Potential Immunocompetence of Proteolytic Fragments Produced by Proteasomes before Evolution of the Vertebrate Immune System

The protease inhibitor N-acetyl-l-leucinyl-l-leucinal-l-norleucinal (LLnL) was purchased from Sigma Chemical Co. (St. Louis, MO); IFN-γ was from Boehringer Mannheim GmbH (Mannheim, Germany). The proteasome inhibitor lactacystin was purified as described (20). The C57BL/6-derived thymoma EL4, the human lymphoblastoid cell line T1, the human erythroblastoid cell line K562, and Drosophila melanogaster Schneider cells were obtained from American Type Culture Collection (Rockville, MD). Monoclonal antibodies were prepared from the hybridoma Y3 (anti-H2 class I K^b heterodimers; 21). Rabbit antiserum specific for sequences encoded by exon 8 of the K^b gene and reactive with free or β₂m-associated K^b heavy chains was a gift from Dr. S. Nathenson (Albert Einstein College, New York).

EL4 cells (10⁷ cells/ml) were incubated for 2 h at 37°C in the presence or absence of proteasome inhibitors in cysteine and methionine-free medium, and for the last 45 min of incubation, (³⁵S) cysteine/methionine (700 μCi/ ml) was added. After metabolic labeling, cells were lysed in 0.5% Nonidet P-40 (ICN Biomedicals Inc., Plainview, New York) and 0.5% Mega 9 (Sigma Chemical Co.). Samples were precleared for 60 min at 4°C with protein A–Sepharose (Pharmacia, Uppsala, Sweden) pretreated with 1 mg/ml bovine serum albumin. For immunoprecipitation, either mAb Y3 (15 μg) or 4 μl rabbit α exon 8 antiserum were added to the precleared lysates for 2 h. For the last 90 min of incubation, protein A–Sepharose was added. Immunoprecipitates were analyzed by SDS-PAGE on 12% gels. Quantitation of gel bands was performed with the aid of a Fujix BAS 1,000 phosphorimager.

Peptides were synthesized by using solid-phase 9-fluorenylmethoxycarbonyl chemistry in a peptide synthesizer (431A; Applied Biosystems, Foster City, CA) and subsequently purified by reverse phase HPLC. The identity of peptides was established by matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI-Tof-MS) and aa sequence analysis on Hewlett-Packard Co. (Palo Alto, CA) instruments. Ribulose 1,5 bisphosphate carboxylase small subunit was purified from intact pea chloroplasts by denaturing continuous electrophoresis (Grimm, R., manuscript submitted).

20S proteasomes were purified from EL4 cells cultured with or without 50 U/ml IFN-γ for 6 d, from K562, T1, and D. melanogaster Schneider cells, as well as from Saccharomyces cerevisiae (strain YRG-2) by fractionated precipitation of the cytosol with polyethylene glycol 6,000 followed by anion exchange chromatography on a Mono Q column (HR 5/5; Pharmacia) as previously described for EL4 cells (22). Modifications of the NaCl gradient (buffer A: 20 mM Tris/HCl [pH 7.2]; buffer B: 20 mM Tris/HCl [pH 7.2], 1 M NaCl) were as follows: S. cerevisiae: 0–38% B in 50 min, 38–48% B in 25 min, proteasomes eluted at 45% B; D. melanogaster: 0–28% B in 50 min, 28–31% B in 25 min, proteasomes eluted at 30% B; IFN-γ–treated EL4 cells: 0–34% B in 65 min, 34–37% B in 40 min, proteasomes eluted at 36% B; K562 cells: 0–37% B in 55 min, 37–41% B in 40 min, proteasomes eluted at 38% B; T1 cells: 0–35% B in 55 min, 35–38% B in 40 min, proteasomes eluted at 37% B. T1 proteasomes were further purified on a Phenylsuperose column (Pharmacia). The proteasomes were recovered in the flow-through fraction. 20S proteasomes from Rhodococcus sp. and recombinant Thermoplasma acidophilum proteasomes were purified as described (23, 24). The purity of proteasomes was assessed by SDS-PAGE followed by silver staining as described (22). Purification of recombinant human red blood cell PA28α is described in references 25 and 26.

Digestions of synthetic peptides (6 μg) and of the small subunit of ribulose 1,5 bisphosphate carboxylase (10 μg) with isolated proteasomes (1 μg) were performed at 37°C, except in the case of Thermoplasma proteasomes at 60°C, in a total volume of 300 μl buffer (20 mM Hepes/KOH [pH 7.0] 1 mM EGTA, 0.5 mM EDTA, 5 mM MgCl₂, 0.5 mM 2-mercaptoethanol, 0.02% sodium azide). Digestions of ovalbumin fragments Ova_37-77 and Ova_239-281, and of the protein substrate were done in the presence of 0.004% SDS. In the experiments shown in Fig. 7, recombinant PA28α was added in a fivefold molar excess over the proteasome. Aliquots of the reaction mixture were separated by reverse phase HPLC on a SMART system equipped with a μRPC C2/C18 SC 2.1/10 column (Pharmacia). Eluent A was 0.1% (vol/vol) TFA/water; eluent B was 80% (vol/ vol) acetonitril/water (0.081% TFA). The identity of peptides in individual HPLC fractions was established by MALDI-Tof-MS and aa sequence analysis by Edman degradation.

For molecular mass determination without separation of the peptide products, the reaction mixtures were desalted with the help of a mini-C18 column. The peptides were then eluted by the addition of 60% acetonitril in water, concentrated, and then directly analyzed with a 62025A Maldi-Tof-MS (Hewlett-Packard Co.) using dihydroxybenzoic acid as the matrix.

For proteasome cleavage site determination, aliquots of the reaction mixtures were directly subjected to aa sequence analysis by Edman degradation (pool sequencing). The sites and efficiencies of cleavage were determined from the sequence cycles and relative yields of amino acids unique in the analyzed sequence. For example, if the unique amino acid X appears in cycle Y of Edman degradation, a cleavage site is revealed Y residues from amino acid X towards the NH₂ terminus; if the unique amino acid appears in more than one cycle, several cleavage sites are revealed, their relative strengths corresponding to the relative yields of the aa in each cycle.

A functional pressure of proteasome-mediated proteolysis on the evolution of MHC class I could be envisaged if proteasomal digestion provided the major source of peptides for assembly with MHC class I. We examined the assembly of MHC class I–peptide complexes in the presence of the proteasome inhibitor lactacystin (27). Except proteasome-related particles in bacteria (28), no other protease has been reported to be inhibited by this compound. The murine thymoma cell line EL4, expressing the class I molecules H-2K^b (K^b) and H-2D^b (D^b), was incubated with or without lactacystin and metabolically labeled with ³⁵S-methionine plus cysteine. Immunoprecipitation was performed either with mAb Y3, detecting a conformational determinant of K^b present only on the assembled trimolecular complex, or with an antiserum (anti–exon 8 antiserum) that recognizes the cytoplasmic tail of K^b and D^b molecules, independent of whether a peptide is bound or not. As shown in Fig. 1, addition of lactacystin caused a marked decrease in the amount of K^b molecules precipitable with the conformation-specific antibody, but not of that precipitated by the anti–exon 8 reagent. Plateau inhibition was at ∼67% as calculated by phosphorimaging (see Materials and Methods), similar or slightly better than that reported for the peptide aldehyde inhibitor LLnL (Fig. 1; reference 10). These data are in agreement with recent results on the inhibition of antigen presentation by lactacystin (29) and further support the notion that proteasomes participate in the generation of the majority of peptides presented by MHC class I.

At early time points in the processing of short polypeptide substrates by isolated mouse 20S proteasomes, single cleavage intermediates can be detected in addition to dual cleavage oligopeptides. After consumption of the original substrate and of the single cleavage intermediates, the reaction appears to slow down and the mixture of peptide fragments approaches a relatively stable state (12). We believe that the stable state in vitro most closely resembles the conditions in vivo. Therefore, although several time points have been analyzed to compare 20S proteasomes at distinct stages of evolution (Fig. 2), we present data on the relatively stable product patterns determined after complete substrate turnover.

We have previously shown that the immunodominant ovalbumin epitope Ova_257-264 (SIINFEKL; reference 30) is the major stable product generated by mouse 20S proteasomes from the 22-mer Ova_Y249-269 as well as from the 44-mer Ova_239-281 (12). Here we show that this octamer is also a dominant product of digestion of Ova_239-281 by 20S proteasomes isolated from D. melanogaster Schneider cells (Fig. 3,A and Fig. 4,A). As a second example, we studied the generation of the nonamer TLWVDPYEV, an endogenous peptide derived from the product of the antiproliferative B cell translocation gene 1 (BTG1) and eluted as a major self-epitope from the human class I molecule HLA-A2.1 (31). Fig. 3, B and C show that this nonamer peptide is the major dual cleavage product generated by yeast (S. cerevisiae) proteasomes of the synthetic 24 mer encompassing this peptide in the sequence of BTG1. Moreover, we studied a 21-mer sequence derived from the tyrosine kinase JAK1 containing the nonamer SYFPEITHI, the most abundant self-peptide presented by mouse H-2K^d molecules of P815 cells (32, 33), and previously shown to be generated by digestion with mouse 20S proteasomes (34). We detected the epitope as the predominant dual cleavage product of the 21 mer with Drosophila proteasomes (Fig. 3 D). Thus, proteasomes from invertebrate eukaryotes have a high potency to generate proteolytic fragments that have been proven to serve as ligands of MHC molecules in the vertebrate immune system.

More than 90% of the peptides so far eluted from MHC class I molecules are 8–10 aa in length (3). We have recently shown that more than half of the dual cleavage proteolytic fragments generated by digestion of Ova_239-281 with mouse 20S proteasomes are 8–11 aa in length (12). In Fig. 4, A and B, we show that the peptides most efficiently produced from this substrate by Drosophila 20S proteasomes are in the size range of MHC class I–binding peptides (Fig. 4,B, shaded area). In addition, we studied a longer and immunologically undefined substrate, the 123-aa small subunit (SSU) of ribulose 1,5 bisphosphate carboxylase from the garden pea (Pisum sativum L.) (Fig. 4, C–E). Digests were prepared with proteasomes from mouse EL4 cells and from yeast. Most of the abundant masses in the digests represented peptide sizes of 5 to 11 aa for EL4 proteasomes or 5 to 13 aa for yeast proteasomes. In both cases, the majority of the peptides were 8–10 mer (Fig. 4, D and E, shaded areas). The masses of a number of abundant products are identical or nearly identical in the digests in Fig. 4, D and E, indicating that many of the peptides produced from SSU by mouse and yeast proteasomes may be identical.

We studied the cleavage site (...P₃P₂P₁ − P_1′P_2′P_3′...) preferences in polypeptides of 20S proteasomes isolated from a variety of organisms including archaebacteria, eubacteria, and nonvertebrate and vertebrate eukaryotes (see Fig. 2). Fig. 5,A shows the results obtained by pool sequencing of digests of the 22-mer Ova_Y249-269 containing the immunodominant Ova_257-264 (SIINFEKL) epitope. The cleavage patterns of all proteasomes of eukaryotic origin, including the murine cell line EL4, the human cell lines T1 and K562, as well as of insects and of yeast are remarkably similar; the predominant cleavage sites reside after the same hydrophobic (L₂₆₄-T₂₆₅) and acidic (E₂₅₆-S₂₅₇) aa. These cleavage sites precisely coincide with the NH₂- and COOH-terminal epitope boundaries. In contrast, the cleavage pattern of archaebacterial proteasomes is clearly different; they prefer to cleave after aromatic (F₂₆₁-E₂₆₂) and aliphatic aa (L₂₆₄-T₂₆₅, L₂₅₅-E₂₅₆), no cleavage after acidic aa is seen, and the major cleavage site destroys the epitope. Analyses of the degradation of longer substrates is shown in Fig. 5, B and C. The 44-mer Ova_239-281 (Fig. 5,B) represents a longer fragment containing the immunodominant SIINFEKL; the 41-mer Ova_37-77 (Fig. 5,C) contains the poorly immunogenic epitope Ova_55-62 (KVVRFDKL), also presented by K^b (22, 35). For bacterial proteasomes, the data in Fig. 5,C highlight the preference for aromatic and aliphatic aa in P₁, whereas cleavage after charged aa is rare but not impossible (Fig. 5,B). Cleavage patterns of all eukaryotic examples, although not fully identical, reflect the same broad but characteristic P₁ specificity spectrum. About 60–65% of the peptide bonds hydrolyzed (e.g., 11/17 in Ova_239-281 by mouse EL4-proteasomes; 11/18 by yeast proteasomes) have an aromatic or a hydrophobic aliphatic aa in the P₁ position. Most of the remaining peptide bonds have either a positively (R) or negatively charged (E, D) aa in the P₁ position. In addition, together with results in Fig. 3 (see above), these data extend to invertebrate eukaryotes our previous finding that proteasomes have a preference for small or polar aa in the P_1′ position of the scissile bond (12). Major cleavage sites are: E₂₅₆-S₂₅₇ and L₂₆₄-T₂₆₅ in Ova_Y249-269 (Fig. 5,A), L₁₀₂-T₁₀₃ and V₁₁₁-S₁₁₂ in BTG1_97-120 (see Fig. 3, B and C), and F₃₅₄-S₃₅₅ in JAK1_348-368 (see Fig. 3 D).

The existence of IFN-γ–inducible proteasomal elements in vertebrates indicates that the proteasomes themselves have evolved by adapting to their novel immunological role. If the specificity of proteasomal proteolysis was drastically and qualitatively altered by the IFN-γ–inducible elements, a significant restricting role of proteasomes in the evolution of MHC would be less likely. Most previous studies on the functional effects of the IFN-γ–inducible β subunits LMP2, LMP7, and MECL1 used short (3–4 aa) fluorogenic substrates, and inconsistent changes in peptidase activities have been reported by different investigators. By digesting polypeptide sequences more likely to resemble physiological proteasome substrates, we compared proteasomes from untreated with that of IFN-γ–treated EL4 cells. Enhanced expression of LMP2 and LMP7 in induced compared to uninduced cells was monitored by Western blot analyses (not shown). Fig. 6, A and B show the HPLC patterns obtained upon digestion of Ova_Y249-269, Fig. 6, C and D that of the BTG1-derived 24 mer. In both cases, we observe that proteasomes from IFN-γ–treated cells generate increased amounts of fragments with hydrophobic COOH termini (YVSGLEQL is peak 4 in Fig. 6, A and B, and TLWVDPYEV is peak 4 in Fig. 6, C and D) and decreased amounts of fragments with acidic COOH termini (YVSGLE is peak 2, and YVSGLEQLE is peak 3 in Fig. 6, A and B; TLWVDPYE is peak 2 in Fig. 6, C and D). This is in line with general preferences of MHC class I molecules, although it does not necessarily result in improved production of each individual epitope; production of BTG1_103-111 (TLWVYPDEV, peak 4 in Fig. 6, C and D) is improved, whereas production of Ova_257-264 (SIINFEKL, contained in peak 5, theoretical mass: 963.14, in Fig. 6, A and B) is impaired. In spite of these quantitative changes, however, the same set of major proteolytic fragments is produced by proteasomes isolated from uninduced and from IFN-γ–induced cells.

The IFN-γ–inducible PA28 enhancer, existing in two homologous forms, α and β (25, 36), binds as a ring-like hexa/heptameric structure to the α endplates of the 20S proteasome (37). PA28-capped 20S proteasomes exhibit enhanced activity towards short fluorogenic model substrates (26, 38–40). Here we analyze the effects of recombinant PA28α on the digestion of the 22-mer Ova_Y249-269 by 20S proteasomes (Fig. 7). Addition of PA28α leads to a slightly increased turnover of the substrate and reverses the ratio between HPLC peak 4 (SIINFEKL and ESIINFEKL) and HPLC peak 5 (SIINFEKLTEWTS and ESIINFEKLTEWTS). The peptides in peak 4 represent dual cleavage fragments; those in peak 5 are produced by a single cleavage only. Thus, in the presence of PA28α, the rate of accumulation of dual cleavage peptides is relatively increased, in line with results recently reported for native PA28 presumably consisting of both α and β isoforms (34). However, the dual cleavage peptides that are more efficiently produced in the presence of the activator are generated also in its absence. Together, these results support the notion that the IFN-γ–inducible proteasomal elements, including PA28 as well as the inducible β-type subunits, modify the 20S-proteasome in such a way that the generation of immunocompetent peptides is quantitatively improved, without drastic alterations of the specificity of proteasomal proteolysis.

The work presented in this paper was stimulated by our observation that each of three proven MHC class I ligands was found as a major proteolytic fragment upon digestion of precursor polypeptides with yeast and/or insect proteasomes. One possible way to rationalize this observation was that the preexisting proteolytic fragments of proteasomes provided an important evolutionary force in shaping the peptide binding groove of MHC class I molecules. Such a mechanism would predict that proteasomal digestion would be the major source of MHC class I ligands in immunologically competent vertebrates. The evidence in favor of this notion is in part circumstantial and is a matter of continuing debate. We show that lactacystin, specific proteasome inhibitor until otherwise demonstrated, inhibits the assembly of about two-thirds of newly synthesized MHC class I heavy chains with β₂m. Inhibition of recognition of a panel of cytotoxic T lymphocyte epitopes by lactacystin has recently been reported (29). Although protease inhibitors cannot provide definitive proof, the combined data on this and other (10) proteasome inhibitors suggest that proteasomes are critically involved in the generation of peptides for assembly with MHC class I. In addition, data from this and other laboratories (12, 34, 41) suggest that proteasomes often are involved in the final proteolytic steps of epitope generation.

A given MHC class I molecule can accommodate large arrays of different peptides. Typically, peptide specificity is constrained by two binding pockets, whereas the rest of the peptide sequence may vary. This is reminiscent of many endoproteases whose active sites can accommodate many different peptide sequences. In many of these enzymes, the primary determinant of substrate specificity is the S₁ subsite, accommodating the P₁ residue of the scissile bond P₁–P_1′. The F pocket of MHC class I molecules accommodates the side chain of the aa at the COOH terminus of the peptide with a strong preference for aliphatic, aromatic, and charged aa. Except for negatively charged residues, the P₁ preference of proteasomes closely resembles the preferences of the F pocket. The F pocket may therefore represent the structural center of the relationship of MHC class I molecules to the proteasome. Nevertheless, no class I allele has been found so far whose F pocket prefers peptides with acidic COOH termini, although these are also efficiently produced by proteasomes of vertebrates and invertebrates. This may be taken as an indication that MHC class I molecules evolved independently of proteasomes and that their preference for aliphatic, aromatic, or positively charged COOH termini has other reasons. However, it is also possible that there was selection against acidic amino acid side chains in the COOH-terminal ligand position. For example, it is conceivable that an acidic COOH-terminal aa side chain interferes with the formation of the H-bond system between conserved polar residues of class I and the free backbone carboxylate of the ligand. During the approach of a peptide with an acidic COOH-terminal aa side chain, the side-chain carboxylate, instead of the backbone carboxylate, may become engaged in H-bond formation, preventing proper anchoring of the ligand in the pockets of the peptide binding groove. Furthermore, peptides with acidic COOH termini may not be correctly handled by other components of the processing and presentation machinery, such as heat shock proteins (hsp), which have been proposed to shuttle peptides in the class I pathway (42), or TAP. However, at least human TAP has been shown to translocate peptides with acidic COOH termini (43).

Our pool-sequencing data strongly support the high degree of conservation among eukaryotes of the cleavage site usage in polypeptides by 20S proteasomes. Although cleavage efficiencies of individual peptide bonds are not always identical, the P₁ specificities of both vertebrate and invertebrate eukaryotic proteasomes are confined by the same broad, but well-defined, limits. The same applies to the P_1′ position where we frequently see small or polar side chains, particularly in major cleavage sites. For example, in the BTG1-derived sequence, yeast 20S proteasomes most efficiently hydrolyzed bonds that contain a small and/or polar P_1′ residue (L-T, V-S), in addition to a suitable P₁ residue. Similar preferences were seen in Ova_Y249-269 (E-S, L-T) and JAK1_348-368 (F-S). Statistical analysis of so far eluted MHC class I ligands revealed an enrichment of small and polar residues in the NH₂-terminal position (most significantly S, but also G and A) (12). In addition, a recent analysis of D^b-binding ligands has shown a significant enrichment for S, T, and C in the NH₂-terminal position of high affinity peptides (44). Moreover, small polar peptide NH₂ termini favor hydrogen bond formation in HLA-B27 (5). Also, TAP is known to have enhanced binding/translocation efficiency for peptides with small or polar residues (45; P. van Endert, personal communication). We favor the hypothesis that these structural features of the peptide-binding groove of MHC class I, and perhaps also of the peptide-binding site of TAP, may represent evolutionary adaptations to conserved features of proteasomal proteolysis.

A striking property of proteasomes is the defined length of their proteolytic fragments. For archaebacterial proteasomes, peptide length centers around 7 and 8 mers (46). Upon digestion of 22- and 44-mer OVA-derived polypeptides by mouse 20S proteasomes, we have recently observed that the majority of the dual cleavage peptide products were 8–11 mer. Peptides of this length dominated among the dual cleavage products at all time points tested, including early in the time course. Many of the peptides of this length are relatively stable, even upon prolonged digestion by proteasomes (12). Most MHC class I molecules bind 8–10/11-mer peptides (3). Nonamers seem to be preferred by most alleles. In this report, we show that a major proportion of the stable peptides generated by yeast and D. melanogaster proteasomes fall into the size range preferred by MHC class I. Thus, the putative evolutionary relationships between proteasomes and MHC class I may include the length of the peptide-binding groove. In addition, the highly conserved clusters of polar aa at both ends of the peptide binding groove may have evolved to facilitate efficient H-bond formation with the ends of the short peptides produced by proteasomes.

IFN-γ is a pivotal cytokine in the function of the immune system and the incorporation of IFN-γ–inducible elements into the structure of proteasomes is highly suggestive of an evolutionary adaptation to the requirements of the immune system. One of the IFN-γ–inducible β subunits, LMP7, is first detected in the nurseshark (47), i.e., at the same phylogenetic step as most other elements of the vertebrate immune system (48). It was therefore possible that MHC molecules evolved independently of proteasomes followed by a unidirectional adaptation of proteasomes to the requirements of MHC class I. On the other hand, as shown in the present paper, the ability of proteasomes to generate potentially immunocompetent peptides preceded the evolution of the MHC, suggesting the reverse order of adaptation. We were therefore interested in understanding the extent of functional modification inflicted upon proteasomes by IFN-γ–inducible elements.

Based on experiments with fluorogenic tri- and tetrapeptides, IFN-γ–dependent alterations in the substrate specificity of proteasomes were reported. In the first reports on this subject, the authors observed that proteasomes isolated from IFN-γ–treated cells gained about twofold higher chymotrypsin-like (Suc-LLVY-MCA–hydrolyzing) and trypsin-like (Boc-LLR-MCA–hydrolyzing) activities (49, 50), but lost about half of the postglutamyl (Clz-LLE-MNA–hydrolyzing) activity (49). The authors suggested that these functional alterations should favor the degradation of proteins to peptides that terminate in hydrophobic and basic residues that are usually found bound to MHC class I. These results were confirmed by some (51), but not by others (18, 52, 53). Studies using polypeptide substrates have yielded results inconsistent with each other and with that obtained with fluorogenic substrates (18, 41). Taken together, the functional consequences of the incorporation of IFN-γ–inducible β subunits remain incompletely understood.

Using as substrates 22- and 24-mer polypeptide sequences corresponding to natural proteins, we show that the characteristic P₁ specificity spectrum of proteasomes remains qualitatively unchanged in proteasomes isolated from IFN-γ–treated cells. However, as shown for the first time with polypeptide substrates, we see enhanced hydrolysis of individual peptide bonds with hydrophobic P₁ residues by proteasomes of IFN-γ–treated cells compared to that of untreated cells. In addition, we see reduced cleavage of neighboring peptide bonds with acidic P₁ residues. Thus, the data presented here for polypeptide substrates agree with that first reported by Gaczynska et al. with short fluorogenic substrates (49). However, together with our results, the rather mild defects of mice genetically deficient in LMP2 or LMP7 (54, 55), as well as the restoration of antigen presentation in LMP2/LMP7/TAP triple-deficient T2 cells by transfection with TAP alone (56–59), argue against drastic qualitative alterations in the cleavage preferences of proteasomes by IFN-γ–inducible β subunits.

Another IFN-γ–inducible element is the enhancer PA28. Recently, Dick et al. reported for the degradation of 19–25-mer substrates by 20S proteasomes, a substantially enhanced rate of accumulation of dual cleavage products by addition of PA28 (34). Here we show that recombinant PA28α is sufficient to induce this effect. This is in line with the recent finding that transfection of PA28α is sufficient to improve the recognition of virus-infected cells by CTLs (60). However, with and without PA28, the same cleavage sites are used and the same products are generated. Thus, the highly conserved general cleavage specificity of the 20S proteasome remains unchanged in the presence of PA28. Nevertheless, PA28 might have been evolved to optimize the capacity of the 20S proteasome for oligopeptide generation, in particular from short substrates.

Evolutionary relationships have been invoked between MHC class I and the hsp70 family of chaperones (61, 62). However, recent structural studies (63, 64) indicate that the homologies between the peptide-binding regions of hsp70 and MHC molecules are less than originally anticipated. Observations suggesting evolutionary links between proteasomes, MHC, TAP, and perhaps hsp70 were recently reported by Kasahara et al. (65) and Katsanis et al. (66). Both the human and the mouse genomes contain three regions with striking homology to the MHC complex, as they comprise genes coding for proteasome β-type subunits, ABC transporters, hsp70, NOTCH, and complement components. One of these regions, in addition, harbors the gene for CD1. They speculate that the MHC complex and these homologous regions might have been generated by duplication of an ancestral syntenic group in jawless fish, i.e., before emergence of T cell recognition. This implies that MHC class I–like molecules, and perhaps also TAP, may have existed before the adaptive immune system.

Due to their role in the degradation of unfolded polypeptides, proteasomes are adapted to cleave hydrophobic sequences from the inside of proteins. Most of the abundant self-peptides eluted from MHC class I are derived from highly conserved hydrophobic regions of a restricted set of evolutionary conserved ubiquitous intracellular proteins (67). In view of the possible existence of MHC class I before T cell recognition, it is conceivable that the functional cooperation between proteasomes and the precursor of MHC class I was originally designed to present self-peptides and to serve a purpose other than self–nonself discrimination, for example, inhibition of NK killing. MHC polymorphism and T cell recognition may have evolved subsequently, thus accommodating the greater variety of foreign peptides.

We are grateful to Uli Birsner for help with the peptide synthesis and Erika Hug for technical assistance. We also thank Dr. Michael Fischer and Dr. Dieter Wolf for providing yeast proteasomes for initial experiments and Dr. Klaus Früh for providing antibodies.