Antitopes Define Preferential Proteasomal Cleavage Site Usage*

Protein degradation by proteasomes is a major source of peptides presented by major histocompatibility v complex class I proteins. Importantly, interferon γ-induced immunoproteasomes in many cases strongly enhance the generation of antigenic peptides both in vitro and in vivo. Whether this is due to enhanced substrate turnover or to a change in proteasomal cleavage specificity is, however, largely unresolved. To overcome the problems of peptide quantification inherent to mass spectrometry, we introduced the “antitope” as substrate-specific internal standard. The antitope is a non-functional peptide that is generated by proteasomal cleavage within the epitope, resulting in partial overlaps with the functional epitope. Using antitopes as internal standards we demonstrate that the observed enhanced immunoproteasome-dependent presentation of the bacterial listeriolysin O T-cell epitope LLO(296–304) is indeed due to altered cleavage preferences. This method is also applicable to other major histocompatibility class I epitopes as is shown for two potential epitopes derived from Coxsackievirus.

Proteasomes are multisubunit protease complexes that perform most of the non-lysosomal ATP-dependent proteolysis in eukaryotic cells. The 26S proteasome complex is responsible for the degradation of polyubiquitylated proteins and is formed by the so-called 20S proteasome catalytic core that is capped at one or both ends by the 19S regulatory particles (1,2). The catalytic 20S core proteasome itself is responsible for processing of denatured non-ubiquitylated proteins or for further processing of longer peptides of different origin. Overall, products of proteasomal cleavage are peptides of 3-20 amino acids in length (3). Among these cleavage products are peptides that fulfill the criteria for binding to MHC 2 class I molecules with regard to appropriate length and the correct position of anchor residues. Thus, proteasomes generate the adequate C terminus of most investigated MHC class I ligands, whereas the N terminus is often elongated by two or three residues, requiring N-terminal trimming by amino peptidases to allow efficient binding to MHC class I molecules (4 -6).
The 20S proteasome catalytic core is built from 28 subunits arranged as four heptameric rings (7). The two outer rings contain the seven structural ␣ subunits; the two inner rings each contain seven ␤ subunits (␤ 1 -␤ 7 ), of which three (␤1, ␤2, ␤5) exert catalytic activity. Stimulation of cells with IFN-␥ induces the expression of three additional catalytic proteasome subunits. The cytokine-inducible subunits ␤1i, ␤2i, and ␤5i are incorporated into the 20S proteasome core upon its de novo synthesis, thus forming the so-called immunoproteasomes (8,9). Their incorporation alters the catalytic characteristics of the 20S proteasome core. A large number of studies have shown that 20S proteasomes liberate MHC class I epitopes or their N-terminal-elongated precursors out of large polypeptides or denatured proteins in vitro with rates that correlate well with those of the production of antigenic peptides in intact cells (10 -12). Furthermore, detailed studies on the functional importance of immunoproteasomes revealed that in particular the generation and presentation of viral epitopes is strongly enhanced in the presence of immunoproteasomes and that their function is tightly connected with the early phases of an antiviral immune response (13)(14)(15). Recently, we also demonstrated that the efficient presentation of the bacterial Listeriolysin O T-cell epitope LLO(296 -304) requires the presence of immunoproteasomes (16).
Although the biological relevance of immunoproteasomes for an appropriate cellular immune response gradually emerges, the reason for this at the molecular level is by far less clear. In fact, it appears that, depending on the epitope, immunoproteasomes exert their function quite differently. Immunoproteasomes seem to favor cleavage behind hydrophobic residues, which are the predominant anchor residues required for MHC class I binding of antigenic peptides and thus may explain their positive effect on antigen generation (17,18). On the other hand, there exist an increasing number of examples demonstrating that immunoproteasomes also can down-regulate epitope production by enhanced cleavage within the epitope, i.e. epitope destruction or less efficient liberation (19,20). Furthermore, in vitro experiments suggest that immunoproteasomes often degrade substrates considerably faster than their constitutive counterparts. Thus, improved epitope generation might not necessarily be the consequence of altered cleavage specificities but may also be the result of enhanced substrate turnover. The latter argument seems to be supported by the finding that there is, in general, little difference in the overall quality of peptides that are generated by either of the two types of proteasomes (18). In consequence, in most in vitro experiments it is difficult to decide whether the observed positive effects of immunoproteasomes on epitope generation are due to an increased substrate turnover or altered cleavage specificity.
Based on mass spectrometry and the analysis of the bacterial Listeriolysin O T-cell epitope LLO(296 -304) and of three other epitopes, one self-and two virus-derived epitopes, we here report an experimental approach that takes advantage of a substrate-specific non-functional cleavage product, named antitope, that allows one to determine whether predominant generation of a given epitope is due to enhanced substrate turnover or altered cleavage site preference.

MATERIALS AND METHODS
Proteasome Purification from Infected Mice-20S proteasomes were isolated from tissues of wild-type C57Bl/6 mice as well as of IFN-␥ receptor-deficient (IFN␥R Ϫ/Ϫ ) mice. Control mice (day 0) or mice infected with 5 ϫ 10 3 Listeria monocytogenes at days 2, 4, or 6 post infection (p.i.) as indicated in the figures were sacrificed and organs were frozen in liquid nitrogen (16). Further purification steps followed as described (21). Because IFN␥R Ϫ/Ϫ mice died after day 4 p.i., organs were taken from mice at days 2 and 4 p.i.
Proteasome Purification from Cells-T2 and T2.27mp cells were cultured in RPMI 1640 supplemented with 10% fetal calf serum and antibiotics under 5% CO 2 . Proteasomes were isolated from T2 cells for standard proteasomes and from T2.27mp cells that stably express all three immunosubunits for immunoproteasomes. Proteasome isolation was performed as described previously (22). The proteasome was measured at 280 nm and analyzed by SDS-PAGE. The yield was calculated at ϳ90 -95%. Equal amounts of proteasome in assays were adjusted by densitometry of proteasomes in Coomassie-stained gels. Incorporation of immunosubunits was controlled by twodimensional PAGE and Coomassie staining for proteasomes isolated from liver and small intestines (16) or for proteasomes isolated from T2 or T2.27 cells by Western blot using specific antibodies (22).
Peptides-Peptides were synthesized using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) methodology (0.1 mmol) on an Applied Biosystems 433A automated synthesizer. The peptides were purified by HPLC and analyzed by mass spectrometry (ABI Voyager DE PRO).
Peptide Identification, Quantification, and Statistical Analysis-Peptides generated by proteasomes were analyzed by reverse phase HPLC, system HP1100 (Hewlett-Packard) equipped with an RPC C2/C18 SC 2.1/10 column (GE Healthcare). Analysis was performed online with an LCQ ion trap mass spectrometer equipped with an electrospray ion source (ThermoQuest) (21). In Fig. 3, A and B, the kinetic of one representative digest is shown. For analyses of linearity of peptide signals in LC-ESI, different amounts of LLO precursor and epitope peptides were separated on HPLC followed by LC-ESI ion trap MS. For statistical analysis of the LLO and the Hsp60 peptides, ion counts were normalized to the 9GPS standard peptide, which was added in equal amounts to each stopped reaction and the mass of which does not interfere with the masses of any proteasomal cleavage products of analyzed peptides. For LLO(291-317) and Hsp60(171-200) four replicates were averaged each. Statistical significance was attributed to differences in one-tailed, heteroscedastic t-test (p Ͻ 0.05) or in the non-parametric Mann-Whitney t-test (p Ͻ 0.05). The processing of CVB3-derived peptides P2C(1161-1190) and P3D(2158 -2185) was repeated two and three times, respectively.
To determine whether immunoproteasomes display increased cleavage site preference, the generation of indicated antitopes that excludes processing of the correct epitopes was used as reference. Ion counts of the antitope were set as 1 in each reaction, and relative amount of corresponding epitope or precursor to the antitope was calculated (means are shown). The analyzed antitope sequences used are indicated in Figs. 1C, 4A, and 5A.

RESULTS
20S immunoproteasome (i20S proteasomes) -dependent in vitro antigen-processing assays using synthetic polypeptides as substrates containing an MHC class I epitope in many cases exhibit an enhanced liberation of the epitope when compared with that from standard proteasomes (s20S proteasomes). Such an effect could be the result of either an overall faster substrate turnover rate or of altered cleavage site preferences of i20S proteasomes that favor the liberation of the epitopes.
Immunoproteasomes Exhibit a Faster Substrate Turnover Rate-To distinguish between the two above possibilities we used identical amounts of purified s20S and i20S proteasomes for in vitro processing assays and analyzed the turnover rates of LLO-derived 27-mer polypeptides that harbor the LLO(296 -304) T-cell epitope. As shown in Fig. 1A, i20S proteasomes isolated from T2 cells that were transfected with immunosubunits (T2.27mp) as well as from liver of Listeria-infected mice reveal a significantly accelerated degradation of the polypeptide substrate. Analysis of substrate turnover by s20S proteasomes from liver of uninfected control mice (WT) and i20S protea-somes from liver of Listeria-infected mice again showed an accelerated substrate turnover by i20S proteasomes (Fig. 1B). In contrast, 20S proteasomes isolated from livers of IFNR Ϫ/Ϫ mice lacking i20S proteasomes (supplemental Fig. S1) exhibit a diminished substrate degradation rate. This demonstrates that the increased substrate turnover rates are indeed due to i20S proteasomes (Fig. 1B). Importantly, accelerated substrate degradation by i20S-proteasomes is not restricted to the LLO-derived 27-mer polypeptide but is also observed for epitope-containing polypeptides of other origins (Figs. 4B and 5B).
To study whether the quality of peptides generated from LLO 27-mer polypeptides differ between s20S and i20S proteasomes, the peptide fragments generated by the two proteasome types were compared (Fig. 1C). No major differences in the quality of peptides generated by either subtype were observed. Only two additional fragments were identified by mass spectrometry in digests with i20S proteasomes. However, both resulted from the usage of cleavage sites that are in common with those of s20S proteasomes. Remarkably, with the cleavage at Val-122Tyr-13 we detected a strong cleavage site within the epitope whose usage could be crucial for the amount of epitope or epitope precursor peptides generated (Fig. 1C). The peptide fragment Ser-4-Val-12 resulting from cleavage at Val-122Tyr-13 possesses an N-terminal residue (Ser-4) identical to the one of the longest epitope precursor peptide but lacks the correct C terminus of the LLO epitope. We therefore decided to use this peptide fragment (Ser-4-Val-12) as suitable internal standard for the calculation of relative cleavage site preference. Because this peptide fragment precluded the generation of a functional MHC class I epitope it was named antitope.
To determine relative amounts of proteasome-generated peptide fragments by semiquantitative mass spectrometric analysis, an interference of peptides of interest must be excluded. For example, more efficient ionization of one peptide may result in better detection compared with another poorly ionized peptide although identical or even less absolute amounts are generated by the proteasome. Therefore, we first demonstrated the linearity between ion current and peptide amount in the range of up to 200 pmol for the 27-mer substrate, both epitopes (Val-6-Leu-14; Ala-7-Leu-14), the precursor peptide Ser-4-Leu-14, and the antitope Ser-4-Val-12. As shown in Fig. 2, all peptides reveal a peptide concentration-dependent linearity whereby the absolute intensities differ between the peptides. In consequence, an increase in the relative ratio of ESI-MS intensity between two of these peptides will be due to an increase in the relative ratio of respective peptide amounts.
Furthermore, possible interferences between peptides in complex mixtures may quench the signal. Therefore, we analyzed the same peptides in different mixtures containing different ratios of the peptides. Although we observed interferences between the peptides and also a decrease in ion intensities, the mass/signal ratios between the peptides remained the same (data not shown).
Cleavage Site Preference or Degradation Rate-Using the antitope as defined above as internal substrate-specific standard, we next investigated whether observed enhanced liberation of a T-cell epitope by i20S proteasomes is due to an i20S proteasome-dependent higher substrate degradation rate or the result of altered cleavage site usage. To test this, the LLOderived 27-mer polypeptide was digested with s20S liver pro- teasomes of uninfected control mice and with i20S liver proteasomes of Listeria-infected mice. As shown in Fig. 3A, i20S proteasomes isolated from livers 2 and 6 days post infection generated both the Ser-4-Leu-14 precursor fragment and the epitope Ala-7-Leu-14 with increased efficiency. Similarly, the generation of the antitope increased significantly (Fig. 3B). We next calculated the relative ratios of generated antitope Ser-4-Val-12, lacking correct C-terminal cleavage, versus the relative amount of the generated T-cell epitope and precursor. The data shown in Fig. 3C demonstrate that the more efficient generation of both the LLO epitope and the precursor peptides is due to significantly enhanced cleavage site usage of Leu-142Lys-15 by i20S proteasomes compared with the cleavage at Val-122Tyr-13. Remarkably, the usage of both analyzed N-terminal cleavage sites is also enhanced by i20S proteasomes, suggesting that the precursor, the antitope, and the epitope are generated by concerted dual cleavages (Fig. 3A). These data also demonstrate that frequently observed cleavages within an epitope do not necessarily have to result in a concomitant down-regulation of epitope production.
To control that the observed up-regulation is not a side effect of liver i20S proteasomes in infected mice, we also tested proteasomes from small intestines that constitutively contain high levels of i20S proteasomes that are not further increased after infection (16) and 20S proteasomes isolated from livers of infected IFNR Ϫ/Ϫ -deficient mice that contain mainly standard proteasomes. As shown in Fig. 3D, the elevated epitope/antitope and precursor/antitope ratios in digest with small intestine proteasomes is not changed during the infection and no elevation of the epitope/antitope and precursor/antitope ratio was observed in IFNR Ϫ/Ϫ -deficient mice that produce only negligible amounts of i20S proteasomes (supplemental Fig. S1).
To further assess the strength of the antitope approach, we extended our study to three other substrates and generated epitopes of different origin. Previous data had indicated that the generation of the murine Hsp60 epitope is not affected by i20S proteasomes. Digestion of a murine Hsp60-derived 30-mer polypeptide containing the epitope resulted in the generation of the correct T-cell epitope (Lys-10-Asp-18), a precursor peptide (Ser-5-Asp-18), and an antitope (Ser-5-Ile-16), the latter being the result of cleavage between Ile-162Ser-17 (Fig. 4A). As observed for the LLO substrate, degradation of the Hsp60 30-mer polypeptide is considerably accelerated in the presence of i20S proteasomes (Fig. 4B). Concomitant with the enhanced degradation rate, the production of the epitope (Lys-10-Asp-18), the precursor peptide (Ser-5-Asp-18), and the antitope (Ser-5-Ile-16) is also enhanced by i20S proteasomes from Listeria-infected mice livers or from small intestine (Fig. 4C). However, standardization of the Hsp60 epitope and precursor peptides against the Hsp60 antitope reveals that there is no change in the preference of cleavage site usage by i20S proteasomes (Fig. 4D). Thus, in contrast to the LLO epitope, increased Hsp60 epitope generation is closely connected with accelerated substrate turnover.
In the last set of experiments we studied the generation of two predicted viral MHC class I epitopes derived from proteins of the Coxsackie virus B3 (CVB3) (Fig. 5A). Again, i20S proteasomes reveal an accelerated turnover of both the CVB3 P2Cand CVB3 P3D-derived polypeptide substrates (Fig. 5B). Despite the fact that the differences in epitope precursor and antitope production by i20S proteasomes or s20S proteasomes are not striking (Fig. 5C), standardization with the CVB3 P2C antitope Gln-13-Tyr-22 demonstrates that immunoproteasomes exhibit a clear preference for the usage of the correct C-terminal Phe-232Ala-24 cleavage site compared with Antitope fragments generated after 2 h of incubation with proteasomes from liver of control WT mice (d0) and infected WT mice (d2 and d6 p.i.) are compared in the right panel. The generation of the antitope Ser-4-Val-12 was significantly (*, p Ͻ 0.05) favored by i20S proteasomes as calculated for four independent experiments by one-tailed heteroscedastic t-test. C, the epitope (diamonds) and the precursor (triangles) generated by liver proteasomes of Listeria-infected WT mice (d0, d2, and d6 p.i.) were normalized to antitope intensities for each digest (antitopes set as 1, gray dotted line and circles). Standard proteasomes are displayed as white and immunoproteasomes as black symbols. The values are means of four independent in vitro digests. D, proteasomes derived from small intestines (s.i.) of WT control (d0) and infected (d2 and d6) mice that contain constitutively high levels of i20S proteasomes showed no differences in the generation of epitope and precursor (left panel). Similarly, fragment generation by proteasomes derived from infected IFN␥R Ϫ/Ϫ mice (d0, d2, and d4 p.i.), which contain only marginal amounts of i20S proteasomes, yielded no significant differences (right panel) in epitope generation. Symbols are as described in C.
In slight contrast to the data obtained with the P2C polypeptide, we detected two antitopes in the P3D-derived polypeptide (Fig. 5A). One of them possesses the same N-terminal residue as the epitope (Thr-13-Phe-17). Determination of the ratio between the produced epitope (Thr-13-Leu-20) versus the corresponding antitope shows a strong preference for cleavage behind the C-terminal residue of the epitope by i20S proteasomes (Fig. 5, E and F). The other antitope (Pro-15-Leu-20) was used as additional internal standard (Fig. 5, A and E). This peptide possesses the C-terminal residue of the epitope and precursor (Leu-20), whereas the N terminus of the peptide is generated by an epitope destroying cleavage between Leu-142Pro-15. Normalization of the generated epitope against the peptide Pro-15-Leu-20 (Fig. 5F) indicates that the generation of the correct N terminus of the epitope is also enhanced in the presence of i20S proteasomes. The experiments shown in Fig. 5F demonstrate that the antitope approach can also be applied for the determination of relative N-terminal cleavage site usage.

DISCUSSION
Previous in vitro analyses have shown that the presence of immunosubunits can alter the preference of proteasomal cleavage site usage in a given substrate, implying this to be the major reason for the frequently observed increase of i20S proteasome-dependent antigen presentation. In fact, in most cases studied (both in cells and in vitro) this conclusion seemed to be supported by the observed increase in i20S proteasome-dependent specific MHC class I antigen presentation. However, the finding that substrate degradation in vitro is in many cases accelerated in the presence of i20S proteasomes also raised the possibility that a general increase in peptide quantity due to enhanced substrate turnover may also serve as valid explanation for improved antigen presentation.
Thus, only if there exists a clear yes or no answer for the i20S proteasome dependence of an epitope, as has been shown for the hepatitus B virus core antigen and others, is a conclusion concerning cleavage site preferences straightforward (10,13,24,25). Regarding the hepatitus B virus core antigen it became for the first time evident that incorporation of immunosubunits induces structural changes in the 20S proteasome core and that alterations in cleavage site preference do not necessarily require a novel catalytic site (10).
However, in most cases it remained difficult to distinguish whether improved CD8 ϩ T-cell epitope presentation is due to increased substrate turnover or to specifically enhanced generation of the epitope itself. Part of the problem is inherent to mass spectrometry, which does not allow differential quantification of proteasomal peptide products.
Studying the generation of the Listeria LLO epitope we were faced with exactly this problem. In vivo experiments showed that i20S proteasomes enhanced the presentation of the LLO(297-304) (Ala-7-Leu-14) epitope in target cells; in agreement with this, in vitro digests revealed an apparently preferential generation of the LLO epitope or its precursor Ser-4-Leu-14 (LLO(294 -304)) peptides. The fact that i20S proteasomes degraded the LLO substrate much faster than standard proteasomes made it, however, difficult to discern whether improved epitope generation is due to altered C-terminal cleavage site preference or to accelerated turnover of the LLO substrate. We therefore analyzed the generation of a substrate-specific, but epitope-independent, peptide that is cleaved by both standard and immunoproteasomes and thus competes with the generation of a functional epitope. As such, we first defined the Ser-4-Val-12 (LLO(294 -302)) peptide as an antitope whose generation is due a cleavage within the epitope.
Interestingly, i20S proteasomes from T2.27 cells or from the liver of infected mice also generated the antitope with increased efficiency. Similar observations from other antigens, i.e. cleavage within the epitope sequence, were previously interpreted as being a sign for preferential epitope destruction by immunoproteasomes. However, when the amount of LLO epitope and precursor was normalized in correlation to the amount of antitope generated by both s20S and i20S proteasomes, it became evident that the correct C-terminal cleavage was enhanced in the presence of i20S proteasomes by a factor of approximately three in comparison to the antitope. Thus, the increased production of the LLO(294 -304) T-cell epitope by i20S proteasomes was indeed the consequence of altered C-terminal cleavage site usage and not a consequence of accelerated substrate turnover. Our experiments also show that utilization of an antitope peptide for the discrimination of proteasomal cleavage site preferences is not restricted to the LLO substrate but can be extended to other epitopes, as demonstrated here for the two CVB3-derived epitopes, i.e. P2C and P3D.
For P3D, two antitope peptides could be selected for cleavage site analysis. One of the antitope peptides shared the N-terminal residue and the other peptide the C-terminal residue with epitope sequence. In consequence, the partial overlap of the antitope with the C and N termini of the epitope allowed the determination of the cleavage site preference for either the C-terminal or N-terminal cleavage required for epitope liberation.
An important and often neglected issue is that enhanced C-terminal cleavage alone as usually assigned to i20S proteasomes does not sufficiently explain the improved epitope liberation, because this also requires a concomitant N-terminal cleavage. As such, the widespread opinion that immunoproteasomes support epitope generation due to their capacity to cleave more efficiently behind hydrophobic, i.e. potential anchor residues, falls short of the real requirements for antigen processing. The analysis of the LLO substrate revealed that i20S proteasomes also enhance N-terminal cleavage. Increased and coordinated dual cleavage events were previously shown for the function of PA28/20S proteasome complexes, whereby PA28 seems to induce subtle structural changes on the 20S core proteasomes without affecting the active site themselves (26,27). In addition, induction of structural changes by the ␤5i subunit was shown to result in altered cleavage site preferences of the 20S proteasome core complex. It is unclear how far double cleavage events by i20S proteasomes and those induced by PA28 are mechanistically related, but it is interesting to note that apparently both the IFN-induced i20S proteasomes and PA28 exert structural changes onto the 20S core that allow for more efficient double cleavage events and concomitant enhanced epitope production.
Acknowledgment-We thank Ilse Drung for excellent technical assistance. FIGURE 5. Generation of two theoretical predicted Coxsackie virus-derived epitopes reveals an immunoproteasome dependence. A, the chosen sequences derived from the CVB3 polyprotein, P2C(1161-1190) and P3D(2170 -2177), are shown in the schemes. The epitopes were predicted with SYFPEITI program. Epitopes, precursors, and the corresponding antitopes that were identified in digests by ESI-MSMS are displayed. B, in vitro degradation of the synthetic substrate peptides was performed with s20S proteasomes (open symbols) and with i20S proteasomes (filled symbols). The substrate degradation was analyzed by HPLC-ESI-MS. The remaining substrates in digests for both analyzed peptides are shown as means (ϮS.E.) for two (P2C) and three (P3D) independent digests. C, comparison of the generation of the CVB3 P2C precursor peptide Gln-13-Phe-23 and the antitope Gln-13-Tyr-22 within a 2-h incubation with standard (white bars) and immunoproteasomes (black bars). The means (ϮS.E.) of the ion current intensities from two analyses are shown. The predicted epitope was only marginally produced and is not shown. D, the precursor/antitope ratio (open circles, generated by s20S; black circles generated by i20S proteasomes) suggests an i20S proteasome-favored processing of the precursor peptide. The antitope is set as 1 (dotted gray lines). Means of the calculated quotients are shown. E, the generation of CVB3 P3D-derived epitope Thr-13-Phe-20 and the antitopes Thr-13-Phe-17 and Pro-15-Leu-20 by s20S (white bars) and i20S (black bars) proteasomes is shown as means (ϮS.E.) of three independent experiments. F, the ratios of epitope to N-terminal antitope (epitope Thr-13-Leu-20/antitope Thr-13-Phe-17, open diamonds by s20S and black diamonds by i20S proteasomes) and epitope to C-terminal antitope (epitope/antitope Pro-15-Leu-20, squares) reveal an i20S proteasome-favored generation of the epitope. Antitopes (dotted gray line) were set as 1. Generated peptides in C and E are shown as means (ϮS.E.) of two and three independent experiments, respectively; the calculated quotients are shown as means of the three experiments.