant) with the following primer sequences: IL-21R mutant 5’-GCAGCGCATCGCCTTCTATC-3’ and 5’-GAAGTTCTGCACAGTGTCTAGC-3’; and the absence of a ~280bp PCR product for the IL-21R wild type locus 5’-CTCCAAAGGGAGGGATCAGAAC-3’ and 5’-GAAGTTCTGCACAGTGTCTAGC-3’. OT-II mice were screened for the presence of the TCR transgene by detection of a ~bp PCR product with the following primer sequences: OTII 5’-GCTGCTGCACAGACCTACT-3’ and 5’-CAGCTCACCTAACACGAGGA-3’. SMARTA mice were screened for the presence of the TCR transgene by detection of a ~170 bp PCR product with the following primer sequences: Vα2 5’-ATAAAAAGGAAGATGGACGATT-3’ and 5’-TGGGGCTGACTGATACCG-3’.

Adoptive Transfers and Immunizations

Hapten-specific B cells were enriched from the spleens of B1–8⁺ Jκ-^/- mice by immunomagnetic purification with the EasySep Negative Selection Mouse B Cell Enrichment Kit (StemCell Technologies). T cells were isolated from the spleens and lymph nodes of OT-II mice with the EasySep Negative Selection Mouse CD4⁺ T Cell Enrichment Kit (StemCell Technologies). For all flow cytometry, sorting and histology experiments, 3 × 10⁶ B cells and 1 × 10⁶ T cells were transferred prior to immunization. To ensure consistency in the cell transfer of the donor cell populations, all recipients of a time course experiment received their transferred cells on the same day. Cells were injected intravenously into SMARTA recipients that were immunized 1 – 5 days later such that the tissue harvest, staining and flow cytometric analysis for the entire time course study were begun on the same day. In some experiments, wild type C57BL/6 mice were used as recipients instead. For immunizations, NP-OVA was precipitated in alum at 0.25 mg/mL and 50 μg injected i.p. The succinic anhydride ester of Nitrophenyl (Biosearch Technologies) was conjugated to ovalbumin (Sigma) in-house. In some transfer studies, 3mg BrdU in 200 μL PBS was injected I.V. in the tail vein 6 hrs before sacrifice.

Antibody Reagents

Antibodies with the following specificities were used for flow cytometry and histology: GFP-FITC (goat, Rockland), CD38-PE/Cy7 and -Pacific Blue (90, BioLegend), CD45R-PE, -BV421, -PE/Cy7, and -APC/Cy7 (RA2–6B2, BioLegend), Bcl6-AL647 (K112–91, BD Pharmingen), Bcl6-AL647 (7D1, Santa Cruz), Activated Caspase-3-PE (C92–605, BD Pharmingen), BrdU-bi and –AL647 (3D4, Phoenix), CD4-BV421 (GK1.5, BioLegend), ICOS-APC (C398.4A, BioLegend), CD62L-APC/Cy7 (Mel-14, BioLegend), CCR7-PE (4B12, BioLegend), CD86-BV605 (GL1, BD Bioscience), CD40-bi (HM40–3, eBioscience), CXCR4-PE (L276F12, BioLegend), CD83-bi (Michel-19, BioLegend), Mouse IgD-V450 (11–26c.2a, BD Bioscience), and CXCR5-PE (2G8, BD Pharmingen). The following antibodies were purified and conjugated in-house: CD4-Pacific Blue and –AL680 (GK1.5), CD45R-AL647 (RA2–6B2), IgD-bi (11–26c), CD35-bi (8C12), and CD23-AL680 (B3B4). Anti-FITC-AL488 (goat polyclonal, Invitrogen), streptavidin-BV421, and –BV605 (BioLegend), and streptavidin-AL555, and –PECy7 (Invitrogen) were used as secondary reagents.

Immunofluorescent Histology

Portions of excised spleens were fixed with 1% paraformaldehyde-lysine-periodate solution and frozen in OCT (TissueTek) after passage through sucrose gradient solutions. Cryostat sections (7μm) were blocked in PBS containing 1% BSA, 0.1% Tween-20, and 10% rat serum prior to staining with combinations of the antibodies described above. Prior to staining with anti-BrdU, some sections were permeabilized with 0.1% Triton-X in 0.1% sodium citrate buffer, followed by DNase (Sigma D5025, Bovine Pancreas) digestion to reveal BrdU epitopes. Stained sections were mounted with Prolong Gold anti-fade mounting medium (Invitrogen). Separate images for each fluorochrome were acquired with an automated wide-field microscope (Nikon Eclipse Ti) and a CCD camera (QImaging Retiga 2000R) with NIS Elements software. Emitted light was collected through 440/40 or 460/50, 525/50, 605/70, and 700/75 nm bandpass filters. Final processing to overlay single-channel images was performed with Adobe Photoshop.

Flow Cytometry

Spleen cell suspensions were initially blocked with an anti-FcγR (CD16/32 2.4G2) in PBS containing 3% FCS before further incubation with a combination of the indicated reagents. Dead cells were excluded on the basis of EMA staining or Live/Dead Fixable Aqua (Molecular Probes). Intracellular staining of Bcl6 and Activated Caspase-3 was done using the BD Cytofix/Cytoperm Fixation/Permeabilization Solution Kit (BD Biosciences). Staining with anti-BrdU required that samples were fixed and permeabilized in a 3-step process: an initial fixation with 1% paraformaldehyde, a second fixation and permeabilization in ethanol, and a final fixation and permeabilization in 1% paraformaldehyde containing 0.05% Tween-20. This was followed by DNase digestion to reveal BrdU epitopes for staining. Flow cytometry was performed on a LSRII cytometer (Becton Dickinson) and analyzed with FlowJo software (Treestar). 1 × 10⁶ events were collected per sample. After doublet discrimination and gating on live cells, populations, as indicated in the figure legends, were assessed. Fluorescence Minus One (FMO) controls were used to guide further analysis of viable cells after doublet discrimination.

Quantitative Polymerase Chain Reaction

RNA was isolated (RNeasy Plus Kit, QIAGEN) from cells sorted based on expression of the classical Tfh cell markers CXCR5 and ICOS such that an Ag specific Tfh population (GFP⁺ CD4⁺ CD45R^- CXCR5^hi ICOS^hi) and the non-responding endogenous T cell population (GFP^- CD4⁺ CD45R^- CD62L^hi) were collected separately for each sample. After cDNA was generated (Superscript VILO cDNA Synthesis Kit, Invitrogen), the IL-4 and IL-21 genes relative to the β-Actin genes were amplified (M 3000p, Stratagene) with primers as described: IL21 (46); IL4 (7). Relative gene expression was calculated as the log 2 difference between the Ct value for the gene of interest and β-actin.

Statistical Analysis

Prism software (Graphpad) was used to graph the data and to calculate statistical significance using an unpaired Student’s t test with a two-tailed 95% confidence interval, unless otherwise indicated. ANOVA followed by Student’s t test with a Bonferroni correction was applied to multiple comparisons.

Results

IL-21 gene expression by the Tfh population precedes migration into the B cell follicle; Elevated IL-4 expression is most prominent at mature GC time points.

To assess the influence of IL-21 and IL-4 on the initial stages of GC B cell differentiation, we examined the kinetics of Tfh cell cytokine expression during the early stages of an adaptive immune response. For these studies, we employed a well-characterized adoptive transfer system. In order to track individual Ag-specific T cells in vivo, purified hapten-specific B cells obtained from B1–8^+/+ Jk−^/− mice were adoptively transferred together with ovalbumin (OVA)-specific (OTII) T cells that expressed green fluorescent protein (GFP) under the direction of the ubiquitin promoter. SMARTA Tg mice that harbor an irrelevant TCR-transgene were used as recipients to eliminate the endogenous T cell response to OVA (45). Recipients received an intraperitoneal (i.p) immunization with NP-haptenated OVA (NP-OVA) in alum and splenocytes were prepared for sorting 1 to 11 days later. T cells were sorted based on expression of the classical Tfh cell markers CXCR5 and ICOS such that an Ag specific Tfh population (GFP⁺ CD4⁺ CD45R^- CXCR5^hi ICOS^hi) and the non-responding endogenous T cell population (GFP^- CD4⁺ CD45R^- CD62L^hi) were collected separately.

IL-21 gene expression (Fig. 1A) was significantly increased as early as 2 days post-immunization (p.i.) in the differentiating Tfh population when compared to non-responding endogenous T cells. This increased further 3 days p.i. and remained elevated through day 11 p.i., consistent with previous reports on IL-21 expression kinetics (37, 47). By contrast, IL-4 gene expression (Fig. 1B) was delayed relative to IL-21 and could only be detected 3 days p.i. and, similar to prior observations (36, 37, 40), becomes predominant only at a mature GC time point. The dominance of IL-21 at the early d2 time point suggests that IL-21 may play a larger role in influencing early B cell differentiation while IL-4 may be more important within the mature intra-follicular GC. Our previous report on the microanatomic location of Tfh cells using this same adoptive transfer system found that the initial development of Tfh cells began at the periphery of B cell follicles 2–3 days p.i., prior to their migration into the follicle interior shortly thereafter (7). Based on this timing with our transfer system, IL-21 expression initiates when Tfh cells are completing their residency at the follicle periphery, whereas an increase in IL-4 expression by Tfh cells becomes more prominent after T cell migration to the follicle interior.

Figure 1. Kinetics of Tfh cell IL-21 and IL-4 gene expression.

(A - B) GFP⁺ OVA-specific T cells and non-fluorescent NP-specific B cells were transferred into non-responding SMARTA recipients, which were subsequently immunized i.p. with NP-OVA in alum. Spleens were harvested at indicated times post-immunization and two populations of T cells, Tfh (GFP⁺ CD4⁺ CXCR5^hi ICOS^hi) and non-responding host naive T cells (GFP^- CD4⁺ CD62L^hi) were sorted. cDNA was generated from mRNA and the expression of the IL-21 (A) and IL-4 (B) genes relative to β-Actin and normalized to naïve T cell levels was determined by quantitative PCR. Each data point represents an individual mouse except for the day 1 and day 2 Tfh populations. Due to the recovery of very small numbers at these early time points, two mice were pooled for each data point. Shown is a representative experiment of three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001.

Nascent GC B cell development occurs in the combined absence of IL21R and STAT6 signaling.

In order to evaluate potential differences in the B cell intrinsic roles of IL-21 and IL-4 signaling during initial GC development, B1–8+/+ Jk−/− mice were crossed to IL-21R−/− mice or STAT6−/− mice, the latter of which are unable to signal through IL-4 or IL-13 receptors that uniquely require STAT6 for downstream signaling (48). WT, IL21R−^/−, STAT6−^/−, or IL21R−^/−STAT6−^/− (DKO) NP-specific B cells were adoptively transferred with OVA-specific T cells into either SMARTA or wild type C57BL/6 recipients. Similar results were observed with either type of recipient (data not shown). Recipients were immunized with NP-OVA in alum i.p. 2–6 days prior to analysis.

The kinetics of Bcl6 expression by transferred Ag-specific B cells was determined by flow cytometry of splenocytes (Fig. 2A and S1). Two days p.i. Bcl6 expression levels within the NP-specific B cell population was slightly but universally shifted (~2X fold increase, Fig. S1), unrelated to an increase in cell size (data not shown). It was not until 3d p.i. that a distinct Bcl6hi population became apparent, presumably representing GC-committed B cells. Indeed this coincided with the onset of CD38 downregulation on Bcl6hi but not Bcl6lo wild type NP-specific B cells (Fig. S2). Importantly, Bcl-6^hi cells were evident within IL-21R−^/− STAT6−^/− (DKO) B cell populations at this early stage of the response (Fig. 2A). Therefore, IL-21 and IL-4 signaling is not required for early formation of pre-GC B cells prior to follicular localization. By d4 p.i., significantly reduced numbers of Bcl-6^hi cells were observed in the IL-21R−^/− and DKO responses, while the defect in the STAT6−/− B cell response was less severe (Fig. 2A and Table I). By d6 p.i, a timepoint at which WT GCs are established, the STAT6−^/− response was also significantly defective compared to wild type, although not to the level of the IL-21R−^/− or DKO response (Fig. 2A). Therefore, expansion of the antigen specific GC population is compromised among STAT6−^/− B cells and more severely in the IL-21R−^/− and DKO populations. Together, these results suggest that B cell differentiation to the Bcl6^hi state can initially occur in the combined absence of IL-4 and IL-21, but that further expansion is limited beyond the nascent state.

Figure 2. The generation of Bcl6^hi Ag specific B cells is reduced in the absence of either IL21R or STAT6 dependent signals due to reduced proliferation.

(A - D) Flow cytometry analysis of splenic Ag specific B cells. GFP⁺ hapten-specific B cells (WT, IL21R−^/−, STAT6−^/−, or IL21R−^/−STAT6−^/−) and non-fluorescent OVA-specific T cells were transferred into non-responding SMARTA or C57BL/6 recipients, prior to i.p. immunization with NP-OVA in alum. Transfers and immunizations were timed so that analysis was performed on the same day for all time points. Recipients received an intravenous injection of 3 mg BrdU 6 hrs before harvest to label cells actively proliferating. The legend shown in A applies to data plots C - D. (A) Shown is the percentage of Bcl6^hi Ag-spec. B cells (CD45R⁺ CD4^- GFP⁺). Each data point represents an individual mouse. Shown is a representative experiment of three independent experiments. (B) Ag-specific B cell expansion is reduced in the absence of either IL21R or STAT6 signaling at the onset of the follicular germinal center response, 4 days p.i. Shown is the average number of GFP⁺ Ag-specific B cells per 1×10⁶ total splenocytes as determined by flow cytometry. Data shown is pooled from three separate experiments. (C) Ag-specific B cell proliferation is reduced in the absence of IL21R and STAT6 signaling. Shown is the percentage of BrdU⁺ CD38^lo Ag-specific B cells. Shown is a representative experiment of three independent experiments. (D) The absence of either IL21R or STAT6 dependent signaling does not affect the amount of apoptotic cell death present during the early GC response as determined by activated caspase-3 staining. When both IL21R and STAT6 dependent signaling is knocked out, however, cell death is found to increase in the CD38^lo subpopulation. *p < 0.05, **p < 0.01, ***p < 0.001, n.s not significant.

Table 1.

Bcl6^hi antigen specific B cell Numbers

Genotype	# of Bcl6hi Antigen Specific B cells per 103 Total Antigen Specific B cellsa
	Naive	Day 2b	Day 3b	Day 4b	Day 6b
WT	9.9 ± 6.3	6.6 ± 2.2	37.8 ± 9.6	230.1 ± 37.5	363.7 ± 26.8
IL21R-/-	6.1 ± 3.4	3.9 ± 1.2	15.2 ± 3.2	66.6 ± 13.9**	46.7 ± 10.1***
STAT6-/-	3.4 ± 1.2	2.4 ± 0.9	23.1 ± 4.3	224.8 ± 47.5	106.3 ± 9.0***
IL21R-/- STAT6-/-	3.5 ± 1.8	1.7 ± 2.2	14.5 ± 5.3	41.9 ± 3.9**	30.0 ± 4.4***

^aAntigen specific B cells were defined as GFP⁺ B220⁺ CD4^- splenic mononuclear cells

^bDays post-immunization with NP-OVA in alum I.P.

Indicated significance based on comparison to WT population at each given time point. Shown is a representative experiment of three independent experiments. Data shown as mean ± SEM. n = 4 or 5.

^*p < 0.05,

^**p < 0.01,

^***p < 0.001.

Proliferation rate and extent of apoptosis of Ag specific B cells in the absence of IL21R and STAT6 dependent signaling.

The total numbers of Ag-specific B cells (Fig. 2B) were comparable regardless of genotype from the naïve time-point through 3 days p.i. This indicates that survival of naïve B cells in the recipient host was not affected in the absence of IL21R- or STAT6-dependent signals, consistent with previous reports (25, 49) and that initial expansion of responding B cells occurs independent of these signals at the follicle periphery. To determine if the reduction in Ag-specific B cell numbers observed 4–6 days p.i. was due to a change in proliferation rate within the GC compartment, the extent of incorporation of the nucleotide analog BrdU during DNA replication was assessed by flow cytometry. Consistent with the timing of the diminution of responding B cell numbers, a decrease in BrdU⁺ CD38lo B cells was evident earlier among IL21R−^/− B cells (d4), than the STAT6 deficient population (d6) (Fig. 2C).

We also evaluated the amount of apoptotic cell death based on activated caspase-3 staining by flow cytometry. The extent of apoptotic cell death within the CD38^lo population of IL21R or STAT6 deficient B cells was not significantly different (Fig. 2D), suggesting that the decrease in Ag-specific B cell numbers more likely results from a reduced proliferative rate rather than an increased rate of cell death. However, CD38lo DKO B cells showed a significant increase in the amount of cell death, indicating that either IL21R or STAT6 dependent signaling is sufficient to at least transiently promote survival independently. That the rapid proliferative capacity typical of d6 WT GC B cells is not achieved unless B cells are sufficient in both the IL-21 and IL-4 signaling pathways indicates that these cytokines synergize in a B cell intrinsic fashion to evoke a response that is distinct from that observed with either independently.

To further assess the physical location of proliferating B cells, immunofluorescence histology of splenic tissue sections was performed (Fig. 3). Based on BrdU incorporation, proliferative Ag-specific B cells appropriately homed to the periphery of follicles regardless of genotype 2 days p.i. (Fig. 3A,C). By 3 days p.i. Ag-specific B cells, regardless of genotype, could be found with upregulated Bcl6 at the follicle periphery (Fig. 3A,D). By 6 days p.i., GFP⁺ Bcl6^hi B cells deficient in either IL21R or STAT6 were observed within follicles harboring newly formed GCs. Whereas the GCs observed in the recipients of WT B cells were dominated by the transferred GFP⁺ Ag-specific B cells, the small GCs that did form with IL21R or STAT6 deficient B cells had a mixed composition that included some recipient-derived B cells that may have gained a competitive advantage under these conditions (Fig. 3B). Thus, while an initial increase in Bcl6 levels can occur independently of IL21R and STAT6 signaling, propagation within GCs requires both.

Figure 3. Appropriate Ag specific B cell homing and increased Bcl6 expression can occur in the absence of IL21R and STAT6 dependent signals.

(A - D) Immunofluorescence staining of spleen sections from C57BL/6 mice that received an adoptive transfer of GFP⁺ NP-specific B cells (WT, IL-21R−^/−, STAT6−^/−, or IL-21R−^/−STAT6−^/−) and non-fluorescent OVA-specific T cells. Recipients were immunized 1 – 5 days after transfer and tissue was harvested for analysis 2–6 p.i. Mice received an intravenous injection of 3 mg BrdU 6 hrs before harvest. Sections were stained for B220 to highlight B cell follicles, GFP, BrdU to identify actively proliferating cells, and Bcl6. Scale bars represent 100 μm. Representative images from one of three independent experiments for each time point are shown. (A) Histology time-course to evaluate arrival of Ag-specific B cells to the B cell follicle after adoptive transfer (naïve), homing to bridging channels (day 2), increased levels of Bcl6 (day 3), and GC formation (day 6). (B) Higher magnification images of germinal centers, indicated in day 6 column by white boxes. (C) Higher magnification images of bridging channels, indicated in day 2 column by white boxes. Circles highlight examples to demonstrate proper homing of Ag-specific B cells to the area and active proliferation regardless of genotype. (D) Higher magnification images from B cell follicles, indicated in day 3 column by white boxes. Circles highlight examples of high levels of nuclear Bcl6 in Ag-specific B cells regardless of genotype and their proliferation status.

CD38 and CD23 expression differences of emergent GC B cells.

To assess the extent of maturity of emerging GC B cells, the extent of suppression of CD38 and CD23 surface membrane levels were assessed by flow cytometry. CD38 and CD23 are expressed at high levels by naïve follicular B cells and these expression levels are maintained upon initial Bcl6 elevation during pre-GC B cell formation. At mature GCs timepoints within WT mice, CD38 levels are reduced on both centroblasts and centrocytes, while CD23 levels are typically higher in centrocytes (50, 51). An increase in Bcl6 levels coincides with reduced CD38 levels in WT GC B cells, regardless of CD23 levels (Fig. 4A,B). Consistent with previous reports on the effect of IL-21 on GC B cells (24, 25), the Bcl6hi IL21R deficient and DKO populations, while expressing levels higher than naïve B cells nevertheless displayed lower Bcl6 MFI levels compared to WT GC B cells at 6 – 11 days p.i. However, Bcl6 MFIs are not significantly different in STAT6 deficient B cells (Fig. 4C).

Figure 4. CD23 and CD38 expression levels of Bcl6^hi Antigen-specific B cells.

(A – E) Flow cytometry analysis of splenic Ag-specific B cells. Cell transfers and immunizations were performed as described in Fig. 2. Plots shown are representative of three independent experiments with each data point representing an individual mouse. (A – B) CD38 down-regulation directly correlates with an increase of Bcl6 in responding Ag-specific B cells. Transferred GFP⁺ B cells were analyzed based on subset gating with CD23 and CD38 (A). One representative plot from a naïve and day 6 time point is shown. A histogram of Bcl6 expression (B), color coordinated with the subset-gating scheme from panel A, shows that increased Bcl6 protein levels are directly coordinated with a reduction in CD38 levels at this time point. (C) In the absence of IL21R dependent signaling, expression of Bcl6 in Ag-specific B cells is reduced. Shown is the mean fluorescene intensity (MFI) of Bcl6 staining for CD38^lo GFP+ Ag-specific B cells. Day 9 and 11 time-points are from an independent experiment. (D) In the absence of IL21R dependent signaling, expression of CD23 is elevated in Bcl6^hi Ag-specific B cells. Shown is the MFI of CD23 staining for Bcl6^hi Ag-specific B cells. The day 6 and 11 time-points are from an independent experiments. (E) In the absence of either IL21R or STAT6 dependent signaling, CD38 down-regulation in Bcl6^hi Ag-specific B cells is reduced. Shown is the MFI of CD38 staining for Bcl6^hi Ag-specific B cells. The day 6 and 11 time-points are from independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001.

The phenotype of the signaling deficient Bcl6^hi B cells indicates that the GC B cells that form, do so aberrantly. We found that in the absence of IL21R signaling the distribution of CD23^hi vs CD23^lo CD38^lo B cells was skewed in favor of a CD23^hi phenotype (Fig. 4D), consistent with a previous report of CD23 down-regulation after exposure to IL-21 in vitro (31). By contrast, the distribution of CD23 subsets among STAT6−^/− GC B cells remained comparable to WT GC B cells. Analysis of the CD23 MFI of CD38^lo B cells revealed that IL21R−^/− GC B cells have supra-elevated levels of CD23 that are significantly higher than that of naive B cell populations (Fig. 4D). Interestingly, in the DKO B cells CD23 levels remain comparable to that of non-responding B cells, indicating that IL-4 signaling in the absence of IL-21 results in CD23 overexpression, but the absence of both does not during GC development. Notably, the opposite relationship was observed with CD38 expression levels. CD38 levels among STAT6−^/− Bcl6^hi B cells was significantly higher than that of WT Bcl6^hi B cells (Fig. 4E). Thus, an absence of either IL-21 or IL-4 signaling inhibits normal GC B cell differentiation, indicating that these cytokines play non-redundant roles that are B cell intrinsic.

Aberrant intra-follicular GC formation and skewing of centroblast and centrocyte phenotypes.

The elevated CD23 levels in the IL21R−^/− B cells are reminiscent of a phenotypic feature of centrocytes, causing us to question the relative proportion of centrocytes and centroblasts in the absence of normal cytokine signals. Centroblasts within the dark zone are predominantly CXCR4^hi and CD86^lo, while centrocytes in the light zone are typically CXCR4^lo and CD86^hi. Although the phenotype can vary within a zone, these characteristics can be used as a crude measurement of their relative proportions by flow cytometry (Fig. 5A). Based on these criteria and using the gating shown in Fig. 5A, we found there were small but statistically significant differences between STAT6−/− and IL21R−^/− Bcl6hi B cells. IL-21R−/− GC B cells have a small shift toward a centrocyte-like phenotype and are somewhat more likely to be physically located within the light zone (Fig. 5B-D). By contrast, STAT6−^/− B cells are slightly more centroblast-like in their phenotype yet maintain a physical distribution comparable to WT B cells (Fig. 5B-D). At these later time points, the majority of Bcl6^hi B cells derived from the adoptive transfer are found within intra-follicular germinal centers (Fig 5DE, 96.4% +/− 1.7% of WT, 93.4% +/− 1.9% of IL21R−/− and 93.0% +/− 0.5% of Stat6−/− Bcl6^hi B cells).

Figure 5. Skewing of centroblast and centrocyte populations in the absence of IL21R and STAT6 dependent signaling.

(A,B,E-G) Flow cytometry analysis of splenic Ag-specific (GFP⁺) B cells. Transfers and immunizations were performed as described in Fig. 2. The legend shown in B applies to data plots E - G. Shown is a representative experiment of three independent experiments.(A) Gating strategies for identifying the Bcl6^hi GFP⁺ B cell subset, CXCR4^hi and CXCR4^lo populations, and centroblasts and centrocytes populations based on CXCR4 and CD86 expression. (B) Percentage of centroblast and centrocyte populations based on the gating of flow cytometry data shown in A. Each data point represents an individual mouse. (C) Immunofluorescence staining of spleen sections. Sections of tissue obtained 9 days p.i. were stained for IgD+ follicular mantle (blue), GFP, CD35 to identify follicular dendritic cells in the GC light zone (red), and Bcl6 (greyscale). Scale bars represent 100 μm. Shown are representative images from one of three independent experiments. (D) Quantification of GC B cell zonal residence and numbers. Shown are the percentages of Bcl6^hi GFP+ GC B cells that reside within the LZ 9 days p.i., counted within spleen sections stained as in panel C. The number of Bcl6^hi Ag-specific B cells per mm² of germinal center were quantified from the same sections with each data point representing an individual germinal center. Shown is a representative experiment of three independent experiments. (E) Shown is the percentage of Bcl6^hi Ag-spec. B cells (CD45R⁺ CD4^- GFP⁺) 9 days p.i. as determined by flow cytometry. Each data point represents an individual mouse. (F) CD40 MFI levels of centrocyte and centroblast populations as gated in panel A. Each data point represents an individual mouse. (G) CD86 MFI levels of CXCR4^hi and CXCR4^lo populations as gated in panel A. Each data point represents an individual mouse.

However, with either signaling deficiency, the GC B cells that form fail to properly adopt other classical characteristics. Those CXCR4^hi IL21R−^/− centroblasts that do form overexpress CD40 and CD86, typically centrocyte features (Fig. 5F,G). Conversely, CXCR4^lo STAT6−^/− centrocytes under express CD86, typically a centroblast features (Fig. 5F,G). Thus, the proper generation of these GC subsets cannot be achieved without B cell intrinsic reception of both cytokines. A schematic summarizing these results is shown in Fig. 6.

Figure 6. Schematic of skewed GC B cell transitions during self-renewal in the absence of IL-4 or IL-21 dependent signaling.

An initial increase in Bcl6 expression by B cells can occur independently of these cytokines, however proper subsequent GC formation requires responsiveness to both. In the absence of either, intra-follicular GC B cells instead aberrantly form centroblast and centrocyte subsets with uncharacteristic phenotypic features. Whereas IL-21 signaling plays a larger role in instructing the centrocyte to centroblast transition, STAT6 dependent signaling has a greater influence on centrocyte formation.

Discussion

After engagement with cognate antigen, activated B cells clonally expand through multiple rounds of cell division, undergoing a process of differentiation that ultimately results in generation of one of several B cell lineages: early memory B cells, short-lived antibody secreting cells (ASC) or germinal center (GC) B cells. This lineage fate choice is a complex process, occurring over the course of days during a time period when B cells are in close contact with activated T cells that are themselves concurrently undergoing differentiation to T helper cell subsets (5–7). The formation of both the GC B cell and Tfh subsets requires expression of the transcriptional repressor Bcl6 (8, 9), however the process leading to B cell commitment to the GC lineage is not fully understood. According to contemporary paradigms, B cell lineage fate is the direct result of instructive CD40L and cytokine exposure via a sustained immunological synapse with fully differentiated T helper cells, specifically IL-4 and IL-21 derived from Tfh in the case of GC formation (35–37, 52–54). Other innate cell types are also known to produce IL-4, including eosinophils and basophils, and can function as antigen presenting cells to differentiating Th2 cells within mucosal tissues and draining lymph nodes (55–58). Although they have not been observed within intrafollicular GCs (56), it is unclear whether perifollicular innate cell types have the ability to provide IL-4 directly and specifically to activated B cells during early differentiation after immunization.

The evidence presented here suggests that GC B cell formation is a staged process that can be developmentally arrested. In the absence of both IL-21 and IL-4 derived signals, Ag-specific B cells retain a comparable proliferation rate and form similar numbers of B cells expressing higher levels of Bcl6 during the first few days of an adaptive immune response. However, their further expansion and survival is greatly stunted, indicating that this first phase of initial divergence to the GC lineage can occur without an obligate subsequent proliferative expansion. Regardless of their capacity to respond to these cytokines, in our experimental system Bcl6^hi Ag-spec. B cells are enabled to re-enter the follicle interior after their differentiation to this nascent GC state at the follicle periphery.

The results of this study indicate that B cell signaling by these cytokines may influence GC B cells at different stages. IL-4 and IL-21 are produced with distinct kinetics by T cells with a Tfh phenotype. As shown here and similar to previous observations, IL-21 expression levels increase 2–3 days after immunization, a time period when Tfh cells are predominantly located at the periphery of follicles, while an increase in IL-4 transcription levels is less apparent in our experimental system until a point in time when Tfh cells are primarily intra-follicular (36, 37, 40, 47). Consistent with this, the transition from the peri-follicular pre-GC B cell state to the intra-follicular stage is unaffected in STAT6−/− B cells. IL-21 signaling therefore has a greater influence on differentiation at the time of pre-GC B cell re-entry to the follicle interior. However, proper GC formation during a subsequent phase within follicles is highly reliant on both cytokines. The behavior of IL-21R−/− B cells during the delayed intra-follicular expansion phase further stresses the importance of IL-4 at a later point.

As the GC matures, B cells undergo many rounds of cell division that are interspersed and propelled by contacts with Tfh cells and follicular dendritic cells within the light zone (22) A report by Bannard et. al. suggests that the transition of light zone resident centrocyte differentiation to that of dark zone centroblast results from a timed program that is instructed through Tfh cell cytokine delivery and may be resolved through cell division (59). The results presented here suggest than an efficient transition between the centroblast and centrocyte states is compromised when the germinal center matures under the influence of one cytokine without the other. Our results suggest that IL-21 and IL-4 signaling potentiates different aspects of this transition. In the absence of either, intra-follicular GC B cells instead aberrantly form centroblast and centrocyte subsets with uncharacteristic phenotypic features. IL21R deficient GC B cells have supra-elevated CD23 levels and a slightly higher percentage of CXCR4lo centrocyte-like cells within maturing GCs. However, the IL-21R−/− CXCR4hi centroblasts that do form overexpress CD40 and CD86, typically centrocyte features, suggesting that IL-21 signaling plays a large role in instructing the centocyte to centroblast transition.

By contrast, IL-4 dependent signaling has a greater influence on formation of the centrocyte stage. Although STAT6−/− GC B cells have a normal distribution across the zones of their smaller GCs, STAT6−/− CXCR4lo centrocytes underexpress CD86. These results are consistent with the idea that newly generated centroblasts formed without IL-4 may be unable to regain the complete centrocyte program thereafter. Thus, while an initial increase in Bcl6 expression by B cells can occur independently of these cytokines, proper subsequent GC formation and propagation requires the ability to respond to both to evoke a developmental program that is distinct from that obtained separately. This is consistent with a prior report of a disorganized GC structure observed in hematoxylin and eosin stained tissue of mice deficient in both IL-4 and IL-21 (30).

Experimental systems can vary tremendously in their use of adjuvants, or antigen quantity or persistence, factors that can all directly or indirectly effect the kinetics and/or size of GC B cell responses. Here using a model antigen system we correlate the kinetics of Tfh IL-4 and IL-21 expression with the characteristics of emerging GC B cells that are deficient in the receptors for these cytokines in a B cell intrinsic fashion. Our results with adoptively transferred STAT6−/− B cells are similar to a study of IL4 deficient mice infected with helminths in that GC formation is likewise reduced, but differs in another regard (27). The Turqueti-Neves et. al. study observed the formation of GCs with underdeveloped follicular dendritic cells and a preponderance of DZ phenotype GC B cells in mice that systemically lacked IL-4, consistent with our conclusions on a role for IL-4 in promoting the centroblast to centrocyte transition. The introduction of IL-4 sufficient BM in mixed BM chimeras increased the relative percentage of centrocytes compared to IL-4 knockout mice, even when these cells were CXCR5 deficient, albeit in smaller GCs, suggesting that perifollicular IL-4 delivery can contribute to centrocyte formation during helminth infection (27). Although we can not rule out that IL-4 is produced by some perifollicular Tfh in our experimental system, there is significantly less IL-4 expression during early time points and as observed in other studies (36,40). The persistence or burden of pathogen associated antigens might contribute to a larger role of IL-4 at the follicle periphery during helminth infections.

The results presented here and in another recent study (54) indicate that under physiologic conditions in vivo, Bcl6 expression is not sufficient to ensure complete GC B cell development. Although over-expression studies have suggested that Bcl6 acts as a “master regulator” of the GC response, it is known that transcriptional regulation by Bcl6 requires the assembly of co-factors to form a repressive complex, the composition of which influences target gene repression (11, 13). Although composition of the Bcl6 repressor complex cannot be addressed in this study, it is tempting to speculate that these cytokines might influence its nature.

Based on these findings we favor a phased model of GC B cell development in which the initial progression toward this lineage can occur in the absence of IL-4 and IL-21. Although the Bcl6-expressing B cells that form under these conditions transiently persist in vivo, their proliferation is limited, revealing a stage of lineage divergence at which GC B cell development can be arrested. Interruption of the progression along the GC lineage would offer a means to potentiate this arm of B cell responses without risk of a needless clonal expansion of an inappropriate specificity. Such a hiatus in GC B cell development would be advantageous during adaptive immune responses that invoke small numbers of Tfh cells, or otherwise poorly coordinate T and B cell responses temporally.