Follicular Dendritic Cells Carry MHC Class II-Expressing Microvesicles at Their Surface¹

Follicular dendritic cells (FDCs)³ are accessory cells of the immune system essential for the germinal center reaction (1, 2, 3, 4). During germinal center reactions, FDCs present Ags to B lymphocytes and T lymphocytes (5). B cells recognize native Ags held in immune complexes at the surface of FDCs. To this aim, FDCs express a set of several different complement receptors (6, 7). The binding of B lymphocytes to these Ags is highly competitive and an essential selection step for memory B lymphocytes. A prerequisite for stimulation of T helper cells is MHC class II expression, which has been reported for FDCs in several studies using light microscopy (summarized in Ref. 8). However, the suggested role in presentation of processed Ag in the form of peptide-loaded MHC class II is not in line with the fact that FDCs characteristically do not have phagocytic activity (9). Furthermore, FDCs do not synthesize MHC class II molecules themselves (5). This was demonstrated in bone marrow transplantation studies of Gray et al. (5), who showed that murine FDCs passively acquire peptide-loaded MHC class II molecules from the donor. The mechanistic basis of this phenomenon remained an enigma up to now.

In B cells, MHC class II molecules predominantly reside in MHC class II compartments (MIICs) (10), which are the subcellular site of peptide loading onto MHC class II molecules (11). MIICs are part of the endosomal/lysosomal pathway in APC and often have a multivesicular phenotype (12, 13). The internal vesicles of multivesicular MIICs are probably formed by inward budding from the limiting membrane (14), a process that seems to require phosphatidylinositol 3-kinase activity (15). Multivesicular MIICs can fuse with the plasma membrane in an exocytic manner, thereby releasing their internal vesicles, which have been shown to carry MHC class II molecules at their surface and to be capable of specifically stimulating CD4-positive T cells in vitro. The released internal vesicles of MIICs have been called exosomes (16), a term that was introduced for exocytosed vesicles in reticulocytes (17). Exocytosis of MIICs or multivesicular bodies has also been found in other APC, such as dendritic cells (18, 19), tonsil B cells (M.J.K. and H.J.G., unpublished data), and other hemopoietic cells, such as mast cells (20) and platelets (21).

We have previously shown that the internal vesicles of MIICs and exosomes are enriched in members of the tetraspan protein superfamily, such as CD37, CD53, CD63, CD81, and CD82 (22). Tetraspan proteins have been shown to be involved in Ag presentation, T cell signaling, T cell activation, motility, and adhesion. They are able to form oligomeric complexes with other tetraspan proteins (23, 24, 25), HLA-DR (23, 26), HLA-DM and -DO (26) and interact with integrins (25, 27) and the T cell coreceptors CD4 and CD8 (28). By immunoelectron microscopy (IEM), tetraspan proteins are confined to the internal vesicles of MIICs and exosomes, whereas other MIIC markers, such as Lamp-1, Lamp-2, and HLA-DM, reside primarily in the MIIC-limiting membrane (22) and are absent from exosomes.

In this paper, we present an IEM analysis of the microvesicles attached to the FDC surface in situ and show that these vesicles represent exosomes, possibly derived from B cells. The specific accumulation of exosomes on FDCs might play a role in fine tuning of the immune response.

The EBV-transformed B cell line RN was maintained as described (16). Commercial mouse mAbs were anti-CD37 (S-B3; Sanofi, Montpellier, France), anti-CD107b/Lamp-2 (H4B4; PharMingen, San Diego, CA), anti-CD63/Lamp-3 (CLB-gran1/2, 435), anti-CD53 (MEM-53), and anti-LFA-1 (CLB-LFA-1/1, CD18) from CLB (Amsterdam, The Netherlands). The mAbs against CD81 (M38) and CD82 (C33) were a gift of Dr. O. Yoshie (Osaka, Japan); mAb 5C1 (HLA-DM) was obtained from Dr. J. Trowsdale (Cambridge, U.K.). The mAb 6C4 (against lysobisphosphatidic acid (LBPA)) was provided by Jean Gruenberg (Geneva, Switzerland). The rabbit polyclonal Abs used in this study were: anti-MHC class II from Dr. H. L. Ploegh (Massachusetts Institute of Technology, Cambridge, MA), anti-biotin (Rockland, Gilbertsville, PA), anti-CD35/CR1 (3D9) obtained from J. A. Schifferli (Basel, Switzerland) (29), polyclonal anti-Ii (ICN-1 and ICC5) from Dr. P. A. Morton (Monsanto Co., St. Louis, MO) (30), and anti-mouse IgG from Dako (Glostrup, Denmark).

Fresh human tonsils were obtained from patients undergoing routine tonsillectomy. B cells were purified as previously described (31, 32). In short, T cells were removed by two cycles of rosetting with neuraminidase-treated sheep erythrocytes. The nonrosetting cells were further depleted of adherent cells by incubation in tissue culture flasks in HEPES-buffered RPMI 1640 (Life Technologies, Paisley, U.K.) supplemented with 50% heat-inactivated FCS for 1 h at 37°C. The nonadherent cells were recovered and subjected to centrifugation through a discontinuous 65, 52.5, and 42.5% Percoll gradient (Amersham Pharmacia Biotech, Uppsala, Sweden). At the interface of 42.5%–52.5% Percoll, the low density fraction was harvested which contained in vivo activated B cells and several FDCs. Cells were fixed in 2% paraformaldehyde and 0.2% glutaraldehyde in phosphate buffer and processed for ultracryotomy as described below.

FDCs were isolated from tonsils as described by Parmentier et al. (33). Tonsils were cut into pieces and treated with a collagenase (200 U/ml collagenase IV, Worthington Biochemical, Lakewood, NJ)-DNase (10 U/ml DNase I, Boehringer Mannheim, Mannheim, Germany) solution in IMDM (Life Technologies), followed by density sedimentation on a cold discontinuous BSA gradient (Path-o-cyte 4, bovine albumin, Instruchemie, The Netherlands) in HBSS consisting of layers of 1.5, 2.5, and 5%. The cells at the 2.5%–5% interface were harvested and washed in HBSS and IMDM. After incubating the cells with a mAb against human LFA-1 for 30 min at 4°C, LFA-1 expressing B cells were depleted by sheep anti-mouse Ig-coated Dynabeads (Dynal, Oslo, Norway). The resulting FDC-enriched cell suspension was cultured for 12–14 h at 37°C in IMDM supplemented with 10% FCS. After culture, it contained FDC/B cell clusters and other cell types as small lymphocytes, plasma cells, macrophages, dendritic cells, erythrocytes, and exceptionally mast cells and basophils.

Exosomes were isolated by differential centrifugation as previously described (16). Briefly, RN cells were washed by centrifugation and recultured in fresh medium for 18 h. Cell culture media (35 ml) containing about 5 × 10⁷ cells were centrifuged once for 10 min at 200 × g (pellet P₁), twice for 10 min at 500 × g (pellet P₂), twice for 15 min at 2,000 × g (pellet P₃), once for 30 min at 10,000 × g (pellet P₄), and finally once for 60 min at 70,000 × g (pellet P₅), using a rotor SW27 (Beckman Instruments, Fullerton, CA). To remove the excess of serum proteins, P₅ was washed once in 35 ml PBS and centrifuged again for 60 min at 70,000 × g. For in vitro binding studies with isolated human FDCs, this pellet was resuspended in 1 ml PBS and biotinylated by adding EZ-Link sulfo-NHS-biotin (Pierce, Rockford, IL) at a final concentration of 1 mg/ml for 30 min at room temperature. The reaction was blocked by resuspension in 34 ml PBS/15 mM ammonium chloride and sedimented for 60 min at 70,000 × g. The pellet was washed once in 35 ml PBS and resuspended in 100 μl PBS, 1% BSA, 5× penicillin/streptomycin. The resulting homogenous suspension of small vesicles contained only small amounts of plasma membrane and is referred to as biotinylated exosomes. Biotinylation efficiency was >95% as checked by whole mount immunoelectron microscopic analysis (22).

Isolated FDC/B cell clusters were allowed to attach overnight in a 35-mm petri dish. After the cells were washed three times with 1 ml HBSS at 37°C, the removal of cells in suspension was checked microscopically and 900 μl HBSS, 10% FCS was added. Parallel plates were incubated for 1 h at 37°C with 100 μl PBS, 1% BSA, 5× penicillin/streptomycin with or without biotinylated exosomes. Cells were fixed in 2% paraformaldehyde and 0.2% glutaraldehyde in phosphate buffer and processed as described below.

Cells were processed for ultrathin cryosectioning and immunolabeling as described previously (34, 35, 36, 37). Briefly, cells were fixed in a mixture of 2% paraformaldehyde and 0.2% glutaraldehyde in phosphate buffer. After being washed with PBS-glycine and PBS, cell pellets were embedded in 10 or 12% gelatin, cut into small blocks, which were infiltrated with 2.3 M sucrose overnight at 4°C. Finally, the blocks were frozen in liquid nitrogen for cryosectioning. Ultrathin cryosections were picked up from the knife with a mixture of 2.3 M sucrose and 2% methylcellulose. The sections were indirectly immunolabeled with 10-nm protein A-gold particles in single labelings or 10- and 15-nm particles in double labelings. Finally, the ultrathin sections were embedded in a mixture of 2% methylcellulose and 0.4% uranyl.

Ultrathin cryosections of cells from the in vitro binding experiment were immunolabeled by use of a polyclonal rabbit Ab against biotin and subsequently visualized by 10 nm protein A-gold. Electron micrographs were randomly taken from FDC-B cell clusters and various other cell types at magnifications of ×10,000 and ×5,000–10,000, respectively. The plasma membrane length of FDCs forming a cluster and solitary other cells was determined according to the method of Weibel (38). The number of biotin-positive vesicles attached to it was determined by counting vesicles with a distance of <40 nm from the plasma membrane. Finally, a Student t test was performed.

FDCs coisolated with in vivo activated B lymphocytes were morphologically characterized by the presence of long and branched dendrites with desmosomes at contact sites between neighboring dendrites. Immunogold labeling showed that FDCs were heavily labeled for CD35 (complement receptor 1), whereas different Abs against other FDC-specific markers such as the long isoform of CR2/CD21 (39) failed to react in IEM. Interestingly, numerous 40- to 70-nm membrane vesicles were seen attached to the surface of FDCs (Fig. 1), but not those of B cells. These microvesicles were different from the iccosomes described by Szakal et al. (40), which are ∼5 times larger. Microvesicles were found on FDCs from several donors and with three different isolation procedures.

FDCs have been shown to exhibit MHC class II at their cell surface (8) but are unable to synthesize it themselves (5). Therefore cells must acquire MHC class II passively from other sources (5), a process that has not yet been clarified. As shown in Fig. 1 B, MHC class II molecules at the FDC surface are confined to the microvesicles attached to their plasma membrane. In accordance with the functional data of Gray et al. (5), the FDC plasma membrane as well as the intracellular compartments were devoid of MHC class II by IEM. These observations suggest that the microvesicles represent MHC class II carrier vesicles derived from other cells that provide the FDCs with exogenous MHC class II at their surface.

The size of the microvesicles, as well as their labeling for MHC class II, prompted us to investigate whether these structures represent exosomes. We have previously shown that B cells secrete ∼50-nm MHC class II-expressing exosomes during exocytosis of multivesicular endosomes or MIICs (16). B cell-derived exosomes are enriched in members of the tetraspan family of proteins, expose MHC class I and II at their surface, contain the costimulatory molecule CD86, but are poor in the lysosomal membrane proteins Lamp-1 and Lamp-2, as well as HLA-DM (22). Microvesicles attached to FDCs exhibit a remarkably similar marker profile as characterized by IEM (Table I). They contain MHC class I and class II molecules. In addition, they are positive for the tetraspan proteins CD37, CD53, CD63, CD81, and CD82 (Fig. 2), but are poor or negative for Lamp-1 and Lamp-2 and proteins associated with Ag presentation such as Ii and HLA-DM. Importantly, the microvesicles contained lysobisphosphatidic acid (Fig. 3,A), a lipid of the internal vesicles of multivesicular endosomes (Fig. 3 B) (41). Taken together, these marker patterns strongly suggest that the microvesicles are exosomes secreted by APC and that they adhere but not fuse with the plasma membrane of FDCs.

Because FDC/B cell clusters isolated with an enzymatic treatment (as described in Materials and Methods) did not show many FDC-attached microvesicles, we decided to use these FDC-enriched cell populations for an in vitro binding assay with exogenous biotinylated exosomes derived from human B cells. Exosomes were isolated as described before (16). Binding of exosomes to the various cell types present in the FDC-enriched cell population was expressed as the number of biotin-labeled exosomes per unit plasma membrane length (Table II). Biotin-positive microvesicles were found attached to the FDC surface (Fig. 4). On average, the FDC plasma membrane showed 39 attached biotin-positive vesicles per 100-μm plasma membrane. This was in marked contrast to other cell types present in the FDC preparation like lymphocytes, plasma cells, macrophages, dendritic cells, erythrocytes, and occasional mast cells and basophils. For these cells, which served as internal controls, the number of biotinylated microvesicles attached to their surface did not exceed 7 per 100-μm plasma membrane. We conclude that B cell-derived exosomes exhibit a high preference to bind to FDCs.

To exclude the possibility that binding of exosomes to FDCs is merely an in vitro artifact, we analyzed FDCs in situ in germinal centers of human tonsils by cryoimmunogold electron microscopy. In their physiological state, FDCs form a follicular reticulum together with B lymphocytes. A typical example of such a binucleated FDC surrounded by several B lymphocytes is shown in Fig. 5,A. We frequently observed microvesicles labeled for the exosomal marker CD82 and MHC class II attached to the plasma membrane of FDCs in situ (Fig. 5 B). This labeling pattern could be followed along the branching processes of FDCs (not shown). These observations indicate that exosome binding to the surface of FDCs is not due to experimental manipulations, but is a physiological phenomenon.

In this study, we have shown for the first time the occurrence of numerous microvesicles attached to the FDC surface. They may have been observed earlier by Parmentier et al. (33), but these authors referred to them as peripheral villous extensions of FDCs rather than free vesicles as we found. The microvesicles we describe here contain MHC class II molecules as well as other proteins not expressed by FDCs. The striking differences in protein composition between the microvesicles and the FDC plasma membrane make it very unlikely that the microvesicles originate from FDCs, but rather resemble exosomes (22). In an in vitro binding assay, we show that isolated B cell exosomes specifically bind to FDCs.

FDCs have been reported to passively acquire MHC class II molecules (5), and the authors suggested that shed material from surrounding B cells (42) might bind to the FDC surface. However, the morphological basis for this phenomenon was not further investigated. We now show that FDCs bear MHC class II-containing microvesicles with a size of 40–70 nm at their surface. These microvesicles were different from FDC-derived iccosomes (40). Iccosomes are pinched off from the dendritic processes of FDCs during the early phase of a germinal center reaction and therefore should contain the same proteins as the FDC plasma membrane. This is not the case for the microvesicles. Furthermore, the diameter of iccosomes is 0.25–0.4 μm, which is ∼5 times larger than the microvesicles attached to FDCs described in this study. The presence of MHC class II molecules at the cell surface of mouse FDCs has been demonstrated before at the ultrastructural level using the immunoperoxidase technique (43). This technique resulted in a diffuse staining at the FDC surface and did not allow the discrimination between peroxidase deposits on the FDC plasma membrane and adjacent vesicles. On the basis of our findings that microvesicles attached to the FDC surface contain MHC class II, we suggest that these earlier data show staining of exosomes rather than surface staining of FDCs. This interpretation is more consistent with bone marrow transplantation experiments published by Gray et al. (5), showing that FDCs passively acquire MHC class II molecules from the donor.

B cell apoptosis is an abundant event in lymphoid follicles and may result in membrane fragments that in some way or another might bind to FDCs. However, the FDC attached microvesicles do not represent B cell remnants. They are of rather uniform size and are enriched in lysosomal markers and LBPA, a lipid highly enriched in internal membranes of multivesicular endosomes (41).

We have previously demonstrated that B cells secrete microvesicles called exosomes (16) that are characterized by the presence of MHC class II, the costimulatory molecule CD86, and several tetraspan proteins (CD37, CD53, CD63, CD81, CD82) (22). The protein composition of microvesicles attached to FDCs turned out to be identical with that of B cell-derived exosomes (22) (Table I). All tetraspan proteins present on exosomes were also found on the microvesicles. Together, these data strongly suggest that the microvesicles on FDCs represent exosomes. Because FDCs are in close contact with B cells, these latter are likely candidates for exosome delivery to FDCs. Although the molecular basis of this interaction has not yet been elucidated, we found that exosomes bind specifically to FDCs as shown in vitro (Table II). Tetraspan proteins are enriched on exosomes and microvesicles attached to FDCs. Their precise function is largely unknown, and it is tempting to speculate that they might play a role in the attachment of microvesicles to FDCs. Tetraspan proteins are able to form oligomeric complexes with other tetraspan proteins (23, 24, 25), HLA-DR (23, 26) and to interact with integrins (25, 27) and the T cell coreceptors CD4 and CD8 (28). CD81 is part of the CD19/CD21/CD81-signaling complex, which is proposed to lower the threshold for B cell responses (44). CD81 acts as a costimulatory molecule, which is important for the optimal stimulation of Ab production (45). CD82 and CD63 have been reported to form complexes with MHC class II, HLA-DM, and HLA-DO, which reside in MIICs, suggesting an important role for these tetraspan proteins in peptide loading of MHC class II (26). HLA-DM is restricted to the limiting membrane of MIICs and is not present on the internal vesicles of MIICs and their secreted equivalent, the exosomes (22, 26). The same holds true for the lysosomal membrane proteins Lamp-1 and Lamp-2. How internal vesicles of MIICs are formed and how proteins are sorted into or excluded from the vesicles is not known, but recent data showed that multivesicular body morphogenesis seems to require phosphatidylinositol 3-kinase activity (15).

Exosomes have been discussed earlier to have a costimulatory function in vitro (16), and in previous studies we have shown that the MHC class II molecules on exosomes are in a SDS-stable configuration, implying that they are peptide loaded (16). Indeed, they can specifically stimulate CD4-positive T cells in vitro (16), but their physiological target has remained undetermined. We have shown that FDCs display microvesicles at their surface, which strongly resemble exosomes. What could be the function of exosomes attached to FDCs? Ready-to-use peptide-loaded MHC class II molecules on exosomes might play a role in Th cell recruitment into the apical light zone of the germinal center. In this microenvironment, the last check for Ag specificity of B cells during germinal center reaction takes place, namely, the isotype switching and differentiation into either plasma cells or memory B cells (4), which also involves T cell help. It has been demonstrated that CD4-positive memory T cells are present in the outer zone of germinal centers (46) and that coculture of these T cells with germinal center B cells rapidly induces a memory phenotype in B cells (47). However, it is unclear how Ag-specific T cells manage to find the selected candidate memory B lymphocytes in the three-dimensional structure of the germinal center. We hypothesize that exosomes, which display peptide-loaded MHC class II molecules, attach to the highly mobile processes of FDCs, which could provide guidance for Ag-specific T cells from the outer zone of the follicle into the apical light zone. These exosomes could be derived from B cells in the germinal center, but might also originate from DCs (18, 19). If exosomes on FDCs have the function to guide Ag-specific T cells for costimulation of B cell differentiation to the place of action, one should assume that this step must be tightly regulated to avoid constant stimulation. Clearance of exosomes and their half-life in vivo are therefore interesting subjects of further investigation. According to observations in our in vitro binding experiment, macrophages might play a role in removing exosomes, because we found biotin-labeled microvesicles throughout their endocytic system (not shown). Due to the fact that half-life determination in humans is impossible, further functional experiments are crucially dependent on the availability of a murine B cell line that generates sufficient exosomes. Future experiments in the mouse system will probably give more insight in the role of exosomes in vivo.

We thank Dr. H. L. Ploegh for providing the anti-HLA-DR polyclonal Ab and anti-MHC class I heavy chain polyclonal Ab, Dr. P. A. Morton for the generous gift of anti-Ii Abs, Dr. J. Gruenberg for the monoclonal anti-LBPA Ab. Special thanks to Jean-Michel Escola for his introduction into the exosome work and to Janice Griffith for technical support. R. M. C. Scriwanek, T. van Rijn, and M. K. Niekerk are gratefully acknowledged for their photographic work.