Germinal Centers Regulate Human Th2 Development¹

Naive CD4⁺ T lymphocytes recirculate between blood and T cell zones of lymphoid tissues, where they continuously survey MHC class II/peptide complexes displayed on APCs (1). This encounter gives rise to distinct migratory properties. Some T cells rapidly up-regulate tissue-selective adhesion molecules and receptors for inflammatory chemokines and consequently migrate to inflamed peripheral target sites, where they are critically involved in for example delayed-type hypersensitivity reactions (2, 3). Other T cells fail to induce expression of tissue-selective molecules but instead acquire high-level expression of chemokine receptor CXCR5, which confers responsiveness to CXCL13 that is exclusively produced within the B cell areas (4, 5). Consequently, these T cells migrate into the B cell follicles, where they provide help for germinal center (GC)³ development (3) and B cell differentiation (2). The rapid and selective migration of recently activated T cells appears to precede their acquisition of effector functions, including secretion of Th1- or Th2-associated cytokines (6). Cytokine-producing T lymphocytes can, therefore, be detected only at very low frequencies within lymphoid tissues in contrast to extralymphoid tissue to which the cytokine-producing effector cells ultimately home (7, 8, 9, 10, 11).

Affinity maturation of B cell Igs and development of B cell memory occur within GCs and require help provided by CD4⁺ T cells (12). T cells supporting B cell differentiation have been widely accepted to belong to the Th2 subset, because Th2 clones support in vitro Ab production better than Th1 cells do, and also secrete a pattern of cytokines with well documented effects in B cell growth, differentiation, and isotype switching (13, 14). More recent results do, however, suggest that follicular Th cells do not produce Th1 (IFN-γ) or Th2 (IL-4) cytokines and can therefore not be classified as either of these effector subsets (2). Thus, despite being functionally specialized in providing help for GC-associated B cell differentiation, the follicular Th cells appear to conform to a more general pattern of a nonpolarized and cytokine-deficient phenotype of CD4⁺ T cells within lymphoid tissues (15).

In contrast, B cells do themselves play an important role in the development and maintenance of CD4⁺ T cell memory (16, 17), induction of peripheral tolerance (18), and in regulation of Th cell polarization (19, 20). Although B cells are inadequate APCs for priming naive T cells (21), they can present Ags for primed or memory T cells and thereby elicit cytokine secretion (20, 22). Under conditions being optimal for induction of strong humoral immune responses, cognate interaction between B and T cells takes place at the edge of the follicles ∼2 days after primary immunization and is followed by a GC formation at day 4–5. Thereafter, both B and T cells continue to divide inside the growing GC (23, 24). Accordingly, follicular CD4⁺ T cells, initially primed on T cell zone-localized dendritic cells (DCs), are likely to receive a second boost of differentiation signals during the interaction with GC B cells. We have recently shown that GC B cells produce IL-4 in vivo, suggesting that GCs do provide an appropriate microenvironment for Th2 development (25). In this study we demonstrate that although CD4⁺ T cells from human tonsil lack immediate effector functions, they develop into cytokine-secreting cells after polyclonal activation in vitro. In contrast, the fraction of CD4⁺ T cells in peripheral blood, which lack apparent secretion of both Th1- and Th2-associated cytokines, retain their cytokine-deficient phenotype after polyclonal activation in vitro. Focusing on the Th2 development, we further show that this discrepancy between the tonsil and blood relates to the presence of a CD45RA⁺ precursor population for Th2 cells in tonsils but not in peripheral blood. These Th2 precursors, which are identified by the chemoattractant receptor-homologous molecule expressed on Th2 cells (CRTH2) (26, 27), do not produce detectable amounts of cytokines when analyzed immediately ex vivo, but develop a robust Th2-associated cytokine phenotype after in vitro activation. Furthermore, similar to the previously recognized follicular T cells, characterized by the expression of CD57 (28, 29, 30, 31), the CD4⁺ CRTH2⁺ tonsil T cells express high levels of CXCR5 and consistently are localized within or close to the GCs. However, functionally these two subsets are entirely different, because we identify the CD57⁺ GC T cells as being anergic. In contrast, the CRTH2⁺ GC T cells may represent recently activated T cells receiving instructions for their Th2 development within the microenvironment of the GCs. In support for this suggestion, we demonstrate that GC B cells can induce IL-4-dependent production of IL-4 and IL-13 from allogenic CD4⁺ T cells. Taken together, these results emphasize a reciprocal triggering of B and T cells, and define the GC as an anatomical microenvironment not only being essential for B cell differentiation, but also necessary for induction of effector functions and perhaps maintenance of homeostasis among CD4⁺ T cells.

Cell culturing was performed in complete medium consisting of RPMI 1640, supplemented with 10% FCS, 2 mM l-glutamine, nonessential amino acids, and 50 μg/ml gentamicin (all from Life Technologies, Paisley, U.K.). The following murine mAbs were purchased from BD Biosciences (San Jose, CA): Anti-CD4-FITC or -PerCP (SK3, IgG1), anti-CD45RO-PE (UCHL1, IgG2a), anti-CD57-FITC or -biotin (HNK-1, IgM), anti-CD69-PE (L78, IgG1), and anti-IL-4-PE (3010.211, IgG1). PE-conjugated mAbs against CD4, CD27 (M-T271, mouse IgG1), IL-13 (JES10-5A2, rat IgG1), and IFN-γ (4S.B3, mouse IgG1), as well as a primary mouse mAb to CD28 (CD28.2, IgG1) were obtained from BD PharMingen (San Diego, CA). Primary mouse mAbs to CD45RA (4KB5, IgG1) and IgD (IgD26, IgG1), polyclonal rabbit anti-CD3ε as well as secondary reagents, including PE- or PECy5-conjugated goat F(ab′)₂ anti-mouse IgG and PE- or PECy5-conjugated streptavidin, were purchased from DAKO (Copenhagen, Denmark). The OKT3 mAb against CD3 was purchased from Ortho Biotech (Raritan, NJ). Mouse anti-CXCR5 (51505.111, IgG2b) was obtained from R&D Systems (Minneapolis, MN). Purified or biotinylated anti-CRTH2 mAb (BM16, rat IgG2a) was kindly provided by Dr. K. Nagata (Saitama, Japan).

Human tonsils were obtained from children undergoing routine tonsillectomy at Lund University Hospital (Lund, Sweden) or Malmö Academic Hospital (Malmö, Sweden). Tonsils were minced in RPMI 1640, supplemented with 10% FCS and 50 μg/ml gentamicin, and mononuclear cells isolated by Ficoll-Paque (Pharmacia Fine Chemicals, Uppsala, Sweden) density centrifugation. PBLs were also isolated by Ficoll-Paque density centrifugation from buffy coats (Lund University Hospital Blood Bank). CD4⁺ T cells were isolated using MACS microbeads (Miltenyi Biotech, Bergisch Gladbach, Germany). Briefly, tonsillar lymphocytes or PBL were incubated with anti-CD4-conjugated microbeads in PBS, containing 1% BSA (w/v) and 2 mM EDTA (PBS/BSA/EDTA), for 15 min at 8°C. CD4⁺ cells were then purified on a LS⁺/VS⁺ column (Miltenyi Biotech), according to recommendations of the manufacturer. Positively selected cells were passed over an additional LS⁺/VS⁺ column to obtain increased purity. If downstream applications involved staining of purified CD4⁺ cells with primary mouse mAbs (e.g., anti-CXCR5 and anti-CD45RA) followed by secondary anti-mouse reagent, CD4⁺ T cells were instead purified with a CD4 Positive Isolation kit from Dynal Biotech (Oslo, Norway), according to the enclosed instructions. Briefly, CD4⁺ T cells were purified with anti-CD4-conjugated dynabeads and thereafter the anti-CD4 mAb-bead conjugates were released by use of DETACHaBEADS. In this way, unlabeled CD4⁺ T cells were obtained and could be used for subsequent FACS sorting or analysis. For both protocols, isolated CD4⁺ T cells had routinely a purity of 96–99%.

CD4⁺CD27^+/− and CD4⁺CD57^+/− subsets were isolated using flow cytometric cell sorting. Briefly, MACS-purified CD4⁺ T cells were incubated for 30 min at 4°C with anti-CD27-PE or anti-CD57-FITC mAbs diluted in PBS/BSA/EDTA. After washing, cells were sorted into CD27^+/− and CD57^+/− fractions, respectively, using a FACSVantage SE (BD Biosciences).

Fractionation of CD4⁺ T cells into CRTH2^+/− subsets was also performed by FACS sorting. MACS purified CD4⁺ T cells from tonsil were incubated for 30 min with a biotin-conjugated anti-CRTH2 mAb at 8°C. After washing, cells were further incubated for 15 min at 4°C with PECy5-labeled streptavidin and sorted into CRTH2^+/− subsets using the FACSVantage SE cell sorter. PBS/BSA/EDTA was used as dilution and washing buffer.

For purification of CXCR5^highCRTH2⁻CD57⁻ and CXCR5^dimCRTH2⁻CD57⁻ subsets, unlabeled CD4⁺ T cells purified with dynabeads/DETACHaBEADs were used. Cells were first incubated for 30 min at 4°C with a primary mouse IgG2b anti-CXCR5 mAb. After washing, cells were labeled for 15 min with PECy5-conjugated goat F(ab)₂ anti-mouse IgG, washed and thereafter incubated for an additional 15 min in 10% mouse serum to block remaining anti-mouse reactivity. Without washing, biotin-conjugated anti-CRTH2 and FITC-conjugated CD57 mAbs were added and cells were further incubated for 30 min at 8°C. After washing, this was followed by a 15 min incubation with PE-labeled streptavidin. Labeled cells were then sorted into CXCR5^high and CXCR5^dim subsets on the FACSVantage SE cell sorter and gates were set to delete FITC- (CD57) and PE- (CRTH2) positive events. Sorted populations were >95% pure in relation to CXCR5 expression (see Fig. 4 for definition of the CXCR5^high and CXCR5^dim phenotypes) and contained <0.5% of CD57⁺ and CRTH2⁺ events, respectively. PBS/BSA/EDTA was used for all incubation and washing steps.

For purification of GC B cells and CD14⁺ macrophages (Mφ), tonsil leukocytes were layered on 60% isotonic Percoll (Pharmacia Biotech) and centrifuged at 750 × g for 20 min. The buoyant fraction was recovered and used for subsequent purification steps. GC B cell were isolated by negative selection. The low-density tonsil cells were incubated for 30 min at 4°C with primary mouse IgG mAbs to CD3 (T cells), to IgD and CD39 (naive and memory B cells), and to CD4, CD11c, CD11b, CD14, CD16, CD163 and DC-specific ICAM-grabbing nonintegrin (SIGN; myeloid cells and DCs). All mAbs were mouse IgG and obtained from Dako, except from anti-DC-SIGN, which was purchased from R&D Systems. Labeled cells were thereafter depleted by two rounds of selection, using pan-mouse IgG magnetic beads (Dynal Biotech). Negatively selected cells contained >97% CD20⁺ B cells and >90% CD20⁺CD38⁺ GC B cells as assessed by FACS analysis. Macrophages were isolated by positive selection, using CD14-conjugated MACS microbeads (Miltenyi Biotech) and an LS⁺/VS⁺ column. Over 96% of the purified CD14⁺ cells expressed high levels of HLA-DR, as determined by FACS analysis. Purified macrophages were fixed in 1% paraform aldehyde (PFA) for 5 min at room temperature and thereafter washed three times in large volumes of complete medium.

FACS-sorted CD4⁺CXCR5^dim T cells (2 × 10⁵) were cultured in complete medium with an equal number of allogenic GC B cells or PFA-fixed Mφ in flat-bottom 96-well tissue culture plates (Costar, Cambridge, MA). Cultures with T cells and GC B cells were initiated in the absence or presence of 2 μg/ml of a neutralizing mAb to IL-4 (MP4-25D2, rat IgG1; BD PharMingen). After 5 days of priming, cells were recovered and thoroughly washed and thereafter recultured for 8 days in the presence of 10 ng/ml of recombinant IL-2 (R&D Systems). At this time point, cultures contained >98% CD3⁺ T cells. Established polyclonal T cell lines were analyzed for cytokine production by flow cytometry, as described below.

Cytokine production was analyzed from CD4⁺ T cells and the various CD4⁺ T cell subsets either directly after isolation or after in vitro activation and expansion into polyclonal short-term T cell lines. For the latter purpose, purified T cells were activated for 5–7 days with plastic adsorbed anti-CD3 mAb (OKT3, 5 μg/ml in PBS for 1 h at 37°C) and 0.5 μg/ml of soluble anti-CD28 mAb (BD PharMingen). In some experiments, neutralizing mAbs against IL-4 and IFN-γ (BD PharMingen) were added at a concentration of 2 μg/ml during these first 5–7 days of culture. Thereafter, T cells were further expanded with 10 ng/ml recombinant IL-2 (R&D Systems) for 5–7 days. Primary T cells and established polyclonal short-term T cell lines were analyzed for cytokine production by flow cytometry and/or ELISA, as described below.

Flow cytometry analysis was performed on a FACScan flow cytometer (BD Biosciences). Live lymphocytes were gated based on forward and side scatter properties and 10,000–50,000 events were collected and analyzed, using the CellQuest software (BD Biosciences). Isotype controls relevant for each Ab were used for background staining. For CXCR5 and CD45RA labeling, unlabeled CD4⁺ lymphocytes were stained with primary mouse mAb, according to the protocol for CXCR5 staining previously described. For staining with direct-conjugated mAbs, cells were incubated with the mAbs for 30 min on ice. PBS/BSA/EDTA was used for all labeling and washing steps.

CD57-positive and CD57-negative T cells were assayed for proliferation during anti-CD3 and anti-CD28 stimulation. A total of 20,000–40,000 CD4⁺CD57^+/− were seeded on anti-CD3 mAb precoated wells (OKT3, 5 μg/ml, 100 μl/well) in the presence of soluble anti-CD28 (0.5 μg/ml). In some cultures either IL-2 (10 ng/ml) or anti-IL-2 (4 μg/ml; R&D Systems) was added during the culture period. The cells were cultured for 5 days, and for the last 16 h of the culture period 0.5 μCi [³H]thymidine (Amersham, Arlington Heights, IL) was added to the culture. The cells were then harvested and incorporated radioactivity was counted in a Wallac 1450 MicroBeta Liquid Scintillation Counter.

For detection of intracellular cytokine production, freshly purified T cells or polyclonal short-term T cell lines were cultured in complete medium containing PMA (50 ng/ml) and ionomycin (1 μg/ml) for 5 h. During the last 3 h, brefeldin A (10 μg/ml) was added to the cultures. Cells were fixed in 2% PFA, permeabilized with PBS/BSA containing 0.5% saponin, and stained with PE-labeled mAbs to IL-4 or IL-13 and a FITC-labeled mAb to IFN-γ in the presence of saponin. Flow cytometry analysis was performed on a FACScan flow cytometer (BD Biosciences).

For detection of cytokine production by ELISA, freshly purified T cells or polyclonal short-term T cell lines were stimulated at 10⁶ cells/ml with plate-bound anti-CD3 (OKT3, 5 μg/ml, 100 μl/well) and soluble anti-CD28 (0.5 μg/ml) for 24 h. Culture supernatants were recovered and concentrations of IL-4, IL-5, IL-10, IL-13, and/or IFN-γ were analyzed in specific ELISAs with matched Ab pairs (R&D Systems), according to the manufacturer’s recommendations. Recombinant cytokines (R&D Systems) were used for generation of appropriate standard curves.

Cryostat sections (8 μm) of human tonsil were simultaneously fixed and quenched for endogenous peroxidase activity with 0.6% H₂O₂ in methanol. After 20 min treatment, sections were air-dried and then rehydrated in PBS. Tissue sections were further blocked for 30 min with 5% goat serum diluted in PBS and, after that, endogenous biotin was blocked with a biotin blocking kit (DAKO). Primary rabbit anti-CD3ε Ab, mouse anti IgD mAb and rat anti-CRTH2, or an isotype-matched control mAb, were applied in 5% goat serum in PBS, and sections were incubated at 4°C overnight. After extensive washing in PBS, primary Abs were targeted with Alexa 488-conjugated goat anti-rabbit IgG (Molecular Probes, Eugene, OR), Cy5-conjugated goat F(ab)₂ anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) and biotin-conjugated goat anti-rat IgG (Jackson ImmunoResearch Laboratories). Tissue samples were incubated with these secondary reagents diluted in 5% goat serum for 60 min. Subsequently, anti-CRTH2-dependent tissue deposition of biotin was increased by using a tyramide signal biotin amplification system (NEN, Boston, MA) and visualized by Alexa 568-conjugated streptavidin (Molecular Probes). For Imaging, a laser-scanning confocal device (model MRC-1024; Bio-Rad, Hercules, CA) equipped with a 15 mW krypton/argon laser, and attached to an Eclipse E800 microscope (Nikon, Melville, NY) was used. The isotype-matched control for the CRTH2 staining, did not give rise to any detectable signal with the utilized settings for the laser, detectors, and optical filters.

The human palatine tonsils are lymphoepithelial tissues, which are involved in the induction of both local and systemic immune responses (32). Positioned in the oropharynx, they are continuously exposed to bacteria, fungi, viruses, and other inhaled or ingested substances (33). The high antigenic burden is reflected in a comprehensive B cell activation with numerous germinal centers and only very few primary B cell follicles (data not shown). Consistent with an abundant T cell-dependent B cell response, most tonsil CD4⁺ T cells also display an activated phenotype, expressing CD69 and chemokine receptor CXCR5 (Fig. 1 and Refs. 29 , 34). A substantial fraction of the CD4⁺CD45RA⁺ tonsil T cells are indeed also positive for CD69 and/or CXCR5, demonstrating that tonsils contain a relatively large number of recently activated CD4⁺ T cells (Fig. 1). This finding is in agreement with a previous report of a relatively large number of tonsil CD4⁺ T cells displaying a transitional CD45RA⁺/CD45RO⁺ phenotype, generally not present in nonlymphoid tissues (35).

To address the influence of the tonsil microenvironment on CD4⁺ T cell differentiation, we measured cytokine production directly from purified CD4⁺ T cell or from polyclonal short-term T cells lines established in the absence of polarizing cytokines. The CD4⁺ T cells exhibited a poor immediate cytokine-producing capability (Fig. 2). Nevertheless, stimulation with anti-CD3 plus anti-CD28 followed by a 7-day expansion with IL-2, gave rise to a substantial fraction of cytokine-producing cells, displaying a Th1 (IFN-γ but not IL-4/IL-3), a Th2 (IL-4/IL-13 but not IFN-γ) or a Th0 (IL-4/IL-13 and IFN-γ) phenotype (Fig. 2). The relative distribution among these subsets varied considerably between different donors (Table I). Thus, even though the majority of CD4⁺ T cells in the tonsil display an activated phenotype (see Fig. 1), the frequency of cells immediately producing IFN-γ, IL-4, or IL-13 in response to ex vivo polyclonal stimulation is remarkably low. However, the comprehensive induction of cytokine production after in vitro expansion indicate that at least a proportion of these apparently quiescent CD4⁺ T cells represent cells progressing toward a cytokine-producing phenotype in vivo. Such T cells, being in the process of differentiation, may be contained in both the CD45RA⁺CD45RO⁺ transitional (35) and the CD45R0⁺ memory subsets (8).

To determine whether this observed onset of cytokine production during the in vitro expansion of tonsil CD4⁺ T cells actually reflect an in vivo imprinted potential of these cells, we performed experiments in which nonpolarized T cells from tonsil and peripheral blood, respectively, were compared. Expression of CD27 is irreversibly lost after terminal T cell differentiation (36), and T cells producing cytokines were indeed highly enriched within the CD27⁻ subset in both peripheral blood and tonsils (Fig. 3). After deletion of the CD27⁻ effector cells by FACS sorting (which represented typically <3% of CD4⁺ tonsil T cells), short-term polyclonal CD4⁺ T cell lines established from blood preserved the cytokine-deficient phenotype (Fig. 3). In marked contrast, lines derived from tonsillar CD4⁺CD27⁺ T cells were induced to secrete large amounts of both Th1- and Th2-associated cytokines. These experiments clearly show that the method being used for in vitro activation and expansion of the T cells does not by default induce cytokine production in CD4⁺ T lymphocytes. Therefore we conclude that a substantial number of CD4⁺ T cells in tonsil, but not in peripheral blood, indeed represent recently primed cells that are progressing toward cytokine-secreting phenotypes.

We have recently shown that IL-4 production in human mucosa-associated lymphoid tissues is highly confined to GCs where B cells constitute a predominant cellular source (25). In response to platelet activating factor, GC B cells also secrete relatively large amounts of IL-4 in vitro (37). Based on the important role of IL-4 in Th2 development (38, 39, 40), we asked whether GCs might provide a microenvironment, which favors Th2 differentiation. A subset of GC T cells, defined by expression of the NK cell marker CD57, homogeneously express CXCR5 (28). FACS analysis of CXCR5 expression revealed, however, that this marker divides the tonsil CD4⁺ T cells into three populations: one negative (CXCR5⁻), one displaying intermediate levels (CXCR5^dim), and one expressing high levels of CXCR5 (CXCR5^high). The CD57⁺ GC T cells were indeed strongly enriched in the CXCR5^high subset (Fig. 4 A). These data demonstrate that CD4⁺CD57⁺ T cells, which are proven to reside within GCs but not representing typical Th2 effector cells (28), display a CXCR5^high phenotype. However, they also reveal that additional CD4⁺ T cells, which lack CD57, also are contained in this CXCR5^high population. Consistently, immunohistochemical examination of tonsils reveals that only a proportion of CD4⁺ T cells within GC express CD57 (data not shown and Ref. 30).

Next, we took advantage of an mAb recognizing the described recently chemoattractant receptor for PGD₂, CRTH2, which is exclusively expressed on Th2 and T cytotoxic type 2 cells among peripheral blood T cells (26, 27). Among tonsil CD4⁺ T cells, CRTH2 was only expressed on cells in the CXCR5^high subset, suggesting that Th2 cells indeed are mainly localized to the GCs in the tonsil (Fig. 4,B). In contrast, peripheral blood CD4⁺ CRTH2⁺ T lymphocytes did not express detectable amounts of CXCR5, ruling out the possibility that Th2 cells intrinsically express CXCR5 and hence by default migrate into B cell areas. A GC-associated localization of the tonsil CRTH2⁺ Th cells was confirmed by immunofluorescence analysis of tissue sections. Although anti-CRTH2 reactivity was visualized in various compartments of the tonsil tissue, costaining with an anti-CD3 Ab revealed that T cells expressing CRTH2 were mostly present within GCs or in narrow interfollicular regions close to the mantle zones (Fig. 4, C and D). In contrast, anti-CRTH2 reactivity in T cell zones was mostly associated with non-T cells, including endothelial cells (Fig. 4 E). This clearly distinguished a CXCR5^high follicular Th cell subset expressing CRTH2.

Besides CXCR5 expression, tonsil CD4⁺CRTH2⁺ T cells and their circulating counterparts also differed in their expression of CD45 isoforms. In tonsil, CRTH2 defines CD4⁺ T lymphocytes, which mostly display the transitional CD45RA⁺/CD45RO⁺ phenotype but also cells that completely lack the CD45RO⁺ memory phenotype (Fig. 5,A). In blood, essentially all CD4⁺ CRTH2⁺ T lymphocytes were single-positive for the CD45RO isoform and, in addition, expressed higher levels of CRTH2. Therefore, CRTH2⁺ Th2 cells in tonsil appear to represent recently activated naive T cells. In further support for this suggestion, FACS-sorted CD4⁺ CRTH2⁺ tonsil T cells did not produce substantial amounts of cytokines. However, as polyclonal short-term T cell lines they secreted large quantities of IL-4, IL-5, and IL-13, but importantly, no IFN-γ (Fig. 5,B). The in vitro induction of cytokine production was also accompanied by increased expression of CRTH2 (Fig. 5 C). Taken together these results demonstrate that 1) the CRTH2⁺ Th2 cells in tonsil are functionally immature and consistently display a “naive to memory” transitional phenotype, 2) like the CD57⁺ GC T cells, CRTH2⁺ Th2 cells are only present among CXCR5^high T cells and hence reside within or close to GCs, and finally 3) the tonsil is virtually deficient in functionally mature Th2 effector cells.

Because expression of both CRTH2 and CD57 to a large extent is restricted to CD4⁺ T cells also displaying a CXCR5^high phenotype, we next addressed the functional relationship between T cells defined by these two phenotypic features. CRTH2 and CD57 did not jointly define one distinct population of tonsil CD4⁺ T lymphocytes, but displayed only a minor overlap in their cellular distribution (Fig. 6,A). We purified the CD4⁺CD57⁺ T cells by FACS sorting without including CRTH2 labeling in this procedure (i.e., CD57⁺ CRTH2⁺ cells were not excluded from sorted cells). In response to anti-CD3/CD28 stimulation, CD57⁺ GC T did not secrete detectable amounts of cytokines (Table II). Moreover, when trying to establish short-term T cell lines from the CD4⁺CD57⁺ T cells, we found that these cells did not proliferate in response to anti-CD3/CD28 triggering, unless also IL-2 was added to cultures (Fig. 6, B and C). The IL-2-dependent proliferation still required the mAbs against CD3 and CD28, as IL-2 could not by it self mediate cell-division. (Fig. 6,D). Furthermore, anti-CD3/CD28-induced proliferation of the CD4⁺CD57⁻ tonsil T cells was clearly dependent on endogenous IL-2 production in these cultures (Fig. 6,E). Taken together, these results suggest that the CD57⁺ T lymphocytes fail to express the IL-2 gene when activated by TCR engagement. Because membrane-proximal uncoupling of signaling from the TCR/CD3 complex to the IL-2 locus represents a key feature of anergized T cells (41), these findings indicate that the CD57⁺ GC T cells are anergic. To address this issue, we analyzed IL-2 production from CD57⁺ and CD57⁻ T cells, respectively, either after anti-CD3/CD28 stimulation or after activation with PMA and ionomycin. These pharmacological agents mimic stimulation through the TCR and CD28, but bypass the membrane-proximal signaling components and instead directly target the membrane-distal part of the signaling pathways (42). As expected, anti-CD3/CD28 stimulation induced IL-2 production only in the CD57⁻ subset. In marked contrast, PMA and ionomycin promoted strong IL-2 expression in both subsets (Fig. 6 F). Hence CD57⁺ GC T cells display all consistent features of anergic T cells, when stimulated with anti-CD3/CD28, including deficiency in production of effector cytokines, inability to secrete IL-2 and therefore also to enter the cell cycle, and restored responsiveness if IL-2 is provided (42). This unresponsiveness, thus, also includes the fraction of the CD4⁺CD57⁺ cells that coexpress CRTH2.

The experiments described so far do not reveal whether acquisition of CRTH2 and high levels of CXCR5 expression are integrated features of a broader functional reprogramming of T cells during their initial interaction with DCs, or whether elevated CXCR5 expression and hence follicular homing precedes the developmental instructions leading to a Th2 phenotype. Trying to answer this question, we first compared the CXCR5^dim Th cells with the remaining fraction of CXCR5^high Th cells, lacking expression of CRTH2 and CD57, in terms of in vitro acquired cytokine-producing phenotypes. We performed cell sorting of CD4⁺ T cells in which CRTH2⁺ and CD57⁺ cells were deleted and remaining cells sorted into one CXCR5^high and one CXCR5^dim fraction (according to Fig. 4). Again, both subsets secreted only low amounts of all measured cytokines when directly stimulated with anti-CD3/CD28. After prolonged stimulation under nonpolarizing conditions and further expansion with IL-2, extensive cytokine production was detected for both subsets when re-stimulated with anti-CD3/CD28 (Fig. 7). CXCR5^high-derived lines produced IL-4, IL-10, IL-5 but lower levels of IFN-γ, whereas the CXCR5^dim-derived lines secreted a reverse pattern of cytokines. Accordingly, although not giving rise to completely IFN-γ deficient cell lines, the CRTH2⁻ CXCR5^high T cells were clearly biased toward a Th2-phenotype, as compared with the CXCR5^dim T cells. Hence, it appears as CD4⁺ T cells displaying a phenotype being permissive for follicular homing/localization generally tend to develop toward Th2 effector cells. The CD4⁺ CXCR5^dim T cells, which reside mostly in nonfollicular regions of the tonsil, instead seem to represent Th1 precursor cells.

Next we took advantage of the poor capacity of the CD4⁺CXCR5^dim T cells to spontaneously develop into Th2-like cells and investigated how coculture with allogenic GC B cells influenced their acquired cytokine production. In this way, we artificially relocated T cell zone-localized cells into a GC-like microenvironment. First, in agreement with previous results (43), GC B cells were poor in eliciting proliferation of naive T cells from peripheral blood. However, MLR performed with GC B cells and CD4⁺CXCR5^dim T cells isolated from tonsils obtained from two different donors displayed substantial proliferation where the T cells gradually outnumbered the B cells (data not shown). To provide a relevant control for possible polarizing effects imposed by the allogenic B cells, we used CD14⁺ macrophages. Macrophages and B cells were isolated from the same tonsil specimen but the former were briefly fixed in 1% PFA to prevent secretion of Th1 polarizing cytokines, such as IL-12. This treatment did not prevent T cell proliferation. After an initial 5 days priming culture, T cells were further expanded for 8 days with IL-2 and after that analyzed with flow cytometry for IL-4, IL-13, and IFN-γ production. In line with their cytokine profile when propagated with anti-CD3 plus anti-CD28 (Fig. 7), the CXCR5^dim T cells stimulated with PFA-fixed macrophages produced only IFN-γ (Fig. 8). GC B cells did not decrease this IFN-γ production but induced the T cells to also produce IL-4 and IL-13. Finally, this B cell-dependent Th2 polarization was clearly dependent on endogenously produced IL-4, because addition of a neutralizing anti-IL-4 mAb severely reduced the B cell-induced production of IL-4 and IL-13 form T cells. These data support the interpretation that the GCs are not only important for the evolution of the Ab repertoires, but also for the development of Th cell phenotypes.

In the current study we demonstrate that the human tonsil contains a subset of CD4⁺ T cells expressing CRTH2 and that these CRTH2 Th cells represent recently activated naive T cells progressing toward the Th2 effector phenotype. This conclusion derives from two separate observations. First, in marked contrast to the CD4⁺ CRTH2⁺ T cells in peripheral blood (27), the majority of tonsil equivalents could not produce significant amounts of cytokines when immediately analyzed ex vivo. Instead they secrete large quantities of IL-4, IL-5, and IL-13, but no IFN-γ, when allowed to develop into effector cells in vitro (without addition of polarizing cytokines). Thus, as polyclonal short-term T cell lines, the CD4⁺ CRTH2⁺ T cells in tonsil display a cytokine profile similar to that of freshly isolated blood CD4⁺ CRTH2⁺ T cells (27). Importantly, although our experimental procedure involved propagation of polyclonal short-term T cell lines, this prolonged activation and expansion of purified T cell subsets cannot account for the induction of a skewed and Th2-polarized phenotype, because sorted tonsil CRTH2⁻ T cells gave rise to lines producing large quantities of IFN-γ but only low levels of IL-4, IL-5, and IL-13. We also demonstrate that nonpolarized T cells obtained from peripheral blood maintain their deficiency in cytokine production, when subjected to the same kind of stimuli. Consequently, CRTH2 expression defines Th2-committed cells also in the lymphoid organ but is confined to functionally immature cells. Secondly, the T cell-associated CRTH2 expression in tonsil is largely restricted to cells still displaying a transitional CD45RA⁺RO⁺ phenotype. The naive to memory conversion of CD4⁺ T cells has previously been thoroughly studied by Picker et al. (35) and was found to take place in up to 10% of T cells within lymphoid tissues, but essentially not at all in peripheral blood. In general, we found that several of the CD4⁺CD45RA⁺ tonsil T cells also express the activation marker CD69 and different levels of CXCR5 (see Fig. 1), suggesting that they are progressing toward memory/effector cells within separate compartments of the tonsil.

Based on previous observations of a GC-associated B cell production of IL-4 (25, 37), we next address whether these Th2 precursors preferentially are evolving within the microenvironment of the GCs. In particular, because B cells can modulate CD4⁺ T cell functions (20) and also appear to be important for development of CD4⁺ T cell memory (17), we considered the possibility that GC B cells may be as important for the functional development of CD4⁺ T cells as the T cells are for the GC-associated B cell differentiation. Indeed, several previous reports have demonstrated an important role for B cells in the Th2 development (19, 44, 45). To answer this question we have taken advantage of a differential expression of the chemokine receptor CXCR5 among tonsil T cells, where the migratory properties conferred by this receptor have during the past years been extensively studied. In mice with targeted disruption of the CXCR5 gene (46), inguinal lymph nodes are absent and Peyer’s patches are either aborted or display severe morphological alterations with no defined B cell follicles. Also the spleen lack follicles and exhibit impaired GC formation after immunization of these animals. Consistently, transferred lymphocytes from CXCR5-deficient mice fail to enter B cell follicles in these organs of wild-type animals, demonstrating an essential role for CXCR5 in follicular homing. Importantly, CXCR5 expression is not likely to target lymphocytes to other anatomical compartments because production of the ligand CXCL13 is highly restricted to this particular area of lymphoid organs and is not detectable in nonlymphoid tissues of mice or humans (4, 5). Accordingly, expression of CXCL13 in human tonsil is entirely restricted to B cell follicles and GCs (34). Naive T cells, which have not encountered Ag, do not express CXCR5 and are excluded from B cell areas (47). Human naive T cells are induced to express CXCR5 and acquire responsiveness to CXCL13 within 2 days after in vitro activation with LPS-matured DCs (6). This does not, however, necessarily translate into in vivo migration toward and into follicles, because the vast majority of memory CD4⁺ T cells in the human tonsil express CXCR5, whereas only a minor fraction actually reside in GCs. Consistently, Ansel et al. (47) demonstrated that immunization of mice without adjuvant, a protocol that does not support follicular T cell-homing, still gave rise to CXCR5 expression among specific T cells, albeit at relatively low levels. In contrast, when adjuvant was used, specific T cells acquired high-level expression of CXCR5 and migrated into follicles. Therefore, the relative levels of CXCR5 expression appear to be critical in the decision of CD4⁺ T cells whether to enter B cell areas or not. In the present study, we demonstrate that CXCR5 staining of CD4⁺ T cells in human tonsil divides these into three subsets, exhibiting negative, intermediate, and high levels of CXCR5 expression. Consistent to the need for a solid CXCR5 expression for T cells to enter follicles, the previously described CD57⁺ GC T cells homogeneously express high levels of CXCR5. Therefore, we have used this CXCR5^high phenotype as a criteria to define CD4⁺ T cells that are likely to reside within GC, and by analogy, the CXCR5^dim expression to identify activated T cell mostly present in interfollicular areas. In this aspect, our work differs from two recent studies in which follicular homing was assigned as a property of the entire population of CD4⁺ CXCR5⁺ T cells (34, 48). It also differs from the work published by Kim et al. (28), because we do not only consider the CD57⁺ T cells, but also include the CD57⁻CXCR5^high subset, when defining GC-localized CD4⁺ T cells. Indeed, the CD57⁺ T cells only compose a fraction of CD4⁺ T cells residing within the GCs (30).

Using the differential CXCR5 expression among the tonsil T cells, we demonstrated that essentially all the CRTH2⁺ Th2 precursor cells in tonsil are CXCR5^high in phenotype and some of them also coexpress the GC-restricted marker CD57. Consistent to this phenotype, they preferentially are localized within or close to GCs. Furthermore, whereas the CXCR5^high subset contain all CRTH2⁺ Th2 precursors, and even if these are deleted, still display a Th2-biased phenotype, the CXCR5^dim T cells clearly are enriched for IFN-γ-producing cells. DCs are potent producers of IL-12, a cytokine, which trigger Th1 differentiation and IFN-γ production (49). The ability of DCs to efficiently present Ags for T cells in the T cell zones depends on an appropriate activation of these cells, which can be mediated by proinflammatory agents or signaling via Toll-like receptors (TLRs) recognizing evolutionary conserved motifs from various microorganisms (50). Myd88-deficient mice, in which signaling from the TLRs is blocked, cannot produce IL-12 or mount Th1-responses when challenged with pathogenic microbes (51, 52, 53). However, these mice still display a perfect Th2-oriented immune response, suggesting that other mechanisms than the TLRs are involved in the regulation of Th2 development (51). In this context, we show that the CD4⁺ CXCR5^dim T cells, which in vitro not spontaneously develop substantial production of Th2-associated cytokines, are induced to produce IL-4 and IL-13 if cultured with allogenic GC B cells. These results provide direct evidence that GC B cells can direct Th2 development and suggest that T cells acquire instructions for Th2 differentiation within the GC. We also demonstrate that these instructions are likely to include B cell-produced IL-4.

A large fraction of GC T cell express the CD57 marker, which has allowed their isolation and in vitro characterization. Cytokine production from the CD57⁺ GC T cells has been addressed by several groups, but published results are not entirely consistent (28, 29, 54, 55). We demonstrate that previous discrepancies probably reflect different in vitro methods used to activate these cells, because they in fact are anergic and do not respond to anti-CD3 plus anti-CD28 stimulation. With pharmacological agents such as PMA and ionomycin, membrane-proximal signaling from the TCR/CD3 complex is bypassed and cytokine-producing responses observed. Interestingly, in response to PMA and ionomycin, the CD57⁺ T cells preferentially produce IL-10 (28). Rather than representing a Th2-selective cytokine, IL-10 is produced by different subsets of regulatory T cells, including the peripherally derived Tr1 cells (56). Indeed, an important role for regulatory T cells within GCs was described recently in the mouse where these T cells were shown to be selectively recruited/retained within the follicles by their expression of CCR5 (57). Because IL-10-producing T cells with regulatory function also can be anergic (58), the human CD57⁺ GC T cells may be analogous to these murine CCR5⁺ GC T cells. In any way, the high content of anergic T cells within GCs suggests that these anatomical compartments also play a crucial role in controlling the duration and/or the magnitude of CD4⁺ T cells responses. Activation of T cells requires both antigenic and costimulatory signals and in the absence of one of these they instead become anergized (59). A comprehensive expression of CD57 on purified CD4⁺ T cells from blood can be obtained if these are stimulated with anti-CD28 mAb only (60). Thus, an alternative explanation to a selective recruitment/retention, as described for the murine CCR5⁺ GC T cells, is that the CD57⁺ phenotype in GCs may arise as a consequence of a limited access to Ag. This may be the situation during late phases of immune responses when Ags gradually are cleared from the host.

In summary, our results suggest an important role for GCs in development of Th2 effector cells. The CRTH2⁺ Th2 cells may arise from nonpolarized CD4⁺ T cells, which after initial priming on T cell zone-localized DCs fail to acquire a capacity for peripheral tissue homing but instead migrate into follicles in a CXCR5-dependent manner to provide help for B cell differentiation (2). Once located in the follicles/GCs, they are triggered by B cells and receive signals that promote their Th2 differentiation. However, rather than secreting large quantities of cytokines in the GCs, they acquire expression of chemotactic receptors conferring responsiveness to chemokines produced in peripheral tissues and start to migrate toward these locations. One such chemotactic receptor is CRTH2 (61, 62) that indeed has been used in this study to enable the identification of these Th2 precursor cells. We also found that ∼50% of the CD4⁺ CRTH2⁺ T cells had up-regulated CCR3 (data not shown), a receptor being involved in the early recruitment of allergen-specific Th2 cells to inflamed airways (63). The sequential priming, first on DCs and than on B cells, is in accordance to a later appearance of Th2 cells as compared with Th1 cells, during primary immune responses (64). In addition, Th2 cells have also been implicated in late-phase suppression of Th1-orientated immune responses (65), and their delayed appearance during these circumstances also fit well with a requirement for a detour via GCs to develop the Th2 phenotype.