Surface LAMP-2 Is an Endocytic Receptor That Diverts Antigen Internalized by Human Dendritic Cells into Highly Immunogenic Exosomes

Lysosome-associated membrane protein-2 (LAMP-2) is a member of a family of structurally related type 1 membrane proteins that also includes LAMP-1, DC-LAMP, BAD-LAMP, LIMP-2, and CD68 (1–4). LAMP family proteins are characterized by a heavily glycosylated luminal (extracellular) domain; these contain a flexible hinge-region and a unique pseudo β-prism fold, and are linked to a single-spanning intramembranous domain and a short cytoplasmic tail, which contain retrieval and targeting signatures (5–7). These structural characteristics in combination with their surface expression make LAMPs suitable to transport exogenous molecules from the plasma membrane to lysosomes.

The canonical members, LAMP-1 and LAMP-2, are prominently expressed in lysosomes whose integrity they maintain by forming a glycocalix that protects the limiting membranes from enzymatic degradation (4, 8, 9), and by facilitating lysosomal fusion with other vesicles (6, 7, 10). However, LAMP-2 has additional critical functions, because its genetic deficiency causes severe disease both clinically (Danon disease) and in mice (11–13), whereas LAMP-1–deficient mice are healthy (14). LAMP-2 has three splice variants (A, B, and C) with isoform specific differences in their cytoplasmic tails that affect their function (5–7, 15): LAMP-2A is essential for chaperone mediated autophagy (CMA) (16, 17), a process necessary for MHC class II presentation of cytoplasmic Ags to CD4 T cells (18, 19); LAMP-2B (together with LAMP-1) stabilizes the ABC transporter TAPL in the lysosomal membrane (20); and LAMP-2C is an endosomal receptor for free cytoplasmic RNA and DNA that also inhibits CMA (21, 22).

Although most abundant in lysosomes, LAMP-1 and LAMP-2 are both expressed at low levels on the cell surface as a result of both direct trafficking of nascent protein to the plasma membrane and the transfer from late endosomes and lysosomes that fuse with it (23–25). The intensity of surface expression depends on the cell type and state of activation (26), and also varies between species. LAMP-1 and 2 are especially prominent on the surface of activated leukocytes (where they have been designated CD107a and CD107b, respectively), including primary monocytes, macrophages (27), as well as on THP-1 cell line (28). Finally, LAMP-1 and LAMP-2 are also incorporated into the extracellular vesicles (EV) released from murine dendritic cells (DC) (29, 30).

The function of surface LAMP-1 (CD107a) and LAMP-2 (CD107b) on leukocytes is uncertain but they have been invoked in cell-cell adhesion (27, 28, 31), and in protecting the plasma membrane from enzymes released after leukocyte degranulation (32, 33). By contrast, other LAMP family members have well-characterized roles as endocytic receptors (1–3), for example, DC-LAMP facilitates Ag uptake by DC, trafficking to the MHC class II Ag loading compartment (MIIC) and augmenting CD4 T cell responses (1, 34), whereas CD68 is a surface receptor for polyanionic ligands on monocyte-derived macrophages and DC (3). Notably, the absence of LAMP-2 abrogates the ability of B cell lines to present soluble human serum albumin and induce proliferation of a specific CD4 T cell hybridoma (18, 19); a phenomenon distinct from the effect of LAMP-2 on CMA. This raises the question of whether, like other family members, surface LAMP-1 and LAMP-2 are endocytic receptors that affect Ag presentation.

We addressed this using specific Abs to LAMP-1 and LAMP-2 and Ags conjugated to them, a strategy previously applied successfully to characterize DEC-205, DC-SIGN, and DCIR as immune receptors on DC (35–37). As with known DC endocytic receptors, surface ligation of LAMP-2, but not LAMP-1, induced rapid internalization and trafficking to the MIIC, as did Ag associated with it. Despite this, monocyte-derived dendritic cells (MoDC) pulsed with Ag conjugated to H4B4—an mAb to LAMP-2—expressed significantly fewer Ag-derived peptides in the HLA class II peptidome and evoked significantly less T cell proliferation. Instead, Ags internalized with LAMP-2 are selectively routed into highly immunogenic exosomes (EXO) that stimulate robust CD4 T cell responses. Thus, we have identified LAMP-2 as an endocytic receptor that routes exogenous Ags into a novel pathway that generates highly immunogenic EV. This has implications for the development of immune responses both locally and at distant sites, and for diseases in which autoantibodies to LAMP-2 have been implicated.

PBMC were isolated from healthy volunteers and an individual with genetic deficiency of LAMP-2 (Danon disease) using protocols approved by the Medical University Vienna Ethics Committee (ECS 1089/2012). Monocytes and CD4⁺ naive T cells were purified with microbeads (130050201-130094131; MACS Miltenyi Biotec). Monocytes were differentiated into MoDC by culturing them for 5–7 d in GMP serum-free DC cell growth medium (20801-0100; CellGenix) and 1% PenStrep (15140122; Life Technologies) in the presence of 50 ng/ml IL-4 (200-04; PeproTech) and 50 ng/ml GM-CSF (400-011 DC; CoaChrom diagnostic). MoDC were matured in the presence of 50 ng/ml IFN-γ (300-02; PeproTech) and 100 ng/ml ultrapure LPS (ALX-581-007; Enzo Life Sciences) for 48 h.

Resident splenic DC were isolated from C57BL/6 mice by 15 min digestion with collagenase (CLS-2; Worthington) before stopping the reaction with PBS/EDTA (BE02-017F; Lonza). Splenic DC were isolated from a single-cell suspension by positive selection with CD11c microbeads (130-097-059; MACS Miltenyi Biotec). Mouse bone marrow–derived DC were generated by culturing single-cell suspensions isolated from bone marrow for 8 d in RPMI 1640 supplemented with 20 ng/ml recombinant GM-CSF (315-03; PeproTech), the medium was replaced every other day. Both types of DC were activated by incubation for 24 h with 100 ng of ultrapure LPS (Tlrl-3pelps; Eubio).

The ldl receptor–deficient Chinese hamster ovary cell line (ldlD) is one of the standard cell lines used for studying glycoproteins expressed on the plasma membrane (38, 39). Cells stably transfected with full-length human LAMP-2 cDNA with a tyrosine to histidine mutation in the cytoplasmic motif leading to exclusive LAMP-2 expression on the cell surface (ldlD /hLAMP-2H) have been previously described (40–42).

The mouse IgG1 mAb H4B4 specific for human LAMP-2 was purchased from Developmental Studies Hybridoma Bank and an isotype-matched control IgG1 from BioLegend (400153). Flow cytometry mAbs were as follows: αHLA-DR (130-095-295) was from MACS Miltenyi Biotec; αCD4 (120048), αLAMP-2 (354304) and αCD3 (300420) were from BioLegend; αCD80 (MHCD8004), αCD14 (47-0149-42), αCD83 (25-0839-42), αCD70 (46-0701-82) αCD86 (25-0869-42) were from eBioscience; and αLAMP-2 (ab178546) was from Abcam.

Primary Abs for immunofluorescence and immunoblotting were as follows: αHLA-DM (EPR7981), αLAMP-1 (1D4B, ab25245), αEEA1 (ab2900), α-mouse-LAMP-2 (ABL-93, Ab25339), CD63 (ab118307), HLA-DR (ab92511) from Abcam; αHLA-DR (MCA71R) from AbD Serotec; αLAMP-2 (PAB13805) from Abnova; αKLH (H0892) from Sigma-Aldrich; ICAM-1 (NB110-60977) from Novus Biologicals; and rabbit anti human LAMP-2 (932b) (41).

MoDC were seeded on to adhesion slides (0900000; Marienfeld), fixed with 3% PFA, and permeabilized with methanol before incubation with primary and secondary Abs. As fixation and permeabilization of MoDC renders surface LAMP-2 undetectable, H4B4 Fab fragments incubated with live MoDC on ice were used for its detection before fixation, permeabilization, and detection of intracellular LAMP-2 as described above. Confocal images were acquired with Zeiss LSM 700, with a 63× objective. Colocalization was analyzed with an ImageJ plug-in and Just another Colocalization plug-in. Integrated intensity signals were calculated by ImageJ using the formula: (mean fluorescence intensity of selected cells − background intensity) × number of stacks. The colocalization was quantified independently by two investigators (D.A.L. and M.G.).

Cells were harvested, resuspended in MACS buffer (130-091-221; MACS Miltenyi Biotec), and kept on ice to minimize cell death and prevent endocytosis. The cells were incubated for 15 min with Fc-block (130-059-901; MACS Miltenyi Biotec) before being washed and stained with the indicated Abs for 30 min; the samples were acquired with a BD Canto II using Diva software and analyzed with FlowJo. Duplets were excluded using FSC-A/FSC-H, whereas DAPI was used to exclude dead cells.

Total RNA was extracted from immature MoDC (iDC) and mature MoDC (mDC) with RNeasy mini kit (74104; Qiagen), quantified by NanoDrop spectrophotometer and used to prepare cDNA with SuperScript III reverse transcriptase (18080-051; Invitrogen). Quantitative PCR was performed using KAPA SYBR FAST master mix (KR0389; KAPA BIOSYSTEM) in combination with forward (5′-TGAAGGATCCCTGAACATCACTCAGGATAAGGT-3′) and reverse (5′-AAACTCGAGCCATTAACCAAATACATG-3′) primers specific for LAMP-2 (VBC Biotech) and GADPH (forward: 5′-GAGATGGCACGGGACACTACCT-3′ and reverse: 5′-GTGGTGGTCTGACAGTTCGC-3′) (VBC Biotech). The samples were analyzed on a qTOWER (3) G device (Analytik Jena) and the relative level of LAMP-2 expression was normalized to the level of GADPH expression using the double δ cycle threshold (2^−ΔΔCT) method.

Abs to LAMP-1, LAMP-2, and the isotype control mouse IgG1 were incubated at 20 μg/ml with MoDC for various time periods, either at 37 or 4°C. The fraction of internalized Ab was calculated by subtracting the membrane signals obtained in the samples at 4°C from the signal obtained at 37°C. Where indicated, the cells were preincubated with Fc-block (130-059-901; MACS Miltenyi Biotec) for 20 min.

The Ags keyhole limpet hemocyanin (KLH) and Bet v 1 were conjugated to the monoclonal Abs using the heterobifunctional cross-linker maleimide. Modified mcKLH (77610; Thermo Fisher Scientific) was incubated with SATA-modified Ab. Due to the size differences between the two molecules—KLH (450 kDa) and IgG1 (150 kDa)—and to ensure an efficient delivery of the conjugate through LAMP-2, a 3-fold excess of Ab was used in the reaction. The superficial charge and the diameter of the KLH conjugates were analyzed by dynamic light scattering (DLS) (43) using a Malvern Zetasizer Nano ZS instrument fitted with a 532 nm laser at a fixed scattering angle of 173°, using the Dispersion Technology Software 7.02 to ensure their uniformity and to exclude the presence of aggregates. Bet v 1 (110707; Biomay) was modified with maleimide and covalently attached to the SATA-modified Abs using a controlled protein-protein cross-linking kit (23456; Thermo Fisher Scientific). Due to the size of Bet v 1 (48 kDa) and IgG1 (150 kDa), three times more Ag was used.

BSA over-labeled with boron-dipyrromethene (BODIPY) (12050, BSA-DQ; Molecular Probes) was used to monitor the proteolysis of BSA in MoDC pretreated with H4B4 or ISO, and a similar strategy was used to measure proteolysis of H4B4*KLH and ISO*KLH. The conjugates were labeled with an excess of amine-reactive BODIPY FL NHS dye (D2184; Molecular Probes) to generate a DQ substrate that was then used to quantify their proteolysis in a pulse and chase assay.

Naive CD4 T cells were stained with Cell Proliferation Dye 670 (65-0840-85; BioLegend) and cocultured with MoDC at a 4:1 ratio in the presence of 2% autologous serum together with IFN-γ at 50 ng/ml (300-02; PeproTech), native KLH (5 μg/ml) (77605; Thermo Fisher Scientific), H4B4*KLH, or ISO*KLH. The cells were harvested after 6 d and stained with Abs to CD3, CD4, and CD11c together with DAPI to quantify live proliferating CD4⁺ T cells. The absolute number of proliferating cells was calculated using Absolute Bright Count beads (C36950; Life Technologies) (44).

We quantified CD4 T cell proliferation to the major birch pollen allergen Bet v 1 using an optimized T cell proliferation assay (45, 46). Briefly, MoDC were incubated with Bet v 1 (110707 at 10 μg/ml; Biomay), ISO*Bet v 1 or H4B4*Bet v 1 in the presence of autologous Bet v 1–specific T cell lines (10 × 10⁴) (47). After a 16 h pulse with 0.5 μCi of thymidine, the cultures were harvested and radionuclide uptake was measured by scintillation counting.

The ProPresent assay was performed by Proimmune (Oxford, U.K.) to identify unique KLH peptides bound to HLA-DR molecules in Ag-pulsed MoDC, using Ags prepared by us. The MoDC were generated from monocytes isolated from four normal healthy blood donors with different HLA types (Table II). The levels of MHC complexes present in the DC lysates (pre- and postimmunoaffinity isolation) were assessed by ELISA (Table II), peptides were eluted from the purified HLA-DR molecules and analyzed by high resolution sequencing mass spectrometry. Within the pool of HLA-DR–bound peptides, protein sequence analysis software referencing the Uniprot Swiss-Prot Complete Human Proteome Database1 was then used to identify all possible unique KLH-derived peptides, as well as peptides derived from two endogenous control proteins, myeloperoxidase (MPO) and Heat Shock Protein 70 (Supplemental Fig. 2).

EV were harvested from the supernatants (sn) of MoDC that had been pulsed with various stimuli for 24 h. They were purified according to previously described methods by differential centrifugation (29) and by discontinuous density gradient (48). In each case, the pellets were resuspended in PBS, washed once, and the total protein content was measured by NanoDrop. Preparations containing equal amounts of EV proteins were used as an Ag source for T cell proliferation assays or were loaded on to 4–20% gradient SDS-PAGE gels for immunoblotting (EC6025; Thermo Fisher Scientific).

Total protein contents of bulk EV, microparticles (MP), and EXO were measured with NanoDrop to normalize their contents before loading on to an SDS-PAGE electrophoresis gel (XP04200; Thermo Fisher Scientific), followed by transfer on to a polyvinylidene difluoride membrane (Immobilon-FL, IPFL00010). Western blotting was performed as described by the manufacturer (LI-COR Bioscience) and images were acquired with a LI-COR Odyssey and quantified using ImageJ.

MP- and EXO-enriched fractions were fixed in 2% PFA and a 5 μl drop was placed on formvar-coated 200 mesh Ni-grids (S162N; Agar Scientific). KLH was detected in EV by immuno–electron microscopy after incubating the grids overnight in a humid chamber at 4°C with a specific αKLH Ab produced in rabbit (H0892; Sigma-Aldrich), which was detected with a gold-labeled secondary Ab (EM:GAR15, goat anti-rabbit 15 nm immunogold conjugate; BBI Solution). After air drying, the grids were examined in a JEOL JEM 1010 (JEOL, Japan) transmission electron microscope at 60 kV.

Confocal microscopy demonstrated abundant cytoplasmic LAMP-2 in iDC together with obvious expression on the cell surface (Fig. 1A). Quantitation of surface LAMP-1 and LAMP-2 by flow cytometry revealed they were similarly abundant on iDC (Fig. 1B), but that surface LAMP-2 was 8-fold greater (p = <0.0001) in IFN-γ/LPS mDC, whereas there was no significant change in surface LAMP-1 expression (Fig. 1B); maturation increased the proportion of total cellular LAMP-2 on the cell surface from 18% in iDC to 30% in mDC, as assessed by flow cytometry. Increased maturation had no effect on the LAMP-2 mRNA concentration measured by RT-PCR, which was not significantly different in iDC and mDC (quantitative PCR data – 2^−ΔΔCT of mDC/iDC = 0.6, p = 0.3). Accordingly, changes in the rate of newly synthesized LAMP-2 is unlikely to contribute to the observed differences between iDC and mDC surface expression.

Maturation induces extensive lysosomal remodeling (49) and this affected the cytoplasmic distribution of LAMP-2 within intracellular compartments, most notably within MIIC (Fig. 1C, 1D). Using combinations of Abs to LAMP-1, LAMP-2, HLA-DM, and HLA-DR, we showed that regardless of the maturation state, around 50% of LAMP-1 colocalized with LAMP-2 in MoDC (Pearson coefficient 0.5) (Fig. 1E) and around 70% of LAMP-1 colocalized with HLA-DM (Fig. 1F), confirming its value as an MIIC marker (34, 50, 51). By contrast, colocalization of LAMP-2 with HLA-DM never exceeded 40% in iDC but was consistently around 70% in mDC (Fig. 1G). The images and the quantification relate solely to cytoplasmic LAMP-2 because H4B4 does not detect surface LAMP-2 after fixation and permeabilization. Together, these results show that LAMP-1 and LAMP-2 occupy distinct but overlapping intracellular compartments in MoDC and are differentially affected by maturation.

Confocal microscopy showed that iDC internalized small amounts of the isotype control mAb (Fig. 2A) that was abrogated when Fc receptors were blocked (Fig. 2B). By contrast, MoDC internalized obviously larger amounts of H4B4, an mAb specific for LAMP-2, despite Fc-blockade and regardless of activation status; quantitation by flow cytometry confirmed these results (Fig. 2C, 2D), which are consistent with receptor-mediated uptake rather than pinocytosis (52). We confirmed the specificity of H4B4 uptake using MoDC from a LAMP-2 deficient individual with Danon disease caused by complete LAMP-2 deficiency due to a single base-pair deletion in exon 2 (c.179delC) that introduces a premature stop codon (12). Immature LAMP-2–deficient MoDC showed significantly more nonspecific uptake of mouse IgG than control MoDC, but again this was abrogated by maturation. By contrast, specific uptake of H4B4 was confined to LAMP-2–sufficient cells and was unaffected by maturation (Fig. 2E, 2F): EBV-transformed B cells from the Danon patient also failed to take up H4B4 (Fig. 2G). Notably, the mAb to LAMP-1 (1B4D) was internalized no better than the isotype control, indicating that surface LAMP-1, unlike LAMP-2, is not specifically taken up after ligation (Fig. 2H). Ligand-mediated cross-linking is commonly required for receptor internalization, and we tested whether this was the case for LAMP-2 by comparing the uptake of monovalent Fab fragments of H4B4 with that of divalent whole Ab. Human MoDC were incubated for various time periods with either H4B4 or its Fab fragments and then fixed and permeabilized before intracellular Igs were quantified by flow cytometry using either Fab-specific (αFab-Alexa Fluor 488) or Fc-specific (αFc-Alexa Fluor 488) Abs. Results for αFab-Alexa Fluor 488 show identical uptake of intact H4B4 and the Fab fragment preparations (Fig. 2I), whereas staining with αFc-Alexa Fluor 488 failed to detect intact Ab in the MoDC pulsed with the Fab fragment preparations (Fig. 2J). This confirms that monovalent fragments, rather than residual whole Ab, had been taken up from these preparations and that cross-linking of LAMP-2 on the cell surface was not required for internalization.

In contrast to human MoDC, LAMP-2 was not detected on the surface of CD11c⁺ murine splenocytes, which confirms an early report using an mAb to MAC3 (CD107b) (47) which is LAMP-2. Moreover, the rat anti-mouse mAb ABL-93 was not internalized by splenic- or bone marrow–derived mouse DC (data not shown).

In pulse-chase experiments, H4B4 trafficked to early endosomes after 30 min as indicated by colocalization with EEA1 (Fig. 3A), and to MIIC after 60 min as demonstrated by colocalization with HLA-DM (Fig. 3B, Supplemental Fig. 1A); quantitation of these results is shown in Fig. 3C. However, colocalization with HLA-DM was transient and barely detectable after 3 h (Fig. 3B). The uptake and subsequent trafficking to MIIC was not unique to H4B4 as we obtained identical results with two other αLAMP-2 Abs: the mouse anti-human mAb CD3 (4) (Supplemental Fig. 1B), and the rabbit polyclonal Ab 932b (53) (Supplemental Fig. 1C). The kinetics of internalization and trafficking of Abs to LAMP-2 mimics those of Abs specific for known endocytic receptors including DEC-205, DC-SIGN, and DCIR. Ags conjugated to monoclonal Abs specific for these receptors have enhanced ability to stimulate CD4 T cell proliferation (1, 50, 54, 55), so we tested whether conjugation to H4B4 had the same affect.

These studies were performed using KLH, which is a large primary Ag (molecular mass 450 kDa) that evokes specific proliferation of naive CD4 T cells (56), and was previously shown to be suitable for characterizing endocytic receptors on MoDC (36, 37). KLH was conjugated to H4B4 or isotype-matched control mAb using a hetero-bifunctional cross-linker. The resulting conjugates were homogeneous in size and free from aggregates, as shown by DLS (Table I), and bound specifically to recombinant human LAMP-2 stably expressed in the plasma membrane of a well-characterized cell line (ldlD /hLAMP-2H) (40, 41), but not to the surface of the untransfected parent ldlD cell line (Fig. 4A). We confirmed that H4B4*KLH conjugate bound specifically to LAMP-2 on the surface of MoDC because pretreatment with free H4B4 reduced binding by over 80%, whereas binding of ISO*KLH was unaffected (Fig. 4B). MoDC pulsed with KLH or ISO*KLH stimulated vigorous proliferation of naive CD4 T cells, whereas MoDC pulsed with H4B4*KLH did not, with the absolute number of proliferating T cells being similar to nonstimulated controls (Fig. 4C). By contrast, MoDC incubated with native KLH plus free H4B4 proliferated normally, indicating the inhibitory effect of H4B4*KLH was due to the specific properties of the conjugate and not simply to the presence of Abs to LAMP-2 (Fig. 4C).

To ensure that the phenomenon was not unique to KLH, we repeated the experiments with a second Ag, Bet v 1, which is a 48 kDa protein responsible for allergy to birch pollen that we have previously used to characterize Ag specific T cell lines generated from allergic subjects (45, 46). Like KLH, Bet v 1 was conjugated to H4B4 (H4B4*Bet v 1) or to an isotype control Ab (ISO*Bet v 1). H4B4*Bet v 1 bound specifically to ldlD/hLAMP-2H cells (Fig. 4D), and the appropriate size of the conjugates was confirmed by immunoblots (data not shown). In this case, we measured T cell proliferation with a standardized thymidine incorporation assay routinely used for monitoring specific T cell lines isolated from individuals allergic to birch pollen (57). Bet v 1–specific T cell lines derived from three different highly allergic donors were stimulated with MoDC pulsed with native Bet v 1, ISO*Bet v 1, or H4B4*Bet v 1. The native and the isotype-conjugated Ags induced robust proliferation, whereas proliferation induced by MoDC pulsed with H4B4*Bet v 1 was significantly less (Fig. 4E). Thus, conjugation to the LAMP-2–specific mAb H4B4 significantly reduced proliferation of both peripheral blood CD4 T cells in response to a large primary Ag and short-term T cell lines specific for a secondary Ag.

To gain an insight into why Ags internalized by LAMP-2 in MoDC do not induce robust immune responses, we compared uptake and trafficking of H4B4*KLH and ISO*KLH in pulse-chase experiments. KLH was internalized significantly faster and trafficked more rapidly to the MIIC when MoDC were pulsed with H4B4*KLH than with ISO*KLH (Fig. 4F). However, colocalization with HLA-DM was also more transient: after 30 min it was greater in H4B4*KLH-pulsed cells but thereafter the difference decreased steadily until at 6 h its colocalization was significantly less than in ISO*KLH pulsed cells. Consequently, 2.5 times more KLH from H4B4*KLH accumulated over the first 30 min in MIIC (58, 59) than from ISO*KLH; the amounts were similar after 60 min (Fig. 4G); there was significantly more ISO*KLH after 6 h.

We next determined whether the markedly different kinetics of Ag accumulation in MIIC affected loading of KLH-derived peptides on to HLA-DR molecules displayed at the cell surface using the ProPresent assay (Proimmune, Oxford, U.K.). Separate aliquots of MoDC from four healthy donors (Table II) were pulsed with KLH and H4B4*KLH before HLA-DR molecules were purified from them and bound peptides analyzed by mass spectrometry. Specifically, we characterized all the KLH-derived peptides within this peptidome as well as peptides from two endogenous proteins used as controls, MPO and Hsc70 (the complete list of peptides identified is provided in Supplemental Fig. 2). In all four donors, fewer unique KLH-derived peptides were recovered from MoDC pulsed with H4B4*KLH than from those pulsed with native KLH, whereas the number of MPO- and Hsc70-derived peptides were similar (Table III). When the results from all four donors were combined, there was a highly significant specific reduction in the number of HLA-DR–bound KLH-derived peptides in MoDC pulsed with H4B4*KLH compared with those pulsed with KLH alone (KLH peptides: H4B4*KLH pulsed 8; KLH pulsed 32; control peptides: H4B4*KLH pulsed 12; KLH pulsed 11, p = 0.0093, Fisher exact test).

The specific reduction in H4B4*KLH-derived KLH peptides could either be due to the reduced time of residence of KLH in MIIC (Fig. 3C), or alternatively to a generalized inhibition of proteolytic activity in the MoDC induced by anti–LAMP-2 Ab activity in the H4B4*KLH conjugate. To exclude the latter, we tested the effect of H4B4 on the proteolysis of an independent exogenous protein internalized by MoDC, namely a commercially available substrate in which BSA is overstained with BODIPY, generating a self-quenched dye (BSA-DQ) that becomes fluorescent upon proteolysis. Pretreatment of MoDC with H4B4 or ISO had no effect on proteolysis of BSA-DQ, excluding a general effect on enzymatic digestion (Fig. 4H). We next used the same strategy to measure proteolysis of the two conjugates by over-labeling them with BODIPY and using the resulting DQ substrate to quantify their proteolysis in pulse and chase experiments. There was little difference in the degree of digestion after 1 h, but by 3 h twice as much ISO*KLH had been digested as H4B4*KLH (Fig. 4I). These results suggest that the reduced proteolysis is due to the transient localization of KLH to MIIC after LAMP-2–mediated uptake (Fig. 3C) and raises the question of whether trafficking of H4B4*KLH deviates from MIIC and is rerouted into EV after internalization.

After peptide loading in MIIC, most HLA-DR molecules traffic to the cell surface, although some are incorporated into EV. To determine whether conjugation to H4B4 influences this partitioning, MoDC were incubated with either H4B4*KLH or ISO* KLH in the presence of IFN-γ for 24 h. Incubation with H4B4*KLH did not increase the number of apoptotic or necrotic MoDC, as assessed by annexin-V and DAPI staining respectively, or the expression of costimulatory molecules (Supplemental Fig. 3). Notably, surface HLA-DR expression by these MoDC was similar to the unstimulated control and significantly reduced compared with ISO*KLH cells, as was surface LAMP-2 that was quantified using a rabbit anti-human LAMP-2 Ab (932b) that does not cross-inhibit H4B4 (Fig. 5A). Instead, bulk EV purified from the sn of MoDC (48, 60) pulsed with H4B4*KLH contained significantly more mouse IgG and KLH as well as HLA-DR and ICAM-1 (Fig. 5B). Further experiments performed with EBV-transformed B cell lines from normal and LAMP-2–deficient individuals confirmed that enrichment of KLH in EV from H4B4*KLH-treated cells was LAMP-2 dependent and not unique to MoDC (Supplemental Fig. 4).

EV carrying Ags as cargo and expressing MHC class II on their surface can stimulate CD4 T cell responses (61–63), so we tested the potential of EV from H4B4*KLH-pulsed MoDC to directly stimulate T cells in contact proliferation–independent assays. MoDC were incubated with H4B4*KLH or ISO*KLH either in a conventional proliferation assay or in a transwell system in which the stimulating MoDC in the lower chamber were separated by a 0.4 μm membrane from responding CD4 T cells in the upper chamber. As before, T cell proliferation in the contact-dependent assay was significantly attenuated in the H4B4*KLH-stimulated cultures (Fig. 5C); and the T cells in them did not upregulate surface HLA-DR, used as a late T cell activation marker (Fig. 5D). The contact-independent assay gave the opposite result and T cell proliferation as well as HLA-DR upregulation occurred exclusively in the H4B4*KLH-stimulated cultures (Fig. 5E, 5F). Identical results were obtained with three different donors in independent experiments. To ensure that the proliferation was induced by EV rather than soluble proteins, such as cytokines, in the sn we differentially centrifuged the sns from H4B4*KLH- and ISO*H4B4-pulsed MoDC to separate EV in the pellet from soluble molecules in the sn. These fractions were then used to stimulate unpulsed MoDC cocultured with naive CD4 T cells. Both the resuspended EV and the unfractionated sn (sn + EV) from H4B4*KLH-induced T cell proliferation, whereas the response to the sn depleted of EV was no greater than background. None of the fractions from MoDC pulsed with ISO-KLH induced proliferation (Fig. 5G).

EV are highly heterogeneous but fall into two broad classes that differ in origin, size, and function: MP that bud directly from the plasma membrane, and intraluminal vesicles formed within the MIIC and released as EXO (60, 64–69). Ag-pulsed DC release both types of EV, which contain HLA-DR and internalized Ags (70, 71). To define their respective contributions to the proliferative response, we differentially centrifuged sns from H4B4*KLH- and ISO*KLH-pulsed MoDC to obtain fractions enriched for either MP or EXO (Fig. 6A). The MP fraction contained abundant HLA-DR but little CD63—a classical EXO marker—whereas the EXO fraction was strongly positive for CD63 but contained less HLA-DR (Fig. 6B). Notably, the EXO fraction from H4B4*KLH-pulsed MoDC contained significantly more KLH and LAMP-2 and relatively more CD63 and HLA-DR on immunoblots than the EXO fraction from cells pulsed with ISO*KLH or native KLH alone or incubated with free H4B4. Furthermore, centrifugation of bulk EV over a discontinuous OptiPrep density gradient showed that KLH was concentrated in the EXO fractions with a density of 1.13–1.17 g/ml (Fig. 6C). Finally, we used immuno–electron microscopy to confirm the presence of KLH in EXO and its enrichment in those released by H4B4*KLH-treated MoDC (Fig. 6D).

We next tested the effectiveness of MP and EXO as a source of Ag in standard proliferation assays, showing that EXO from H4B4*KLH-pulsed MoDC induced naive CD4 T cell proliferation, whereas EXO from ISO*KLH did not (Fig. 7A); and that MP from these cells consistently induced low levels of proliferation, possibly due to a contamination from small numbers of EXO (Fig. 7A). Finally, we confirmed that the CD4 T cell response was specific for KLH in a restimulation assay in which naive CD4 T cells and MoDC were incubated with KLH and IFN-γ for 6 d before the CD4 cells were sorted into two separate groups: those that had proliferated (T-KLH) and those that had not (T-nil) (Fig. 7B). The T-KLH or the T-nil were cocultured with unpulsed MoDC and restimulated with MP or EXO from pulsed MoDC with H4B4*KLH or ISO*KLH, or with native KLH as a positive control (Fig. 7B). Restimulation with native KLH induced around a 2-fold increase of T-KLH over T-nil, and there was an equivalent difference when MP or EXO were used as restimulants, thus confirming the specificity of the response (Fig. 7B). Notably, EXO from H4B4*KLH-pulsed MoDC induced significantly greater proliferation of T-KLH than the other stimuli including MP from the same cells. Under the conditions used, neither EXO nor MP induced proliferation of T-KLH in the absence of unpulsed MoDC (Fig. 7C).

LAMP-2 is critical for the presentation of cytoplasmic Ags by human EBV-transformed B cell lines through a CMA-dependent process, and has a less well-defined role in presentation of exogenous Ags (18, 19, 22). In this study we show that LAMP-2 on the surface of human MoDC is upregulated after maturation and is rapidly internalized after ligation, a receptor-mediated process that is not inhibited in mDC (52), and traffics to the MIIC—properties it shares with established endocytic receptors (1, 37, 54, 72). However, unlike DC, Ags internalized by LAMP-2 fail to stimulate T cells in conventional proliferation assays and Ag-derived peptides are poorly represented in the HLA-DR peptidome present on the cell surface. Instead, they are rerouted into EV that are particularly effective at inducing CD4 T cell proliferation. This identifies a novel pathway whereby internalized Ags are incorporated into EV capable of widespread dissemination and potentially able to stimulate T cells locally and at distant sites.

Ag presentation is regulated by the activation status of the DC, a process that includes lysosomal remodeling and translocation of peptide-loaded class-II molecules to the cell surface. Maturation also increases surface expression of endocytic receptors on the DC and their abundance in MIIC (1, 35, 54). Once ligated, such receptors are rapidly internalized and traffic to MIIC, as does LAMP-2 when ligated with all three LAMP-2–specific Abs tested in this study; by contrast, LAMP-1 is not internalized by antibody after ligation. The three isoforms of LAMP-2 (A, B, and C) differ in function and distribution (6, 7). Unfortunately, it has proved impossible to determine which isoforms were expressed on the surface of MoDC because of cross-reactivity of the available isoform-specific Abs to LAMP-2. However, this is potentially important because in a model system utilizing chimeric LAMP-2, LAMP-2A and LAMP-2B were rapidly internalized after ligation by Ab, whereas LAMP-2C was not (11).

Ags conjugated to Abs directed against endocytic receptors traffic to the MIIC (35–37), as did H4B4*KLH (37, 52, 73). Despite this, H4B4*KLH and H4B4*Bet v 1 consistently evoked less CD4 T cell proliferation than native or isotype-conjugated Ags (ISO*KLH, ISO*Bet v 1). The inability of H4B4*KLH to stimulate CD4 T cells in conventional contact-dependent assays is most simply explained by the marked reduction of HLA-DR molecules displayed on the cell surface, combined with the specific reduction of KLH-derived peptides bound to them. By contrast, H4B4*KLH had no effect on the number of HLA-DR–bound peptides from two endogenous proteins, MPO and Hsc70, used as controls. Thus, on their surface MoDC pulsed with H4B4*KLH express HLA-DR molecules loaded with peptides derived from self-proteins, and only low levels of KLH-derived peptide, constituting a potentially tolerogenic signal for naive CD4 T cells. It is likely that H4B4*Bet v 1 also reduced the abundance of Bet v 1–derived HLA-DR and peptides, but we suggest the number was still sufficient to induce a low-level proliferation of short-term Ag Bet v 1–specific T cell lines generated from subjects highly allergic to birch pollen, unlike the naive T cells from PBMC used in the proliferation assay in response to H4B4*KLH.

An early report documented that EXO released by B cell lines and DC directly stimulated proliferation of T cell clones in vitro (62, 63), as did the EV released from H4B4*KLH-pulsed MoDC in our transwell experiments. More commonly, and despite expressing the molecules required to stimulate CD4 T cells (29, 61, 74), EV have been reported to require APC to induce proliferation unless they are chemically attached to cell-sized latex beads (30, 75). EV are heterogeneous but can be separated into two functionally distinct groups with different origins: EXO (20–100 nm), generated within the multivesicular bodies and released from them after fusion with the plasma membrane; and MP (100–500 nm), which are shed directly from the plasma membrane. MP and EXO differ in their molecular composition that reflect their origins and influence their immunological properties (60, 68). Our separation of EV into EXO- and MP-enriched fractions demonstrates that the EXO fractions are a much greater source of Ag than MP; and that EXO from MoDC pulsed with H4B4*KLH induce significantly more naive CD4 T cell proliferation than MoDC pulsed with other KLH formulations. Lastly, as others have reported (30, 62), the purified EXO in these experiments only induced proliferation when additional MoDC were present, which contrasts with the effect of the heterogeneous unfractionated EV released from H4B4*KLH-pulsed MoDC in the transwell assays; a phenomenon not previously documented. Potential explanations for the difference include the deleterious effects of the purification procedure on the capacity of EV to induce T cell proliferation; the synergistic effects that soluble factors present in the unfractionated samples of EXO and MP; and that the formation of complexes between EXO and MP have effects analogous to those of EXO conjugated to latex beads (75).

mAbs have commonly been used to characterize surface immune-receptors on DC, used either alone, as in the case of DEC-205 (35, 72, 76) and MMR (50), or cross-linked to the KLH as a primary Ag, as with DCIR (37) and DC-SIGN (36). These and many other studies of Ag presentation to CD4 T cells in man used KLH as a model Ag (77–79) because the repertoire of naive circulating T cells contains TCRs that respond specifically to it (56). Nevertheless, using Ag-specific T cell lines generated from subjects highly allergic to birch pollen, including the Bet v 1 Ag, we have confirmed that Ags internalized by LAMP-2 are deviated from surface presentation, thus impairing T cell proliferation. The basal proliferation induced by H4B4*Bet v 1 is likely to reflect a lower threshold for Ag-mediated proliferation of Ag-educated T cells in the short-term T cell lines than a requirement to stimulate a primary response in vitro by naive T cells purified directly from PBMC.

Ligation of Ags to specific Abs has been valuable for characterizing DC endocytic receptors, but identifying their natural ligands remains a formidable challenge. This is well illustrated by DC-SIGN: conjugation of an mAb to DC-SIGN increases the ability of KLH to induce T cell proliferation; Candida albicans (80) and myelin oligodendrocyte glycoprotein (81) are both internalized by DC-SIGN, whereas HIV particles use it as an adherence receptor but remain on the surface to enhance virus dissemination (82). The oligosaccharide chains present in the extracellular domain of LAMPs are obvious candidates as receptor sites able to recognize and discriminate lectin-like molecules expressed either at the surface of microorganisms and self-antigens (83). Currently, galectin-3 is the only well-characterized ligand for the extracellular domain of LAMP-2 (31). This 30 kDa lectin binds to microorganisms and apoptotic neutrophils, increasing their phagocytosis (84–87). It is not yet known whether enhanced uptake in this setting involves LAMP-2 or other galectin-3–dependent endocytic pathways (88).

Lastly, we show that LAMP-2 is not expressed on the surface of murine DC, confirming an early report (47), and appears to be part of a pattern emerging for the LAMP family proteins because both DC-LAMP (89) and BAD-LAMP (2, 90) are expressed on the surface of human conventional DC but not of mouse DC. The difference in LAMP-2 expression precludes the use of mouse models for analyzing the in vivo consequences of the in vitro effects we describe in this study.

In summary, we have identified LAMP-2 as an endocytic receptor on human MoDC that routes proteins into unusual Ag-processing pathways together with nascent HLA-DR and peptides away from the cell surface and into highly immunogenic EXO. This novel pathway has implications for the development of immune responses both locally and at distant sites, as well as for diseases in which autoantibodies to LAMP-2 have been implicated.

We thank Johannes Huppa (MedUni Vienna) for helpful discussions, Thomas Felzmann and Maria Carmen Visus-Miguel (Activartis) for providing elutriated monocytes and lymphocytes, Mariano Licciardi (Dipartimento di Scienze e Tecnologie Biologiche, Chimiche e Farmaceutiche, Università degli Studi di Palermo) for the analysis of the conjugates by DLS, and Irina Grigorieva for the setting for the RT-PCR experiment.

The authors have no financial conflicts of interest.