Tripeptidyl Peptidase II Is the Major Peptidase Needed to Trim Long Antigenic Precursors, but Is Not Required for Most MHC Class I Antigen Presentation¹

Antigen-specific cytotoxic CD8⁺ T lymphocytes recognize short peptides that are bound to MHC class I molecules (MHC class I) on the cell surface. The great majority of the peptides associated with MHC class I are derived from cell proteins, most of which are degraded by the proteasome (reviewed in Refs. 1, 2, 3). Purified 26S and 20S proteasomes degrade proteins to peptides of 3–22 aa long whose length follows a log normal distribution (4, 5). Although 70–80% of proteasome products are too small to serve as MHC class I epitopes, ∼10% of peptides are 8–10 residues long, the size required for strong binding to MHC class I complexes. Another 10–15% of peptides are too long to bind directly to MHC class I molecules but may serve as precursors for MHC class I-binding peptides (4, 5, 6). Several lines of evidence have shown that proteasome cleavages generate the C-terminal residue of MHC class I-binding peptides (7, 8). However, mature epitopes can be efficiently generated in vivo from N-extended precursors (7, 8). This trimming process requires a free N terminus (8) and is therefore mediated by aminopeptidases. One such aminopeptidase which plays an important role in Ag presentation in vivo is endoplasmic reticulum (ER)⁴ aminopeptidase 1 (ERAP1) (9, 10) (also known as ER aminopeptidase associated with Ag processing or ERAAP (11)), an IFN-γ-induced aminopeptidase in the ER. Trimming of N-extended precursors can also occur in the cytosol. Although there are many cytosolic aminopeptidases, whether they perform specific functions in Ag presentation, and their relative importance, is presently unclear.

Recently, several studies have also suggested an important role in Ag processing for tripeptidyl peptidase II (TPPII). This enzyme is an exceptionally large (2–9 megadaltons) cytosolic peptidase that removes groups of three residues from the N terminus of peptide substrates (tripeptidyl exopeptidase activity) and has also been reported to have a weak endoprotease activity (12). Recent studies have suggested that TPPII may play two roles in MHC class I Ag presentation. One reported function is to make the endoproteolytic cleavages necessary to generate a presented peptide, as has been reported in the processing for HIV Nef_73–82 (13). However, it is unlikely that TPPII frequently generates the C-terminal residues, a function normally served by proteasomes. Early studies had suggested that TPPII may even substitute for the proteasome in the degradation of cell proteins and generation of most antigenic peptides (14); however, these suggestions have not been substantiated.

A second role proposed for TPPII is in trimming the N-terminal extensions from long precursor peptides (15). It has been suggested, based on studies in cultured cells with inhibitors, that among cytosolic peptides, only TPPII can trim peptides longer than ∼16 aa (16), consistent with previous biochemical studies establishing that other cytosolic peptidases (e.g., thimet oligopeptidase (TOP) (16) and various aminopeptidases (15)) have little activity against peptides longer than ∼13–15 residues. Remarkably, in these experiments inhibiting TPPII reduced Ag presentation to the same extent as did inhibitors of the proteasome. Moreover, inhibiting both TPPII and proteasomes had no greater effect, which led to the surprising conclusion that TPPII was essential for MHC class I Ag presentation (16, 17). Moreover, because inhibiting TPPII only affected the trimming of peptides longer than 15 residues in length, Reits et al. (16) concluded that proteasomes must be generating predominately precursor peptides that are 16 residues or longer, in clear contrast to findings on the size distribution of peptides produced by purified proteasomes (4, 5, 6). It was therefore postulated that intracellular proteasomes behave differently in vivo than they do after purification. These findings led to a model in which proteasomes make the initial cleavages in Ags to generate oligopeptides of 16 or more residues, and then TPPII plays an essential role in shortening the N-terminal extensions on these peptides to a size where other aminopeptidases could trim the products of TPPII down to mature epitopes or to individual amino acids (16, 17).

In this study, we examine the role of TPPII in trimming peptides for MHC class I Ag presentation. Initial studies on the purified enzyme showed that TPPII is in fact not selective for long peptides, and that it can trim both long and short ones. In intact cells, however, TPPII plays an important role in trimming N-extended peptide precursors longer than 14 residues for MHC class I Ag presentation, presumably because other peptidases cannot efficiently digest such long substrates. Elimination of TPPII by siRNA is shown to result in only a small reduction in the generation of SIINFEKL from OVA and does not reduce the overall supply of MHC class I-presented peptides from cellular proteins. Therefore, TPPII is the primary enzyme that processes peptides that are longer than ∼14 residues, but these are only a small fraction of those released by the proteasome. Thus, its role contrasts sharply with that of the proteasome which generates the great majority of MHC class I-presented peptides.

Hela-Kb cells have been described previously (10). The mAb W6/32 recognizes human MHC H chain (HC) only when associated with β_2-microglobulin (β₂-m) (18). HC10 recognizes HC only when not associated with β₂-m (19). B8.24.3 recognizes H-2K^b associated with β₂-m (20). A polyclonal rabbit antiserum that recognizes H-2K^b not associated with β₂-m was a gift from S. Nathenson (Albert Einstein College of Medicine, New York, NY). 25.D1.16 recognizes H-2K^b only in association with SIINFEKL (21).

Recombinant adenovirus expressing the HSV TAP-blocking protein ICP47, and control virus AdBHG10, have been previously described (22).

siRNA was obtained from Qiagen. TPPII-specific siRNA was 5′-AAGCAACTCACTGGCCAAATT-3′ and 5′-AATTTGGCCAGTGAGTTGC-3′. As a control, we used siRNA directed against murine TOP, which does not have a target sequence in human cells, as previously described (10). siRNA directed against ERAP1 has been previously described (10). RNA oligos were annealed and prepared, and cells were transfected with siRNA using Oligofectamine (Invitrogen Life Technologies), as previously described (10). Based on preliminary experiments, cells were analyzed 3 or 4 days after transfection. To confirm the efficacy of knockdown by PCR, mRNA was collected from cells (RNeasy kit; Qiagen) and PCR was performed using TPPII-specific primers 5′-GGTGGGCAAGTCTCAGTGAT-3′ and 5′-CATCAAAGCGGTTGATTCCT-3′, or control primers specific for ERAP1: 5′-GGGAGCTGGAGAGAGGCTAT-3′ and 5′-CTTGCTTTGAAGGCAGGTTC-3′.

Plasmid pUG1 was constructed using ubiquitin cDNA (provided by L. Whitton, The Scripps Research Institute, La Jolla, CA), modified to have SfoI and BamHI sites at the 3′ end. The internal ribosome entry sequence (IRES) and GFP sequence from the murine stem cell virus (23) was cloned downstream of this sequence in a pcDNA6 (Invitrogen Life Technologies) backbone. Oligonucleotides encoding various peptides were cloned into the SfoI-BamHI sites to produce a ubiquitin fusion protein. The sequences of the peptides expressed in this way are listed in Table I. To produce plasmid pIG-FLOVA, the full-length OVA gene, was subcloned into pUG1 replacing ubiquitin so that the IRES-GFP cassette was downstream of OVA. Cells were transfected using HeLa Monster (Mirus) as previously described (10).

Presentation of the immunodominant H-2K^b-binding epitope from chicken OVA was measured by flow cytometry. Hela-Kb cells were transfected with siRNA and, 3 days later, the cells were transfected with plasmids expressing various SIINFEKL precursors. Twenty-four hours later, the cells were analyzed by flow cytometry, gating on GFP-expressing cells. Transfection efficiency generally was between 10 and 35%, depending on the construct; samples with very low transfection efficiency (<2%) were not included in analysis.

Samples were compared using two-tailed Student t analysis, and differences were deemed statistically significant for p values <0.05.

Cells were starved for 1 h in cysteine/methionine-free medium, metabolically labeled with ³⁵S, and chased with a 10-fold excess of normal medium. In some cases, the proteasome inhibitor MG132 (Calbiochem) was added to medium at 10 μM throughout the starve, label, and chase periods. Cells were lysed with 1% Nonidet P-40 and 0.5% deoxycholic acid in TBS, with a protease inhibitor mixture (Roche). Lysates were incubated overnight at 4°C, then clarified and immunoprecipitated with appropriate Abs. Because folding is inefficient in Hela-Kb cells (as with many nonlymphoid cells), three times as much lysate was use for the “Folded” immunoprecipitate (mAb W6/32) as for the “Unfolded”) (mAb HC10). Immunoprecipitates were separated using SDS-PAGE and autoradiographs were scanned and quantified using Quantity One software (Bio-Rad). “Percent folded” MHC class I was calculated by adding the intensity of the “Folded” band, to the intensity of “Unfolded” band multiplied by three.

Cells were transfected with siRNA for 3 days, and with plasmid expressing either full-length OVA or SIINFEKL for 2 days. Cells were trypsinized and washed, and then treated with acid (1:1 mixture of 0.163 M citric acid and 0.32 M NaH₂PO₄ (pH 2.6)) for 60 s, then transferred to a 100-fold excess of RPMI 1640 medium with 10% FCS, buffered with HEPES. Cells were pelleted and resuspended in warm medium, and incubated for various times before analysis by flow cytometry. In some cases, MG132 (10 μM) was added to the medium for 90 min before the trypsinization and subsequently through the recovery period.

TPPII was purified from young rabbit muscle, using high concentrations of DTT and ammonium sulfate for stabilizing its structure as reported for the Drosophila homolog (24). After the muscles were minced to small pieces, they were homogenized in a buffer containing 50 mM sodium phosphate (pH 7.5), 0.25 M sucrose, 5 mM DTT, 5 mM MgCl₂, 2 mM ATP, and 10% glycerol. The homogenate was centrifuged for 15 min at 10,000 × g to remove cell debris and then was centrifuged for 1 h at 100,000 × g to remove membranous fractions. The supernatants were loaded onto a DE-52 column and after washing, were eluted with a linear gradient of 0–400 mM ammonium sulfate. The TPPII peak was identified by butabindide-sensitive activity against Ala-Ala-aminomethyl coumarin (AAF-amc). Active fractions were collected, loaded on a resource Q column (Bio-Rad) after diluting the ammonium sulfate concentration to 10 mM and eluted with a linear gradient of 300–0 mM ammonium sulfate. Fractions from the ResourceQ column containing TPPII (i.e., AAF-amc activity that was inhibited by 4 nM butabindide) were concentrated to 1 ml and loaded on a 38-ml glycerol gradient (23–37% glycerol in 25 mM HEPES (pH 7.5), 5 mM DTT, 0.5 mM ATP, 5 mM MgCl₂). After centrifugation for 22 h at 100,000 × g, the gradient was fractionated, and the fractions active against AAF-amc were pooled and concentrated. The resulting preparations were free of proteasomes as shown by the lack of activity against the proteasomal substrate Succinyl-Leu-Leu-Val-Tyr-aminomethyl coumarin (Suc-LLVY-amc), and of aminopeptidase activity because the activity against AAF-amc was not affected by the general aminopeptidase inhibitor, bestatin.

To measure TPPII activity, cell extracts were prepared by lysing HeLa cells in ice-cold 50 mM sodium phosphate (pH 7.5), 5 mM MgCl₂, 2 mM ATP, 1 mM DTT, and 10% glycerol using Dounce homogenization. This homogenate was spun at 10,000 × g for 15 min to remove nuclear and membrane fractions, and contained active proteasomes and TPPII, as tested by activity against specific fluorogenic substrates, Suc-LLVY-amc and AAF-amc. The resulting supernatant fractions were aliquoted, snap-frozen in liquid nitrogen and stored at −80°C until use.

Ten micrograms of Hela extract, in the absence or presence of butabindide (4 nM), were incubated with 100 μM of each peptide at 37°C for 2 h, and the new N termini of the peptide products generated were assayed by using fluorescamine as described previously (25, 26). The extracts were preincubated with or without butabindide for 20 min at room temperature. Percent inhibition in each case is calculated from the amount of N termini generated (nanomoles) in the absence of butabindide.

To measure activity of TPPII against specific peptides, 100 μM of each peptide was incubated with 4 nM of purified TPPII, at 37°C for 1 h, and the new N termini of the peptides generated were assayed by using fluorescamine. One set of peptides used for these experiment have five identical N-terminal residues and three C-terminal residues (except for the six-residue peptide that has the same five N-terminal residues and the final common C-terminal residue) (Table II). Another set of peptides consisted of the OVA epitope SIINFEKL with 0, 3, 6, or 9 residues of natural N-terminal flanking residues (Table II).

It has been previously suggested that TPPII preferentially degrades peptides greater than 16 residues in length (27). However, this conclusion was based on indirect data obtained with inhibitors and fluorescent peptides injected into cells (16). The size range of peptides that are hydrolyzed by TPPII has not been carefully analyzed. To investigate this issue, we analyzed the ability of purified mammalian TPPII to hydrolyze peptides of varying lengths. We first compared the ability of purified TPPII to degrade a homologous series of peptides that vary in length from 6 to 30 residues, but contain the same 5 (or more) N-terminal residues. TPPII is a tripeptidyl peptidase that releases with each cleavage tripeptides from the N terminus of the substrate. Therefore, all the peptides tested contain the same initial cleavage site. In addition, all except the shortest substrates share the same three C-terminal residues. The rate of hydrolysis of these substrates was assayed by measuring the initial rate of the appearance of new N termini using fluorescamine (25).

TPPII was able to digest peptides in this series that were 16–30 residues in length, in agreement with the in vivo findings of Reits et al. (16). However, the enzyme showed no consistent preference for these long substrates (Fig. 1 A). Although there were some differences in cleavage rates, they were presumably due to the substrates’ internal sequences. The important point is that all peptides ranging from 6 to 30 residues were cleaved reasonably well by TPPII.

To confirm and extend these results, we also analyzed a series of N-extended precursors of the MHC class I-presented OVA epitope SIINFEKL that contained 0–9 residues of the natural flanking sequence. Again, TPPII was able to trim precursors that were 17 residues in length but also ones that were shorter, and showed no consistent preference for longer peptides (Fig. 1 B). In fact, with both these series of substrates, a moderate preference for peptides 8–14 residues long was observed.

To study the role of TPPII in MHC class I Ag presentation, we prevented its expression with siRNA. Hela-Kb cells were treated with control siRNA or with siRNA specific for TPPII. RNA levels were markedly reduced by 3 or 4 days after transfection (Fig. 2,A). Although Abs suitable for measuring protein levels of TPPII were not available to us, we showed that TPPII’s enzymatic activity, assayed with the fluorometric substrate AAF-amc, was reduced by ∼90% (Fig. 2,B). (This siRNA-mediated reduction in TPPII was substantially greater than that achieved by Reits et al. (16) in their study of Ag presentation.) The cells remained viable and continued to grow over the 4-day period, and the activities of two other critical cellular proteolytic activities, the proteasome, (measured by Suc-LLVY-amc), and TOP (assayed using McC-PLGPK-Dnp), were not affected by the treatment with TPPII-specific siRNA (Fig. 2 B).

Previous studies in cultured cells suggested that TPPII plays a critical role in processing peptides longer than ∼16 residues (16) to MHC class I-presented peptides. To determine whether TPPII trimming of these long precursors was in fact important in MHC class I Ag presentation, we transfected siRNA-treated cells with plasmids expressing the immunodominant H-2K^b-restricted epitope from OVA, SIINFEKL, extended by 8 or 9 residues on its N terminus to yield 16 and 17 mer, fused to the C terminus of ubiquitin. Ubiquitin C-terminal hydrolases in cells efficiently remove ubiquitin moieties from such fusion proteins, liberating peptides with defined N termini (28). The sequence encoding the ubiquitin-N(x)-SIINFEKL fusion protein was followed by an IRES and GFP, whose intensity was used to identify cells expressing similar levels of peptide. SIINFEKL- H-2K^b complex formation was quantitatively measured by flow cytometry, using the mAb 25.D1.16 (21). We found that, as predicted by Reits et al. (16), TPPII knockdown reduced by 60–80% the presentation of SIINFEKL from N-extended SIINFEKL precursors longer than 16 aa (Fig. 3,A). By contrast, TPPII knockdown had little or no effect on the presentation of SIINFEKL when the mature epitope itself was expressed as a ubiquitin fusion protein (Fig. 3,A). This finding indicates that TPPII knockdown did not affect TAP transport, MHC class I assembly, or other steps in the Ag presentation pathway downstream of peptide production, and also confirms that the siRNA treatment effectively reduced TPPII levels. The observation that TPPII knockdown does not increase the presentation of SIINFEKL even though in vitro TPPII can trim SIINFEKL (Fig. 1 B) suggests that TPPII is not an important (i.e., rate-limiting) enzyme in the degradation of such peptides, which is also consistent with the findings of Reits et al. (16). By contrast, we previously showed that knockdown of the endopeptidase TOP, which is a major enzyme for degradation of SIINFEKL in the cytosol, led to enhanced presentation of SIINFEKL when SIINFEKL or N-extended precursors were expressed in cells (29).

To learn whether TPPII also generates presented peptides from precursors that are shorter than 16 residues, we then tested the effect of TPPII knockdown of the processing and presentation of SIINFEKL from precursors extended by 1 to 6 amino acids at the N terminus (9–15 mer). Although TPPII knockdown had little or no effect on the presentation of SIINFEKL itself (Fig. 3) or of 1+SIINFEKL (Fig. 3,B), surprisingly, TPPII knockdown consistently reduced presentation from precursors that were shorter than 16 residues. Presentation of SIINFEKL from 14 and 15 mer were significantly reduced to about the same extent as 16 mer, and in the absence of TPPII presentation of SIINFEKL from the 12 mer (EQLE-SIINFEKL) was significantly reduced by 30–40% below levels in controls (Fig. 3 B). Interestingly, as with SIINFEKL itself, TPPII knockdown did not significantly increase or decrease the presentation of SIINFEKL from E-SIINFEKL or LE-SIINFEKL.

It was possible that the amino acids upstream of SIINFEKL in these constructs are, for some reason, poorly processed by other cellular aminopeptidases so that TPPII is unusually important in their processing. To address this possibility, we studied presentation of SIINFEKL with N-terminal extensions consisting of the sequence upstream of the immunodominant H-2K^b-restricted peptide RGYVYQGL from the vesicular stomatitis virus (VSV) nucleoprotein. Previous studies have implicated puromycin-specific aminopeptidase and bleomycin hydrolase in generating RGYVYQGL from N-extended precursors (30). However, TPPII knockdown affected the presentation of SIINFEKL from precursors extended by 1–5 aa (L-SIINFEKL through SLSDL-SIINFEKL) to a very similar extent as it did with native SIINFEKL sequence. Presentation of SIINFEKL from 12- and 13-residue precursors LSDL- and SLSDL-SIINFEKL was reduced by ∼30–40% below control levels (Fig. 3 B). Again, TPPII knockdown did not enhance the presentation of SIINFEKL from precursor peptides as short as 10 residues, although it may have slightly reduced the presentation from DL-SIINFEKL for reasons that are unclear. Therefore, while TPPII is particularly important in the processing of peptides longer than ∼14 residues, it also plays a role in processing of shorter N-extended peptides. These findings in intact cells are consistent with our observations that purified TPPII trims both short and long peptides.

Although the data above show TPPII contributes to the generation of presented peptides, it is unclear whether it produces the mature epitopes or simply generates shorter precursors that are then trimmed further by other aminopeptidases. To test this possibility, we compared the effects on SIINFEKL presentation of TPPII knockdown to that of knockdown of ERAP1, the IFN-induced aminopeptidase in the endoplasmic reticulum that trims longer precursors to the 8–9 residue presented epitopes. As previously shown (10), ERAP1 knockdown has little effect when the mature SIINFEKL epitope itself is expressed in cells, but it reduces presentation of SIINFEKL when even a single amino acid is added to its N terminus (Fig. 4). ERAP1 was considerably more important than TPPII in the presentation of peptides 13 residues and shorter, but the effect of ERAP1 knockout on longer peptides (a 14 mer and 16mer) was about equal to knockout of TPPII.

We then knocked down ERAP1 and TPPII simultaneously. The double knockdown reduced levels of ERAP1 and TPPII mRNA about as much as single knockdown (data not shown). Presentation of SIINFEKL from all precursors was slightly less in the double knockdown than when ERAP1 alone was knocked down (Fig. 4). However, the double knockdown reduced presentation from mature SIINFEKL by ∼10%, slightly more than knockdown of either TPPII or ERAP1 alone, although this was not significant (Fig. 4). Because neither knockdown alone affected presentation of SIINFEKL, this suggests that the double knockdown may have more nonspecific effects than either single knockdown, although the nonspecific effects are still quite modest. Even if this is not taken into account, there was no significant difference in the effect of silencing TPPII with ERAP1 over silencing ERAP1 by itself. Therefore, these peptidases almost certainly act sequentially on the same substrate, rather than acting in parallel, redundant pathways. In other words, in the cytosol TPPII must be generating a shorter precursor that is then trimmed in the ER to the mature form by ERAP1. Because ERAP1 poorly trims peptides that are longer than 16 residues (31), these findings are consistent with a model in which TPPII is the major enzyme that processes longer peptides to a length that other aminopeptidases like ERAP1 can then hydrolyze.

Reits et al. (16) interpreted their data to indicate that most MHC class I-presented peptides were initially generated in vivo by proteasomes as long precursors that are predominately longer than 16 residues, and therefore required trimming by TPPII. This conclusion is in sharp disagreement with studies of the peptide products of purified 26S and 20S proteasomes, which mainly yield peptides shorter than 8 residues (4, 5, 32, 33). It was therefore of interest to evaluate the effect of the TPPII knockdown on the presentation of SIINFEKL from full-length OVA. We found that TPPII knockdown only inhibited the presentation of SIINFEKL from the full-length protein by ∼20–30%, as compared with control-transfected cells (Fig. 5,A). These results are consistent with our earlier findings that N-extended SIINFEKL precursors are generated during Ag processing (10) and demonstrate that TPPII does play a role in trimming some of these peptides for Ag presentation. However, the silencing of TPPII has a much smaller effect on the presentation of SIINFEKL from OVA, than it does on presentation from precursors longer than 15 residues (Fig. 3). The important implication of this observation is that the majority of SIINFEKL precursors that are generated from OVA in vivo are <16 residues in length. These results are at variance with the models proposed by Reits et al. (16) and Kloetzel (17), but completely consistent with earlier biochemical studies on isolated proteasomes (4, 5, 6).

The finding that TPPII knockdown had quite modest effects on the presentation of SIINFEKL from full-length OVA prompted us to ask whether TPPII is generally important in generating peptides for MHC class I molecules. MHC class I HC and β₂-m normally associate in the ER before the antigenic peptide is bound; however, this dimeric complex without a bound peptide is relatively unstable, so that HC and β₂-m dissociate at elevated temperatures or after prolonged times (e.g., overnight incubations) at 4°C. After the peptide associates with MHC class I to form a mature trimeric complex, the association between HC and β₂-m becomes much more stable. Therefore, the amount of stable MHC class I can be used as a measure of overall peptide supply.

To quantify peptide supply to MHC class I molecules, we metabolically labeled cells with ³⁵S and lysed them after various chase periods. To allow dissociation of unstable complexes, we incubated the lysates overnight at 4°C (22), and then immunoprecipitated MHC class I with W6/32 (specific for HC associated with β₂-m) or HC10 (specific for HC not associated with β₂-m). As a control to confirm that a reduction in peptide supply could be detected under these conditions, we performed the same procedure on Hela-Kb cells infected for 24 h with a recombinant adenovirus expressing ICP47 (22), a herpes simplex protein that binds to TAP and blocks peptide transport into the ER (22, 34). As expected, in cells expressing ICP47, little stable W6/32-reactive MHC class I was present, even though free HC levels were similar to control-infected cells in which W6/32 reactive (“folded”) MHC class I was readily detected (Fig. 5, B and C). TPPII siRNA and control siRNA-transfected Hela-Kb expressed very similar levels of W6/32-reactive and HC10-reactive HC (Fig. 5, B and C). In fact, in TPPII knockdown cells, a moderate, but consistent, increase in peptide-loaded MHC class I molecules was observed (Fig. 5, B and C), presumably because TPPII normally destroys some peptides (9 residues or longer) that could otherwise serve as antigenic precursors or bind to MHC class I molecules.

It was recently suggested that TPPII was as important as the proteasome in generating MHC class I-presented peptides (16). This suggestion seemed inconsistent with many prior observations indicating a predominant role of the proteasome, as well as with our observation that TPPII has only a modest effect on the presentation of SIINFEKL from OVA (Fig. 5,A), or of peptides from cellular proteins generally (Fig. 5, B and C). To directly evaluate the relative contributions of proteasomes and TPPII in the presentation of SIINFEKL from OVA, we compared the effect of inhibiting proteasomes and of TPPII silencing in HeLa-Kb cells. Because cells cannot be treated with proteasome inhibitors for the lengths of time necessary to measure Ag presentation after transfection (24–48 h) we used an alternate protocol where surface MHC class I molecules were stripped from Ag-transfected cells by acid treatment, and allowed to recover in cells lacking either TPPII or proteasome activity. Hela-Kb cells were treated with TPPII-specific, or control, siRNA for 1 day, then transfected with plasmids expressing various SIINFEKL precursors and incubated for a further 2 days. Cells were then treated with the proteasome inhibitor MG132 followed by citric acid (pH 2.5) to denature pre-existing MHC class I molecules, and allowed to generate new H-2K^b-SIINFEKL complexes.

As expected, the recovery of surface H-2K^b-SIINFEKL on cells expressing a SIINFEKL minigene was not affected by MG132 treatment (Fig. 6,A) (because this minigene does not require proteolysis for presentation), or by TPPII knockdown (which did not affect SIINFEKL presentation: Fig. 3). Therefore, components of the Ag presentation pathway other than the proteolysis machinery were not affected by these treatments. In contrast, MG132 treatment abolished the recovery of H-2K^b-SIINFEKL at the surface of cells expressing full-length OVA, while knockdown of TPPII had little or no effect (Fig. 6 B).

We conclude that generation of SIINFEKL from full-length OVA is absolutely dependent on proteasomes, but that TPPII is not required for this processing. To confirm that this is generally true for most MHC class I-presented peptides, we analyzed the effect on overall peptide supply of proteasome inhibition vs TPPII knockdown. Assembled vs free MHC class I molecules were immunoprecipitated from lysates of radiolabeled cells treated with MG132 or TPPII siRNA and peptide occupancy was tested as described above. Again, TPPII knockdown did not reduce (and in fact somewhat increased) the amount of peptide-loaded MHC class I (Fig. 6, C and D). In contrast, cells treated with MG132 showed a marked reduction in W6/32-reactive MHC class I molecules. In addition to further confirming that this experimental approach can detect reduced peptide supply resulting for formation of MHC class I complexes, this finding demonstrates that proteasomes play a much more important role in generating peptides for MHC class I than does TPPII.

It is well-established that proteasomes catalyze the degradation of most cytosolic proteins, and that most peptides presented on MHC class I molecules are a byproduct of the continual degradation of cell proteins by the ubiquitin-proteasome pathway (1, 2, 3). The great majority of the peptides produced by proteasomes are rapidly hydrolyzed to amino acids by cytosolic peptidases. However, it is still not well-understood how a small fraction escape complete degradation and are processed to the 8–10 residue peptides that bind to MHC class I molecules. Proteasomes can directly generate the mature epitopes, but isolated proteasomes and especially “immunoproteasomes” appear to generate preferentially N-extended precursors (35). A number of studies with model peptides have shown that cellular aminopeptidases are able to process rapidly such N-extended peptides to the mature epitopes, and several aminopeptidases that can carry out this processing in vivo have been identified (9, 11, 36, 37, 38). Nevertheless, it remains unclear how important various cellular aminopeptidases are in generating MHC class I epitopes, or whether they may mainly destroy epitopes that could otherwise be presented.

Because peptides in intact cells are extremely short-lived (39), most studies of peptide generation and hydrolysis have been conducted in cell lysates or with purified enzymes. Such studies have demonstrated that 26S and 20S proteasomes generate peptides ranging from 3 to ∼24 residues long, with the great majority being <8 aa long and ∼90% smaller than 15 residues. This size distribution appears to result from a kinetic competition between further cleavages by the active sites and the ability of small peptides to diffuse out the gated exit channel in its outer α-ring (40, 41) (4, 5, 6). Recently, Reits et al. (16) reported that inhibitors of TPPII activity reduced surface levels of MHC class I molecules to the same degree as did proteasome inhibitors. The important implication of this finding was that the proteasome and TPPII were both required to generate most presented peptides. Because the same study also found that TPPII was essential in vivo to cleave injected fluorescent peptides that were longer than 16 residues, but not shorter ones, it was also proposed that in intact cells (as opposed to in vitro) proteasomes predominately generate peptides that are longer than 16 residues, which are then trimmed by TPPII for Ag presentation (16, 17). Here, we show that while TPPII may contribute moderately to Ag presentation from some proteins (e.g., OVA), TPPII is not essential for the bulk of MHC class I Ag presentation, and that proteasomes in intact cells most likely generate shorter peptides, as seen with the purified particles.

Consistent with the observations of Reits et al. (16), we find that knocking down TPPII with siRNA does markedly reduce Ag presentation from long precursors: in our experiments, peptides longer than ∼14 residues show significant dependence on TPPII. We do not know whether the remaining presentation of these peptides after TPPII knockdown is due to residual TPPII activity (∼10% of control-treated cells) or whether other peptidases can inefficiently process these peptides. However, peptide supply to newly synthesized MHC class I is not generally reduced when TPPII is knocked down (Fig. 5). In fact, there is a small, but consistent, increase in peptide supply to MHC class I after TPPII knockdown, presumably because TPPII (like other peptidases (10, 29)) destroys some peptides (e.g., 8, 9, or 10 mer) that could otherwise bind to MHC class I by trimming them until they are too short to bind to MHC class I molecules. Therefore, the generation of most peptides does not require TPPII either in place of the proteasome (16, 17) or downstream of this particle.

In addition to analyzing overall peptide supply to multiple HLA class I molecules, we examined the effect of TPPII knockdown on the presentation of a specific peptide, SIINFEKL, the immunodominant H-2K^b-binding peptide from OVA. Reducing TPPII levels reduced the presentation of SIINFEKL from full-length OVA, but the effect was quite modest, in contrast to its large effect on 14 mer and longer precursor peptides. This is consistent with the previous finding that ∼80% of potential SIINFEKL precursors produced by purified proteasomes from OVA are shorter than 15 residues, and virtually all are shorter than 16 residues (35). In contrast, the proteasome inhibitor MG132 dramatically reduced SIINFEKL presentation from full-length OVA. Taken together, these observations strongly suggest that proteasomes in intact cells, as in vitro, generate relatively short peptides, that are not dependent on TPPII for further processing.

Why might our findings differ from those of Reits et al. (16)? One difference is that we tested peptide supply to MHC class I by measuring the stability of newly synthesized MHC class I and by Ag presentation assays, both more direct tests of peptide supply than the measurement of cell surface expression of MHC class I used by Reits et al. (16). Another difference is in the methods of TPPII inhibition. It is possible that the TPPII inhibitor butabindide, used by Reits et al. (16) in most experiments or the treatments to facilitate its entry (e.g., serum deprivation), may affect cells in ways unrelated to TPPII inhibition, especially at the very high concentrations required to treat intact cells. In our hands, multiple experiments using butabindide as described by Reits et al. (16) on Hela-Kb cells did not cause any reduction in the presentation of SIINFEKL from a 16-mer precursor (N8-SIINFEKL) (data not shown), although presentation from this precursor was markedly reduced by TPPII siRNA (e.g., Fig. 3 A), and in the experiments of Reits et al. (16), 16-mer degradation was dependent on TPPII (16). Yet another difference is that we used different cell types. Although it is perhaps possible that proteasomes in different cell types generate peptides with different lengths, this seems quite unlikely and is inconsistent with these particles being highly conserved between cell types and our present understanding of their function.

The substrate size preference of TPPII had not been previously defined. We find that purified mammalian TPPII efficiently digests both short and long peptides. In fact, TPPII hydrolyzes short peptides somewhat better than longer ones (Fig. 1). This contrasts with our findings and those of others (16) that reducing TPPII activity in cells affects the trimming of longer peptides more than shorter ones. This difference probably has nothing to do with the properties of TPPII, but arises because other peptidases in cells can trim rapidly short peptides but not long ones. For example, the primary endopeptidase, TOP, has little activity against peptides longer than 15 mer (42), and known aminopeptidases primarily hydrolyze peptides that are ∼<6 residues long (26, 31, 43). In other words, in vivo TPPII is the rate-limiting enzyme for hydrolyzing long peptides but not shorter ones. These data are consistent with previous findings (16), although it is important to note that these observations do not imply that TPPII preferentially digests longer peptides (17) but only that TPPII is the major enzyme that can trim these peptides, while most cell peptidases strongly prefer shorter ones, consistent with the length of most peptides that are generated by proteasomes.

We also observed some effect of TPPII knockdown on relatively short (11–13 residue; 3–5 residue N-terminal extensions) precursors of SIINFEKL. Loss of TPPII reduced their presentation modestly, and this reduction did not depend on the sequence of the N-terminal extension. This result is consistent with the ability of purified TPPII to hydrolyze peptides in this size range (Fig. 1). We also note a very small effect on precursors with a 2-residue N-terminal extension (a reduction of ∼15%, p = 0.045 for the DL-SIINFEKL precursor; not statistically significant for the LE-SIINFEKL precursor). This is not consistent with studies of purified TPPII: because TPPII removes N-terminal amino acids in groups of three, it converts these 10 mer in vitro to IINFEKL (data not shown), which does not bind to H-2K^b. One possibility is that in intact cells TPPII has different activities that have not been detected as yet, perhaps because of accessory factors that might be lost during purification. It is also possible that TPPII directly or indirectly alters the effects of other peptidases in intact cells. In any case, the magnitude of this effect is quite small.

Somewhat surprisingly also, TPPII knockdown in intact cells expressing SIINFEKL itself did not alter presentation of SIINFEKL itself. Because purified TPPII is able to digest SIINFEKL efficiently, we had expected that knocking down TPPII activity would enhance SIINFEKL presentation by reducing its destruction in the cytosol. The lack of effect argues that TPPII is not a rate-limiting activity for the destruction of SIINFEKL, consistent with the notion that TPPII is not rate-limiting for peptides shorter than ∼14 mer. Previous experiments have shown that TOP normally destroys a significant amount of SIINFEKL (29), and that aminopeptidases (44) may also contribute to SIINFEKL destruction.

The present studies have further documented the important role in processing SIINFEKL precursors played by ERAP1. Although knockdown of ERAP1 did not alter presentation of SIINFEKL from the mature epitope, it drastically reduces presentation from peptides with even a single N-terminal extension (10) (Fig. 4). In fact, ERAP1 knockdown affected presentation of SIINFEKL from a 9 mer and an 11 mer much more than did TPPII knockdown, while the effect on a 13 mer or a 16 mer (Fig. 4) or 17 mer (N9-S8L) (data not shown) was similar for both, consistent with the hypothesis that TPPII is particularly important for trimming relatively long peptides (16), and that ERAP1 only trims peptides from 9 to 16 mer if they are transported into the ER. These experiments also resolved whether TPPII and ERAP1 might act sequentially on longer peptides, with TPPII in the cytosol first converting longer peptides into shorter N-extended peptides that ERAP1 could then trim, and ERAP1 making the final cuts to generate SIINFEKL in the ER or, alternatively, whether TPPII and ERAP1 acted in parallel trimming a distinct set of peptides. In the latter case, knockdown of both ERAP1 and TPPII should almost eliminate presentation of SIINFEKL from precursors such as N5- or N8-SIINFEKL (because knockdown of each enzyme alone reduces presentation by ∼50%). However, we found that the double knockdowns did not reduce presentation much more than knockdown of ERAP1 alone (Fig. 4), suggesting that TPPII mainly acts upstream of ERAP1 generating N-extended precursors rather than the mature epitopes.

The present and previous work together suggest the following model. Peptides generated by proteasomes are predominately <16 aa long. A smaller fraction of peptides produced that are longer than 16 aa are relatively resistant to degradation by most cytosolic peptidases, and are mainly processed by TPPII, until the products are short enough to be degraded by other peptidases. Precursor peptides that are shorter than ∼16 residues, most of which are generated directly by proteasomes or occasionally by TPPII, may be processed in the cytosol to a size appropriate for transport by TAP (8–16 residues) and binding to MHC class I (8–10 residues). The cytosolic peptidases involved in such trimming, and their quantitative importance in Ag presentation, are not yet well-studied and may depend on the sequence of the Ag and its affinity for TAP. TPPII does to some extent trim intermediate-length peptides (11–15 residues in length) for Ag presentation, while other cytosolic aminopeptidases such as leucine aminopeptidase (44) do not seem to play major roles in Ag presentation. Alternatively, these N-extended peptides may be transported into the ER and processed there by ERAP1, which seems to be important for the generation of a significant number of MHC class I-presented peptides (45, 46). The great majority of peptides released by proteasomes, however, are not presented by MHC class I, but are completely degraded in the cytosol, with TOP being particularly important in this destruction.

The authors have no financial conflict of interest.