Soluble heat shock fusion proteins (Hsfp) stimulate mice to produce CD8+ CTL, indicating that these proteins are cross-presented by dendritic cells (DC) to naive CD8 T cells. We report that cross-presentation of these proteins depends upon their binding to DC receptors, likely belonging to the scavenger receptor superfamily. Hsfp entered DC by receptor-mediated endocytosis that was either inhibitable by cytochalasin D or not inhibitable, depending upon aggregation state and time. Most endocytosed Hsfp was transported to lysosomes, but not the small cross-presented fraction that exited early from the endocytic pathway and required access to proteasomes and TAP. Naive CD8 T cell (2C and OT-I) responses to DC incubated with Hsfp at 1 μM were matched by incubating DC with cognate octapeptides at 1–10 pM, indicating that display of very few class I MHC-peptide complexes per DC can be sufficient for cross-presentation. With an Hsfp (heat shock protein-OVA) having peptide sequences for both CD4+ (OT-II) and CD8+ (OT-I) cells, the CD4 cells responded far more vigorously than the CD8 cells and many more class II MHC-peptide than class I MHC-peptide complexes were displayed.
The Ag-stimulated responses of naive T cells are normally driven by peptide-MHC (pMHC)5 complexes displayed on APC, especially dendritic cells (DC). For CD4 T cells, the peptide components of these complexes commonly derive from extracellular (“exogenous”) proteins that are taken up by the APC and proteolytically fragmented in acidic endosomes; the resulting peptides are then loaded on class II MHC molecules to form pMHC-II complexes. For CD8+ T cells, in contrast, the peptide components usually derive from cytosolic proteins synthesized within the APC and cleaved by proteasomes; the resulting peptides are then loaded on class I MHC to form pMHC-I complexes. However, in the process termed cross-presentation, exogenous proteins are processed by DC and displayed as pMHC-I complexes (1, 2).
Cross-presentation probably plays a critical role in initiating CD8+ T cell responses to tumor and virus-infected cells and in establishing tolerance to some Ags (3). It is also often considered in vaccination strategies aimed at raising CD8+ T cell responses. Many of the mechanisms underlying cross-presentation are not well understood, however, particularly as they likely involve translocations of exogenous proteins across cell membranes.
In this study we examined the cross-presentation of some soluble Hsfp by murine DC. In these proteins, the C terminus of a mycobacterial heat shock protein 65 (hsp65) is extended by polypeptides (“fusion partners”), some over 100 amino acids in length. When injected into mice, they stimulate the production of CD8+ T cells that recognize pMHC-I complexes whose peptides derive from the fused polypeptide, implying that in vivo the injected Hsfp are cross-presented by DC to naive CD8 T cells. The potency of the resulting effector CD8+ T cells is evident from their ability to destroy tumor cells that are transfected with the Hsfp or the fusion partner (4, 5, 6, 7, 8).
To analyze cross-presentation of these proteins, we used two approaches. In one we followed the uptake and intracellular distribution in DC of a fluorescein-labeled Hsfp. In the other approach, the activation of naive T cells from TCR transgenic mice was used to monitor DC display of particular pMHC-I complexes in which the peptides were derived from the cross-presented protein. We show that the Hsfp enter DC by receptor-dependent processes via a receptor (or receptors) that likely belong to the scavenger receptor superfamily. Most of the internalized protein is transported to lysosomes, except for a small fraction destined to be cross-presented. This fraction exits early from the endocytic pathway and requires proteasomal activity and the TAP to form pMHC-I complexes. Because pMHC-I complexes that are recognized by the transgenic TCR are known, we were able to use synthetic peptides to estimate the efficacy of cross-presentation. In this way we could also compare the overall efficacy with which DC generated and displayed pMHC-I and pMHC-II complexes whose peptides derived from the same exogenous protein.
Materials and Methods
The hsp used to construct the Hsfp used in this study is Mycobacterium bovis (bacillus Calmette-Guérin) hsp65-2, a homologue of GroEL, the extensively studied hsp65 of Escherichia coli. These proteins are unrelated in amino acid sequence to the mammalian hsp (hsp70, gp96, calreticulin, and BiP) that carry noncovalent-bound hydrophobic peptides (9). The Hsfp used in this study were all produced in Escherichia coli and purified as described (6). In the Hsfp termed 65-P1, the P1 fusion partner is 34 amino acids in length, and contains the SIYRYYGL sequence, referred to as SIY. A CTL clone (2C) that recognizes SIY in association with Kb lyses Kb+ cells (EL-4) transfected with 65-P1 or P1, indicating that the SIY sequence is excised intracellularly and presented with Kb (6). In the Hsfp, 65-OVA, the hsp65 fusion partner is a hen OVA domain, 110 aa in length, containing two sequences of interest. One sequence, SIINFEKL (OVA257–264), is recognized with Kb by the TCR on CD8+ OT-I cells; the other, ISQAVHAAHAEINEAGR (OVA323–339), is recognized with IAb by CD4+ OT-II cells. In the Hsfp partnered with a nucleoprotein (NP) called 65-NP, the fusion partner is a domain (90 amino acids long) of influenza A virus (Puerto Rico/68) nucleoprotein, which contains the sequence ASNENMDAM recognized with Db by CD8+ T cells that express the F5 TCR.
The 65-P1 protein was almost completely excluded from a Superose 6 column, indicating an apparent molecular mass >3 MDa. It was included, however, in a TSK4000 column with a 10 MDa nominal exclusion molecular mass for proteins (data not shown). In view of the monomer’s apparent molecular mass (see below), individual aggregates probably consisted of around 40–120 65-P1 molecules. Nevertheless, the protein was soluble, i.e., it remained in the supernatant after centrifugation for 1 h at 100,000 × g.
To obtain monomeric 65-P1, the aggregated protein was incubated in 8 M urea for 4 h at 37°C and dialyzed against 5 mM sodium phosphate, pH 7.2, at 4°C. On Superose 6, the disaggregated protein retention time was unaffected by additional treatment with either 8 M urea or 6 M guanidine HCl or with 1% SDS or DTT, suggesting that it was monomeric. Stored in 5 mM phosphate, pH 7.2, at 4°C, it remained a monomer over the duration of the experiments described.
In contrast to the Hsfp, the retention time of unmodified (“native”) hsp65 on Superose 6 was unchanged on treating it with 8 M urea or 6 M guanidine HCl, or with 1% SDS, indicating that it was not significantly aggregated; its retention time (corresponding to an apparent molecular mass of 110 kDa) was greater than that of monomeric 65-P1 (apparent molecular mass 80 kDa). This and other differences between Hsfp, seen in preliminary determination of their circular dichroism spectra and thermal denaturation curves (data not shown), indicated that they differed in conformation from native hsp65.
The 65-P1 protein was labeled with 5-FITC (Molecular Probes) at pH 8.1 to an estimated level of 9–10 moles fluorescein/mole protein. Termed F65-P1, the FITC-labeled protein behaved the same as unlabeled 65-P1 in the binding and cross-presentation assays described below.
BSA (Boehringer Mannheim) was incubated for 2 h with an ∼2000-fold molar excess of maleic anhydride in borate buffer, pH 8.5–9, then dialyzed against water. From the molecular mass difference between unreacted and reacted BSA (MALDI-TOF mass spectrometry), over 80% of the 56 lysine residues in BSA appear to have been maleylated.
The MHC haplotype in all mice was H-2b. B6 and TAP−/− mice were purchased from The Jackson Laboratory. Breeder transgenic mice with the 2C TCR on a RAG1-deficient background were obtained from D. Kranz (University of Illinois, Urbana IL) (10); transgenic mice with the OT-I TCR (11) and OT-II TCR (12) were gifts from N. Hacohen, Whitehead Institute, Massachusetts Institute of Technology (MIT), and L. van Parijs (MIT); breeder mice with the F5 TCR (13) were a gift from D. Kioussis (National Institute for Medical Research, London, U.K.). Mice with a targeted deletion of a scavenger receptor (SR), either SR-AI/II−/− (14) or CD36−/− (15) were kindly provided by M. Freeman, Massachusetts General Hospital, and SR-BI−/− mice (16) was a gift from M. Krieger (MIT). All mice were kept in a specific pathogen-free mouse facility. Studies were performed according to institutional guidelines for animal use and care.
Bone marrow-derived DC were generated as described (17). In brief, bone marrow cells from the femur and tibia were cultured after lysing RBC at 106 cells/ml in RPMI 1640 medium (supplemented with 10% heat-inactivated FCS, 2 mM l-glutamine, 10 mM HEPES, 50 μM 2-ME, 100 U/ml penicillin, and 100 μg/ml streptomycin) containing 20 ng/ml murine GM-CSF (R&D Systems). Medium was replaced on day 2 and 4, and on day 6 the cells (immature DC) were harvested for use. As for the DC derived from the spleen (below), bone marrow-derived DC were purified by positive selection of CD11c+ cells using magnetic sorting (MACS system; Miltenyi Biotec). Cell purity was typically >97%, as assessed by CD11c and CD11b staining.
To obtain splenic DC, B6 mice were injected s.c. in the scruff of the neck with 2 × 106 B16-GM-CSF cells (18). After 7–10 days, when the tumor was ∼0.5 cm, spleen were harvested and DC were purified by magnetic sorting as were bone marrow-derived DC (as previously described). Cell purity was typically >95%, as assessed by staining for CD11c, CD11b, TCR, and B220.
Purified DC from spleen or bone marrow were plated at 4 × 105 cells in 0.2 ml/well of a 96-well plate in RPMI 1640 medium with protein or peptides at various concentrations. After incubation at 37°C for 2 h or other times when specified the cells were fixed with 0.5% paraformaldehyde (in PBS) for 15 min (room temperature) and then washed extensively with complete medium. A total of 1 × 105 naive T cells from lymph nodes of 2C, OT-I, or OT-II mice were then added to each well and incubated for 18 h at 37°C. The T cell response was assessed by determining the level of CD69 expression by flow cytometry (Fig. 1 A).
To evaluate the effects of various inhibitors, DC were preincubated with them (30 min at 37°C before adding protein or peptide) at conditions that did not diminish DC viability and function, as determined by two criteria: 1) exclusion of propidium iodide to the same extent in treated and untreated DC; 2) the activation response of naive T cells (2C or OT-1 or OT-II) in presentation assays was similar for inhibitor-treated and untreated DC that had been incubated with the appropriate synthetic peptide (SIY or SIINFEKL or ISQAVHAAHAEINEAGR).
Flow cytometry and Abs
Fluorochrome-conjugated Abs to CD8α, CD69, CD11c, CD11b, panTCR, B220, Vβ2, and Vβ5.1 were purchased from BD Biosciences. The clonotypic Ab 1B2, specific for the 2C TCR (19), was isolated from cultured 1B2 cells culture supernatants using protein A affinity chromatography, and fluorescein-labeled with FITC. In addition to the Abs, purified anti-FcRγIII/II Ab (anti-CD16/CD32; BD Biosciences) was added to each sample to prevent Fc-mediated binding to DC. Cells were stained in PBS containing 0.1% BSA and 0.1% NaN3, and at least 20,000 live cells (propidium iodide-negative) were acquired on a FACSCalibur (Becton Dickinson). Analysis was conducted using CellQuest software (Becton Dickinson). Binding of 65-P1 to DC was measured by incubation with FITC-conjugated 65-P1 (F65-P1) at 4°C for 40 min in PBS containing 0.1% BSA, 0.1% NaN3.
DC at 2.4 × 105 cells/well in eight-well chambered cover glasses (Lab-Tek) were kept at 37°C for 40 min to allow cell adhesion to the glass, and then cooled on ice for 40 min. F65-P1 was added at the indicated concentrations and incubated for 40 min at 37°C. When indicated, DiI low-density lipoprotein (LDL) or Lyso Tracker Red DND-99 (Molecular Probes) was also added. The cells were then chilled on ice for a further 30 min and unbound protein was washed away with ice-cold medium. Fresh medium was added and the cells incubated at 37°C for various times. The cells were then chilled on ice for 30 min and fixed overnight at 4°C in 2% paraformaldehyde in PBS, pH 7.2. When indicated, the plasma membrane was labeled postfixation with 2 μg/ml cholera toxin B labeled with Alexa Fluor 594, and nuclear DNA was labeled with 1 μM TO-PRO3 (Molecular Probes). Cells were mounted in 1 mg/ml p-phenylene diamine, 80% glycerol in PBS pH 7.9 and examined with a Zeiss LSM510 inverted microscope with a ×100 oil immersion objective.
Processing of soluble Hsfp by DC
Cross-presentation of the Hsfp termed 65-P1 (see Materials and Methods) was monitored by following the response of naive 2C cells to DC that had been incubated for various times with different concentrations of the protein. Increased expression of CD69 on 2C cells was observed with increasing protein concentration and longer incubation times (Fig. 1 A,B).
When aggregated and monomeric forms of 65-P1 were compared, cross-presentation was found to be almost 10 times more effective for the monomer (Fig. 1 F). Similar results were obtained with another Hsfp, 65-NP (see Materials and Methods). Like 65-P1, 65-NP was highly aggregated and similarly could be dissociated with 8 M urea. When the aggregated and disaggregated forms of 65-NP were compared, using lymph node cells from F5 mice to monitor display of the cognate pMHC complex (ASNENMDAM-Db, Ref.20), the disaggregated protein was cross-presented at least four times more effectively (data not shown). Based on these findings, monomeric 65-P1 was used for the assays described below, except when otherwise indicated.
LPS has been reported to stimulate cross-presentation by DC (21, 22). 65-P1 was contaminated with very little LPS (6 EU/mg protein) and was cross-presented equally well by DC from LPS-sensitive mice (B6) and LPS-insensitive mice (Tlr4-null B6/ScNCr) (data not shown and Ref.23). Addition of LPS increased cross-presentation of 65-P1 by B6 DC ∼2-fold (Fig. 1 D). However, no LPS was added in any of the experiments described below (beyond what was introduced as trace contaminant in Hsfp).
Activation of the naive T cells was dependent upon MHC haplotype of the DC. Thus, DC from H2b mice (B6) cross-presented 65-P1, but DC from H2k mice (C3H/HeN) did not (Fig. 1 C). This finding and the requirement for the TAP (see below) led us to focus on how 65-P1 enters DC and traffics to compartments involved in the protein cross-presentation.
Binding of 65-P1 to a scavenger receptor
To determine whether cell surface receptors are involved we examined the binding of fluorescein-labeled 65-P1 (F65-P1) by DC. At 4°C in the presence of NaN3, F65-P1 bound rapidly to DC: 70% of maximum binding was reached after 10 min and saturation at ∼60 min (data not shown). The extent of binding depended upon F65-P1 concentration and was maximal at 200–300 μg/ml (Fig. 2,A). Binding specificity was evident in competition experiments: various unlabeled Hsfp, including 65-OVA, mixed with F65-P1, substantially blocked F65-P1 binding to DC (Fig. 2,B), but a 30-fold excess of OVA had a negligible blocking effect. Interestingly, a 30-fold excess of unmodified (native) hsp65 also had no effect, whereas urea-treated native hsp65 had a very limited effect (Fig. 2,B). The findings suggest 65-P1, and the other Hsfp tested, bind to a saturable DC receptor (or receptors) that recognize a structural feature shared by the Hsfp but absent in native hsp65. This difference probably reflects the likely conformation difference between the Hsfp and native hsp65 (see Materials and Methods). Binding to the receptor is crucial for cross-presentation as shown by the competition of presentation of the SIY peptide in 65-P1 by 65-OVA but not by hsp65 (Fig. 2 C).
To identify the receptor, we tested various substances for ability to inhibit the binding of F65-P1 to DC. As seen in Fig. 2, D and E, maleylated BSA inhibited the binding. It also caused a similar reduction in the activation response of 2C T cells, indicating that binding to the receptor was crucial for cross-presentation of the protein (Fig. 2 F). Similar results were obtained with dextran sulfate and fucoidan (data not shown), which are polyanionic inhibitors of scavenger receptors, as is maleylated BSA (24). Taken together, all of these findings indicate that cross-presentation of 65-P1 is mediated by one or more receptors of the scavenger receptor superfamily.
A few members of the superfamily were evaluated with DC from mice having targeted deletion of a scavenger receptor gene: either SR-AI/II or SR-BI, or CD36. These DC, however, cross-presented 65-P1 as effectively as wild-type (B6) DC. We did not test LOX-1, which has been shown to bind human hsp70 (25). CD91, an LDL receptor-like protein, has been reported to serve as a receptor for mammalian hsp70 and gp96 (26, 27), but we observed no decrease in binding of 65-P1 to DC, or in its internalization and cross-presentation, in the presence anti-CD91 Abs or α2-macroglobulin, a CD91 ligand (data not shown). Thus, CD91 is unlikely to be involved in cross-presentation of 65-P1.
Endocytosis and intracellular trafficking of 65-P1
Confocal microscopy was used to track the internalization and intracellular distribution of F65-P1. When DCs were incubated with F65-P1 on ice to inhibit endocytosis, F65-P1 remained bound to the cell surface (Fig. 3,A). When unbound protein was washed away and the temperature was raised to 37°C, the protein entered the cells. At the earliest time points, the internalized F65-P1 was observed in small endocytic vesicles close to the cell surface. After 5–10 min it was seen in larger round endocytic vesicles (Fig. 3,B) and at 25 min the protein had mostly localized in perinuclear vesicles (Fig. 3,C). By 2 h the fluorescence from F65-P1 had faded from most of the cells and diffuse fluorescence could be seen (Fig. 3 D).
Using LDL as a marker for receptor-mediated endocytosis and transport to lysosomes (28), we incubated DC together with F65-P1 and labeled LDL (DiI LDL). After 5 min the proteins were seen in small endocytic vesicles close to the plasma membrane containing either F65-P1 (Fig. 4,A, green) or DiI LDL (Fig. 4,A, red), and also in some larger endosomal vesicles containing both F65-P1 and DiI LDL (Fig. 4,A, yellow). After 25 min most of the endocytosed F65-P1 colocalized with LDL in endosomal structures close to the nucleus (Fig. 4 B). The colocalization was still observed after 120 min at 37°C, at which time LDL had likely reached a mature lysosomal compartment. Similar results were obtained with the fluorescent weak base Lyso Tracker Red in place of Dil LDL (data not shown).
Cross-presentation does not require trafficking of endocytosed 65-P1 to late endosomes or lysosomes
Bafilomycin A1 and chloroquine inhibit endocytic processes that require endosomal acidification, including the development of late endosomes and lysosomes (29, 30). After 30 min at 37°C the amount of F65-P1 internalized by DC that were treated with these inhibitors was greatly reduced (by 64 ± 13% in cells treated with bafilomycin A1 and 54 ± 3% in those treated with chloroquine) (Fig. 4,E). The small amount of internalized F65-P1 in bafilomycin A1-treated cells was confined to small endocytic vesicles near the cell surface, often in a polar distribution; in the case of chloroquine, the reduced amount of internalized protein was confined to a few large endocytic vesicles (Fig. 4,D). Despite these pronounced effects on the total amount of internalized protein and its intracellular distribution, these inhibitors had no effect on the protein cross-presentation (Fig. 4 F).
Cross-presented 65-P1 may exit early from the endosomal pathway
Because cross-presentation of Hsfp did not depend on the protein transport to late endosome/lysosome acidic vesicles, we turned to inhibitors of endocytosis to examine earlier stages of the endocytic pathway. Chlorpromazine interferes with continuous formation of clathrin-coated pits (31). In chlorpromazine-treated DC, F65-P1 remained confined to vesicles resembling those seen in the first few minutes following incubation of DC with F65-P1 in absence of the inhibitor (Fig. 5,B). The total amount of internalized protein was also much reduced: after 30 min, chlorpromazine-treated DC contained 61 ± 10% less F65-P1 than untreated DC (Fig. 5,E). Hyperosmolar sucrose also inhibits clathrin-mediated endocytosis and obliterates coated pits (32), and DC treated with 150 mM sucrose showed a 60 ± 13% reduction in the internalization of F65-P1 after 30 min at 37°C (data not shown), in agreement with the chlorpromazine effect. Notably, however, chlorpromazine had no effect on protein cross presentation (Fig. 5 F). The findings show that receptor-mediated, clathrin-dependent endocytosis resulted in transport of the bulk of the endocytosed F65-P1 to lysosomes. However, the 65-P1 molecules destined for cross-presentation appear to enter DC by a clathrin-independent mechanism. We were unable to explore clathrin-independent mechanisms with agents that disrupt caveolae-dependent and lipid raft-dependent endocytosis, such as filipin, nystatin, and methyl-β-cyclodextrin, because DC treated with these agents had reduced levels of cell surface MHC class I.
DC cross-presentation of 65-P1 requires proteasomal activity and the TAP
Unlike DC from B6 mice, DC from B6 mice with a targeted deletion of TAP (TAP−/−) were unable to cross-present 65-P1 to naive 2C cells (Fig. 6,A). Cells from TAP−/− mice express reduced levels of cell surface MHC class I. However, these low levels were not sufficient to account for the ineffectiveness of TAP−/− DC because the addition of the SIY peptide largely restored the ability of these cells to activate naive 2C cells (Fig. 6 B).
In accord with the requirement for the TAP transporter, we found that several proteasome inhibitors (lactacystin, epoxomycin, and MG132) greatly reduced the ability of TAP+ DC to cross-present 65-P1 (Fig. 6 C). No reduction was seen, however, with a fourth proteasomal inhibitor, z-L3VS. This inhibitor was potent as indicated by its ability to block the presentation of the OVA peptide SIINFEKL by DC infected with a recombinant vaccinia virus (vaccinia-OVA) to naive OT-I cells (data not shown). The differing effects of these inhibitors is not surprising in view of their differential activities on proteasomal catalytic sites (33, 34). The requirements for TAP and proteasomal activity indicate that cross-presentation of 65-P1 depends upon the ability of the endocytosed protein to gain access to the DC cytosol.
Proteasome activity has recently been found to be required for endocytosis mediated by some receptors (IL-2, Fc) (35), and it may be that decreased cross-presentation of Hsfp by proteasomal inhibitor-treated DC is due not only to decreased proteolysis of internalized protein but in part to decreased endocytosis.
Phagocytosis of the receptor-bound protein
Scavenger receptors are known to participate in receptor-mediated endocytosis and phagocytosis. To determine whether cross-presentation of aggregated 65-P1 was dependent on phagocytosis, we examined the effect of cytochalasin D, which blocks actin polymerization and phagocytosis. A comparison of cytochalasin D-treated and untreated DC showed that the treated DC were far less effective than the untreated cells in cross-presenting the aggregated protein, indicating that the receptor-bound aggregated 65-P1 was phagocytosed (Fig. 5 G).
We expected no effect of cytochalasin D on the monomeric protein. However, when we examined shorter time periods, we saw that after 15 min the monomeric as well as the aggregated protein were both cross-presented in a cytochalasin D inhibitable manner (data not shown). After 2 h, however, this inhibitor had no effect on the cross-presentation of monomeric 65-P1, and also had no effect on the binding or internalization of F65-P1, which is slightly aggregated F65-P1 (data not shown).
Processing of hsp65-OVA via the MHC class I and class II pathways
DC that generate peptides from an exogenous protein and display pMHC-I complexes to CD8+ T cells are also likely to generate other peptides from that protein and display some of them as pMHC-II complexes to CD4+ T cells. To compare the overall efficacy of the two processing-display pathways, we used the Hsfp called 65-OVA. In this protein, the OVA domain contains two noncontiguous sequences, OVA257–264 and OVA323–339, which when excised proteolytically, can be presented as peptide-Kb and peptide-IAb complexes to CD8+ T cells that express the OT-I TCR and CD4+ T cells having the OT-II TCR, respectively. Purified CD11c+ splenic DC were isolated from B6 mice and incubated for 4 h with 65-OVA. After fixation and washing, the cells were incubated with naive OT-I or OT-II T cells. As shown in Fig. 7 A, the CD4 T cells (OT-II) were far more extensively activated than the CD8 T cells (OT-I).
To determine whether the disparate responses of OT-I and OT-II cells was due to differences in inherent reactivity of the cells or to the amounts of the respective cognate pMHC complexes displayed, we examined the responses of the cells to DC that had been incubated with various concentrations of the corresponding synthetic peptides. As shown in Fig. 7,B, far higher concentrations of the IAb-bound peptide than the Kb-bound peptide were needed to match the responses of OT-II than OT-I cells. The magnitude of the difference can be appreciated by considering the responses to DC incubated with a particular 65-OVA concentration, say 25 μg/ml. At this concentration, 78% of OT-II cells were activated whereas only 16% of OT-I cells responded; and these responses were matched by incubating DC, respectively, with ∼5 × 10−8 M IAb-bound peptide (OVA323–339) and ∼5 × 10−13 M Kb-bound peptide (OVA257–264) (Fig. 7,B). The IAb-bound peptide may be only one of several overlapping OVA-derived peptides that, with IAb, are recognized by OT-II cells, and perhaps not the most effective one (36). Nevertheless, it is likely that far more pMHC-II than pMHC-I complexes were generated by DC from the same exogenous protein (65-OVA). The difference is consistent with the imaging results, which showed that most endocytosed Hsfp (F65-P1) localized in late endosomal-lysosomal vesicles (Fig. 4, A and B), where exogenous proteins are cleaved into peptides that can be loaded onto maturing class II MHC molecules (37, 38).
The Hsfp 65-P1 represents a group of proteins in which a mycobacterial hsp (bacillus Calmette-Guérin hsp65) is extended at the C terminus by linear polypeptides ranging from 34 to over 100 amino acids in length. As shown in this study, these Hsfps bind to DC. The binding is saturable and can be inhibited by some polyanionic polymers that are known to block scavenger receptors (maleylated BSA, dextran sulfate, fucoidan), suggesting that these Hsfps bind to one or more members of the scavenger receptor superfamily. Inhibition of 65-P1 binding to the receptor(s) by maleylated-BSA and dextran sulfate led to a reduction in cross-presentation, indicating that receptor binding of these proteins is crucial for their cross-presentation.
Different intracellular routes for aggregated and disaggregated (monomeric) Hsfp
Although soluble by conventional criteria, the purified 65-P1 protein was aggregated. It could be dissociated, however, to monomers (Fig. 1 E), and the aggregated and monomeric proteins were internalized differently. Cytochalasin D largely blocked cross-presentation of the aggregated protein, indicating that in this state the protein was phagocytosed. It has been shown that the endoplasmic reticulum contributes to the phagosome membrane (39). The presence in the membrane of the Sec61 complex, which can transfer proteins from the endoplasmic reticulum to the cytosol, suggests a plausible mechanism for the phagocytosed, aggregated 65-P1 protein to gain access to the cytosolic MHC class I-processing pathway (40, 41, 42, 43). This mechanism would also account for our finding that soluble 65-P1 was cross-presented in a cytochalasin D-sensitive manner after very brief incubations (15 min) with DC. At longer incubation times, cross-presentation of the aggregated protein was still sensitive to cytochalasin D but the monomeric protein became insensitive to inhibition by cytochalasin D, indicating a pathway independent of the phagocytic route.
In contrast to aggregated 65-P1, at 2 h cross-presented monomeric protein was not internalized by phagocytosis. A route to the cytosol has been established for several proteins that enter cells by receptor-mediated endocytosis, most notably diphtheria toxin, and also anthrax toxin, Clostridium botulinum C2 toxin, tetanus and botulinum neurotoxins (44, 45), and several growth factors (46, 47). For all these proteins, transfer to the cytosol requires transport to acidic late endosomes-lysosomes. Though the bulk of the endocytosed 65-P1 was, indeed, also transported to lysosomes, this route can be ruled out for cross-presented 65-P1 by the lack of an effect of bafilomycin A1, chloroquine, and also chlorpromazine. Each of these agents blocked 65-P1 transport to acidic endosomes but had no effect on the protein cross-presentation. The absence of an effect with chlorpromazine could also mean that the cross-presented 65-P1 was internalized by a clathrin-independent process.
The 65-P1 molecules destined for cross-presentation could have followed a route to the cytosol that is used by several endocytosed bacterial and plant toxins (cholera and shiga toxins (48, 49) or ricin (50)). In this pathway, proteins that are internalized by either clathrin-mediated or caveolae/raft-mediated endocytosis, or both, converge on early endosomes (51, 52), from which a small fraction follows a retrograde transport pathway to the endoplasmic reticulum (53, 54), involving in some cases, passage through the Golgi to reach the cytosol via the Sec61-peptide complex (55).
That the same protein can be internalized by phagocytosis and endocytosis, depending upon particle size has also been seen with immune complexes that bind to FcγRIIA receptors: small (soluble) complexes were internalized by endocytosis, whereas Ab coated RBC were phagocytosed (35).
Many exogenous proteins gain access to the cell cytosol
Proteins are generally more immunogenic when aggregated than monomeric (56). It was therefore surprising to find that monomeric 65-P1 was more effectively cross-presented than aggregated 65-P1. Disaggregated 65-NP was likewise cross-presented more effectively than aggregated 65-NP. These observations may be related to those made with soluble immune complexes internalized by DC via Fc receptors. Rodriguez et al. (57) found that disaggregation of the endocytosed complexes in acidified endosomes was required for cross-presentation. Although transport to acidic vesicles is not required for cross-presentation of the Hsfp, our findings and others suggest that, in general, retrotranslocation of exogenous proteins across vesicle membranes (phagosomes, lysosomes, pinosomes) to the cytosol may be restricted to individual protein molecules or proteolytic fragments (58).
The Hsfp studied here join the growing number of soluble exogenous proteins that can gain entry by various routes to the cytosolic compartment of living cells. Besides bacterial and plant toxins (59), the group includes fusion proteins derived from some toxins (60), modified arginine-rich polypeptides (61), mammalian hsp (62), secretory RNase (63), and even OVA added at high concentration to DC (21).
One feature that these diverse proteins have in common is the great sensitivity with which their presence in the cytosol can be detected. For some toxins, only one or a few protein molecules per cell can cause cell death. Though less extreme, some CD8 T cells, such as 2C or OT-1, can detect very few pMHC-I complexes per cross-presenting cell. Thus, the incubation of DC for a few hours with 65-OVA or 65-P1 stimulated increased expression of CD69 on naive OT-I or 2C cells, respectively, to a level that was matched by incubating the DC for the same length of time with SIINFEKL (for 65-OVA) at ∼1 pM or SIY (for 65-P1) at ∼10 pM.
Efficacy of cross-presentation
From the on- and off-rate constants determined for the SIINFEKL-Kb reaction (64), it can be estimated that around 10 or fewer SIINFEKL-Kb complexes per DC are present on the cells after 1 h incubation with the peptide at 1 pM. The rate constants for the SIY-Kb reaction have not been determined, but SIY binds to Kb slightly less well than SIINFEKL (65) and the number of SIY-Kb complexes per DC are also likely to be around 10 or less for DC incubated for 1 h with SIY at 10 pM. This pMHC density, generated by cross-presented exogenous (Hsfp) proteins is much less than the cell surface pMHC densities generated from endogenous proteins in cells infected with recombinant vaccinia virus or Listeria. Thus, Princiotta et al. (66) showed that several thousand SIINFEKL-Kb complexes were present on cells a few hours after infecting them with recombinant vaccinia virus expressing full-length OVA; with Listeria-infected cells, around 100–200 copies of an immunodominant peptide were produced per cell from the two major antigenic bacterial proteins 2 h after infection (67). However, in macrophages infected with a recombinant vaccinia virus that expressed β-galactosidase, only 30–40 β-galactosidase-pMHC complexes per cell were found (68).
Estimates of the efficiency of cell surface pMHC production from endogenous cytosolic proteins have ranged with different microbial proteins expressed in infected cells from 2.5% (1 pMHC per 35 copies of a Listeria protein (67)) to <0.025% (68). A value of 0.05% (1 pMHC per 2000 molecules) for OVA expressed by recombinant vaccinia virus in infected DC (66) seems particularly relevant for the present study; it suggests that on the order of 20,000 exogenous Hsfp molecules (65-P1 or 65-OVA) gain entry to the DC cytosol per hour in DC incubated with Hsfp at ∼1 μM. Although this estimate may provide a useful framework for the evaluation of the efficacy of cross-presentation, the DC populations we studied are heterogeneous: thus, a fraction of the cells may display considerably more SIINFEKL-Kb or SIY-Kb complexes than the average values, and this DC subset may be responsible for the naive T cell responses observed.
CD4 T cells (OT-II) responded far more robustly than CD8 T cells (OT-I) to DC-processed 65-OVA. The difference is in accord with the finding that most of the endocytosed Hsfp localized in late lysosomes, where peptides for pMHC-II complexes are generated; it is also in accord with general observation that most proteins injected for immunization purposes stimulate the production of CD4 T cells but not CD8 T cells. That the principal Hsfp studied in this report and some other proteins can stimulate mice to produce cytolytic CD8 T cells suggests that in some cases the cross-presentation of even a very few pMHC-I complexes per activated DC is sufficient to induce a significant CD8 T cell response, which is sufficient to destroy tumor cells that express cognate pMHC-I complexes. Whether the expression of similarly small numbers of pMHC-I complexes per DC account for the well-known ability of some mammalian hsp, such as gp96, to elicit CTL that destroy tumor cells in vivo remains to be determined (69).
Altogether, the number and diversity of exogenous proteins that can gain entry to the cytosol of intact cells has now become large enough to entertain the possibility that most soluble proteins, including many self proteins, can enter the class I MHC-processing pathway in vivo. This possibility is of interest in connection with evidence that naive T cells may become tolerized by interacting with pMHC complexes on “steady state” DC (i.e., nonactivated DC) (70, 71). The few pMHC-I complexes that are likely displayed per DC that cross-present exogenous soluble proteins would probably affect only those CD8 T cells whose TCR have relatively high affinity for those complexes. But it is tolerance of these CD8 T cells that may be particularly critical for avoiding some consequences of autoimmunity (72).
We thank Carol A. McKinley for technical assistance, Glenn A. Paradis for help with flow cytometry, Dudley Strickland for aid in evaluating the CD91 receptor, Nicki Watson of the Whitehead Institute Microscope Facility for aid with confocal microscopy, and Jianzhu Chen, Monty Krieger, and Hidde Ploegh for valuable discussions.
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 in part by National Institutes of Health Grants AI 44477 and CA60686 (to H.N.E.) and by a Cancer Center Core Grant CA14051 (to Tyler Jacks). D.P. was a recipient of a Stressgen Postdoctoral Fellowship.
Abbreviations used in this paper: pMHC, peptide-MHC; DC, dendritic cell; Hsfp, heat shock fusion protein; Hsp, heat shock protein; SR, scavenger receptor; LDL, low-density lipoprotein; NP, nucleoprotein.