Germinal center (GC) responses to T-dependent Ags require effective collaboration between Th cells, activated B cells, and follicular dendritic cells within a highly organized microenvironment. Studies using gene-targeted mice have highlighted nonredundant molecules that are key for initiating and maintaining the GC niche, including the molecules of the ICOS, CD40, and lymphotoxin (LT) pathways. Signaling through ICOS has multiple consequences, including cytokine production, expression of CD40L on Th cells, and differentiation into CXCR5+ follicular Th cells, all of which are important in the GC reaction. We have therefore taken advantage of ICOS−/− mice to dissect which downstream elements are required to initiate the formation of GC. In the context of a T-dependent immune response, we found that GC B cells from ICOS−/− mice express lower levels of LTαβ compared with wild-type GC B cells in vivo, and stimulation of ICOS on T cells induces LTαβ on B cells in vitro. Administration of agonistic anti-LTβ receptor Ab was unable to restore the GC response in ICOS−/− mice, suggesting that additional input from another pathway is required for optimal GC generation. In contrast, treatment with agonistic anti-CD40 Ab in vivo recovered GC networks and restored LTαβ expression on GC B cells in ICOS−/− mice, and this effect was dependent on LTβ receptor signaling. Collectively, these data demonstrate that ICOS activation is a prerequisite for the up-regulation of LTαβ on GC B cells in vivo and provide a model for cooperation between ICOS, CD40, and LT pathways in the context of the GC response.
The germinal center (GC)3 is a dynamic microenvironment where Abs specific for pathogens are generated to protect the host. Within this complex microenvironment, Ag-specific B cells, Th cells, and follicular dendritic cells (FDCs) work together to induce, maintain, and regulate GC responses. Following immunization, Ag-specific B cells migrate to the edges of the T cell zone and form dynamic conjugates with primed CD4+ Th cells (1). Here, Ag-specific B cells receive CD40 ligand (CD40L) costimulation and cytokines from cognate Th cells, and this help is necessary for B cells to expand and colonize emerging GCs (2, 3). Once in this microenvironment, GC B cells undergo intense clonal proliferation, somatic hypermutation, and Ab class switching. As the GC response progresses, Ag-specific GC B cells can interact with FDC networks within the GC, and it is thought that FDCs will not only display Ag, but may themselves be a source of B cell survival factors (4). High-affinity B cell clones are eventually selected and differentiate into memory cells or plasma cells.
ICOS, an inducible costimulatory molecule, has been shown to be important for initiating a GC response. ICOS is a member of the CD28 family of costimulatory molecules that is expressed on T cells following TCR engagement and CD28 signaling (5, 6). Its ligand, B7RP-1 (also known as B7h, B7-H2, GL50, and LICOS), is constitutively expressed on resting B cells, macrophages, and dendritic cells (7, 8, 9, 10, 11). ICOS−/− and B7RP-1−/− mice develop poor IgG1 and IgE Ag-specific titers in response to immunization with T-dependent Ags (12, 13, 14, 15). In addition to these defects in T-dependent responses, the GC microenvironment itself is not sustained in immunized ICOS−/− mice, with GC niches being small and abortive (16). Although it is clear that ICOS stimulation has multiple downstream consequences such as cytokine production (12, 13, 14), the accumulation of follicular Th cells within the GC (17), and the expression of CD40L on Th cells (12, 13), all of which are required for optimal B cell activation and Ab class switching, the reasons for a defect in the architectural elements of the GC itself remain obscure. Because ICOS signaling is upstream of multiple pathways, we have therefore taken advantage of ICOS−/− mice to dissect which downstream elements are sequentially required to initiate the formation of GCs.
To do this, we first explored whether ICOS stimulation is required for optimal expression of members of the lymphotoxin (LT) pathway in the GC because mice deficient in LTα, LTβ, or LTβR all lack FDC networks and form GCs inefficiently, demonstrating an important role for the LT pathway in establishing and maintaining the GC microenvironment (18). Additionally, treatment with a competitive inhibitor of the LT pathway (LTβR-Ig) in adult mice results in rapid de-differentiation of FDCs and the collapse of GC structures (19, 20). LTαβ is expressed on activated lymphocytes and a subset of resting B cells, whereas LTβR is expressed on stromal cells, FDCs, dendritic cells, macrophages, and high endothelial venules (21), suggesting that the pathway plays an important role in communication between activated lymphocytes and accessory cells. Studies using adoptively transferred bone marrow to immunodeficient hosts and cell-specific deletion of LTβ have shown that low levels of LTαβ on naive follicular B cells is important for maintaining the integrity of the splenic microarchitecture and the development of FDCs in the primary follicle (22, 23, 24). Interestingly, GC B cells express particularly high levels of LTαβ, underscoring the function of this TNF family member during GC responses (25), yet it is unclear how LTαβ is regulated on GC B cells. Indeed, although CXCR5 signaling is required for LTαβ expression on follicular B cells, this is not the case for GC B cells (25).
We herein show that following immunization with a T-dependent Ag, GC B cells from ICOS−/− mice express lower levels of LTαβ compared with wild-type (WT) GC B cells, whereas nonactivated ICOS−/− follicular B cells expressed normal levels of LTαβ, demonstrating a novel mechanism of local regulation of LTαβ expression within the GC microenvironment. Injection of agonistic Ab to both the LTβR and CD40 pathways allowed us to identify which signals from these TNF family members precede GC formation. Indeed, we found that signals delivered by both CD40 and LTβR are required for restoration of GC in ICOS−/− mice. These results reveal a sequence of events for the local regulation of LTαβ expression within the GC niche and provide an explanation for the GC defect observed in ICOS−/− mice.
Materials and Methods
WT C57BL/6 mice were either bred in-house or obtained from The Jackson Laboratory. LTβ−/− mice were generated as previously described using C57BL/6 embryonic stem cells (26) and were provided by B&K Universal. ICOS−/− mice were a kind gift from Dr. Tak Mak and were backcrossed to the C57BL/6 background (n = 9) (13). CD40L−/− mice were obtained from The Jackson Laboratory. All mice were maintained in our animal facility in pathogen-free conditions and, in all cases, age-matched mice were used for comparative studies. All experiments were conducted according to University of Toronto approved animal use protocols.
In vivo immunization and treatments
Six- to 8-wk-old mice were immunized i.p. with of 100 μg of NP16-CGG ((4-hydroxy-3-nitrophenyl)acetyl chicken γ-globulin; Biosearch Technologies) precipitated in a 1:1 volume of alum adjuvant (Pierce). In some experiments, at the time of immunization, mice were treated at days 0, 3, and 7 with 100 μg of either anti-LTβR (a gift from Dr. C. Ware, La Jolla Institute of Allergy and Immunology, La Jolla, CA), 50 μg of anti-CD40 (FGK45.1; a gift from Dr. R. Mittler, Emory University, Atlanta, GA), or 100 μg of LTβR-Ig (a gift from Jeff Browning, Biogen-Idec) at days −1 and 7. Rat anti-mouse or hamster IgG Ab (Jackson Immunoresearch) was used as a matched control. Blood was collected from the lateral vein of the leg before immunization and at various time points postimmunization (days 3, 6, 9, and 11). GC responses in treated/control mice were assayed at the peak of the response, that is, 12 days postimmunization.
Immunohistochemistry was performed as described (27). Briefly, fresh spleens were collected and embedded in OCT compound (Tissue-Tek, Sakura Finetek) and flash-frozen in an ice-cold bath of 2-methylbutane. Cryostat sections (5 μm) were obtained using a Leica 1035S cryostat, fixed in acetone for 7 min, stained with biotinylated peanut lectin (agglutinin) (PNA, Vector Laboratories), CD35, or GL7 (the latter two from BD Pharmingen), and counterstained with FITC-labeled B220. Sections were also stained with mAb FDC-M2 (ImmunoKontact) or FDC-M1 (BD Pharmingen) to label primary FDC networks or secondary GC-FDC networks, respectively. Secondary biotin-labeled anti-rat IgG (Jackson Immunoresearch) was used to detect FDC-M1 or FDC-M2. Sections were then stained with secondary anti-rat IgG biotin and then counterstained with FITC-labeled IgD (SouthernBiotech). All sections were stained with Abs conjugated to alkaline phosphatase (AP) (Roche) or HRP (Proenzyme). VECTOR NovaRED (HRP) and AP substrate kits were used for HRP and AP colormetric development (Vector Laboratories). Images of stained sections were obtained using a Leica DM-R digital fluorescence microscope and OpenLAB software.
In vitro restimulation
Single spleen-cell suspensions from LTβ−/−, wild-type, and ICOS−/− mice were prepared. Cells were plated in wells containing RPMI-1640 medium (Sigma-Aldrich), 10% FBS, 1% 2-ME, 1% l-glutamine, and 1% penicillin-streptavidin. To restimulate cells, 50 μg/ml NP-keyhole limpet hemocyanin (KLH) was added to plated cells. As a positive control, 0.5 ng/ml PMA and 0.5 ng/ml ionomycin were used to stimulate cells. Cell cultures were incubated overnight at 37°C, 5% CO2. For in vitro ICOS stimulation, 24-flat well plates were coated with 5 μg/ml anti-CD3 (145–2C11) and 5 μg/ml anti-ICOS (clone C398.4A) (28) or 5 μg/ml control whole rat IgG Ab overnight at 4°C. CD4+ and B220+ cells from single spleen-cell suspensions of LTβ−/−, WT, ICOS−/−, and CD40L−/− mice were purified and isolated using CD4+ T cell enrichment and B cell enrichment kits for RoboSep automated cell selection (StemCell Technologies). Flow cytometry analysis was performed to check for purity. A total of 5.0 × 105 purified CD4+ cells and 5.0 × 105 purified B220+ cells were cocultured together in media as described. Cells were analyzed by flow cytometry after 24, 48, and 72 h of incubation at 37°C. To sustain T cell survival and activation, 10 μg/ml IL-2 was added before incubation. Purified CD4+ T cells were also stimulated with 5 μg/ml anti-CD3 and 5 μg/ml anti-CD28 (clone 37.51, a gift from Dr. Watts), and, in other cases, a blocking anti-CD40L Ab (clone MR1, a gift from Dr. Watts) was added and B cells were analyzed at 48 h of incubation.
Abs and staining
Abs to B220, CD4, CD25, and CD95 (Fas), as well as all streptavidin fluorescent conjugates, were purchased from eBioscience. Ab to GL7 was purchased from BD Biosciences. PNA was purchased from Biomeda, and anti-IgD was obtained from Southern Biotech. LTβR-Ig fusion protein was used to detect surface LTαβ and LIGHT (LT-like inducible protein that competes with glycoprotein D for binding herpesvirus mediator on T cells) (18), while anti-LTβ hamster mAb BBF6 was used to detect LTαβ specifically (both were kind gifts from Dr. J. Browning). Ab to CD69 was a kind gift from Dr. T. Watts. Anti-ICOS agonistic Ab was provided by Dr. U. Dianzani (University of Eastern Piedmont, Novara, Italy), and anti-LTβR agonistic Ab was a kind gift from Dr. Carl Ware.
Analysis of ligand expression on in vitro restimulated cells was performed using anti-B220, anti-CD4, anti-CD69, anti-CD25, LTβR-Ig, and anti-LTβ BBF6. LTβR-Ig staining was detected using PE-conjugated anti-human Ig (Jackson Immunochemicals), and anti-LTβ BBF6 staining was detected using FITC-conjugated anti-hamster Ig (Jackson Immunochemicals). For ex vivo ligand expression, fresh splenocytes from immunized WT, LTβ−/−, and ICOS−/− mice were stained with PNA and Abs against CD4, CXCR5, B220, IgD, Fas, GL7, and LTβ (BBF6) or LTβR-Ig. Cell staining was assessed by flow cytometry on FACSCalibur (BD Biosciences) and analyzed using FlowJo software.
Immunoplates (96-well round-bottom MaxiSorp, Nunc) were coated overnight with NP30-BSA to quantify the total anti-NP IgG1 response and with NP3-BSA to detect high-affinity anti-NP IgG1 response. BSA/TBS blocking solution (2%) was used to block nonspecific binding before adding serum in wells. Supernatants of P1–5 hybridoma were used as a standard (29). Anti-NP IgG1 in serum was detected using goat anti-mouse IgG1 conjugated to HRP (Southern Biotech) and 2,2′-amino-bis-3-ethylbenzthiazoline-6-sulfonic acid substrate (Sigma-Aldrich) with H2O2. OD was read at 405 nm using Multiskan Ascent ELISA reader and software (Thermo Scientific). Relative affinity of anti-NP-specific responses was determined by comparing relative units to the standard curve and calculating the relative unit ratio of NP3-BSA/NP30-BSA.
Defective germinal center formation in ICOS−/− and LTβ−/− mice in response to NP-CGG
ICOS−/− mice have morphological defects in GC formation that are similar to LTβ−/− mice (12, 13, 16). We first wanted to compare the surface phenotype of GC B cells and the location and size of GCs in both ICOS−/− and LTβ−/− mice with each strain being immunized with the same Ag and evaluated at the same time point. We therefore immunized ICOS−/−, LTβ−/−, and WT mice with 100 μg of NP16-CGG, a well-characterized thymus-dependent Ag (30). On day 12, at the peak of the GC response, we prepared splenocytes for flow cytometric (FACS) analysis. B220+ B cells were evaluated for their expression of GL7 and Fas, and double-positive cells were classified as GC B cells. As expected, a substantial increase in the percentage and numbers of GL7+Fas+ GC B cells was observed in immunized WT mice compared with mice that were not immunized (Fig. 1, A and Bi). In contrast, LTβ−/− and ICOS−/− mice had significantly reduced percentage and numbers of GL7+Fas+ GC B cells (p ≤ 0.02), although these cells were not completely absent, and there remained a small population within the GL7+Fas+ gate. Similar findings were observed when GC B cells were defined as PNA+IgD−, and GC B cells were found to be significantly reduced, although not completely absent, in both LTβ−/− and ICOS−/− mice (p ≤ 0.02) (Fig. 1, A and Bii). Interestingly, however, a large proportion of B cells in both LTβ−/− and ICOS−/− mice expressed GL7 but failed to up-regulate Fas (p ≤ 0.03) (Fig. 1, A and C), suggesting that GC B cell differentiation may be incomplete in the knock-out strains.
Subsequent histological examination of spleen sections stained with either PNA or GL7 (red) to detect GC structures revealed defective GC formation in ICOS−/− and LTβ−/− mice with only few, small organized PNA+ and GL7+ cell clusters (red) observed within the splenic B cell follicles (blue) (Fig. 1 D). These results confirm previous reports (16, 31) demonstrating that the GC response is impaired in ICOS−/− and LTβ−/− mice. Because GL7 is a marker of B cell activation (32), we postulate that the appearance of GL7+Fas− B cells may be the result of an accumulation of B cells initially activated via their BCR but failing to undergo subsequent negative selection mediated by an intact Ag-bearing FDC network within the GC niche.
Expression of LTαβ on ICOS−/− T and B cells in vitro
Given the similar GC phenotypes observed in both ICOS−/− and LTβ−/− mice, we reasoned that the expression of LTαβ, which is critical for the formation and maintenance of the GC niche, may be influenced by ICOS signaling. To investigate this possibility, we first examined whether expression of LTαβ in vitro is dependent on ICOS signaling. To test this, we stimulated fresh splenocytes from naive ICOS−/− mice with PMA/ionomycin or anti-CD3 in culture overnight, and analyzed the surface expression of LTαβ by FACS on activated CD4+ T cells and B220+ B cells. These conditions have been shown to promote LTαβ expression on lymphocytes (33). Early activation markers CD25 and CD69 were used to identify activated cells (gating not shown) and, as expected, activated WT CD4+ T and B220+ B cells expressed elevated levels of LTαβ using either the soluble LTβR (Fig. 2,A) or a mAb (Fig. 2,B) to detect LTαβ. Subsequent analysis on activated ICOS−/− T and B cells revealed no significant difference in LTαβ expression compared with activated WT T and B cells. The same results were seen with anti-CD3 stimulation on ICOS−/− T cells (Fig. 2 C).
We also examined LTαβ expression in response to Ag. To do this, ICOS−/−, LTβ−/−, and WT mice were immunized with NP-KLH. At 12 days postimmunization, spleens were harvested and splenocytes were isolated and restimulated in culture with NP-KLH. After 24 h, FACS was performed to evaluate LTαβ expression on activated KLH-specific CD4+ T cells and NP or KLH-specific B220+ B cells. From this in vitro recall experiment, small proportions of total CD4+ T cells and B220+ B cells were activated by NP-KLH restimulation in all groups of immunized (but not naive) mice based on the up-regulation of CD69 and CD25 (data not shown). Analysis of these activated CD4+ T cells and B220+ B cells revealed up-regulation of LTαβ on both WT and ICOS−/− B and T cells; however, we did notice a trend toward decreased LTαβ expression in the ICOS−/− B cells (Fig. 2 D, statistically nonsignificant). Thus, ICOS−/− lymphocytes have no intrinsic defect in their ability to express LTαβ in response to PMA/ionomycin, anti-CD3, or Ag stimulation in vitro.
Expression of LTαβ on ICOS−/− lymphocytes in vivo
Although our in vitro experiments did not reveal a significant difference in LTαβ expression, these experiments did not take into account that LTαβ may be regulated differently on follicular vs GC B cells in an in vivo context. The mild reduction of LTαβ levels on ICOS−/− Ag-stimulated B cells prompted us to examine a potential interaction between ICOS and the LT pathway during the GC response by evaluating LTαβ expression on B cell subsets from ICOS−/− mice directly ex vivo. Previous studies have shown that follicular B cells express low levels of LTαβ. This basal level of LTαβ is essential for maintaining aspects of the splenic architecture, including the “primary” FDC network as well as the correct positioning of B cells within the follicle (25). We investigated whether there was abnormal expression of LTαβ on ICOS−/− follicular B cells (B220+GL7−Fas−) and found no significant difference in expression levels of LTαβ on WT vs ICOS−/− follicular B cells compared with staining on similarly gated LTβ−/− cells or isotype control stained cells (Fig. 3,A). The same was true whether we gated on B220+IgD+PNA− follicular B cells (data not shown). Consistent with normal levels of LTαβ on follicular B cells, the primary FDC network was found to be intact in ICOS−/− mice (Fig. 3,B). Specifically, primary FDC networks in naive mice were identified with an Ab to CD35 (red), which is expressed at high levels on FDC, and spleen sections were counterstained with Ab to B220 (blue) to locate B cell follicles. We also examined the primary network using the FDC-M2 Ab (red), which recognizes complement C4 deposited on FDC (34), and FDC-M2 staining was likewise found to be normal in ICOS−/− mice (Fig. 3,B). As expected, LTβ−/− spleens lacked FDC networks, which is evident by the absence of CD35 and FDC-M2 staining within B cell follicles (Fig. 3 B). Taken together, expression of LTαβ on ICOS−/− follicular B cells is normal, and the primary FDC network is intact.
We next investigated the level of LTαβ expression on GC B cells from ICOS−/− mice. Following immunization with NP-CGG, we examined LTαβ expression on PNA+IgD−B220+ or GL7+Fas+B220+ B cells, both of which we defined as GC B cells. Previous studies have shown that LTαβ is expressed at higher levels on GC B cells compared with follicular B cells, suggesting the possibility of additional regulation of LTαβ levels within the GC microenvironment itself (25). We detected LTαβ on WT PNA+IgD− GC B cells when compared with similarly gated LTβ−/− cells or isotype control stained cells (Fig. 4,A). Additionally, we observed that the level of LTαβ is consistently greater on GC B cells compared with follicular B cells (PNA−IgD+) when we compared the staining on both populations from the same animal (Fig. 4, A overlay and Ci). As in Fig. 3, ICOS−/− follicular B cells (PNA−IgD+) also expressed detectable levels of LTαβ compared with similarly gated LTβ−/− controls or isotype stained controls. However, when the small population of ICOS−/−PNA+IgD− GC B cells was assessed, these gated cells exhibited a statistically significant reduction in LTαβ expression compared with WT controls (p ≤ 0.03) (Fig. 4,A), and, indeed, the level of LTαβ on ICOS−/−PNA+IgD− GC B cells was similar to what was found on ICOS−/− follicular B cells (Fig. 4, A overlay and Ci). Using a different staining mixture that included the LTβR-Ig reagent, we examined LTαβ expression with respect to GL7 staining of B cells from each strain of mice and found that while WT vs ICOS−/− follicular (GL7−) B cells exhibited equivalent staining, ICOS−/− GL7+ B cells expressed significantly less LTαβ than did similarly gated WT GL7+ B cells (Fig. 4, B overlay and Cii) (p ≤ 0.004). Although their frequency is low in ICOS−/− mice, we also assessed LTαβ levels on the remaining GL7+Fas+ B cells and similarly found a statistically significant decrease in LTαβ expression using this gating methodology (Fig. 4, B and Cii) (p ≤ 0.0004). Taken together, our results show that despite normal expression of LTαβ on follicular B cells, there is a specific reduction in LTαβ expression on IgD− B cells that express PNA, or B cells that express GL7 in ICOS−/− mice.
ICOS crosslinking on T cells induces LTαβ expression on B cells
To test whether ICOS signaling can condition T cells so that they promote LTαβ expression on B cells, we designed an in vitro assay whereby we could manipulate signals delivered to T cells. To this end, we obtained purified CD4+ T cells and B220+ B cells from WT, ICOS−/−, and LTβ−/− mice and cocultured B and T cells in wells plated with anti-CD3, agonistic anti-ICOS, or control rat IgG to stimulate CD4+ T cells. To sustain T cell survival and activation in culture, IL-2 was added into wells. Cells were stimulated in culture for 24, 48, and 72 h, and expression of LTαβ on the B cells was assessed by FACS analysis. For controls, we used either ICOS−/− T cells or LTβ−/− B cells, and in both of these scenarios, no specific staining for LTαβ on the surface of the B cells was observed (Fig. 5,A). However, in the case of coculture of WT B and T cells, stimulation of ICOS on CD4+ T cells was found to induce significant LTαβ expression on the cocultured B cells. Expression of LTαβ was specific to ICOS stimulation, as WT T cells stimulated with anti-CD3 and a rat IgG isotype control failed to express LTαβ. Elevated LTαβ expression was only seen after 48 h of coculture and was not evident at 24 h (Fig. 5,B), presumably because ICOS expression is not constitutive but rather is induced on the surface of activated T cells (5). Moreover, LTαβ expression on cocultured B cells at this 48 h time point could not be achieved using conventional anti-CD28 costimulation, but was rather specific to ICOS costimulation (Fig. 5,A, 3rd panel), even though these conditions readily induced LTαβ expression on T cells themselves (Fig. 5 A, 4th panel). Therefore, these experiments demonstrate that ICOS signaling renders T cells capable of inducing the expression of LTαβ on B cells in vitro.
Treatment of ICOS−/− mice with agonistic Abs to CD40 and LTβR
We next examined whether ICOS-induced LTαβ expression on GC B cells is sufficient for GC formation. We evaluated this in vivo by injecting ICOS−/− mice with agonistic anti-LTβR Ab following immunization to bypass the reduced level of LTαβ on GC B cells in these mice. On day 12 postimmunization, we analyzed treated animals for recovery of GC B cells and GC networks. FACS analysis of anti-LTβR-treated ICOS−/− mice showed that there was no significant recovery of GL7+Fas+ GC B cells compared with control-treated ICOS−/− mice (Fig. 6, A and B). Additionally, histological examinations of spleen sections from treated ICOS−/− mice revealed few, small GL7+ clusters that were similar in size and number to control-treated ICOS−/− mice, although the GL7+ clusters from the anti-LTβR treated mice did appear to be more tightly packed/organized (Fig. 6,C, see arrow). Morphometric analysis of spleen sections from agonist-treated vs control-treated ICOS−/− mice revealed that there was no significant increase in the area of GL7+ clusters indexed over total area of the spleen in ICOS−/− mice treated with the agonistic Ab (Fig. 6,D). Serum from individual mice was analyzed to determine whether agonist Ab treatment resulted in recovery of the high-affinity anti-NP IgG1 response in ICOS−/− mice, which is impaired (Fig. 6,E). Comparing the ratio of high- to low-affinity anti-NP titers, we found no significant increase in high-affinity anti-NP IgG1 Ab titers in ICOS−/− mice treated with control vs anti-LTβR agonistic Ab (Fig. 6 F). Collectively, our results show that restoration of LTβR signaling in the context of the GC response is not sufficient for rescue of the GC reaction in ICOS−/− mice.
One important caveat to these experiments is that follicular Th cells, which display CD40L on their surface, are significantly decreased in ICOS−/− mice (17). Indeed, optimal expression of CD40L on Th cells is enhanced (12) and is dependent (13) on ICOS triggering. Therefore, despite enforced LTβR signaling, delivery of CD40-mediated signals to GC B cells is still lacking. Because the LT pathway is not sufficient to restore the GC response in ICOS−/− mice, we assessed the cooperative function of CD40 signaling in this context by treating ICOS−/− mice with an agonistic anti-CD40 Ab on days 3 and 7 postimmunization to restore T cell help to B cells in ICOS−/− mice. Others have shown that anti-CD40 treatment of ICOS−/− mice results in a rescue of Ab class-switching (12), but the effects on the GC niche itself or on affinity maturation have not been measured. Accordingly, on day 12 postimmunization, we analyzed the spleens of anti-CD40-treated ICOS−/− mice and assessed the formation of GC structures. Our FACS analysis revealed that anti-CD40-treated ICOS−/− mice showed significant recovery of GL7+Fas+ GC B cells compared with control treated mice (p ≤ 0.001) (Fig. 7, A and B). Additionally, spleens from treated ICOS−/− mice exhibited increased size and numbers of GL7+ clusters within B cell follicles compared with control-treated mice (Fig. 7,C). This corresponded to an increase in the FDC network itself, as evidenced by staining with FDC-M1 to identify the “secondary” GC-FDC network (Fig. 7,C). Serum from these treated mice showed elevated levels of high-affinity anti-NP IgG1 Ab compared with control-treated ICOS−/− mice (p ≤ 0.05) (Fig. 7,D). Interestingly, when we examined LTαβ expression on GL7+ GC B cells, there was an increase in the level of LTαβ on GL7+ GC B cells from treated ICOS−/− mice such that it was comparable to what was observed on WT GL7+ GC B cells (Fig. 7,Ei). Quantitative analysis of LTαβ expression on GL7+ GC B cells over basal expression on follicular B cells showed that anti-CD40-treated ICOS−/− mice had a statistically significant (p ≤ 0.01) increase in fold expression of LTαβ compared with control-treated ICOS−/− mice (Fig. 7 Eii).
To examine whether the ICOS-dependent up-regulation of LTαβ on B cells is indeed CD40 dependent, we returned to our in vitro assay system and substituted CD40L−/− T cells for WT T cells. Although the population of CD25+-activated B cells is diminished when cocultured with CD40L−/− T cells, there remains a small population that can be analyzed for LTαβ expression (Fig. 7,F dot plots). Consistent with a role for CD40 signaling in ICOS-dependent LTαβ expression on B cells, we failed to observe the induction of LTαβ on B cells cocultured with CD40L−/− T cells (Fig. 7,F, upper histogram with blue trace representing WT cells). The same results were obtained if we blocked CD40L using the MR1 Ab (Fig. 7 F, lower histogram). Therefore, our results show that anti-CD40 treatment in ICOS−/− mice restores GC structures concomitant with the recovery of LTαβ expression on GC-B cells, and that CD40 signaling is likely responsible for the ICOS-dependent up-regulation of LTαβ on GC B cells.
Rescue of GC reaction in ICOS−/− mice with anti-CD40 is dependent on LTβR signaling
Although these data link the expression of LTαβ with CD40 stimulation on GC B cells in vivo, they did not prove whether the restoration of LTαβ expression on anti-CD40-treated ICOS−/− GC B cells was responsible for the recovery of the GC microenvironment in ICOS−/− mice. To determine whether CD40-mediated rescue of the GC response is dependent on LTβR signaling in ICOS−/− mice, we doubly treated these mice with the anti-CD40 agonist Ab together with a competitive inhibitor of the LT pathway (LTβR-Ig). As before, anti-CD40 treatment combined with a matched control Ab succeeded in restoring the numbers of GL7+Fas+ cells in ICOS−/− mice. Indeed, the number of GC B cells appears greater, likely due to the increased splenic cellularity observed in anti-CD40-treated ICOS−/− mice. Nonetheless, the representation of GC B cells as a frequency of the B cell compartment is comparable (Fig. 8, A and B). In contrast, ICOS−/− mice that received LTβR-Ig did not show evidence of a recovered GC B cell population upon treatment with anti-CD40 agonistic Ab (p ≤ 0.0009) (Fig. 8, A and B). Likewise, although affinity maturation of the IgG1 response is restored in anti-CD40-treated ICOS−/− mice, this restoration is slightly but significantly blunted in the presence of LTβR-Ig (p ≤ 0.03) (Fig. 8,C). Finally, the restoration of the GC network that was observed with anti-CD40 treatment in ICOS−/− mice was nullified in the presence of LTβR-Ig, as evidenced by a lack of GL7 clusters and FDC-M1 expressing cells within the IgD-negative GC niche (Fig. 8 D). These results imply that CD40 and LTβR signaling are both required to restore GC responses in ICOS−/− mice.
Within the GC microenvironment, GC B cells make protracted contacts with FDC-M1+ cells (35). Additionally, Th cells also make contact with GC B cells within this niche, presumably to provide a source of CD40L and cytokines, and this help may be limiting (36). The sequence of events that leads to the formation of a functional GC niche depends on molecules expressed by each of these three cell types: follicular Th cells, GC B cells, and FDCs. Because ICOS−/− mice exhibit multiple downstream defects that culminate in the inability to form a GC niche, we took advantage of these mice to dissect the key requirements for GC formation. We have shown herein that ICOS stimulation is required for optimal expression of LTαβ on GC B cells; however, engagement of the LTβR on its own was insufficient to restore GCs in these mice. Because ICOS stimulation induces CD40L expression on Th cells beyond what is observed with conventional CD28 costimulation (12), and ICOS−/− T cells have impaired CD40L expression (13), we examined the consequences of enforced CD40 signaling in ICOS−/− mice. Administration of agonistic anti-CD40 Ab to ICOS−/− mice resulted in the recovery of the GC niche concomitant with the restoration of LTαβ levels on GC B cells. As these effects were nullified in the presence of a LT pathway inhibitor, input from both CD40 and LTβR is required downstream of ICOS stimulation to promote GC formation.
Maintenance of the FDC network in primary splenic follicles requires the continual expression of low levels of LTαβ on a subset of resting follicular B cells. Constitutive LTβR engagement on resident stromal cells and FDC within the white pulp follicle is important for maintaining homeostatic levels of CXCL13, which serves two major functions: to attract CXCR5+ resting B cells into the follicle, and to induce/maintain the expression of LTαβ on the B cell by signaling through CXCR5 (37). Little is known regarding how LTαβ functions to maintain “inducible” stromal cell niches that are provoked during an immune response, such as the GC. Clearly, these networks rely on LTβR signaling, but it is not understood how the expression of LTαβ on activated GC B cells is regulated. Indeed, although LTαβ expression on follicular B cells depends on CXCR5 signaling, the level of LTαβ on GL7+ GC B cells is normal in CXCL13−/− mice (25). We have shown that ICOS signaling on T cells is required to regulate LTαβ expression on GC B cells in a CD40-dependent manner. This provides a novel mechanism for the local regulation of LTαβ expression in the inducible GC niche. In contrast, ICOS−/− follicular B cells express normal levels of LTαβ (Figs. 3 and 4) and exhibit an intact primary FDC network (Fig. 3) as well as normal T-B segregation and marginal zone organization (data not shown), all of which depend on LTβR signaling outside of the GC. Thus, depending on the anatomic location, LTαβ expression is regulated by different sets of molecules.
Administration of agonistic anti-LTβR Ab to ICOS−/− mice did not significantly restore GC structures, although we observed improved aggregation of GL7+ cells into a tightly organized structure. Thus, expression of LTαβ on GC B cells may induce a GC organization sequela, provoking LTβR+ FDCs, which normally form a loose network within the primary follicle, to come together in an organized niche within what will become a secondary follicle. In agreement with this notion, treatment of Cynomolgus macaques with LTβR-Ig results in the disappearance of FDCs; however, upon withdrawal of this agent, FDCs gradually reemerge, but in diffuse networks (19). Presumably once the inhibitor is fully cleared and maximal LTβR signaling is restored, chemokine gradients become reestablished in such a way that FDC networks coalesce completely to form a dense FDC network scaffold within the GC. This would be analogous to the situation of administration of agonistic Ab in our experiments. Interestingly, however, the anti-LTβR Ab failed to augment the total number of GL7+Fas+ GC B cells in ICOS−/− mice, nor did this treatment improve affinity maturation of the anti-NP response. We reason that other ICOS-dependent factors (such as CD40L) are required for promoting the survival of GC B cells as well as affinity maturation and Ab class-switching activities in vivo. In other words, because ICOS−/− mice have impaired CD40L expression (13), and because CD40 activation is essential for B cell activation (38), it makes sense that LTβR signaling on its own would be insufficient to restore GCs. Thus, while anti-LTβR treatment can act on resident stromal cells and FDCs inducing their organization into GC structures, enforced LTβR signaling is not sufficient to completely restore the GC response in ICOS−/− mice where T cell help is largely absent.
Administration of agonistic anti-CD40 Ab, however, restored the defect in T cell help in ICOS−/− mice. Importantly, we found that recovery of GC correlated with the recovery of LTαβ expression on ICOS−/− GC B cells. Although supraphysiological levels of CD40 signaling may supersede many subtle effects of a normal immune response, these results demonstrate that CD40 signaling can promote LTαβ expression within the GC niche in vivo, which is consistent with our finding that ICOS-dependent LTαβ expression on B cells is dependent on CD40 signaling in vitro. These data provide a mechanism for the dependence on ICOS signaling for optimal LTαβ expression on GC B cells. Moreover, the restoration of LTαβ on ICOS−/− GC B cells in anti-CD40-treated mice suggested to us that CD40L and LTαβ were working in concert to induce proper GC formation. Indeed, our data using dual treatment of agonistic anti-CD40 with LTβR-Ig blocking reagent demonstrate that the CD40-dependent rescue of the GC reaction is dependent on LTαβ expression.
We do not have direct evidence that LTαβ on the GC B cells themselves is responsible for the rescue of the GC niche, and in the absence of markers for GC-resident follicular Th cells (for example, CD57 identifies these cells in humans but not in mice) (39), we were unable to evaluate whether anti-CD40 treatment likewise induced LTαβ expression on GC-resident T cells indirectly. For example, anti-CD40 treatment may activate dendritic cells such that they may costimulate Ag-specific T cells to become activated, express LTαβ, and migrate to the GC. Bone marrow chimera experiments have shown that B cells are the relevant source of LTαβ for maintenance of the FDC network (22, 23), although in the absence of B cell-expressed LTαβ there is some compensation from the T cell compartment (24). In the context of ICOS−/− mice, we think that T cells are an unlikely source of LTαβ for sustaining the GC niche during in vivo priming because LTαβ expression on activated T cells is transient (40) and is therefore not likely to sustain GC architecture several days postimmunization. Additionally, conventional costimulation (CD28) is adequate for up-regulation of LTαβ on T cells (Fig. 5), and thus this would not explain the GC defect in ICOS−/− mice where CD28 costimulation is intact. Therefore, we favor the hypothesis that LTαβ on the GC B cell represents the most likely reservoir of LTβR ligand responsible for the formation of the secondary FDC network within the GC niche. The consequences of LTβR stimulation within the GC niche via GC B cell-expressed LTαβ are likely numerous, including the restoration of chemokine gradients and adhesion molecules such as VCAM. In the future, it would be interesting to determine what downstream consequences of LTβR signaling are blocked in the absence of LTαβ specifically expressed on GC B cells.
Although treatment with anti-CD40 agonistic Ab in ICOS−/− mice can restore class switching (12), affinity maturation can occur independent of the GC microenvironment (41), and the effect of anti-CD40 on affinity maturation in ICOS−/− mice has not been evaluated. We found that anti-CD40 treatment was able to restore affinity maturation of the T-dependent NP-specific response. Additionally, affinity maturation still occurred in anti-CD40 + LTβR-Ig dual-treated mice. LTβR−/− mice exhibit a profound defect in affinity maturation at low doses of Ag (1 μg NP-CGG); however, at high doses (200 μg NP-CGG) the defect is only mild (42). Because our study also used high doses of Ag, we were not surprised that LTβR-Ig treatment did not completely prevent affinity maturation in dual-treated mice, although there was a slight but statistically significant reduction in high-affinity IgG1 anti-NP titers in dual-treated ICOS−/− mice compared with mice treated with anti-CD40 alone. Thus, our observations reveal a partial dependence on LTβR signaling for optimal affinity maturation of the anti-NP IgG1 response in ICOS−/− mice.
Our model is that ICOS stimulation on activated T cells results in the acquisition of a follicular Th phenotype (expression of CXCR5) to enable migration into the B cell follicle (17). In agreement, in our system CXCR5+ Th cells were also significantly decreased in ICOS−/− mice (data not shown). Follicular Th cells express very high levels of CD40L (39), and their proximity with Ag-specific B cells provides a means for CD40 triggering in the B cell. Because CD40 signaling has been shown to induce the expression of LTαβ on splenic B cells in vitro (25), and enforced CD40 signaling restores high levels of LTαβ expression on GC B cells in vivo (Fig. 7,D), we hypothesized that CD40 signaling is responsible for the elevated levels of LTαβ expression that we and others have observed on the surface of GC B cells as compared with resting follicular B cells. Indeed, we found that ICOS-activated CD40L-deficient T cells could not provoke LTαβ expression on cocultured B cells (Fig. 7), thus placing CD40 activation downstream of ICOS in the regulation of LTαβ on B cells. A concentrated source of LTαβ on GC B cells in turn may provide a strong signal via LTβR on the FDC so that FDC coalesce into a tight secondary network. Taken together, the GC microenvironment requires both: ① a LTβR-dependent FDC network sustained by LTαβ expressed on GC B cells, which serves as a necessary scaffold for B cells and ② CD40-dependent signals delivered to B cells to provoke and sustain survival/proliferation and seeding of the emerging GC niche (Fig. 9). ICOS-mediated cytokines are also likely important for an optimal GC reaction; however, we think that ICOS-dependent cytokines such as IL-4/IL-10 play a secondary role in regulating LTαβ expression within the GC as compared with CD40L. Indeed, attempts at blocking Th2 cytokines IL-4, IL-5, IL-10, and IL-13 did not affect ICOS-dependent LTαβ expression on B cells in vitro (data not shown).
In summary, our studies have provided insight into the regulation of LTαβ expression within the GC niche. Additionally, the failure to up-regulate LTαβ on GC B cells in ICOS−/− mice provides a potential explanation for the abortive GC structures that form when these animals are immunized with T-dependent Ag. These results reveal a sequence of molecular events, beginning with ICOS activation, that integrate to initiate GC formation.
We acknowledge Cheryl Smith and Dionne White for management of the Faculty of Medicine flow cytometry facility. We also acknowledge Drs. Tania Watts and Mariana Vidric for experimental advice, and Drs. Jeff Browning and Evangelia Notidis for critical input and for providing LTβR-Ig and BBF6 reagents.
The authors have no financial conflicts of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by an operating grant from the Canadian Institutes of Health Research (Grant 165683 to J.L.G.), an Ontario Graduate Studentship award (to F.V.), the National Institutes of Health (Grant R37A133068 to C.F.W.), and Associazione Italiana Ricerca sul Cancro (to U.D.).
Abbreviations used in this paper: GC, germinal center; AP, alkaline phosphatase; CGG, chicken γ-globulin; FDC, follicular dendritic cell; KLH, keyhole limpet hemocyanin; LT, lymphotoxin; LTβR, lymphotoxin β receptor; MFI, mean fluorescence intensity; NP, (4-hydroxy-3-nitrophenyl)acetyl; PNA, peanut lectin (agglutinin); WT, wild type.