Cytotoxic T lymphocytes (CTLs) recognize peptides from foreign antigens that are presented by major histocompatibility complex (MHC) class I molecules. Antigens are degraded in the cytosol by proteasomes and the resulting peptides are translocated by the transporter associated with antigen presentation (TAP) into the endoplasmic reticulum (ER) where they bind to MHC class I molecules1. Only properly assembled MHC class I–peptide complexes are transported to the cell surface. Peptide binding appears to be facilitated by a loading complex consisting of MHC class I, tapasin, TAP, the ER chaperone calreticulin2 and the thiol oxidoreductase ER603,4,5. Tapasin is a 48-kD transmembrane protein2 that is encoded by a gene at the centromeric end of the MHC6,7. Tapasin appears to be essential for the loading complex as it forms a bridge between TAP and MHC class I molecules, so that four tapasin class I–calreticulin complexes are bound to each TAP1-TAP2 heterodimer8. After peptide loading, class I molecules undergo conformational changes9, triggering their release from tapasin and export to the cell surface.

In vitro work with the tapasin-mutant B cell line 721.22010 and with transfected insect cells has suggested a role for tapasin in MHC class I surface expression and peptide loading8,11,12,13,14,15, but the requirement for tapasin appears to vary depending on the HLA allele investigated12,16,17. In insect cells tapasin was found to retain class I molecules in the ER13, but this function alone does not explain why tapasin enhances peptide loading because soluble tapasin devoid of the ER-retention motif can still support loading18.

Because most functional data on tapasin have been generated with the use of the above-mentioned insect cells or the human mutant cell line 721.220, which contains residual expression of a truncated tapasin product19, it is still not clear how tapasin would effect the complex interactions of an immune response. To understand the function of tapasin in vivo we generated, by homologous recombination, mice with deficient tapasin expression. We show that tapasin-negative mice are not only compromised in their CTL and natural killer (NK) cell responses but also in their ability to optimally select and load peptides.


Phenotype of tapasin-deficient mice

The tapasin gene construct used for generation of knockout mice is schematically shown in Fig. 1a. Tapasin-deficient (Tpn−/−) mice developed and bred normally but tapasin protein was not detectable by immunoblotting (Fig. 1b). Cell surface staining of splenocytes revealed a 90% reduction of MHC class I H-2Kb and H-2Db expression, which indicates a requirement for tapasin in class I assembly, but this reduction was not as profound as the one observed in TAP1−/− mice20 (Fig. 1c). Similar results were obtained for lymph node cells and peripheral blood lymphocytes (data not shown). The number of CD8+ T cells in the periphery, but not of CD4+ T cells, was reduced (Fig. 1d) suggesting impaired positive selection due to low class I expression in the thymus. Indeed, the number of single positive CD8+ T cells in the thymus was reduced (data not shown).

Figure 1: Phenotype of Tpn−/− mice.
figure 1

(a) Gene construct used to generate Tpn−/− mice. White boxes indicate tapasin gene exons. Neomycin resistance gene (Neor) inserted at the XhoI 6921 site of exon 4 is indicated by black box. (b) Western blot of spleen lysates using the PAV serum against tapasin. (c) H-2Db (thick lines) and H-2Kb (thin lines) expression on splenic lymphocytes from Tpn+/+, Tpn+/− and Tpn−/− littermates, and TAP1−/− mice. Controls were isotype-matched antibodies to TNP (dotted lines). Data shown are representative of four independent experiments. The distinct staining intensity of H-2Db and H-2Kb antibodies is due to different biotinylation. (d) CD4/CD8 ratio of splenic lymphocytes. Staining of Tpn+/− cells was similar to Tpn+/+ cells (data not shown). Results are representative of several independent experiments.

Impaired CTL responses and antigen presentation by DC

To assess immune responsiveness, CD8+ T cell responses to virus and protein antigen were investigated. Tpn−/− mice were infected with influenza virus and CTL activity determined by lysis of peptide-loaded RMA target cells. Influenza-specific CTL activity was reduced in Tpn−/− mice (Fig. 2a). This was probably due to two factors: the diminished frequency of CD8+ T cells and the inability of Tpn−/− cells to efficiently present influenza antigen via the MHC class I pathway. Because dendritic cells (DCs) are more readily infected with influenza virus than lymphocytes, we studied their ability to present influenza antigens. Infected DCs from Tpn−/− mice showed reduced lysis by influenza-specific CTLs obtained from infected wild-type mice, suggesting a presentation defect (Fig. 2b).

Figure 2: Reduced influenza-specific CTL responses and antigen presentation by Tpn−/− bone marrow–derived DCs.
figure 2

(a) Influenza virus-specific CTL responses of Tpn−/− (), TAP1−/− (▪) and wild-type () mice against influenza NP(366–374)–loaded RMA cells. (b) Presentation of influenza antigen by infected DC from Tpn−/− and wild-type DC to influenza-specific C57BL/6 (B6)-derived CTL. For control, noninfected Tpn−/− () and wild-type () DC were used. (c) Cross-presentation of soluble ovalbumin (OVA) by Tpn−/−, TAP1−/− and wild-type DC to H-2Kb–OVA-specific T cell hybridoma B3Z. Decreased absorbancy values indicate reduced cross-presentation. Symbols as in a. (d) Presentation of exogenous OVA peptide SIINFEKL by DCs to B3Z. Results are representative of several independent experiments. Symbols as in a.

We also investigated the ability of DC to cross-present exogenous protein via the MHC class I pathway21. There is evidence that this cross-presentation mechanism is crucial for antiviral immunity22 and tolerance induction to organ-specific antigens23 thus we investigated the role of tapasin in cross-presentation. After bone marrow–derived DC were pulsed with OVA only wild-type DCs, and not Tpn−/− or TAP1−/− DCs, could present OVA via MHC class I molecules to the OVA– H-2Kb–specific hybridoma B3Z (Fig. 2c). There was no difference in presentation of the H-2Kb–binding OVA peptide SIINFEKL (Fig. 2d). Thus, tapasin was required for peptide loading in the cross-presentation pathway of DCs.

Altered NK cell repertoire in Tpn−/− mice

Splenic NK cell numbers, as assessed by flow cytometry using the DX5 monoclonal antibody (mAb), were similar in C57BL/6 (B6) and Tpn−/− mice (3.76±0.93%; 3.73±0.51%, respectively). Interleukin 2 (IL-2)–activated NK cells from B6 mice readily killed Tpn−/− target cells, whereas B6 targets survived (Fig. 3a). In contrast, IL-2–activated NK cells from Tpn−/− mice did not kill Tpn−/− target cells indicating that Tpn−/− NK cells had an altered specificity. The latter result was not a mere consequence of a general impairment of NK cell–mediated cytotoxicity in Tpn−/− mice, as NK cells from Tpn−/− mice readily killed YAC-1 targets (58% specific lysis at an effector:target ratio of 100:1). In all respects, NK cells from B6 mice and Tpn+/− mice displayed similar patterns of reactivity, excluding the possibility that the effects observed with Tpn−/− NK cells were merely related to the 129/Sv background genes in the Tpn−/− mice. In line with the in vitro cytotoxicity data, Tpn−/− bone marrow cells were readily rejected in B6 mice but were not in NK1.1+ cell–depleted B6 mice (Fig. 3b). This demonstrates an NK cell–dependent rejection response against Tpn−/− bone marrow grafts in vivo.

Figure 3: Altered NK cell repertoire in Tpn−/− mice.
figure 3

(a) Cytotoxic activity of IL-2–activated NK cells from wild-type (left panel) and Tpn−/− (right panel) mice against wild-type () and Tpn−/−-derived () Concanavalin (Con A) blast targets. (b) In vivo rejection responses against wild-type or Tpn−/− bone marrow grafted to B6 mice. Rejection of wild-type and Tpn−/− grafts by B6 mice (left panel). Rejection of Tpn−/− grafts by untreated (NK1.1+) or NK1.1-depleted (NK1.1−) B6 recipients (right panel). Low levels of incorporated 125I–5-iodo-2′–deoxyuridine is indicative of rejection of grafted bone marrow.

Impaired peptide selection in Tpn−/− mice

Tapasin may have a quantitative effect on the peptide loading process as suggested by the impaired antigen presentation by DCs from Tpn−/− mice (Fig. 2). Consequently we investigated whether or not tapasin could influence the selection of the class I–bound peptides. Because peptides are crucial for stabilization of MHC molecules, their thermostability reflects the average affinity of the spectrum of bound peptides24,25. Cell lysates of spleen cells were incubated for various time periods at 37 °C and remaining H-2Kb molecules precipitated with the Y3 antibody, which binds only to conformationally intact α12 domains of H-2Kb, and not to denatured molecules26,27. The amount of precipitated H-2Kb was determined by western blotting using antiserum to the H-2Kb cytoplasmic tail28. The upper H-2Kb bands were endoglycosidase H (endoH) resistant and therefore reflected post-Golgi material. The lower bands were endoH sensitive and represented H-2Kb molecules that were still in the ER (Fig 4a and data not shown).

Figure 4: Altered peptide repertoire in Tpn−/− mice.
figure 4

(a) Thermostability of H-2Kb molecules. 1% NP-40 lysates of splenocytes were incubated at 37 °C for various time periods. H-2Kb molecules were precipitated with conformation-dependent antibody Y327,28 and probed by western blotting with antiserum P8 to the cytoplasmic tail of H-2Kb28. (b) Stability of cell surface H-2Kb and H-2Db molecules. De novo transport of MHC class I molecules was blocked with brefeldin A (BFA) and, after various time periods, the remaining H-2Kb and H-2Db molecules on the cell surface of splenocytes were determined by staining with conformation-dependent antibodies K10-56.1 and B22-249. 1. At time 0 class I molecules were set as 100%. (c) Stability of cell surface H-2Kb and H-2Db molecules after proteasome inhibition with lactacystin. Results are representative of several independent experiments.

The H-2Kb molecules in lysates of tapasin-deficient cells decayed rapidly at 37 °C and were clearly less stable than wild-type–derived H-2Kb (Fig. 4a). As T cell recognition depends on MHC peptides displayed at the cell surface, we directly measured the effect of tapasin on the thermostability of class I molecules on the cell. De novo transport of MHC class I to the cell surface was blocked with the drug BFA29 in Tpn−/− and Tpn+/+ lymphocytes, and cells were stained with H-2Kb and H-2Db antibodies that were specific for intact MHC class I conformation. Surface MHC class I molecules on tapasin-negative cells decayed faster than those of Tpn+/+ cells (Fig. 4b). In several experiments the half-life for H-2Kb ranged between 0.5–2.5 h in Tpn−/− cells versus 9–10 h in wild-type cells whereas the half-life for H-2Db ranged between 2–4 h in Tpn−/− versus 10-12 h in wild-type cells.

Although conformation-specific antibodies were used for detection of MHC class I molecules, it was not clear from these results whether the H-2Kb and H-2Db molecules on Tpn−/− cells were empty or occupied with peptide. To investigate the peptide dependency, Tpn−/− cells were incubated for various times with lactacystin, an inhibitor of the proteasome that is the major supplier for MHC class I–binding peptides. Lactacystin led to reduced H-2Kb and H-2Db cell surface expression (Fig. 4c). As lactacystin prevents the formation of new peptide-loaded MHC class I molecules30, the results should represent the decay of preexisting class I molecules at the cell surface. These decay curves were similar to the ones obtained with BFA (Fig. 4b) and so we concluded that the majority of H-2Kb and H-2Db molecules on tapasin-negative cells were loaded with peptides that were generated by the proteasome. The reduced half-life of class I molecules on tapasin-negative compared to wild-type cells indicated that tapasin influenced the quality of the peptide repertoire by selecting more stably binding peptides. The amount of class I on tapasin-negative cells was too low for isolation and sequencing of peptides.

Tolerance to wild-type MHC class I antigens

To see whether the peptide repertoires of Tpn−/− and wild-type mice were completely or only partially different, we investigated the ability of CTLs from these mice to respond to syngeneic H-2b antigens. Tpn−/− mice failed to generate CTLs after immunization with H-2b wild-type cells (Fig. 5a), suggesting tolerance due to partially overlapping peptide repertoires. In contrast, TAP1−/− mice mounted a strong CTL response against syngeneic H-2b antigens from wild-type mice31,32 (see Fig. 5a), probably because TAP1−/− mice are not tolerant to the set of cytosol-derived and TAP-dependent peptides that normally constitute the major pool of MHC class I–bound peptides. For comparison, the response against H-2k alloantigens was measured. The primary H-2k–specific CTL response generated by Tpn−/− splenocytes was reduced (Fig. 5b) but after in vivo priming the response was comparable to that of Tpn+/+ splenocytes (Fig. 5c). These results indicate that the few CD8+ T cells present in tapasin-negative mice were fully functional and could be expanded by stimulation with alloantigens.

Figure 5: Syngeneic and allogeneic CTL responses.
figure 5

(a) Secondary CTL response against H-2b syngeneic cells. Specific killing of 129/Sv (H-2b) targets (filled symbols) CBA (H-2k) targets were used as controls (open symbols). (b) Primary CTL response against MHC alloantigens. Specific killing of CBA (H-2k) target cells (filled symbols), nonspecific killing of syngeneic 129/Sv cells (open symbols). (c) Secondary CTL response against MHC alloantigens.


The results presented here suggest distinct functions for tapasin. Tapasin appears to be critical for peptide loading onto both class I alleles expressed by the H-2b haplotype, H-2Kb and H-2Db. The loading defect leads to reduced selection of CD8+ T cells and to an impairment of CTL viral responses. Likewise, cross-presentation of protein antigen by DC is severely affected. The few CD8+ T cells in Tpn−/− mice, however, appeared to be fully functional suggesting that tapasin does not directly interfere with T cell function. In other studies, tapasin has been considered to be dispensable for presentation of OVA peptide by the H-2Kb molecule13,14, although this may have been due to peptide overexpression by the vaccinia virus systems used. These studies and our work, as presented here, suggest a quantitative rather than an absolute role for tapasin in loading of particular antigenic peptides.

We have also demonstrated that a deficiency in tapasin expression is sufficient to render target cells susceptible to NK cell–mediated killing in vitro and to mediate NK cell–dependent rejection in vivo. The results also show that NK cells from Tpn−/− mice have an altered specificity and are self-tolerant. This adaptation of the NK cell repertoire most likely occurs during development of NK cells and probably serves to ensure self-tolerance. In this respect, our data resemble results obtained in TAP1−/− and β2-microglobulin–deficient (β2M−/−) mice33. However, the fine specificity of Tpn−/− NK cells may differ from that seen in TAP1−/− and β2M−/− mice as well as that of H-2Kb−/− H-2Db−/− mice34.

In addition to its quantitative role in peptide loading, tapasin appears to be crucial for peptide selection. In the absence of tapasin, peptides bind less stably to H-2Kb and H-2Db molecules, resulting in lower stability of class I molecules. Thus, tapasin influences the peptide repertoire by selecting more stably binding peptides. Reduced class I stability has also been proposed for HLA alleles expressed in 721.220 cells11,17. The precise mechanism that tapasin uses for selection of more stably binding peptides is not clear, but it may be comparable to that proposed for the HLA-DM molecule, which influences peptide loading and peptide selection in the MHC class II pathway. It has been suggested that HLA-DM keeps “empty” MHC class II molecules in a peptide-receptive, and probably more open, conformation35,36. Likewise, within the MHC class I–loading complex, tapasin may stabilize “empty” MHC class I molecules in a peptide-receptive open conformation. In agreement with this assumption it has been reported that tapasin-associated and peptide-loaded H-2Ld class I molecules display different conformations9. In such a scenario, which may be similar to the quality control model proposed for HLA-DM35,36, optimally fitting peptides would more efficiently close the peptide binding groove than weakly binding ones and thereby may initiate dissociation of peptide-loaded MHC class I molecules from tapasin and subsequent transport to the cell surface. Although loading pathways and compartments are different, both class I and class II molecules seem to be similar in their requirement for the assistance of specialized accessory molecules for optimal peptide loading and selection.


Generation of tapasin-deficient mice.

To establish knockout mice, published procedures were followed37. A 9-kb fragment containing the mouse tapasin gene was isolated from a strain 129 λ phage library by hybridization with a mouse tapasin cDNA. The tapasin gene was disrupted by insertion of the neomycin resistance gene in reverse orientation into the XhoI site of exon 4. Homologous recombination using the strain 129 (H-2b)–derived embryonic stem cell line E14.1 and generation of knockout mice was done as described37. Identification of Tpn−/− mice was done by Southern blotting of EcoRI plus BglII digests, resulting in fragments of 8.4 kb and 6.2 kb from Tpn+/+ and Tpn−/− mice, respectively.

Antibodies, cytofluorometric analysis and western blotting.

For H-2 cell surface staining, cells were incubated with biotinylated H-2Kb–specific mAb K10-56 and H-2Db–specific mAb B22-24927, followed by phycoerythrin-labeled streptavidin (Pharmingen, San Diego). CD4 and CD8 staining was done with fluorescein isothiocyanate–conjugated CD4 and phycoerythrin (PE)-conjugated CD8 antibodies (Pharmingen) using a FACScan (Becton Dickinson, San Jose). NK cells were stained with DX5 antibody (Pharmingen). For assessment of tapasin expression 107 splenocytes were lysed in SDS-sample buffer and run through 15% SDS-PAGE. Tapasin was visualized by western blotting using rabbit serum PAV generated against amino acids 2–20 of human tapasin, which is crossreactive with mouse tapasin, following described procedures38. For thermostability studies, NP-40 lysates of spleen cells were incubated at 37 °C for various time periods, H-2Kb molecules were immunoprecipitated with conformation-dependent antibody Y326,27 and probed by western blotting using rabbit serum P8 against the cytoplasmic tail of H-2Kb molecules28. Bands were visualized by enhanced chemiluminescence and quantified using the Lumi-Imager system (Roche Diagnostics, Mannheim, Germany). Where indicated, spleen cells were cultured at 37 °C with 20 μM lactacystin (Sigma) or BFA (Sigma). Decreasing concentrations of BFA (10, 5, 2.5 and 1.25 μg/ml) were used at various time points as described29. To increase the staining intensity of surface H-2Kb molecules for stability analysis, incubation with purified K10-56 antibody was followed by biotinylated goat anti-mouse (Pharmingen) and then PE-labeled streptavidin

CTL and T hybridoma responses.

Antiviral CTL responses were performed as previously described29. Briefly, mice were immunized intraperitoneally with 400 HAU live A/Japan/305/57 influenza virus. After 10 days, splenocytes were restimulated in vitro with 2000 HAU influenza virus. CTL activity was assessed 5 days later using RMA cells loaded with 10 ng/ml of influenza NP(366–374) in a standard 51Cr-release assay. For target DCs39, bone marrow–derived DCs were incubated with 1500 HAU A/Japan/305/57 influenza virus in serum-free medium for 10 min on ice followed by 30 min incubation at 37 °C. After removal of excess virus by washing, bone marrow–derived DCs were 51Cr-labeled and employed as targets in a standard 51Cr-release assay using influenza specific CTL. For CTL responses against syngeneic MHC (H-2b), spleen cells of mice immunized 8 days earlier and restimulated in vitro for 6 days with syngeneic B6 spleen cells were tested against 51Cr- labeled 129/Sv (H-2b) target cells. For allogeneic MHC (H-2k) CTL responses, mice were immunized (only for secondary responses) and spleen cells stimulated in vitro for 5 days with CBA (H-2k) target cells and CTL activity tested by using 51Cr-labeled CBA target cells. Target cells were splenocytes cultured with 5 μg/ml Con A for 48 h. For presentation of soluble OVA, DCs were generated from bone marrow as described40 and pulsed with soluble OVA, which had been tested for the absence of free peptides. OVA presentation was evaluated using the OVA–H-2Kb–specific T cell hybridoma B3Z with spectrophotometrical quantification of T cell activity41.

NK cell cytotoxicity and bone marrow transplantation assays.

Generation of IL-2–activated NK cells was performed as previously described39. NK cell cytotoxicity was tested in a standard 51Cr-release assay. Target cells were obtained by culturing single spleen cell suspensions in minimum essential alpha medium supplemented with 10% fetal calf serum, antibiotics and 2.5 μg/ml of Con A for 48 h. Bone marrow transplantation experiments were performed essentially as previously described42. 106 bone marrow cells in PBS buffer were intravenously grafted into groups of six to seven irradiated (800 rad) B6 recipients. On day five the mice were intraperitoneally inoculated with 3 μCi of 125I–5-iodo-2′–deoxyuridine in 0.2 ml PBS buffer, killed 24 h later, and radioactivity in the spleen measured in a gamma-counter. For depletion of NK cells, 24 h before bone marrow grafting, mice were inoculated with 200 μg of NK1.1 PK136 mAb.