Introduction
Cytotoxic T lymphocytes recognize complexes of polymorphic major histocompatibility complex (MHC) class I heterodimers and short peptides of 8–10 amino acids, which are derived from intracellular antigens (
Rammensee et al., 1993). The different subsets of peptides bound by individual MHC class I alleles are produced in the cytosol through proteolytic breakdown of antigens by the multisubunit proteasomal complex present in all eukaryotic cells (
Goldberg and Rock, 1992). From the total pool of proteolytic breakdown intermediates, peptides of at least seven amino acids are translocated selectively across the membrane of the endoplasmic reticulum (ER) by the dedicated peptide transporter TAP (transporter associated with antigen processing) (
Heemels and Ploegh, 1995). The transmembrane‐spanning TAP consists of the two MHC‐encoded subunits TAP1 and TAP2, which belong to the ATP‐binding cassette (ABC) superfamily of transporters (
Townsend and Trowsdale, 1993). In the lumen of the ER, peptides are loaded onto newly synthesized MHC class I heterodimers formed by class I heavy chains (HCs) and β
2‐microglobulin (β2m). Only properly assembled trimeric complexes are transported along the secretory pathway to the cell surface. Each step of MHC class I assembly is assisted by ER‐resident chaperones. Following translocation, folding of HCs is controlled by calnexin (
Vassilakos et al., 1996). Free HCs have also been found in association with Bip (
Noessner and Parham, 1995). Upon β2m association, they associate with calreticulin (
Sadasivan et al., 1996) and subsequently with TAP, from which they dissociate upon peptide binding (
Ortmann et al., 1994;
Suh et al., 1994). In addition, the thiol‐dependent reductase ER‐60 recently has been shown to be part of this peptide loading complex (
Hughes and Cresswell, 1998;
Lindquist et al., 1998;
Morrice and Powis, 1998).
Essential for the formation of TAP–MHC class I complexes is the 48 kDa glycoprotein tapasin which was identified initially by co‐precipitation with TAP and MHC class I (
Sadasivan et al., 1996). Tapasin restores the defect in the mutant human cell line .220, in which MHC class I molecules do not interact with TAP and fail to present antigen to T cells (
Grandea et al., 1995;
Ortmann et al., 1997). The stoichiometry of affinity‐purified TAP–MHC class I complexes suggests that each TAP heterodimer associates with four MHC class I–tapasin–calreticulin complexes (
Ortmann et al., 1997). The primary structure of tapasin revealed a type I transmembrane protein belonging to the immunoglobulin superfamily with no close relatives (
Li et al., 1997;
Ortmann et al., 1997). The tapasin gene is located within 500 kb of the TAP genes at the centromeric end of the MHC (
Grandea et al., 1998;
Herberg et al., 1998). The putative cytoplasmic portion of tapasin contains a functional ER‐retention motif (
Jackson et al., 1990), and removal of the transmembrane and cytoplasmic domains results in a secreted molecule (
Lehner et al., 1998). Interestingly, this truncated version of tapasin restored antigen presentation in .220 cells, although the class I–TAP association was no longer detectable (
Lehner et al., 1998). These results question whether the primary function of tapasin is to form a bridge between MHC class I molecules and TAP as previously hypothesized. We now demonstrate that tapasin functions as a molecular chaperone, which retains empty MHC class I heterodimers in the ER until they acquire peptides.
Discussion
By using
D.melanogaster cells as an experimental system to rebuild the MHC class I pathway, we demonstrated that tapasin retains empty MHC class I molecules in the ER. Unlike the slower exit observed upon co‐transfection of calnexin (
Jackson et al., 1994) or calreticulin (D.Williams, personal communication) into
D.melanogaster cells, intracellular transport of empty class I molecules was completely inhibited by tapasin. Peptide import into the ER either by TAP or by signal sequence‐dependent translocation released MHC class I molecules from tapasin retention, suggesting that tapasin is predominantly responsible for the ER retention of MHC class I molecules in TAP‐deficient cells (
Townsend et al., 1989). Thus, tapasin seems to be able to discriminate between the peptide‐receptive and the peptide‐bound conformation of MHC class I molecules. Furthermore, tapasin seems to monitor the quality of bound peptides because release from tapasin retention was dependent on the availability of peptides binding with high affinity to Kb, as shown by the fact that overexpression of the minigene Ova8 resulted in a higher proportion of transported Kb molecules when compared with peptides produced in the cytosol. We conclude that tapasin chaperones the peptide binding process by retaining MHC class I molecules until they acquire stabilizing peptides. Such a chaperone function for tapasin had been hypothesized earlier (
Li et al., 1997;
Lehner et al., 1998), but experimental support was missing until now. Since this quality control feature of tapasin also operates in the absence of TAP, we further conclude that the retention/release function of tapasin represents a distinct step in the antigen processing pathway which accounts for the observation that MHC class I molecules dissociate from TAP upon import of specific peptides (
Ortmann et al., 1994;
Suh et al., 1994) as well as for the finding that Ld molecules do not co‐precipitate with calreticulin and tapasin after addition of specific peptides to lysates of TAP‐deficient cells (
Harris et al., 1998). Furthermore, the observation that HLA‐B27 molecules transported to the cell surface in the absence of tapasin contained a different set of peptides rendering them unstable (
Peh et al., 1998) is consistent with a quality control function of tapasin.
It might be that tapasin acts in a similar fashion to the HLA‐DM molecule which monitors the peptide loading of MHC class II molecules in the endosomal compartment. HLA‐DM (or H2‐M in mouse) stabilizes MHC class II in a peptide‐receptive state, thereby catalyzing the replacement of the invariant chain‐derived low affinity peptide CLIP by peptides of higher affinity (
Denzin and Cresswell, 1995;
Weber et al., 1996). Similarly, tapasin might stabilize MHC class I during exchange of low affinity with high affinity peptides.
How tapasin senses the conformational changes occurring in the MHC class I molecule upon binding of stabilizing peptides currently is not known. However, some conclusions about residues involved in the class I–tapasin interaction can be drawn from studies investigating class I–TAP association. Co‐immunoprecipitation experiments have indicated that residues within the peptide‐binding groove may be involved in TAP association (
Neisig et al., 1996). Moreover, a point mutation in the α2 domain of HLA‐A2.1 molecules abrogates their ability to associate with TAP and accelerates their transport to the cell surface (
Lewis et al., 1996;
Peace‐Brewer et al., 1996). Although the intracellular transport of these mutant molecules was dependent on TAP, they did not contain stabilizing peptides at the cell surface (
Lewis and Elliot, 1998). Upon inhibition of intracellular transport by brefeldin A, however, stabilizing peptides were acquired, thus indicating a missing interaction with a retention molecule. Our demonstration that tapasin retains MHC class I implicates tapasin for this role. The observed lack of association of HLA‐A2.1 T134K with calreticulin (
Lewis and Elliot, 1998) could be secondary to a missing tapasin interaction. The peptide‐binding site of MHC class I molecules is in a molten globule state in the absence of peptides (
Bouvier and Wiley, 1998), undergoing a conformational change upon peptide insertion (
Rigney et al., 1998). By interacting with the peptide‐binding domain, tapasin might monitor these changes during peptide loading.
For murine MHC class I alleles, the α3 domain has been implicated in the association with tapasin–TAP (
Carreno et al., 1995;
Suh et al., 1996;
Harris et al., 1998;
Kulig et al., 1998). Whether or not human class I molecules also interact with tapasin via the α3 domain remains to be investigated. Since murine tapasin failed to promote peptide loading of some human class I alleles, it seems that there are species‐specific differences in this interaction (
Peh et al., 1998). Similarly, human tapasin binds less efficiently to the murine molecule Kb in insect cell transfectants (our unpublished observations). Moreover, Yewdell and co‐workers did not observe an effect of human tapasin on murine MHC class I molecules expressed by vaccinia virus in a mosquito cell line (
Deng et al., 1998). Failure of human tapasin to retain murine MHC class I might cause empty murine MHC molecules to be transported to the cell surface of TAP‐deficient human cells (
Alexander et al., 1989). In addition to species differences, it seems that different alleles within one species (human) show a distinctive dependence on tapasin for surface expression and antigen presentation (
Peh et al., 1998). Thus, it could be that tapasin binds with different affinity to various alleles. However, murine MHC alleles did not differ markedly in their retention by tapasin in insect cells (our unpublished observations). Alternatively, the availability of peptides binding to the various MHC alleles could determine their individual dependence on tapasin for obtaining stabilizing peptides. The latter assumption is supported by our observation that high levels of Ova8 expression obliterated the need for tapasin to promote assembly with peptides.
This observation also shows that house‐keeping chaperones such as calnexin and calreticulin, which have
Drosophila homologs (
Christodoulou et al., 1997), are sufficient for MHC class I molecules to fold into a peptide‐receptive conformation even in the absence of tapasin. By contrast, invertebrates lack the MHC class I‐specific retention/release function of tapasin. Insect cells also lack the ability to promote MHC class I association with TAP, similarly to tapasin‐deficient mammalian cells (
Ortmann et al., 1997). The finding that tapasin can perform this function in the invertebrate folding environment suggests that additional chaperones found in the TAP–tapasin complex (
Sadasivan et al., 1996;
Hughes and Cresswell, 1998;
Lindquist et al., 1998;
Morrice and Powis, 1998) are either not necessary for complex formation or can be replaced by eukaryotic house‐keeping chaperones. By contrast, tapasin was not only necessary but sufficient for complex formation. In addition, tapasin enhanced TAP1–TAP2 association in insect cells, which is consistent with the previous finding that tapasin association with TAP increased the peptide transport activity in .220 cells (
Lehner et al., 1998). The multiple functions of tapasin seem to be localized to separate parts of the tapasin molecule. Whereas the interaction with MHC class I molecules involves the luminal domain of tapasin, it seems that the interaction with TAP localizes in the transmembrane domain of tapasin, since a truncated version of tapasin bound to MHC class I but not to TAP (
Lehner et al., 1998). Interestingly, tail‐deleted tapasin was still able to promote peptide loading despite its secretion due to the missing ER retention signal (
Lehner et al., 1998). It seems likely that tail‐deleted tapasin resides long enough in the ER to chaperone peptide loading, particularly since transfection ensured a constant supply of highly expressed truncated tapasin in these experiments. Thus, the molecular chaperone activity of tapasin seems to be less dispensable than the tapasin‐dependent promotion of TAP association and peptide transport.
The importance for tapasin‐mediated editing of peptide loading is also stressed by the finding that tapasin was required to reconstitute antigen processing in invertebrate cells. Retention/release by tapasin was as crucial for antigen processing as peptide transport by TAP and peptide display by MHC class I. In contrast to these dedicated molecules which were developed during the evolution of the MHC (
Herberg et al., 1998), the ability to generate peptides from antigens seems to be already present in lower eukaryotes. The observation that Ova8 presentation could be inhibited by lactacystin implicates the proteasome in the processing of ovalbumin in
D.melanogaster cells, as observed in mammalian cells (
Craiu et al., 1997). The ability of invertebrate proteasomes to generate the correct epitopes from precursor peptides has also been demonstrated
in vitro (
Niedermann et al., 1997). Taken together, these data support the hypothesis that the MHC class I system has taken advantage of the pre‐existing ubiquitin–proteasome system to display the by‐products of protein turnover at the cell surface. Peptide generation for MHC class I was optimized during vertebrate evolution due to the development of interferon‐γ‐induced subunits, some encoded in the MHC, as well as modulators of the proteasome (
Früh and Yang, 1999). However, our data suggest that these modulatory subunits are not absolutely required for antigen processing. Thus, the minimal specific components which are needed for MHC class I antigen presentation are MHC class I, TAP and tapasin.