The proteasome has been shown to make the proper C-terminal cleavage for the generation of several immunodominant class I-presented peptides whereas aminopeptidases generate their proper N termini. In this study, we show that these two distinct proteolytic processes are also involved in generating a subdominant OVA peptide KVVRFDKL (K-L). Moreover, proteasome inhibitors did not enhance the presentation of any K-L construct, suggesting that destruction of this peptide by proteasomes, if any, does not limit its presentation. We have further examined in intact cells the influence of residues flanking this epitope on these proteolytic processes. When the N-terminal flanking residues of K-L are fused to an immunodominant OVA peptide SIINFEKL (S-L), the presentation of S-L is reduced as compared with a construct with its natural flanking sequence and was not inhibited (or enhanced) by proteasome inhibitors. Similarly, a reduction in presentation was observed when the C-terminal flanking residues of the subdominant epitope were attached to S-L. A detailed analysis revealed that the Pro at the P1′ position of K-L was responsible for this reduction, and presentation of these C-terminally extended constructs was sensitive to proteasome inhibitor. The study suggests that both the N- and C-terminal flanks of the subdominant peptide are suboptimal for Ag presentation. Moreover, three of four C-terminal residues that flank other subdominant or cryptic epitopes in OVA reduced the presentation of S-L. Therefore, the residues that flank the C termini of several subdominant and cryptic epitopes are often suboptimal for cleavage and may contribute to the phenomenon of immunodominance.
Peptides that are derived from cellular and viral proteins are bound by MHC class I molecules and displayed on the surface of cells (1, 2). Through this mechanism, CD8 T cells can detect and eliminate cells that are synthesizing foreign or abnormal proteins. Most proteins contain many sequences that are predicted to bind to MHC class I molecules (3). However, upon immunization with whole Ags, typically only one (or a few) of their potential epitopes elicit strong (immunodominant) cytotoxic T cell (CTL) responses. A few others may elicit weak (subdominant) responses while the majority elicit no responses (cryptic epitopes) (4). For example, there are 26 potential Kd-binding peptides in the proteins of influenza virus, yet only two generate a strong immune response and another three stimulate a weak response (5). Also, among the 51 potential Kb- and Db-binding peptides in the nucleoprotein and glycoprotein of lymphocytic choriomeningitis virus, only three generate strong primary responses; while responses to another three can be detected in secondary responses (6). Moreover, none of the 18 potential Dd-binding peptides generates a response in H-2d mice (7). Similarly, of the six potential Kb-binding peptides in OVA, only one (SIINFEKL) (S-L) stimulates a strong immune response whereas the others elicit a weak response or none at all (8).
The generation and presentation of peptides on MHC class I molecules, the availability of responsive T cells, and poorly understood immunoregulatory effects can all influence whether immune responses are generated to a particular epitope (4). Where examined, one of the most important factors is whether a specific peptide is correctly generated and presented on MHC class I molecules. The majority of Kd-restricted epitopes from influenza Ags (16 of 26) do not bind to class I molecules with sufficient affinity to be presented (5). Of the 10 peptides that bind to Kd class I molecules, immune responses are only generated to 5. Within this group of peptides, binding to MHC molecules with high or low affinity does not necessarily correlate with the strength of the response (5).
In many of these cases, the subdominant or cryptic epitopes stimulate weak or no responses because of a failure of APCs to generate the antigenic peptide efficiently (4). However, the reason that a particular peptide is not efficiently produced is generally not known. The presented peptide must be of a precise size (typically eight or nine residues) to bind to a class I molecule (9, 10) and, conceivably, cellular proteases may make the appropriate cleavages inefficiently, if at all. In addition, it is possible proteases may destroy potential epitopes by cleaving them internally. For example, purified proteasomes appear to make the appropriate cleavages to generate the subdominant OVA peptide KVVRFDKL (K-L) inefficiently and instead tend to cleave within the epitope itself (11). To understand these phenomena, it is necessary to identify the proteases that make the cleavages that generate or destroy the antigenic peptide and to characterize their specificity.
The bulk of cellular proteins are degraded by proteasomes. The proteolytic core of the proteasome is a 20S cylindrical particle (700 kDa) composed of two outer α rings and two inner β rings (12). In mammalian proteasomes, each β ring contains three active sites that are classified based on their specificity for the P1 residue (the residue on the N-terminal side of the scissile bound) of small fluorogenic substrates. One site (chymotryptic) cleaves preferentially on the carboxylic side of hydrophobic amino acids, another site (tryptic) cleaves after basic residues, and the third site (peptidylglutamyl hydrolyzing or caspase-like) cleaves after acidic ones (13, 14). The specificity of these sites for sequences in proteins or larger oligopeptides is not well defined; however, the proteasome clearly has specificity for additional residues (e.g., in the P4 and P5 positions (Ref. 15)).
Inhibitors of the proteasome can block the generation of a majority of MHC class I-presented peptides in APCs (17), indicating that these proteolytic particles are needed to make at least some of the cleavages that yield antigenic peptides. However, it has not been clear whether proteasomes make the precise cleavages that generate antigenic peptides or whether other proteases are needed to trim the proteasomal products to the appropriate size for presentation. Purified proteasomes can generate antigenic peptides of the proper size as well as many oligopeptides that are too long or too short (11, 18, 19, 20, 21, 22). However, it is still uncertain whether results obtained from in vitro biochemical studies can be extrapolated to in vivo situations (reviewed in Ref. 2). Recently, it has been shown with intact APCs that proteasome inhibitors block the presentation of four immunodominant antigenic peptides containing a single or more C-terminal flanking residues (23, 24, 25). Therefore, proteasomes are responsible for the cleavage that generates the proper C terminus of presented peptides. In contrast, the removal of N-terminal flanking residues from these immunodominant peptides is resistant to proteasome inhibitors (23, 24, 25). These residues are efficiently removed by aminopeptidases both in vitro (25, 26) and in vivo (25). In this report, we investigate whether similar mechanisms are involved in generating a subdominant epitope and whether the N-terminal and/or C-terminal cleavages are limiting its presentation in vivo.
Materials and Methods
A vaccinia construct that contained a full-length cDNA sequence for chicken OVA (V-FLOVA) was obtained from Drs. Jon Yewdell and Jack Bennink (National Institutes of Health, Bethesda, MD), and a vaccinia construct that encoded T7 RNA polymerase (vTF7-3) was obtained from the American Type Culture Collection (VR-2153; Manassas, VA). Both recombinant viruses were propagated in HeLa cells and stored at −80°C.
Cell lines and hybridomas
E36.12.4 APCs were originally derived from E36 cells (hamster lung carcinoma cells) and were stably transfected with murine H-2Kb and ICAM-1 molecules (27). The T hybridoma 1G8 specific for K-L was obtained from Dr. James McCluskey (Flinders Medical Center) (28). The T cell hybridoma RF33.70 specific for S-L was described previously (29). These T cell hybridomas are highly specific and do not recognize unrelated peptides. All cells were maintained in RPMI 1640 medium with 5% FCS (Atlanta Biologicals, Norcross, GA).
Peptides and proteasome inhibitor
The peptide S-L, corresponding to OVA residues 257–264, and the peptide K-L, corresponding to OVA residues 55–62, were purchased from Molecular Resources (Colorado State University, Fort Collins, CO). These two peptides were purified by reverse phase HPLC and were at least 95% pure. The proteasome inhibitor clasto-lactacystin β-lactone (β-lactone) was kindly provided by Dr. Julian Adams (ProScript, Cambridge, MA) and was dissolved at 10 μM in DMSO and stored at −80°C.
Synthetic minigenes encoding the antigenic peptides corresponding to chicken OVA OVA55–62 (K-L) or OVA257–264 (S-L) with or without either natural or mutated N- or C-terminal flanking residues were constructed from synthetic oligonucleotides. The coding sequences for all of the plasmid constructs were preceded by a translation initiation codon (ATG), and therefore all of the translated products start with Met, which should be rapidly removed by cytosolic methionine aminopeptidase. The oligonucleotides were cloned into pBluescript SK under the control of the T7 RNA promoter as described previously (23). The various Ag constructs are shown in Fig. 3⇓B.
The cDNA of OVA was cloned into pBluescript SK under the control of the T7 promoter using HindIII and XbaI sites. cDNAs of OVA with position 265 mutated from Thr → Pro (p.OVA.P) and from Thr → Leu (p.OVA.L) were produced using the Clontech transformer site-directed mutagenesis kit (Clontech Laboratories, Palo Alto, CA). Briefly, the target plasmid (p.OVA) was denatured and then a selection primer (which converts an ScaI site to a DrdI site) GTGACTGGTGAGTCCTCAACCAAGTC and a mutagenic primer CAACTTTGAAAAACTGCCTGAATGGACCAG were annealed to the ssDNA, and the second strand was synthesized with T4 DNA polymerase and ligase. Repair-deficient competent Escherichia coli were transformed, and recombinant isolated plasmids were selected by cutting nonrecombinant plasmids with ScaI. DH5α-competent E. coli were then transformed, and plasmid minipreps were digested with DrdI. Positives were then sequenced to check for mutagenic insertion.
Ag presentation assay
Ags were introduced into H-2Kb APCs (E36/Kb) either by recombinant vaccinia infection or by infection plus plasmid transfection. E36 APCs were first infected with V-FLOVA (multiplicity of infection (MOI),5 10) for 2 h or infected with vTF7-3 first and then transfected with Lipofectin-bound plasmid, as described previously (23), for various lengths of time as described in the figure legends. These Ag-expressing APCs were fixed to stop further Ag processing, washed, and then incubated with T cell hybridomas (1 × 105/well) for 20 h. The amount of plasmid transfected and the length of incubation before fixation were titrated to obtain limiting conditions of Ag presentation and these limiting conditions were then used in all experiments. The expression level of MHC class I-peptides complex on the surface of APCs was then evaluated by their ability to stimulate the peptide-specific T hybridomas to produce IL-2. Levels of IL-2 in the supernatant were measured by the CTLL cell proliferation (cpm) (30). For all of the experiments, the peptide-specific T hybridomas generated a very low level of IL-2 (cpm <2500) when no Ag (APC# = 0) was present.
Generation of subdominant and dominant class I-presented peptides from full-length and short oligopeptide constructs of OVA
In initial experiments, we analyzed the presentation on Kb of K-L (OVA55–62) and S-L (OVA257–264) peptides from OVA. E36 (H-2Kb) APCs were infected with a vaccinia recombinant containing the full-length OVA sequence (V-FLOVA) and fixed after 2 h. The presentation of K-L or S-L peptides was assayed by coculturing these cells with specific T cell hybridomas and measuring their production of IL-2. As shown in Fig. 1⇓A, the response of the K-L-specific hybridoma (1G8) was much lower than that of the S-L-specific clone (RF33.70) (Fig. 1⇓A). In contrast, when the APCs were incubated with a titration of K-L or S-L peptides added to the medium, the K-L-specific hybridoma responded to slightly lower concentrations of peptide than the S-L-specific hybrid (Figs. 1⇓B and 2, B and D). Since the K-L-reactive hybridoma was at least as sensitive as the S-L-specific one, our results indicate that K-L is presented more poorly than S-L in cells expressing the full-length OVA protein (Fig. 1⇓A). These results confirm the findings of earlier reports in which the K-L-specific response was much lower than the S-L-specific response measured by the GA4.2 T hybridoma when OVA protein was introduced into the APCs by electroporation (28).
We also compared the presentation of these same epitopes from minigenes encoding the eight-residue K-L and S-L peptides. E36/Kb APCs were infected with a vaccinia recombinant expressing T7 polymerase and were then transfected with increasing amounts of plasmids encoding the K-L or S-L peptides (preceded by Met) for 75 min under the control of a T7 promoter. After this incubation, further synthesis and presentation of these constructs were prevented by fixing the APCs. As we have previously described, the level of Ag presentation can be controlled by varying the dose of plasmid (Fig. 2⇓) or length of incubation (23). As shown in Fig. 2⇓, when S-L and K-L peptides were expressed endogenously in the APCs, the K-L-specific hybridoma responded to somewhat lower numbers of these APCs than the S-L-specific hybrid, even under conditions of limiting Ag (Fig. 2⇓, A and C). Therefore, once the K-L and S-L peptides are generated, they are presented with similar efficiency in this system. These results, along with those in Fig. 1⇑, indicate that the responses of the K-L-specific 1G8 and S-L-specific RF33.70 T hybridomas can be compared to determine the relative amount of K-L and S-L being generated and presented on Kb. The lower presentation efficiency of the K-L epitope from OVA protein (Fig. 1⇑A) but not from the minimal short construct (8-mer) (Fig. 2⇓) further indicates that this peptide is generated less efficiently than S-L.
For other Ags, both adjacent and distant sequences can influence the efficiency with which a particular peptide is generated. To examine whether sequences immediately flanking the K-L and S-L peptides were influencing their presentation, we constructed minigenes encoding S-L with its natural flanking sequences ((M)VSGLEQLES-LTEWTS) or the flanking sequences of K-L ((M)DSTRTQINS-LPGFGD) which were cloned into plasmids under the control of the T7 promoter (Fig. 3⇓B). These constructs were then expressed in APCs by the T7 polymerase transfection method, and Ag presentation was measured. When S-L was flanked by its natural flanking residues ((M)VSGLEQLES-LTEWTS), strong presentation was observed with 7 μg of plasmid (Fig. 4⇓A). However, when S-L was surrounded by the flanking residues of K-L, S-L presentation was low even with 20 μg of plasmid (Fig. 4⇓B). These data indicate that the flanking sequences of K-L negatively affect peptide generation and are consistent with earlier results (11).
Distinct proteolytic processes are involved in generating the N and C termini of the subdominant OVA epitope
Given the results described above, it is important to understand whether the residues that flank S-L and K-L are removed by similar mechanisms. Recently, we and others have demonstrated that in APCs distinct proteolytic processes are involved in the cleavages that generate the proper N and C termini of several immunodominant peptides, including S-L (23). The C-terminal cleavage is made by proteasomes and can be blocked by proteasome inhibitors whereas the N terminus can be trimmed by aminopeptidases, and this process is resistant to proteasome inhibitors (23, 24, 25). We therefore examined whether the removal of N- or C-terminal residues that flank K-L was also blocked by proteasome inhibitors. A cDNA for full-length OVA or minigene construct consisting of the K-L peptide with five additional residues of natural flanking sequence ((M)K-LPGFGD) was constructed and expressed in APCs. The presentation of the K-L peptide was subsequently detected using the T cell hybridoma assay. When the APCs were treated with the proteasome inhibitor β-lactone, presentation of K-L from full-length OVA (Fig. 5⇓A) and from the C-terminally extended construct (M)K-LPGFGD (Fig. 5⇓B) was markedly reduced. In contrast, this proteasome inhibitor did not block the presentation of the octamer K-L peptide expressed from a minigene (which does not require proteolysis for presentation) (Fig. 5⇓D). These results indicate that the proteasome inhibitor blocks presentation by interfering with the proteolytic removal of the C-terminal residues of (M)K-LPGFGD and is not affecting other steps in the class I pathway. Interestingly, the proteasome inhibitor also did not enhance the presentation of the K-L minigene (Fig. 5⇓D). This is surprising because earlier reports using purified 20S proteasomes found multiple cleavages within the K-L epitope itself, with the major cleavage between Arg (R) and Phe (F) (KVVR-cleavage-FDKL) (11), resulting in the destruction of this peptide. These in vitro studies concluded that destruction of this peptide by proteasomes was one of the reasons that led to its subdominance; however, our findings that the presentation of K-L was similar to that of S-L (Fig. 2⇑) and that proteasome inhibitors did not enhance the presentation of K-L from the minimal 8-mer construct (Fig. 5⇓D) in intact cells do not support this conclusion.
Similar experiments were performed analyzing the presentation of the K-L peptide with an eight-residue extension of natural flanking residues from its N terminus ((M)DSTRTQINK-L). In contrast to the results with the C-terminally extended construct, the proteasome inhibitor did not block the presentation of the (M)DSTRTQINK-L peptide (Fig. 5⇑E). It is again notable that the proteasome inhibitor also did not enhance the presentation of the (M)DSTRTQINK-L peptide, further suggesting that proteasome-mediated destruction of the K-L epitope does not limit its presentation in vivo. Similar results were obtained using even higher concentrations of β-lactone (20 μM; data not shown). The failure of the proteasome inhibitor to block presentation of this N-terminally extended construct was similar to findings reported earlier for S-L and three other viral immunodominant peptides (23, 24, 25). Therefore, the generation of the proper N and C termini of both subdominant and immunodominant peptides involve distinct proteolytic processes. Given our results described above, one or both of these processes must remove the flanking residues of K-L less efficiently than the sequences surrounding S-L.
N-terminal flanking sequences can affect the efficiency of Ag presentation
To address whether the N-terminal flanking sequence of the subdominant K-L peptide impairs Ag presentation, we constructed a chimeric cDNA encoding these residues (DSTRTQIN) as amino-terminal flanking sequence on the immunodominant S-L peptide ((M)DSTRTQINS-L). This construct was expressed in APCs and its presentation was compared with an S-L construct with natural flanking residues ((M)VSGLEQLES-L). The construct with the natural flanking residues was presented considerably better than the chimeric construct with the subdominant flanking sequences (Fig. 4⇑C). These results demonstrate that compared with (M)VSGLEQLE, the (M)DSTRTQIN residues are trimmed less efficiently from the N terminus of S-L.
C-terminal flanking sequences can also affect the efficiency of Ag presentation
We also examined whether the C-terminal flanking residues of K-L contributed to its suboptimal presentation. Plasmids were made that encoded S-L with its natural C-terminal flanking residues (S-LTEWTS) or with those from K-L (S-LPGFGD) (Fig. 3⇑B). These constructs were expressed in APCs and S-L presentation was measured. S-L presentation was decreased significantly when its C-terminal flank (S-LTEWTS) was replaced with that of K-L (S-LPGFGD) (Fig. 6⇓A). Similar experiments were done comparing the presentation of K-L constructs with natural flanking residues (K-LPGFGD) or with the flanking sequence from the immunodominant peptide S-L (K-LTEWTS). In this case, the flanking residues from S-L markedly augmented the presentation of K-L (Fig. 6⇓B). The generation of the presented peptides from all of these constructs was still dependent on the proteasome because proteasome inhibitors blocked their presentation (data not shown). These results indicate that the removal of the C-terminal flanking residues of K-L is also less efficient than for those flanking S-L and thus limits Ag presentation.
The result that the C-terminal flanks of K-L limited Ag presentation led us to further characterize which flanking residues were unfavorable. In preliminary experiments we found that removal of the P5′ (D), P4′ (G), and P3′ (F) residues did not significantly alter the efficiency of presentation (data not shown). Given these results, and since previous data have suggested that the P1′ residue can affect presentation (11, 31), we replaced the P1′ residue (Pro) of K-L with the P1′ residue of S-L (Thr) and studied the generation of K-L. The K-L presentation was enhanced when the P1′ residue was changed from Pro to Thr (K-LPGFGD→K-LTGFGD) (Fig. 6⇑C). The generation of the K-L peptide from this Thr-containing construct was also blocked by proteasome inhibitors (Fig. 5⇑C), indicating that the proteasome has specificity for this P1′ residue. Consistent with this result, the enhanced presentation of the chimeric construct of K-L with the C-terminal flanking residues of S-L (K-LTEWTS) was markedly reduced when the P1′ Thr was mutated to Pro (K-LPEWTS) (Fig. 6⇑B). We also performed the same kind of analysis for the S-L epitope. Mutation of its natural P1′ residue (Thr) to the P1′ residue of K-L (Pro) reduced its Ag presentation (Fig. 6⇑A).These results show that the P1′ Pro residue contributes to the suboptimal removal of the K-L C-terminal flanking residues.
It is uncertain whether short oligopeptide constructs will always be cleaved identically to the same sequence in the corresponding full-length protein, e.g., oligopeptides are not substrates for ubiquitination, whereas full-length proteins (including OVA) are polyubiquitinated. Moreover, Eisenlohr’s group (32) has described one example where the generation of presented peptides from an oligopeptide precursor is different from a protein. We therefore further investigated whether the P1′ residue would affect the generation of the S-L peptide from full-length OVA. We made plasmids that encoded the full-length OVA protein (Fig. 3⇑A) with the C-terminal P1′ residue of S-L changed from T265→P265 (p.OVA.P) and from T265→L265 (p.OVA.L), a residue that was also shown to decrease S-L generation from a 9-mer S-L.L (31). As shown in Fig. 7⇓, S-L presentation from the mutated OVA protein OVA.P was significantly reduced when compared with that of wild-type OVA OVA (Fig. 7⇓A). Mutation of the P1′ residue to L (OVA.L) also significantly decreased S-L generation from the whole OVA protein (Fig. 7⇓B). These results further generalized the importance of the P1′ residue for the generation of class I-presented peptides.
Influence on Ag presentation of C-terminal flanking residues from cryptic OVA epitopes
In a final set of experiments, we explored whether C-terminal flanks of cryptic peptides within OVA may also limit Ag presentation. OVA has four more peptides that contain the proper motif for binding to H-2Kb molecules but that fail to stimulate immune responses (cryptic epitopes) (8, 33). Because of the lack of T cell hybridomas specific for these subdominant peptides, we constructed plasmids that encoded chimeric oligopeptides, which had the C-terminal flanks of OVA12–19 (KVHHA), OVA25–32 (AIMSA), OVA107–114 (PEYLQ), and OVA176–183 (WEKTF) linked to S-L. S-L generation from these constructs was then studied using the T hybridoma RF33.70. Interestingly, two of these C-terminal flanks, PEYLQ and WEKTF, markedly decreased S-L generation (Fig. 8⇓, A and B). It is notable that one of these has a Pro at the P1′ position, like K-L, but the other does not. In addition, the sequence KVHHA also consistently lowered the presentation of S-L as compared with that of TEWTS (Fig. 8⇓C), although the magnitude of this effect was modest. Surprisingly, one of the flanks, AIMSA, enhanced the S-L generation (Fig. 8⇓D). Therefore, it seems that the majority of cryptic epitopes have suboptimal sequences flanking their carboxyl-termini.
There are many sequences within Ags that are predicted to bind to MHC class I molecules (3); however, the majority of these fail to stimulate immune responses or are only weakly immunogenic (4). In many cases, this is due to a failure of cells to generate the appropriate presented peptide (5). To better predict what epitopes in proteins are presented, it is necessary to define in intact cells the proteases that produce and/or destroy antigenic peptides and then to understand their specificity. We have examined this issue for the weakly immunogenic (subdominant) K-L epitope in OVA.
We confirmed that the K-L epitope is presented less efficiently from OVA than the immunodominant epitope S-L, as has been described previously (11). In our assay system, when eight-residue (plus the initiating Met) peptides K-L or S-L are expressed from minigenes, the K-L peptide is presented with similar efficiency as the S-L peptide (Fig. 2⇑). However, K-L presentation is weaker than S-L when they are generated from full-length OVA. Therefore, flanking residues reduce the efficiency of K-L presentation. In addition, the K-L peptide has been reported to bind to Kb with lower affinity than S-L (28), which would further contribute to its weaker presentation (although the affinity of these peptides would not be measured in our assays which use T cell hybridomas that have similar reactivity to a titration of the K-L and S-L peptides (Figs. 1⇑B and 2, B and D).
Earlier studies have shown that purified proteasomes preferentially cleave octamer K-L or extended K-L constructs within the K-L sequence, thereby destroying the presented peptide (11). This was suggested to contribute to the poor presentation of this epitope. However, we find no evidence to support this mechanism in intact cells. In our assays, the K-L peptide sequence is as efficiently presented as S-L when introduced as minigenes that encode their sequences. More important, the presentation of K-L is not enhanced by proteasome inhibitors even when they are present at very high concentrations (which are sufficient to inhibit protein degradation (34) and Ag presentation). Similarly, proteasome inhibitors do not enhance the presentation of K-L from constructs that have N-terminal flanking residues that might theoretically be needed for binding to proteasomes and cleavage within the K-L sequence. Therefore, destruction of the K-L epitope by proteasomes is not limiting its presentation. The inefficient generation of the K-L epitope is instead due to impaired removal of its flanking residues.
Earlier studies had shown that flanking residues could influence the presentation of class I-presented epitopes (35, 36). However, the proteases that were responsible for removing these sequences were not known, and, therefore, it was uncertain how these flanking sequences were exerting their effects. In most earlier studies proteolytic mechanisms were not examined, or in one case which examined whether the proteasome was involved (for the generation of the influenza nucleoprotein (NP) 147–155 peptide), proteasome inhibitors did not block the presentation of this epitope (37). Therefore, we sought to define the proteolytic mechanisms that removed the N-terminal and C-terminal flanking residues from K-L and determine whether these processes were influenced by these flanking sequences.
The presentation of K-L from OVA was blocked by proteasome inhibitors (Fig. 5⇑A). In contrast, the presentation of the eight-residue K-L peptide from a minigene was not inhibited by these agents. Therefore, the proteasome is required for generating the K-L epitope from longer constructs. Proteasome inhibitors also blocked the presentation of K-L from constructs that had carboxyl-terminal flanking residues (Fig. 5⇑, B and C). These results indicate that the proteasome is responsible for removing the carboxyl-terminal flank of this epitope. In contrast, proteasome inhibitors did not block the presentation of K-L from constructs with N-terminal extensions (Fig. 5⇑E). These results are identical to those with several immunodominant epitopes where aminopeptidases were found to trim N-terminal extensions while proteasomes removed the C-terminal flanking residues. We conclude that the subdominant K-L epitope is not generated by fundamentally different mechanisms than the immunodominant peptides.
Since two distinct proteolytic processes are involved in removing the N- and C-terminal flanking residues (2), we examined whether one or both of these processes was responsible for the suboptimal generation of K-L. We found that the presentation of S-L was reduced when its N-terminal flanking residues were replaced with those from K-L. Presumably, aminopeptidases hydrolyze this sequence more slowly than the natural flank of S-L. This is somewhat unexpected because earlier studies had found that alterations in N-terminal flanking sequences did not in general affect Ag presentation (31, 38). This might indicate that the N-terminal trimming mechanism can remove most, but not all, sequences with similar efficiency. The cytoplasm contains several aminopeptidases (39) and perhaps in combination they can remove most residues at similar rates. However, it is also possible that earlier assays were not sufficiently sensitive to detect quantitative differences in presentation. In any case, we provide clear evidence that N-terminal flanking residues can affect the generation of a presented peptide.
We also found that the presentation of an S-L minigene was reduced when its C-terminal flanking residues were replaced with those of K-L. In contrast, the presentation of a K-L minigene was enhanced when its C-terminal flank was replaced with the corresponding S-L sequences (Fig. 6⇑). The suboptimal effect of the K-L flank was due to the Pro residue in the P1′ position. These results are consistent with earlier studies that Pro at this position of extended oligopeptide constructs reduced Ag presentation of S-L (11, 31). Since antigenic peptide generation from oligopeptides may not behave similar to full-length constructs, we extended the studies to the full-length OVA constructs with mutated Thr265 mutated to Pro and found that presentation of S-L was reduced. Pro residues alter polypeptide structure and it has been suggested that proteasomes do not efficiently cleave within Pro-containing sequences (40). However, we also found that mutation of Thr265 → Leu in OVA reduced Ag presentation, which was consistent with earlier results with oligopeptide constructs (31). The results reinforce the importance of P1′ residues for adjacent peptide generation.
Since the proteasome is responsible for making the correct C-terminal cleavage in S-L and K-L, one of the active sites of this proteolytic particle must have specificity for the P1′ residue, a property that has not been previously appreciated. In our study, removal of the P5′, P4′, and P3′ residues did not seem to alter the efficiencies of presenting S-L (data not shown). However, it may be the case that other proteasome active sites have specificity for different P′ residues. For example, we have observed that the removal of the C-terminal flanking residues from an influenza NP-derived peptide NP366–374 by proteasomes requires three C-terminal flanking residues (25). This may indicate that one of the active sites has specificity for the P3′ position.
The TAP transporter has specificity for the C-terminal residue of peptides (41, 42, 43). However, it should be noted that this almost certainly does not account for any limitations on the presentation of our C-terminally extended constructs. We show that the removal of these flanks is dependent on proteasomes and therefore must occur in the cytoplasm. Moreover, we and others have shown that there is no C-terminal trimming activity in the endoplasmic reticulum (23, 44). Our results with K-L suggest that flanking residues probably also contribute to the subdominant status of this epitope, in addition to the low binding affinity of K-L to the MHC molecule as described previously (28). Moreover, we find similar suboptimal flanks at the C terminus of a majority of cryptic epitopes in OVA. Therefore, by affecting proteolytic cleavages, C-terminal flanking sequences may often influence the immunogenicity of potential epitopes. It remains to be determined which of the P′ residues is responsible for the suboptimal removal of the cryptic flanks. However, like K-L, one of these has a Pro at the P1′ position. Interestingly, the C-terminal flank from one cryptic epitope enhanced the presentation of S-L (Fig. 8⇑D). Presumably this particular epitope is cryptic for reasons other than removal of its C-terminal flank or this sequence behaves differently when fused to S-L. This finding does indicate that the sequences flanking immunodominant epitopes may not necessarily be the most optimal one for cleavage.
As we move toward a better understanding of the specificity of the proteases that make the N- and C-terminal cleavages of presented peptides, the ability to predict the epitopes in proteins that get presented should be improved. Moreover, the density of MHC/peptide is critical to activating CTLs and determines the magnitude of CTL responses (16, 45, 46). Therefore, it should be possible to engineer proteins to improve the production in cells of specific epitopes, which might enhance the immunogenicity of vaccines.
We thank Drs. James McCluskey and Wensan Chen (Flinders Medical Center) for the 1G8 T hybridoma. We thank Drs. L. J. Shen and Ian York for critical reading of this manuscript.
↵1 This work was supported by grant from the National Institutes of Health (to K.L.R. A.X.Y.M.) was supported by a training grant from the National Institutes of Health, and S.F.L.v.L. was supported by grants from the Dutch Cancer Society and the Jan Kornelis de Cock Foundation.
↵2 Current address: Children’s Cancer Center, University Hospital Groningen, Groningen, The Netherlands.
↵3 Current address: Department of Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, 300 Brookline Avenue, Boston, MA 02215.
↵4 Address correspondence and reprint requests to Dr. Kenneth Rock, University of Massachusetts Medical School, Worcester, MA 01655. E-mail address:
↵5 Abbreviation used in this paper: MOI, multiplicity of infection.
- Received September 23, 1999.
- Accepted February 2, 2000.
- Copyright © 2000 by The American Association of Immunologists