Visualizing B cell capture of cognate antigen from follicular dendritic cells

To take advantage of the intense fluorescence of PE and the high-affinity hen egg lysozyme (HEL) binding of MD4 Ig transgenic B cells (Goodnow et al., 1988), we generated multivalent HEL-PE antigen using biotinylated HEL and streptavidin (SA)-PE. Mice were passively immunized with polyclonal anti-PE IgG and, 1 d later, immunized s.c. with 10 µg HEL-PE. Analysis of draining LN sections taken at various time points showed that the majority of the HEL-PE became restricted to FDCs 1 d after antigen injection (Fig. 1 A and Fig. S1). Staining of adjacent sections for GC markers GL7 and FcγRIIb confirmed that these were primary follicle FDC networks (Fig. S2). As an approach to favor the likelihood that cognate B cells encountered antigen for the first time on FDCs, MD4 B cells were intravenously transferred into the recipients at 1 d and analyzed at various times by flow cytometry for acquisition of antigen. Within 12 h of transfer, 70% of MD4 B cells in the draining LN were PE⁺, indicating capture of HEL-PE (Fig. 1 B), whereas cotransferred noncognate B cells did not capture the HEL-PE at this time (Fig. 1 B). In control recipients that had been immunized with PE that was not coupled with HEL there was no PE capture (Fig. 1 B), and in mice receiving HEL-PE in the absence of antibody to PE there was minimal antigen capture by cognate B cells (Fig. 1 B).

After antigen capture, cognate B cells up-regulated the costimulatory molecule CD86 (Fig. 1 C) and also CD83 (Fig. 1 C), another early marker of B cell activation (Prazma et al., 2007), whereas the transferred noncognate B cells remained quiescent. In contrast to wild-type MD4 B cells, CXCR5-deficient MD4 B cells that are unable to enter LN follicles (Ansel et al., 2000) showed little antigen capture or evidence of activation at the 12-h time point (Fig. 1, C and D). Defective encounter with FDC-bound antigen may contribute to the reduced antibody responses mounted by CXCR5-deficient B cells (Junt et al., 2005). Indeed, although wild-type MD4 B cells were able to promote proliferation of cotransferred HEL-specific CD4 T cells, CXCR5-deficient MD4 B cells showed minimal ability to promote T cell proliferation (Fig. S3). Moreover, wild-type MD4 B cells were induced to differentiate to GC B cells, whereas very few CXCR5-deficient cells underwent differentiation (Fig. S3). These observations indicate that B cells capturing antigen from FDCs are able to internalize, process, and present the antigen to T cells and to respond to T cell help.

Further supporting the conclusion that antigen encounter at this time point required presentation by FDCs, when CR1/2-deficient mice were immunized very little HEL-PE became deposited on FDCs, as expected (Roozendaal and Carroll, 2007), and antigen capture by wild-type MD4 B cells was significantly reduced (Fig. 1, E and F). The low continued capture that did occur (Fig. 1 F) suggests that some encounter may occur within the CR1/2-deficient follicle, perhaps reflecting residual display by SCS macrophages or low-level FcγRIIb-mediated display within the follicle. Transfer of MD4 B cells to mice that had been preimmunized with duck egg lysozyme (DEL)-PE, an antigen that binds the MD4 B cell receptor with ∼1,000-fold lower affinity (K_a∼10⁷ M^-1) than HEL (Lavoie et al., 1992), provided evidence that low-affinity B cells could capture FDC-associated antigen and become activated (Fig. 1, G and H).

Using the system described in the previous section, LNs were imaged by two-photon microscopy to detect CFSE-labeled MD4 B cells, cyan fluorescent protein (CFP) transgenic (tg) control B cells, and HEL-PE antigen. Within 12 h of B cell transfer to preimmunized hosts, B cells could be observed migrating through the HEL-PE–laden FDC network, and some cells carried HEL-PE at their uropod (Fig. 2 A and Video 1). In some cases the image quality allowed resolution of a punctate or beaded aspect to the HEL-PE distribution along the FDC processes (Fig. 2 B), reminiscent of the IC–coated bodies or iccosomes observed in fixed-tissue ultrastructural studies (Szakal et al., 1988, 1989). A punctate PE distribution within the FDC network was also observed in immunofluorescence analysis of cryosections (Fig. S1), though particle sizes could not be directly compared because of the different methods of image detection. In the absence of cognate B cells, the majority of PE codistributed with the FDC marker CD35 (CR1; Fig. 1 A and Fig. S1), allowing us to use PE fluorescence as a reasonable surrogate for FDC in the imaging analysis. PE ICs and injected FDC-reactive antibodies have previously been used to highlight the GC light zone FDC network in studies tracking the migration dynamics of GC B cells (Allen et al., 2007; Hauser et al., 2007; Schwickert et al., 2007). By tracking individual MD4 B cells over time, acquisition of antigen from primary follicle FDCs could be observed, often involving capture of large antigen aggregates (Fig. 2 C and Video 2). Measurement of contact time with antigen-laden FDC processes showed that cognate B cells had increased contact times (median of 3 min), with some cells maintaining contact for >30 min (Fig. 2 D and Video 3), compared with noncognate B cells where interactions with antigen-labeled FDC processes rarely lasted for more than 1–2 min (Fig. 2 D). These interaction dynamics share similarities with the behaviors observed previously by two-photon microscopy of B cells and FDCs within the GC, where many transient interactions and occasional long interactions were observed (Allen et al., 2007; Hauser et al., 2007; Schwickert et al., 2007). Both cognate and noncognate cells moved with lower mean velocities during periods of contact with antigen-decorated FDCs, though the reduction in velocity was greatest for cognate cells (Fig. 2 E). Cognate B cells also had a reduced velocity when not in contact with the PE⁺ FDCs (Fig. 2 E), which is consistent with the previous finding that B cells recently engaged by antigen moved with reduced velocity (Okada et al., 2005). The cognate B cells also had sharper turning angles than noncognate cells, both during contact and when not in contact with the PE⁺ FDCs (Fig. S4). These observations indicate that B cells interact closely with FDCs and that cognate cells are able to capture large amounts of antigen during such encounters. Of 127 cognate cells that contacted HEL-PE–laden FDCs, 19 cells captured detectable amounts of HEL-PE. The median contact time during the antigen capture event was 6.5 min (Fig. 2 F). The increased interaction time may facilitate integrin-mediated B cell spreading and contraction on the FDC in a process that can facilitate collection and capture of membrane-bound antigen (Fleire et al., 2006).

The observation that cognate B cells often captured large units of antigen together with the beaded aspect of some FDC processes led us to ask whether cognate B cells were capturing fragments of FDCs. Indeed, by flow cytometry about one-third of the PE-bright B cells were also shown to be positive for the follicular stromal marker BP-3 (Fig. 3, A and B; McNagny et al., 1991; Ngo et al., 1999), whereas noncognate B cells showed little staining for this marker (Fig. 3, A and B). To test whether this capture extended to additional FDC markers, Cr2^−/− B cells that lack surface CR1 (CD35) expression were transferred to preimmunized wild-type recipients and the capture of CR1 was examined. Similarly to BP-3, the cognate, but not noncognate, B cells captured CR1 (Fig. 3, C and D). B cells capturing BP-3 or CR1 corresponded to the brightest PE⁺ cells, suggesting a relationship between the extent of antigen capture and the probability of acquiring FDC surface markers. Alternatively, only by capturing an FDC fragment could cognate B cells acquire the highest amounts of antigen. Immunofluorescence microscopy of B cells purified 12 h after transfer revealed cells with single large clusters of PE at one pole (Fig. 3 E). On PE-bright cells where BP-3 could be detected, it was typically located as a single cluster overlapping with or closely adjacent to the PE signal (Fig. 3, E and F). In the case of cells with lower amounts of PE signal, the BP-3 was sometimes situated in a separate cluster (Fig. 3, E and F). Collectively, these observations suggest that cognate B cells are able to acquire FDC surface markers during antigen capture.

Previous work using the MD4 system established that recombinant HEL-C3d-C3d was a much more potent stimulus of B cell activation than HEL alone (Dempsey et al., 1996). We were therefore surprised to find that Cr2^−/− MD4 and wild-type MD4 B cells showed equally efficient activation upon encountering opsonized HEL-PE in the follicle (Fig. S5, A and B), suggesting that activation by multivalent HEL complexes is less dependent on B cell–intrinsic CR2 signaling. Notably, however, when complexes were formed with the low-affinity antigen DEL, CR2 in the cognate B cells made a prominent contribution to the efficiency of both antigen capture and B cell activation (Fig. S5, C and D). This observation adds to the extensive evidence that the CR2 complex acts intrinsically to amplify BCR signaling (Dempsey et al., 1996; Roozendaal and Carroll, 2007) to suggest that it can also augment antigen capture by cognate B cells.

During early studies of antigen deposition on FDCs, it was established that the antigen could be retained in an intact form for periods of weeks (Tew and Mandel, 1978). The importance of this sustained presentation has been emphasized in the context of driving GC responses. However, it has not been determined whether this prolonged retention facilitates ongoing encounter by naive B cells that enter the LN late, perhaps after traveling from a distant site. Given the presence of at least 20 secondary lymphoid tissues in the healthy mouse and the evidence that B cells spend about a day in any given lymphoid tissue (Tomura et al., 2008), if the cell were to sample lymphoid organs randomly, a single B cell has a 50% chance to localize to a given LN within 2 wk (see Materials and methods). The time is likely to be shortened by inflammatory insults that increase entry and decrease exit rates for a draining lymphoid tissue. In time-course experiments, cognate B cells that newly entered a HEL-PE draining LN could be observed capturing FDC-displayed antigen 9 d after the initial immunization (Fig. 4, A and B). These encounters continued to lead to B cell activation as assessed by CD83 and CD86 induction (Fig. 4 B). Thus, FDCs provide a mechanism to increase the chance of antigen encounter even by rare antigen-specific B cells. Starting from day 6, much of the deposited PE was located within GL7⁺ GCs (Fig. S2). These are likely to be GCs generated by endogenous B cells in response to the HEL-PE ICs. Recent work has visualized naive B cell entry to the GC light zone (Schwickert et al., 2007), providing a possible mechanism for naive B cell encounter with antigen in this compartment.

Through the use of two-photon microscopy and brightly labeled high-affinity antigen, we have visualized B cell capture of opsonized antigen from FDCs in primary follicles. In some cases, B cells appear to capture the antigen as a large unit that also involves acquisition of FDC membrane markers, which is consistent with evidence from fixed tissue analysis that B cells can become associated with IC–coated bodies or iccosomes (Szakal et al., 1988, 1989). The frequently beaded appearance of HEL-PE on the FDC may also fit with this concept. The correlation between acquisition of FDC surface antigens and acquisition of the largest amounts of cognate antigen suggests that by capturing FDC fragments, cognate B cells are able to increase the amount of antigen they acquire. Preformed iccosomes may not, however, be necessary to account for the FDC membrane antigen capture, as in vitro studies demonstrated the ability of antigen-binding B cells to extract antigens from cell membranes (Batista et al., 2001) and exchange of membrane between closely opposed cells has been observed in a variety of contexts (Williams et al., 2007). The finding on some of the cells with lower amounts of HEL-PE of separate PE⁺ and BP-3⁺ foci suggests that the antigen and stromal marker can sometimes be captured separately though the possibility that antigen and BP-3 become separated after capture is not excluded. BP-3 is expressed on the entire network of follicular stromal cells and is not limited to the antigen-presenting CR1⁺ FDC (Ngo et al., 1999). Perhaps the integrin activation induced by B cell receptor engagement (Spaargaren et al., 2003; Carrasco et al., 2004) can promote increased adhesive interactions with the follicular stromal cell network that increase capture even of antigen-negative BP-3⁺ membrane fragments. Whether captured as a single unit or separately, these observations suggest that additional signals may be delivered from follicular stromal cells to B cells during cognate antigen recognition.

Many antigens can become opsonized independently of high-affinity antibody, via alternative or lectin pathways of complement activation, and some can become decorated by low-affinity natural antibody, providing pathways by which FDCs can be antigen loaded in the initial phases of a primary response. Collectively, the findings in this paper and the recent work visualizing cognate antigen encounter on SCS macrophages (Carrasco and Batista, 2007; Junt et al., 2007; Phan et al., 2007) can propose that in the first hours after arrival of opsonized antigen in a LN, B cells have the opportunity to encounter antigen displayed by macrophages. At later time points, however, encounter with FDC-bound antigen may predominate, especially in responses involving nonreplicating antigens. In cases where antigen is continuously produced, both antigen display sites may operate in synchrony. Whether the fate of B cells encountering antigen on SCS macrophages and FDCs is the same remains to be investigated. The time course over which naive B cells can encounter antigen on FDCs is likely to vary widely depending on the properties and amounts of antigen, the frequency of specific B cells, and the nature of any prior B cell response. Our finding that encounter could occur 9 d after immunization, a time point when the FDC networks were GC associated, is in accord with the finding that transferred cognate B cells could join a day-9 GC response (Schwickert et al., 2007). Another context where we envision a role for antigen encounter on primary follicle FDCs is after the first antibody production of a primary response. Here, the higher affinity cells may respond first, giving rise to antibody that promotes antigen deposition on FDCs and increasing the chance of productive encounter by low-affinity cells that recognize distinct epitopes. In the case of a secondary response, because we passively immunize with high-affinity IgG specific for PE but not HEL, we suggest our findings may model secondary encounter with a mutated or otherwise modified form of the priming antigen. Finally, although we have visualized the process of naive B cell capture of antigen from FDCs in the primary follicle, the similar migration behaviors of GC B cells with respect to light zone FDCs (Allen et al., 2007; Hauser et al., 2007; Schwickert et al., 2007) suggests that similar antigen capture from FDCs can take place within the GC during antibody affinity maturation.

6–12-wk-old male wild-type and Ly5.2 (CD45.1) congenic C57BL/6 were purchased from either the National Cancer Institute or The Jackson Laboratory. MD4 mice (Goodnow et al., 1988) were from an internal colony. Animals were housed in a germ-free environment in the Laboratory Animal Research Center at the University of California, San Francisco (UCSF) and all experiments conformed to ethical principles and guidelines approved by the UCSF Institutional Animal Care and Use Committee.

HEL- or DEL-PE was produced by incubating HEL- or DEL-biotin in molar excess with SA-PE and removing unbound HEL- or DEL-biotin using the BioSpin 30 column (Bio-Rad Laboratories). ICs were generated as previously described (Phan et al., 2007). In brief, mice were first passively immunized with 2 mg of rabbit anti-PE IgG (Rockland Immunochemicals). Recipient mice were passively immunized in this way unless otherwise indicated. 1 d later, mice received HEL- or DEL-PE s.c. in the flank and tail base to drain the inguinal LN (2.5 µg/site). Unless otherwise stated, mice received transfers of cells 24 h after antigen immunization and, 12 h later, they were analyzed by flow cytometry and sectioning or the LNs were explanted for two-photon microscopy.

Draining LNs were teased apart with microforceps into DME containing 2% fetal bovine serum, penicillin/streptomycin, and Hepes buffer. Single cells were washed and blocked with anti-CD16/32 (2.4G2; UCSF Hybridoma Core Facility) and stained for flow cytometric analysis as previously described (Allen et al., 2007; Phan et al., 2007). Data were acquired on a LSRII flow cytometer (BD) and were analyzed with FlowJo software (Tree Star, Inc.).

Transferred MD4 B cells (CD45.2⁺) were sorted as CD19⁺CD45.1⁻PE^hiBP-3^hi or PE^loBP-3^lo subsets from the LNs of CD45.1 congenic recipients with a FACSAria (BD). Cells were then fixed with 4% paraformaldehyde (Sigma-Aldrich) on 8-well chambered coverslips (Invitrogen) and stained with DAPI (Invitrogen). Images were obtained with an inverted microscope (Axio Observer Z1; Carl Zeiss, Inc.).

Explanted LNs were prepared for imaging as previously described (Okada et al., 2005; Allen et al., 2007). Images were acquired with Video Savant (IO Industries), and maximum-intensity time-lapse images were generated with MetaMorph (MDS Analytical Technologies). Videos were processed with a median noise filter. Excitation wavelength was 910 nm. Emission filters were 440–480 nm for second harmonic, 470–500 nm for CFP, 525–575 nm for CFSE, and 605–675 nm for PE. Cell tracks were made with Imaris 5.01 ×64 (Bitplane), and automated tracks were verified and corrected manually. Contact durations of B cells with FDCs were measured manually for each track by counting time points with a 30-s interval, in which a B cell–derived fluorescent signal was in contact with PE signals. The velocities and turning angles of cells between each imaging frame were calculated with MATLAB (The MathWorks, Inc.). Annotation and final compilation of videos were performed with After Effects 7.0 software (Adobe). Video files were converted to MPEG format with AVI-MPEG Converter for Windows 1.5 (FlyDragon Software).

Prism software (GraphPad Software, Inc.) was used for all statistical analyses. For comparison of two nonparametric datasets, the Mann-Whitney U test was used. Means of two groups were compared with the one-tailed Student's t test. The number of cell recirculation attempts (N) was calculated according to the equation P = (1 − qN), where P is the probability to localize to a particular lymphoid organ (out of 20 equivalent ones) at least once after N trial attempts and q is the probability to localize to any other lymphoid organ in a single trial (for 20 lymphoid organs q = 19/20).

Detailed description of methods can be found in the supplemental text. Fig. S1 is a magnified image of a LN section to depict punctate distribution of HEL-PE in the FDC network. Fig. S2 shows LN sections stained to detect GC markers FcγRIIb and GL7, as well as HEL-PE. Fig. S3 shows T cell proliferation and GC B cell differentiation driven by antigen on FDCs. Fig. S4 shows turning angles of B cells calculated based on two-photon imaging. Fig. S5 shows antigen acquisition and activation of Cr2^−/− MD4 B cells. Video 1 shows real-time imaging of B cell migration in the FDC network. Video 2 shows real-time imaging of antigen capture from FDCs by cognate B cells. Video 3 shows real-time imaging of prolonged interaction of B cells with FDCs.

We thank Matthew Krummel and Andrew Bullen for help with the two-photon microscope and Jesse Green for helpful comments on the manuscript.

K. Suzuki is a Human Frontier Science Program fellow, I. Grigorova is a Cancer Research Institute fellow, and J.G. Cyster is an Investigator of the Howard Hughes Medical Institute. This work was supported in part by National Institutes of Health grant AI45073 and a Sandler New Technology Award.

The authors declare that they have no competing financial interests.