IRAP Identifies an Endosomal Compartment Required for MHC Class I Cross-Presentation

See allHide authors and affiliations

Science  10 Jul 2009:
Vol. 325, Issue 5937, pp. 213-217
DOI: 10.1126/science.1172845


Major histocompatibility complex (MHC) class I molecules present peptides, produced through cytosolic proteasomal degradation of cellular proteins, to cytotoxic T lymphocytes. In dendritic cells, the peptides can also be derived from internalized antigens through a process known as cross-presentation. The cellular compartments involved in cross-presentation remain poorly defined. We found a role for peptide trimming by insulin-regulated aminopeptidase (IRAP) in cross-presentation. In human dendritic cells, IRAP was localized to a Rab14+ endosomal storage compartment in which it interacted with MHC class I molecules. IRAP deficiency compromised cross-presentation in vitro and in vivo but did not affect endogenous presentation. We propose the existence of two pathways for proteasome-dependent cross-presentation in which final peptide trimming involves IRAP in endosomes and involves the related aminopeptidases in the endoplasmic reticulum.

Peptide ligands for MHC class I molecules are produced by intracellular proteases (1). Initial antigen degradation by cytosolic proteasome complexes is frequently followed by N-terminal peptide trimming, which can occur in the cytosol and by endoplasmic reticulum (ER) aminopeptidases (ERAPs) (2). Peptides are transported into the ER by the transporter associated with antigen processing (TAP) for loading of newly synthesized MHC class I molecules. Loading of MHC class I molecules with internalized, cross-presented antigens in dendritic cells (DCs) is thought to play an important role in priming of CD8+ T cell responses to pathogens and tumors, as well as in immune tolerance to self.

While screening crude microsome lysates for peptidases involved in N-terminal trimming of human leukocyte antigen (HLA) class I ligands, we identified insulin-regulated aminopeptidase (IRAP). IRAP was detected as an interferon γ (IFN-γ)–induced activity trimming a fluorogenic Leu–aminomethyl coumarin (AMC) substrate in anion exchange chromatography (Fig. 1A) (3). The peak containing IRAP also trimmed a precursor of the HLA-A2–restricted epitope SLYNTVATL (4, 5). IRAP is a ubiquitous zinc-dependent aminopeptidase closely related to ERAP1 and ERAP2 (6). IRAP localizes to regulated endosomal storage compartments in adipocytes and muscle cells together with the glucose transporter Glut4; these compartments are termed Glut4 storage vesicles (GSVs) (7). Signaling through the insulin or immunoglobulin E (IgE) receptors induces rapid translocation of ~50% of IRAP to the cell surface (7, 8). The function of the compartment storing IRAP in other cell types, such as DCs, remains unknown.

Fig. 1

Identification of an MHC class I–associated aminopeptidase. (A) Hydrolysis of Leu-AMC by fractionated crude microsomal proteins from untreated or IFN-γ–treated HeLa cells. Numbers above peaks indicate relative induction of Leu-AMC hydrolysis by IFN-γ incubation. MHC class I was detected exclusively in peak 4 by immunoblot analysis; ERAP1 and ERAP2 elute exclusively in peak 1 under identical chromatographic conditions (4). (B) Recombinant human IRAP, ERAP1, and ERAP2 were tested for aminoacyl-AMC hydrolysis. For each enzyme, results are expressed as percent of maximum hydrolysis set at 100. (C) Peptide K15I (KIRIQRGPGRAFVTI) (5) was digested with recombinant IRAP. Degradation products are designated by the N-terminal residue followed by peptide length and C-terminal residue. (D) IRAP, MHC class I, and MHC class II molecules were immunoprecipitated (IP) from BMDCs lysed with a mild detergent. Precipitates were split at a ratio of 1:9 for detection of directly precipitated and coprecipitated proteins, and analyzed by immunoblot as indicated. Control DCs were from KbDbβ2m knockout (ko), HLA-B7 transgenic mice; WT, wild type. (E) Subcellular localization of IRAP (red) in human myeloid DCs.

To evaluate whether IRAP qualifies as a trimming aminopeptidase, we tested its substrate specificity and interaction with MHC class I molecules. Human IRAP displayed a broader specificity toward fluorogenic substrates than did its ER-resident relatives (Fig. 1B). IRAP also efficiently converted the 15-nucleotide oligomer epitope precursor K15I to the minimal epitope G9I, which in the ER requires the concerted action of ERAP1 and ERAP2 (4) (Fig. 1C). The amount of IRAP eluted was doubled by IFN-γ incubation of HeLa cells (Fig. 1A), although IRAP mRNA levels were not (fig. S1). Because IRAP eluted in a fraction also containing MHC class I molecules (Fig. 1A), increased recovery of IRAP activity could have resulted from its association with IFN-γ–induced MHC molecules. A fraction of cellular IRAP was coprecipitated with HLA class I molecules from bone marrow–derived DCs (BMDCs) (Fig. 1D), compatible with a role for IRAP in MHC class I antigen presentation.

Because IRAP resides in endocytic vesicles (7), we considered that it may be involved in the MHC class I cross-presentation of exogenous antigens internalized by DCs through phagocytosis or receptor-mediated endocytosis (9). Although cross-presented antigens are believed to be processed mainly by factors also involved in the endogenous pathway, such as TAP, Sec61, or ERAP1 (10), the intracellular routes mediating the junction between the endocytic pathway and the ER have been difficult to decipher (9). In human myeloid DCs and murine BMDCs, ~25% of IRAP colocalized with HLA class I molecules (Fig. 1E and fig. S2), whereas no colocalization was observed with endolysosomal proteins (HLA class II and Lamp-1). In the absence of DC phagocytosis, we could not detect colocalization with three ER-resident proteins: KDEL receptor, ERAP1, and TAP2. This was not due to a failure of antibodies to recognize ER-resident IRAP, because strong colocalization with two ER markers of a hemagglutinin (HA)–tagged IRAP variant carrying a KDEL sequence was readily detectable, whereas ER colocalization was completely absent for HA-tagged IRAP lacking a KDEL sequence (fig. S3). Thus, newly synthesized IRAP molecules must exit rapidly from the ER. However, IRAP showed considerable colocalization with syntaxin 6 (STX6; 51%), a Q-SNARE (soluble N-ethylmaleimide–sensitive factor attachment) known to stain GSVs (11). IRAP colocalized most strongly (76%) with Rab14, which may play a role in preventing phagosome fusion with lysosomes in macrophages (12, 13). Similar data were obtained in murine BMDCs (fig. S4), where IRAP also showed considerable colocalization with the mannose receptor (MR) (57%), reported to mediate efficient cross-presentation of soluble antigen (14).

During phagocytosis, IRAP was strongly enriched in purified early phagosomes but not in late phagosomes, whereas ERAP was not enriched in phagosomes (Fig. 2A), consistent with a recent proteomics analysis of phagosomes (15). In DCs phagocytosing yeast cells, IRAP colocalized preferentially with MHC class I molecules internalized from the surface in vesicles adjacent to phagosomes, or in phagosomal membranes (Fig. 2, B and C, and fig. S5). Internalized MHC class I molecules could not be detected in late Lamp1+ phagosomes (fig. S5). ERAP did not colocalize with internalized MHC class I molecules (<0.5%; Fig. 2B). This suggests that internalized MHC class I molecules may traffic together with IRAP from phagosomes to GSV-like vesicles before phagosome fusion with lysosomes takes place. DC activation by pathogen compounds induces TAP recruitment to MR+ endosomes (14). Phagocytic DC activity did not increase IRAP colocalization with ERAP (Fig. 2D). In contrast, colocalization of IRAP with TAP increased from <2% in resting murine BMDCs to 6.5% (Fig. 2, E and F). Thus, ER-phagosome or endosome fusion events may selectively deliver TAP but not ERAP1 to IRAP+ vesicles.

Fig. 2

IRAP, but not ERAP, is recruited to phagosomes. (A) Serial dilutions of membranes from early (20 min) and late (2 hours) BMDC phagosomes were examined for IRAP and ERAP by immunoblot. Total cellular proteins or membrane proteins, and crude microsomes served as controls. One of two (ERAP blot) or four (IRAP blot) similar experiments is shown. (B) Human myeloid DCs were stained after 20 min of yeast cell phagocytosis for IRAP, ERAP, and HLA-A, B, and C. (C) BMDCs were stained for cell surface Kb molecules, fed yeast cells, and analyzed for colocalization of IRAP with Kb after different intervals. The right panel shows the percent colocalization (mean ± SD) of total cellular MHC class I with IRAP after 15 min. (D) Human myeloid DCs were stained for IRAP and ERAP. (E) Murine BMDCs, before or 20 min after phagocytosis of yeast cells, were stained with antibodies to IRAP and mouse TAP1. (F) The percentage of IRAP colocalizing with TAP1 was calculated using correlation maps.

To examine a potential role of IRAP in antigen presentation, we studied previously generated IRAP knockout mice (8, 16). Deficiency for IRAP, which is normally expressed by all mouse splenocyte subpopulations except granulocytes (Fig. 3A), did not affect MHC class I expression by splenic lymphocytes including CD11chi DCs (Fig. 3B), which also matured normally upon LPS stimulation (fig. S6). BMDCs from IRAP knockout mice expressing full-length ovalbumin (OVA) or the preprocessed Kb-restricted OVA epitope S8L (SIINFEKL, OVA257-64), or pulsed with the synthetic epitope, presented epitope S8L normally to specific OT-I CD8+ T cells (17) (Fig. 3C and fig. S7). Thus, production of epitope S8L in the endogenous processing pathway is proteasome-dependent (Fig. 3C) and ERAP-dependent (18), but does not require IRAP. Endogenous presentation of the male antigen SMCY (19) also required expression of ERAP but not IRAP by DCs (Fig. 3D). These results provide strong evidence against involvement of IRAP in the endogenous MHC class I processing pathway.

Fig. 3

IRAP is required for cross-presentation in vitro. (A) Fluorescence-activated cell sorter analysis of intracellular IRAP expression in mouse splenocyte subpopulations. (B) Expression of H2-Kb on splenic DCs (CD11c+), B cells (B220+), helper (CD4+), and killer (CD8+) T cells. (C) Generation of S8L in the endogenous pathway by vaccinia-infected DCs. The response of OT-I effector T cells was measured by IFN-γ enzyme-linked immunosorbent assay. Epox, epoxomicin (proteasome inhibitor); wt, wild type. (D) Endogenous presentation of the SMCY male antigen by DCs to HY CD8+ T cells. (E) In vitro cross-priming of OT-I cells by DCs incubated with OVA-coated beads. Baf.A1, bafilomycin A1. (F) In vitro cross-priming of OT-I cells with necrotic insect cells containing fusion proteins. CytoD, cytochalasin D; Wortm., wortmannin. (G) DCs were incubated with necrotic insect cells expressing fusion proteins comprising OVA, S8L, or S8L with an N-terminal CSC extension. The experiments shown correspond to one of two [(A), (D), (F), and (G)], three [(C) and (E)], or five (B) experiments. Means + SD are shown in (C) to (G). Splenic DCs were used in (E), BMDCs in all others.

To study the impact of IRAP deficiency on cross-presentation, we first examined the presentation of OVA-coated beads to OT-I cells. The efficiency of bead phagocytosis by BMDCs was not affected by IRAP deficiency (fig. S8). Cross-presentation of beads was abolished by two proteasome inhibitors but was not affected, or even increased, by two inhibitors of lysosomal proteases (Fig. 3E and fig. S9). Cross-presentation also was compromised by TAP deficiency (Fig. 3E). Thus, OVA internalized by phagocytosis was shuttled into the pathway involving antigen degradation in the cytosol, but not into the vacuolar pathway involving acid lysosomal proteases (20, 21). Cross-presentation of particulate OVA was reduced by ERAP deficiency (Fig. 3E) (18, 22). IRAP deficiency resulted in an at least equal reduction of presentation by 50 to 70% (Fig. 3E). This result suggests that precursors of epitope S8L could be trimmed both by an ER aminopeptidase and an endosomal aminopeptidase.

Next, we studied cross-presentation of necrotic insect cells expressing fusion proteins consisting of two or three Ig-binding domains derived from protein G (PrG), ubiquitin (Ub), and OVA or S8L. In the cytosol, OVA or S8L antigen is expected to be removed from these proteins by deubiquitinating enzymes. Cross-priming of OT-I cells required actin and phosphatidylinositol 3-kinase–dependent uptake of antigenic material by live DCs, demonstrating the absence of free peptide in the material (Fig. 3F). Generation of S8L from a phagocytosed fusion protein containing the preprocessed epitope not requiring trimming was not compromised in IRAP or ERAP single- or double-deficient DCs (Fig. 3G). However, absence of IRAP or ERAP reduced generation of S8L from a fusion protein containing full-length OVA. Absence of both peptidases had an additive effect (Fig. 3G), which suggests that the two enzymes may act in independent pathways. S8L generation from a precursor peptide extended by a sequence adapted to the cleavage specificity of IRAP but not ERAP [PrG(2x)-Ub-CSC-S8L] was reduced by >75% in IRAP-deficient DCs but was not affected by ERAP deficiency (Fig. 3G). Thus, IRAP acts as an epitope-trimming peptidase in an endosomal compartment.

Finally, we examined cross-priming of carboxyfluorescein succinimidyl ester (CFSE)–labeled, adoptively transferred naïve OT-I T cells in vivo (Fig. 4). IRAP deficiency did not affect recovery from lymph nodes and spleens of transferred OT-I cells (fig. S10), nor did it alter proliferation of transferred OT-I cells in response to S8L-pulsed DCs (Fig. 4A). In contrast, cross-presentation of cell-associated OVA was reduced (Fig. 4A and fig. S11). Next, we immunoaffinity-purified the fusion protein PrG(3x)-Ub-OVA and targeted it in vivo to cells expressing Toll-like receptor 2 (TLR2) or the MR via binding of its PrG domains to specific antibodies. IRAP deficiency resulted in reduced cross-presentation of receptor-targeted fusion protein (Fig. 4A and fig. S11). Moreover, IRAP deficiency reduced cross-priming of endogenous CD8+ T cells by cell-associated OVA, as revealed by analysis of splenocytes from primed mice with Kb/S8L pentamers (Fig. 4B). Cross-presentation of a second antigen, cell-associated SMCY male antigen, was also compromised in IRAP knockout mice (Fig. 4C).

Fig. 4

IRAP deficiency compromises in vivo cross-priming. (A) Mice were injected with CFSE-labeled OT-I cells, followed 24 hours later by injection of antigen. Three days later, OT-I proliferation (expressed as divisions per precursor OT-I cell) was examined by flow cytometry. Antigens were S8L-pulsed BMDCs, β2m knockout splenocytes electroporated with OVA, or OVA fusion proteins targeted for internalization through TLR2 or the MR. (B) Mice were injected with OVA-loaded β2m knockout splenocytes. Ten days later, priming of endogenous CD8+ T cells was measured by staining with a Kb/S8L pentamer. (C) SV129 mice were injected with CFSE-labeled HY T cells, followed 24 hours later by injection of male MHC-unmatched (Balb/c) splenocytes.

Our findings indicate that the final proteolytic trimming of cross-presented antigens can occur in an endosomal DC compartment sharing several markers associated with regulated endosomal storage vesicles. Recruitment of Rab14 may reduce routing of antigens into an acid lysosomal environment known to be detrimental for cross-presentation (23). Physical association of IRAP with abundant, presumably internalized class I molecules may favor a direct linkage between peptide trimming and MHC class I loading. Cross-presentation of antigens processed in an IRAP-dependent manner required active proteasome but not lysosomal proteases, which suggests that this pathway implicates cytosolic antigen degradation followed by peptide transport into IRAP+ endosomes by TAP recruited upon phagocytosis. However, considering the relatively efficient TAP-independent cross-presentation of S8L (Fig. 3E), the existence of a pathway implicating IRAP together with an alternative endosomal peptide transporter, or together with a role for the proteasome unrelated to antigen degradation, cannot be ruled out entirely.

We found that both IRAP and ERAP are implicated in cross-presentation. Considering the functional redundancy and the complete absence of colocalization between the enzymes, the existence of two parallel proteasome-dependent cross-presentation pathways is the most plausible explanation (fig. S12). According to our model, MHC class I molecules can be loaded with cross-presented peptides in three intracellular compartments: endosomes, ER, and lysosomes/vacuoles.

Supporting Online Material

Materials and Methods

Figs. S1 to S12


  • * These authors contributed equally to this work.

References and Notes

  1. See supporting material on Science Online.

  2. Single-letter abbreviations for amino acid residues: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; Y, Tyr.

  3. We thank Metabolex Inc. for an anti-IRAP serum, N. Shastri for an anti-ERAP serum, B. Rocha for HY mice, O. Lantz for TAP knockout mice, L. Chatenoud for OT-II mice, B. Fouquet and N. Merzougui for technical help, and A. M. Lennon-Dumesnil for critical reading of the manuscript. Supported by grant PROTARVAC of the European Commission (P.v.E., F.G., G.N.), by INSERM fellowships (O.C., R.K.), and by grants from the Deutsche Forschungsgemeinschaft and from the Medical Faculty of the University of Freiburg (G.N.).
View Abstract

Stay Connected to Science

Navigate This Article