Cytotoxic T Lymphocyte Epitopes of HIV-1 Nef: Generation of Multiple Definitive Major Histocompatibility Complex Class I Ligands by Proteasomes

Proteasomes are the major cytosolic proteases of eukaryotic cells and participate in the processing of many antigens presented by MHC class I molecules. However, the precise role of proteasomes in epitope generation is not yet clear. It is particularly a matter of debate whether proteasomes often generate both COOH and NH₂ termini of proteasome-dependent epitopes, or whether they only generate their COOH termini 1,2.

Purified proteasomes degrade polypeptides into a large number of extensively overlapping oligopeptides. Most major peptide products of vertebrate 20S proteasomes have hydrophobic, acidic, or basic amino acids (aa) at the COOH terminus, most frequently leucine. Peptides with other aa at the COOH terminus are produced less frequently and are usually found in small amounts. With the exception of acidic residues, which are found in proteasome products but very rarely in MHC class I ligands, a similar distribution of aa is observed at the COOH termini of MHC class I ligands. Small neutral and polar aa, especially serine, are enriched at the NH₂ termini of proteasomal products, also similar to MHC class I ligands 3,4,5. With a few exceptions, the size range of peptides eluted from typical MHC class I molecules is 7–13 aa, and most alleles prefer nonamers 6. Although some peptides produced by vertebrate 20S or 26S proteasomes from polypeptides or proteins are shorter than 8 aa, only a few are larger than 11 aa 3,4,7, and many major products are in the size range of MHC class I ligands 3,4. Digestion of relatively short polypeptides proceeds partially via single cleavage intermediates; formation of “dual cleavage” peptides, i.e., MHC class I ligands and slightly longer peptides, is accelerated by the IFN-γ–inducible proteasome activator 28 (PA28 [4, 8]). Degradation of most full-length proteins is processive, without the release of longer intermediates 9,10. Of 10 MHC class I ligands examined so far in various laboratories for production by purified proteasomes in vitro, 7 were shown to be generated in definitive form (see Discussion). Collectively, these data support the conclusion that proteasomes produce both the COOH and NH₂ termini of many MHC class I ligands in vitro.

Nevertheless, these data have not led to a general consensus on the precise role of proteasomes in antigen processing. The validity of the in vitro results has been challenged by observations on the generation of the major OVA epitope SIINFEKL from the products of minigenes 11. Although the production of the epitope from COOH-terminally (Ct)-extended versions was inhibited by the proteasome inhibitor lactacystin, production from NH₂-terminally (Nt)-extended versions was not. Moreover, SIINFEKL could be produced from the Nt-extended fragment QLESIINFEKL by leucine aminopeptidase (LAP), a cytosolic peptidase inducible by IFN-γ 12. This was put forward as evidence that the SIINFEKL NH₂ terminus is produced in vivo by trimming enzymes rather than by proteasomes. Although these results can be interpreted in alternative ways (see Discussion), they have led to the general perception that proteasomes mainly release the COOH termini of class I ligands, whereas the NH₂ termini are derived by trimming of longer precursors by aminopeptidases 1,13.

Nef is a key factor in HIV pathogenicity and immunogenicity and a potential candidate for CTL-targeted vaccination. More than 45 CTL epitopes for multiple HLA alleles have been described in HIV Nef by the use of overlapping synthetic peptides and/or “allele-specific” peptide motifs 14. Many of the Nef CTL epitopes are overlapping, and most of them cluster within four regions of the protein. To date, naturally processed peptides corresponding to these epitopes have not been determined. In this study we focused on one of the four immunogenic regions of Nef, Nef_123–152, which is relatively conserved among different HIV subtypes 15. We identified two HLA-A2 and three HLA-B7 peptide ligands in this region. Unexpectedly, both HLA-A2 ligands and two of the three HLA-B7 ligands represented COOH-terminal length variants of one HLA-A2 and one HLA-B7 epitope, respectively. Proteolytic fragments identical to each of the five definitive MHC class I ligands were found in proteasomal digests of the synthetic polypeptide Nef_123–152. In four of these five naturally processed peptides, the NH₂ 123–152 were produced by major proteasome cleavage sites, and Nt-extended precursors were found for only one of the five peptides. Thus, at least four of the class I ligands are loaded onto HLA molecules in the form generated by proteasomes. These results favor the view that proteasomes often produce the peptides finally presented by MHC class I molecules.

The human lymphoblastoid T cell lines T1 (HLA class I typing A2, B5; subtyping for A2 is A*0201) and Jurkat (HLA I typing A9/25, B7/41; subtyping for B7 is B*0702), the HLA-A– and HLA-B–deficient C1R human lymphoblastoid B cell lines transfected with HLA-A2 or HLA-B7 (C1R-A2 and C1R-B7, respectively), and the P815 murine mastocytoma cell lines transfected with HLA-A2 or HLA-B7 (P815-A2 and P815-B7, respectively), have been described elsewhere 16,17,18,19. C1R-A2 and C1R-B7 cells were stably transfected with the pCF⁺ EBV vector containing the sequence encoding the whole Nef protein from the HIV-1 strain LAI under the control of the cytomegalovirus promoter 20. These cells, referred to as Nef⁺ C1R-A2 and Nef⁺ C1R-B7, were used as stimulator cells to induce polyclonal Nef-specific CTL lines and as target cells in experiments with proteasome inhibitors. We were not able to maintain large scale cultures of C1R-A2 and C1R-B7 cells stably expressing Nef for acid elution of naturally processed peptides. For this purpose, we used T1 and Jurkat cells transfected with the heavy metal–inducible vector pSBBRU6e containing the sequence encoding the whole Nef protein from HIV-1 strain LAI, under the control of the human metallothionein IIA promoter 21. These cells are referred to as Nef⁺ T1 cells and Nef⁺ Jurkat cells. Expression of Nef in the transfectants was verified by Western blot analysis. P815-A2 and P815-B7 cells were used as target cells pulsed with peptides in cytotoxicity assays because C1R cells had a high spontaneous ⁵¹Cr release. The human T1 line was used for isolation of proteasomes.

Polyclonal Nef-specific CTL lines were induced as described 20 using PBLs of a healthy HIV-1 seronegative donor (HLA class I typing A2/28, B7/40w60, Cw3/w7; A2 and B7 subtypes are A*0201 and B*0702, respectively) and Nef⁺ C1R-A2 or Nef⁺ C1R-B7 as stimulator cells.

Nef peptide–specific CTL lines were obtained from the same donor 22. In brief, PBLs (4 or 5 × 10⁶) were incubated in 24-well culture plates (Nalge Nunc International) using RPMI 1640 (GIBCO BRL) supplemented with 5% human AB serum (Blood Donor Center Schweizerisches Rotes Kreuz [SRK], Basel, Switzerland), 1 mM sodium pyruvate, 20 mM Hepes, 2 mM glutamine, 100 U/100 μg/ml penicillin/streptomycin (all from GIBCO BRL), 1 × 10⁻⁵ M 2-ME (Sigma Chemical Co.), and 1% MEM nonessential amino acids (GIBCO BRL) for 90 min at 37°C. The nonadherent cells were removed, and the adherent fraction (monocytes) was pulsed with 100 μg/ml of peptide for 4 h to be used as stimulator cells. Adherent cells were then incubated with 1 × 10⁶ autologous PBLs. After 5 d of culture, 100 U/ml proleukin (Chiron Corp.) was added. After an additional 7 d, the PBLs were restimulated with 1 × 10⁵ irradiated autologous PHA (Murex Biotech, Ltd.) T cell blasts, pulsed with 10 μg/ml of peptide.

Cytolytic activity was tested in triplicate in a standard 4-h ⁵¹Cr-release assay 23. Activity was assayed against cells expressing Nef as well as against P815-A2 and P815-B7 cells pulsed with synthetic peptides or reversed phase HPLC (rp-HPLC) fractions containing naturally processed peptides or proteasomal products. To study the effect of proteasome inhibitors, the Nef-expressing cells were incubated in the presence of 10 μM lactacystin (Dr. E.J. Corey, Harvard Medical School, Boston, MA) or 100 μM N-acetyl-leucinyl-leucinyl-norleucinal (LLnL; Sigma Chemical Co.) for 2 h before ⁵¹Cr labeling. The cells were then radiolabeled, and the preexisting peptide–MHC class I complexes on the cell surface were removed by exposure to a quick acid wash (131 mM citric acid, 66 mM disodium phosphate, pH 3.1) at 25°C for 3 min 24. After neutralization in 30 vol of complete RPMI 1640 medium and washing in PBS, cells were plated and used as targets in a 4-h ⁵¹Cr-release assay in either the presence or absence of the inhibitors.

Peptides were either synthesized using solid phase 9-fluorenylmethoxycarbonyl (F-moc) chemistry on an Applied Biosystems 431A peptide synthesizer or purchased from Genosys Biotechnologies. Peptides were purified to >98% homogeneity by rp-HPLC. Amino acid residues are given in single-letter code as follows: C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

Naturally processed peptides were extracted from whole cells with TFA 25 and purified as described previously 26,27. Batches of 10⁹ Nef⁻ T1, Nef⁺ T1, Nef⁻ Jurkat, and Nef⁺ Jurkat cells were lysed by adding 15 ml cold 1% TFA (Sigma Chemical Co.) and disrupted using a hand-held glass homogenizer. Centriprep 10K (Amicon Corp.) centrifugal concentrators were used for isolation of low molecular weight peptides. The ultrafiltrate suspensions (molecular weight <10 kD) were dried by lyophilization and redissolved in 0.1% TFA for fractionation by rp-HPLC.

20S proteasomes were purified from human T1 cells as described previously 4. No impurities were detected by SDS or native gel electrophoresis. Tripeptidyl peptidase II (TPP II) contaminations were excluded by use of the TPP II inhibitors AAF-chloromethylketone (Bachem) and PMSF (Sigma Chemical Co.). Digestion of the synthetic 30-mer polypeptide Nef_123–152 (10 μg) with isolated 20S proteasomes (2 μg) was carried out at 37°C. Digests were incubated in a total volume of 300 μl buffer consisting of 50 mM Tris-HCl (pH 7.8), 1 mM EGTA, 0.5 mM EDTA, 5 mM MgCl₂, 0.5 mM 2-ME, and 0.02% azide. Aliquots of the reaction mixture were fractionated by rp-HPLC.

Aliquots of rp-HPLC fractions were separated by rp-HPLC (Smart System [Amersham Pharmacia Biotech]) equipped with a Sephasil C18 SC2.1/10 column) with the following: eluent A, 0.1% TFA; eluent B, 80% acetonitrile containing 0.081% TFA; flow rate, 100 μl/min; gradient for separation of HLA-A2–restricted peptides, 34.3–35.5% B in 55 min (increase of eluent B, 0.022% per min); gradient for separation of HLA-B7–restricted peptides, 23.5–24.4% B in 28 min (increase of eluent B, 0.032% per min) followed by 24.4–26.5% B in 27 min (increase of 0.078% per min). Fractions were collected from 30 to 85 min (fractions 1–59), and elution was monitored by measuring UV light absorption at 214 nm in a continuous flow detector. TFA/acetonitrile was removed from the eluted fractions by lyophilization. The lyophilized material was redissolved in 200 μl PBS and stored at −70°C. Aliquots of the rp-HPLC fractions were assayed for recognition by specific CTLs in a ⁵¹Cr-release assay. Aliquots of the rp-HPLC fractions from proteasomal digestions were directly analyzed with a G2025A matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometer (Hewlett-Packard) and by Edman degradation on a Hewlett-Packard instrument (model G1000A).

The affinity of peptides for the transporter associated with antigen processing (TAP) was measured essentially as described previously 28, using microsomes purified from Sf9 insect cells expressing human TAP1–TAP2 complexes, iodinated reporter peptide RRYNASTEL, and a dilution series of competitor test peptides. The concentrations required for 50% inhibition of specific binding (IC₅₀) were normalized with respect to the IC₅₀ of the reporter peptide (typically ∼400 nM). The results are expressed as relative IC₅₀ and are mean values from three experiments.

The HLA binding ability of the HLA-B7–restricted peptides was measured by HLA-B7 stabilization on the TAP-deficient hybrid cell line T2-B7 29. In brief, cells were incubated with 100, 50, 25, and 12.5 μM of peptide in serum-free medium at 37°C. Staining for indirect immunofluorescence was performed with ME1 (anti-B7, -Bw22, and -B27) as first antibody, and FITC-labeled anti–mouse IgG (Sigma Chemical Co.) as second antibody. Fluorescence intensities were measured on a FACScan™ cytometer (Becton Dickinson). The fluorescence ratio was calculated as the mean fluorescence of the sample versus the mean fluorescence of the control. The synthetic reference peptide used was the HLA-B7–restricted self-peptide (APRTVALTAL). The peptide Nef₁₉₈LHPEYFKNC₂₀₆, which does not bind to HLA-B7, was used as negative control.

Peptide affinity for immunoaffinity-purified HLA-A2 molecules was measured in a competitive binding assay based on published procedures 30. In brief, HLA-A2 molecules were purified from NP-40 lysates of the homozygous B cell line Jesthorn using Sepharose-immobilized BB7.2 (anti-A2) mAb. Purified HLA-A2 molecules (400 ng) were incubated for 18 h with 2.4 pmol of iodinated reporter peptide hepatitis B virus (HBV) core 18–27 (6Y) and various concentrations of unlabeled competitor peptides in a total volume of 30 μl PBS buffer with 0.05% NP-40 and 1 mM PMSF. Peptide binding was stopped and evaluated by a 4-min centrifugation at 25°C and 1,000 g through gel filtration columns (Micro Bio-Spin 30™; Bio-Rad Laboratories). Bound peptide was quantified by gamma counting of filtrates. The results are expressed as relative IC₅₀ and are mean values from three experiments.

The molar concentrations of MHC ligands in rp-HPLC fractions of acid eluates of Nef-transfected cells were determined by titration of rp-HPLC fractions in a 4-h ⁵¹Cr-release assay, and comparison of the percent specific lysis was obtained with a standard curve of known concentrations of synthetic peptide. The molar amounts of peptide ligands obtained per extraction were multiplied by Avogadro's number and divided by the number of cells that were extracted. Recoveries of control synthetic peptides were determined as described 26.

Lactacystin is an efficient inhibitor of the chymotrypsin- and trypsin-like activities of proteasomes 31, and a weaker inhibitor of the cytoplasmic protease complex TPP II 32. To assess whether proteasomes are involved in the processing of Nef, we studied the effect of lactacystin on the HLA-A2– and HLA-B7–restricted presentation of Nef, using concentrations that discriminate between proteasomes and TPP II. Nef⁺ C1R-A2 and Nef⁺ C1R-B7 cells were incubated for 2 h in the presence of 10 μM lactacystin and then briefly exposed to pH 3.1 to denature and remove surface class I peptide complexes. Acid-stripped target cells were then allowed to reexpress MHC class I–peptide complexes for 4 h during a standard ⁵¹Cr-release assay, in the presence or absence of 10 μM lactacystin. In the absence of lactacystin, target cell lysis by HLA-A2– and HLA-B7–restricted Nef-specific CTLs was restored after acid treatment (Fig. 1), reaching 80–100% of the lysis of untreated target cells (data not shown). Incubation of acid-treated cells with lactacystin completely abrogated the restoration of HLA-A2– and HLA-B7–restricted CTL recognition of Nef epitopes (Fig. 1). CTL recognition of acid- and lactacystin-treated target cells was restored by addition of known HLA-A2– or HLA-B7–binding Nef peptides, excluding nonspecific deterioration of target cells or of CTLs by the experimental procedures (Fig. 1). A similar degree of inhibition was observed with the peptide aldehyde inhibitor N-acetyl-leucinyl-leucinyl-norleucinal, another potent but less specific proteasome inhibitor (data not shown). These results suggested that processing of Nef for presentation by HLA-A2 as well as HLA-B7 MHC molecules was dependent on proteasomes.

One of the four immunogenic regions of Nef (Nef_123–152, Table) was chosen for this study. One HLA-A2 (Nef₁₃₆PLTFGWCYKL₁₄₅) and two HLA-B7 (Nef₁₂₈TPGPGVRYPL₁₃₇ and Nef₁₃₅YPLTFGWCY₁₄₃) restricted CTL epitopes in this region have been identified previously, using polyclonal Nef-specific CTLs from HIV-seropositive donors and target cells pulsed with overlapping Nef peptides or Nef epitopes predicted according to allele-specific motifs 34,35,37. We now wanted to determine the naturally processed peptides corresponding to these epitopes. For acid elution from Nef-transfected cells, large-scale cell cultures of Nef-expressing cells were necessary. Because of the cytotoxic effect of Nef, attempts to prepare large-scale cultures of cells stably transfected with Nef have rarely been successful 41,42. To overcome these problems, we used a regulated Nef expression vector (pSBBRUnef) based on a mutated version of the heavy metal–inducible human metallothionein IIA promoter. T1 (HLA class I typing A2, B5) and Jurkat (HLA I typing A9/25, B7/41) cells transfected with this vector produce low basal levels of Nef compatible with large-scale cultures 21. Several liters of either T1 or Jurkat cells stably transfected with pSBBRUnef were grown, followed by induction of Nef expression for 24 h. Peptides were isolated by acid extraction of cell lysates followed by ultrafiltration (10-kD cutoff).

In parallel, a synthetic polypeptide corresponding to the region Nef_123–152 was digested with 20S proteasomes isolated from T1 cells. Both peptide pools, that of the acid-extracted naturally processed peptides and that obtained upon proteasomal digestion of Nef_123–152, were separated under the exact same conditions by rp-HPLC on an analytical C18 column. Fractions obtained were tested for their ability to sensitize target cells expressing the appropriate MHC class I restriction elements for recognition by Nef peptide–specific CTL lines. To identify the relevant peptides in fractions recognized by CTLs, retention times were compared with those of a series of synthetic Nef-derived overlapping peptides recognized by the same CTLs (Fig. 2 and Fig. 3, arrows). To achieve efficient separation of such closely related peptides, extremely shallow TFA/acetonitrile gradients were used, individually adjusted for the analysis of each of the epitopes under study (see Materials and Methods). To ascertain that peptides identified in acid-eluted fractions were indeed Nef-derived MHC ligands, control lysates of cells not transfected with Nef and from Nef-transfected cells lacking the MHC molecule in question were fractionated using the same gradient and tested by the same CTLs.

Fractions obtained upon rp-HPLC separation of acid extracts of A2⁺, Nef⁺ T1 cells were tested for the ability to sensitize HLA-A2⁺ target cells for lysis by HLA-A2–restricted CTL lines induced against the peptide Nef₁₃₆PLTFGWCYKL₁₄₅ (Fig. 2 A, top). Surprisingly, Nef₁₃₆PLTFGWCYKL₁₄₅-specific CTLs recognized not only fractions coeluting with the synthetic peptide Nef₁₃₆PLTFGWCYKL₁₄₅ (fraction 21/22), but also fractions coeluting with the synthetic peptide Nef₁₃₆PLTFGWCYKLV₁₄₆ (fractions 33–35). Accordingly, a CTL line was generated by stimulation with Nef₁₃₆PLTFGWCYKLV₁₄₆. As shown in Fig. 2 B (top), these CTLs also recognized fractions containing Nef₁₃₆PLTFGWCYKL₁₄₅, as well as fractions containing Nef₁₃₆PLTFGWCYKLV₁₄₆. Neither CTL line recognized peptide material eluted from Nef⁻ T1 cells (Fig. 2A [top] and B [top]) or peptides eluted from A2⁻, Nef⁺ Jurkat cells (Fig. 2 A, top left; data for Nef₁₃₆PLTFGWCYKLV₁₄₆-specific CTLs not shown). These controls indicate that the acid-eluted peptides recognized by both CTL lines were Nef derived and specifically associated with HLA-A2 molecules of the T1 cells. Moreover, and unexpectedly, the results suggest that the two peptides represent two naturally processed versions of the same epitope, differing in length by one COOH-terminal aa.

In parallel, a digest of Nef fragment_123–152 was prepared with isolated proteasomes, fractionated, and analyzed by the same protocol (Fig. 2A [bottom] and B [bottom]). Strikingly, both CTL lines recognized the exact same rp-HPLC fractions as in the Nef⁺ T1 cell extracts (fractions 21/22 and 33–35). Active fractions identified in the Nef⁺ T1 cell extracts and in the products derived by proteasomal digestion and corresponding control fractions were reexamined with serial dilutions of the CTLs, confirming the single E/T ratio results of individual fractions (Fig. 2A [right] and B [right]). These results strongly suggested that the HLA-A2–restricted CTL epitopes detected in material eluted from Nef-transfected cells were identical to those epitopes identified in the proteasomal digest.

To identify HLA-B7–restricted naturally processed Nef peptides, peptides were acid eluted from B7⁺, Nef⁺ Jurkat cells and fractionated by rp-HPLC. Fractions were first screened with an HLA-B7–restricted CTL line generated against Nef₁₂₈ TPGPGVRYPL₁₃₇. Again surprisingly, Nef₁₂₈TPGPGVRYPL₁₃₇-specific CTLs recognized two rp-HPLC fractions of acid-eluted material: fraction 3 corresponding to the elution time of Nef₁₂₈ TPGPGVRY₁₃₅, and fraction 15 corresponding to the elution time of Nef₁₂₈TPGPGVRYPL₁₃₇ (Fig. 3 A, top). Accordingly, CTLs were prepared against the smaller peptide, the octamer Nef₁₂₈TPGPGVRY₁₃₅, and these were found to recognize the same two fractions (Fig. 3 B). These results suggest that the two peptides represent two naturally processed versions of the same epitope, differing in length by two COOH-terminal aa.

Peptides eluted from Nef-transfected Jurkat cells were also screened with a CTL line against Nef₁₃₅YPLTFGWCY₁₄₃. This CTL line recognized a single fraction of naturally processed peptides (fraction 23), corresponding to the elution time of the inducing peptide (Fig. 3 C, top). This suggests that Nef₁₃₅YPLTFGWCY₁₄₃ is indeed a naturally processed peptide, and that this epitope exists only in one HLA-B7–binding version. However, the epitope overlaps extensively with the HLA-A2–binding naturally processed peptides Nef₁₃₆PLTFGWCYKL₁₄₅ and Nef₁₃₆PLTFGWCYKLV₁₄₆ described in the previous section. All three HLA-B7–restricted Nef peptide–specific CTL lines failed to recognize peptide material eluted from Nef⁻ Jurkat cells or HLA-B7⁻, Nef⁺ T1 cells, indicating that the peptides eluted from Nef-transfected Jurkat cells were Nef derived and specifically bound to HLA-B7 molecules (Fig. 3 B, top left).

In parallel, proteasomal digests of Nef_123–152 were separated by rp-HPLC using the exact same gradients. Each of the three HLA-B7–restricted CTL lines recognized fractions identical to that of the acid-eluted peptides (Fig. 3A–C; compare top and bottom). Active fractions identified in the Nef-transfected Jurkat cell extracts and in the products derived by proteasomal digestion and corresponding control fractions were reexamined with serial dilutions of CTLs, confirming the single E/T ratio results of individual fractions (Fig. 3A [right] and B [right]). These results suggest that the peptides recognized by the three HLA-B7–restricted CTLs in acid eluates of Nef-transfected Jurkat cells and in proteasomal digests of Nef_123–152 were identical.

The proteasomal digest of Nef_123–152 was separated by rp-HPLC using a relatively steep gradient to recover the vast majority of all possible fragments (Fig. 4 A). The peptides produced were identified by mass spectrometry and Edman degradation. Fig. 4 B shows a digestion map compiling all identified degradation products and proteasomal cleavage sites, the strength of the latter according to the quantification of the adjacent products by Edman degradation. All of the five naturally processed peptides detected by CTLs (see above) could also be found by protein analytical methods. The HLA-B7–binding octamer peptide Nef₁₂₈TPGPGVRY₁₃₅ is the major dual cleavage product, presumably because it is flanked by predominant cleavage sites and very little cleavage occurs internally. The remaining four peptides suffer more pronounced internal cleavage and are thus produced in smaller, but still considerable amounts. It is remarkable that significant quantities of the HLA-B7 ligands Nef₁₂₈ TPGPGVRYPL₁₃₇ and Nef₁₃₅YPLTFGWCY₁₄₃ are produced upon proteasomal digestion, as both peptides are subject to destruction by the predominant cleavage site in the polypeptide Y₁₃₅-P₁₃₆. This major cleavage site generates the COOH terminus of the abundant HLA-B7–binding octamer Nef_128–135 as well as the NH₂ termini of the two naturally processed HLA-A2 ligands. Most significantly, in four of the five naturally processed peptides (Nef₁₂₈ TPGPGVRY₁₃₅, ₁₂₈TPGPGVRYPL₁₃₇, ₁₃₆PLTFGWCYKL₁₄₅, and ₁₃₆PLTFGWCYKLV₁₄₆), the NH₂ termini correspond to major proteasomal cleavage sites.

The results described above strongly suggest, but do not directly prove, that each of the five acid-eluted peptides are also successfully translocated by TAP and bind to MHC class I. To directly address affinity for TAP, binding assays were performed using microsomes isolated from insect cells expressing human TAP1–TAP2 complexes 28. Relative affinities (IC₅₀ values) measurable in this assay range from 0.1 to 3,000. Peptides with IC₅₀ values >3,000 are excluded from translocation into the endoplasmatic reticulum (ER [43]). The IC₅₀ values suggest a slightly higher TAP-binding affinity for the two HLA-A2 ligands than for the three HLA-B7 ligands (Table). Of note, all five HLA-binding peptides have a proline in either position 1 or 2, a situation thought to be unfavorable for TAP transport 28,44,45. Nevertheless, IC₅₀ values for all five peptides were in the range consistent with moderately efficient TAP transport (Table).

The five HLA ligands were also tested for their ability to bind to their MHC class I restriction elements (Table). For the two HLA-A2 ligands, this was done by a competition assay using immunoaffinity-purified HLA-A2 molecules. For the three HLA-B7 ligands, stabilization of HLA-B7 on the surface of HLA-B7–transfected T2 cells was determined by flow cytometry. As expected, MHC class I binding could be shown for all five peptides. The results for both the HLA-A2 and HLA-B7 ligands suggest binding affinities somewhat below that of the high affinity peptides used as positive control in the assays. Together, the TAP- and MHC-binding data would predict low to moderate efficiencies of presentation for all five of the Nef-derived HLA ligands.

On the basis of the results of the TAP- and MHC-binding assays, low to intermediate copy numbers of presented epitopes were predicted for all five of the peptides. The molar concentrations of the peptides recognized in acid-eluted rp-HPLC fractions were estimated in CTL assays in which each fraction was tested in serial dilutions in parallel with known molar concentrations of the same synthetic peptide. Examples are shown in Fig. 5. Total molecules of recovered peptides were calculated from the molarities, and copy numbers of the peptides per cell could then be estimated (Table). According to these calculations, the HLA-B7 ligand Nef₁₂₈TPGPGVRY₁₃₅ appears to be present in high copy numbers comparable to the most efficiently presented MHC class I ligands known. Intermediate copy numbers are calculated for the second HLA-B7 ligand, Nef₁₂₈TPGPGVRYPL₁₃₇, whereas the remaining three peptides appear at low copy numbers. The exceptionally efficient presentation of Nef₁₂₈TPGPGVRY₁₃₅ cannot be accounted for by exceptionally high values for TAP transport or MHC class I binding. However, Nef₁₂₈ TPGPGVRY₁₃₅ is by far the most abundant dual cleavage fragment in the proteasomal digest (see Fig. 4). This epitope may thus represent another example for a significant influence of proteasomes on epitope hierarchy.

Although it is undisputed that the COOH termini of proteasome-dependent epitopes are predominantly generated by proteasomal cleavage, it is controversial whether proteasomes also contribute significantly to liberation of the NH₂ termini of MHC class I ligands 1. Here we describe five naturally processed MHC class I binding peptides of HIV-1 Nef, corresponding to three different CTL epitopes. These are the first naturally processed peptides ever identified in HIV Nef. In contrast to previous reports on this subject, our study concerns not only a single but also a cohort of determinants derived, in addition, from a highly relevant antigen. Four results in combination strongly suggest the generation of these determinants by proteasomes without assistance of other proteases: (a) all five ligands studied were produced in definitive form by proteasomes in digests of the fragment HIV-1 Nef_123–152 (this paper) as well as of recombinant full-length Nef (Lucchiari-Hartz, M., N. Hitziger, K. Eichmann, and G. Niedermann, manuscript in preparation); (b) the NH₂ termini of four of the five ligands correspond to major proteasomal cleavage sites; (c) Nt-extended proteolytic fragments were found in proteasomal digests for only one of the five definitive ligands; and (d) all five peptides were transported by TAP.

Unexpectedly, four of the five ligands represent COOH-terminal length variants of only one HLA-B7– (8- and 10-mer) and one HLA-A2–binding epitope (10- and 11-mer), respectively. To our knowledge, this is the first description of length variants of CTL epitopes from a non–self-antigen. However, since length variants were seen for two out of three CTL epitopes, this observation might not be exceptional. In both cases, CTL lines induced with the short and the long variant peptide showed CTL cross-recognition of both peptide length variants. However, since we did not analyze CTLs on the clonal level, it cannot be excluded that CTLs with exclusive specificity for the inducing peptide are also induced. The length variants described here are COOH-terminal length variants. Thus, our finding is in agreement with the notion that there is apparently no COOH-terminal trimming activity in the ER 46,47, and also no evidence for effective COOH-terminal trimming activity in the cytosol 12. Recognition of the longer variants by CTLs was not dependent on extracellular trimming by carboxypeptidases present in FCS (data not shown). Two binding modes for class I ligands slightly longer than canonical peptides have been described: they either protrude beyond the COOH-terminal end of the MHC class I peptide binding groove 48 or are fixed at the COOH terminus and bulge out in the middle 49,50. Since we observe extensive CTL cross-recognition, we favor the former binding mode for the longer epitope variants in both of our pairs of class I ligands as being more likely to be compatible with conserved conformation of the central peptide residues critical for TCR recognition.

In four of the five ligands, the NH₂ termini coincide with major proteasome cleavage sites. In the case of all five ligands, not only the COOH-terminal aa, but also the aa in the flanking position N minus 1, are preferred P1 residues of proteasomes (either tyrosine, leucine, valine, or arginine). Four ligands have small or polar aa in the NH₂-terminal (P1′) position of proteasomal cleavage sites. The major cleavage site that creates the COOH terminus of the abundant octamer HLA-B7–binding epitope and the NH₂ terminus of the two HLA-A2 ligands has a valine in position P3. Small and/or polar P1′ residues and hydrophobic P3 residues can promote proteasomal cleavage at the P1–P1′ site 3,4,10,51. The NH₂-terminal two thirds of the HLA-B7 ligand Nef₁₂₈TPGPGVRY₁₃₅, the sequence of which is identical with the NH₂-terminal flanking region of the two HLA-A2 ligands and of the HLA-B7 ligand Nef₁₃₅YPLTFGWCY₁₄₃, contains only proline, glycine, and threonine. These residues are extremely disfavored P1 residues of proteasomes 3,10 conferring protection against proteasomal cleavage.

Four HLA ligands had low to moderate affinities for human TAP, whereas the ₁₃₆PLTFGWCYKLV₁₄₆ HLA-A2 ligand had a slightly higher TAP affinity. All five ligands have proline in either position 1 or 2. It has been recognized previously that prolyl residues in positions 1, 2, or 3 are generally unfavorable for TAP translocation 28,44,45,52. In such cases, it has been proposed that NH₂-terminal extension can improve TAP affinity, and that the final class I ligands may be produced by trimming of epitope precursors in the ER 52,53. However, we have recently presented evidence suggesting that even for presentation by HLA-A2, which is suboptimally adapted to TAP, antigen processing may favor peptides that do not require ER processing 43,45. For the two HLA-A2 ligands and for two of the three HLA-B7 ligands identified in this study, Nt-elongated precursor peptides were not found among the proteasomal products. One of the potential precursors of the HLA-B7–binding octamer Nef₁₂₈TPGPGVRY₁₃₅ was the nonamer YTPGPGVRY. However, this nonamer is generated in lower amounts than the octamer. In addition, it has a proline in position 3 and should therefore not be preferred in TAP transport. The 10-mer NYTPGPGVRY and the 12-mer WQNYTPGPGVRY might have higher TAP affinity because they lack a prolyl residue in positions 1–3. Although these peptides are generated in significantly smaller amounts than the octamer, a contribution of NH₂-terminal trimming of these putative precursors in the ER is presently not excluded and needs to be evaluated. The only proteasomal fragment that could represent an Nt-elongated precursor of the HLA-B7–binding 10-mer ligand Nef₁₂₈ TPGPGVRYPL₁₃₇ was a single cleavage product starting with the NH₂-terminus of the 30-mer substrate. A similar single cleavage intermediate was found in significant amounts for the HLA-B7–binding octamer. However, these single cleavage intermediates were not found in digests of full-length Nef (Lucchiari-Hartz, M., N. Hitziger, K. Eichmann, and G. Niedermann, manuscript in preparation) and most probably result from the limited length of the substrate used here. Together, our data suggest that at least four of the five HLA-A2 and HLA-B7 ligands identified here are translocated into the ER predominantly in their definitive form, despite suboptimal TAP transport. We have shown previously that peptides with similar TAP affinities can even be very efficiently presented when abundantly generated 43. Of note, it has been shown that even the highly restrictive mouse TAP translocates peptides with unfavorable COOH-terminal residues in amounts sufficient for T cell recognition. The selective TAP influence became detectable only at limiting cytosolic peptide concentrations 54.

Several previous studies showing that MHC class I ligands may be direct major products of 20S proteasomes and/or proteasome–PA28 complexes were concerned with highly selected examples, i.e., the high copy self-peptides SYFPEITHI and TLWVDPYEV, derived from the cellular tyrosine kinase Janus kinase (JAK) 1, and the product of the B cell translocation gene 1, respectively 4,8,55, or the immunodominant OVA epitope SIINFEKL 3,4,56. Slightly longer precursor peptides that could be candidates for NH₂-terminal trimming were either not found or were produced in low quantities in these cases. A second group of MHC ligands was shown to be generated by proteasomes as minor products. The L^d ligand YPHFMPTNL derived from the pp89 protein of the murine cytomegalovirus and the subdominant OVA epitope KVVRFDKL are produced by proteasomes, albeit in small amounts, whereas Nt-extended peptides are more efficiently produced 3,8,57,58. A p53-derived and a β-galactosidase–derived class I ligand could also be detected in proteasomal digests of a source polypeptide and the source protein, respectively, albeit only with the highly sensitive use of specific CTLs 27,56. Three further reports describe unsuccessful attempts to detect epitopes in proteasomal digests: the KSPWFTTL peptide derived from the p15E protein of the AKV/MCF type of murine leukemia virus 53, the vesicular stomatitis virus nucleoprotein–derived epitope RGYVYQGL 59, and the peptide IPGLPLSL derived from the c-akt protooncogene 60. In these cases, only slightly longer peptides were found by protein analytical methods. However, in one of these studies, minor products have not been analyzed 60. Furthermore, the exact NH₂ terminus of human melanoma antigen (MAGE)3_271–179 is liberated by proteasomes. Nevertheless, the epitope is cryptic, because its COOH terminus is normally not liberated by proteasomes. However, crypticity of this epitope is abolished in the presence of lactacystin, since lactacystin-treated proteasomes generate the epitope COOH terminus in addition to its NH₂ terminus 61. Together, these data suggest that proteasomes often produce the final MHC class I ligands, albeit in varying amounts. Therefore, the available data have not led to a general consensus on the role of proteasomes in antigen processing.

The notion that epitope NH₂ termini may frequently result from nonproteasomal cleavage in vivo stems primarily from a study on the generation of SIINFEKL in cells transfected with minigenes. Although production of the epitope from Ct-extended versions was inhibited by 2 or 20 μM lactacystin, that from Nt-extended versions was not 11. This finding may have an alternative explanation. The peptide bond at the SIINFEKL NH₂ terminus is hydrolyzed efficiently by purified 20S proteasomes and proteasome–PA28 complexes 3,4,56. Since the SIINFEKL NH₂ terminus is directly preceded by glutamic acid (E), cleavage of the E₂₅₆-S₂₅₇ bond is most likely dependent on the postglutamyl activity of the proteasome. This activity is only marginally and competitively inhibited by lactacystin, i.e., only partially even at excessive lactacystin concentrations 62. Thus, it is to be expected that proteasomal generation of the SIINFEKL NH₂ terminus is poorly inhibited by lactacystin. Moreover, it is unlikely that the extraordinarily rapid hydrolysis by proteasomes at this site is efficiently blocked by any competitive inhibitor.

Along the same line, it was proposed that the SIINFEKL NH₂ terminus is produced by LAP, a cytosolic aminopeptidase inducible by IFN-γ. Small amounts of SIINFEKL were shown to be produced upon prolonged incubation of synthetic QLESIINFEKL with cytosol preparations from IFN-γ–treated cells, or by purified LAP 12. We think that this study potentially overemphasizes a minor mechanism in the generation of SIINFEKL. This epitope is excised by purified proteasomes from partial or total OVA in 7–10-fold greater amounts than QLESIINFEKL 3,56. In addition, SIINFEKL is resistant against proteasomal attack once generated 3. In contrast, SIINFEKL is degraded when incubated with LAP (our unpublished data). Thus, the production of SIINFEKL via QLESIINFEKL by a two-step digestion involving aminopeptidases is likely to be a minor pathway. In addition, TPP II, which was inadvertently depleted from the cytosol in the protocol used by Beninga et al. 12, is a strong candidate for the generation of SIINFEKL via the QLESIINFEKL precursor.

We do not mean to exclude that precursor trimming (at least NH₂-terminal trimming) also contributes to MHC class I ligand formation. The steric constraints on peptide extensions at the NH₂ terminus of the class I peptide binding groove appear to be stricter than on peptide extensions at the COOH terminus 63. In accordance with that fact, alignments of eluted class I ligands 6 and MHC–peptide affinity measurements 64 suggest that stable complex formation between class I MHC molecules and Nt-extended peptides may not occur normally, although there may be exceptions 65. These stringent requirements for correct NH₂ termini in MHC class I ligands are compatible with a trimming activity generating suitable termini. This activity may act in cases where proteasomes do not generate the correct NH₂ terminus of a class I ligand. Moreover, trimming may often contribute to ligand formation when elongated peptides are produced by proteasomes in addition to the minimal ligands, especially when such precursors have significantly higher TAP affinities than the minimal epitopes. Indeed, experimental evidence for NH₂-terminal trimming capacity has been reported in the cytosol 12 and in the ER for both signal peptide–coupled epitope precursors 11,66,67,68 and a TAP-translocated peptide 43. However, it remains to be determined whether the generation of an epitope by two distinct proteases is as efficient as the generation of an epitope by proteasomes alone.

Nef is a major virulence factor of HIV and simian immunodeficiency virus (SIV), and appears to be critical for the development of AIDS 69,70,71. Among other biological effects, Nef downregulates MHC class I expression in HIV-infected and Nef-transfected cells 72, and was suggested to partially protect HIV-infected primary CD4⁺ T cells against recognition by HLA-A2–restricted HIV-Gag and HIV reverse transcriptase–specific CTL clones 73. On the other hand, there is also strong evidence that CD8⁺ lymphocytes play an important role in controlling viremia in SIV and HIV infections 74,75. An effective HIV vaccine should therefore be designed to elicit CTL responses. Before the present report, only two naturally processed CTL epitopes of HIV were known, one for Gag and the other for reverse transcriptase 26. In view of the MHC downregulation by Nef, it may be mandatory to target CTL vaccines to epitopes that are presented in high copy numbers, such as the HLA-B7–restricted naturally processed Nef peptide, ₁₂₈TPGPGVRY₁₃₅, identified in this paper. Nef is synthesized at the earliest stage of viral gene expression and is abundantly expressed 76. High anti-Nef CTL responses have been detected in the acute and asymptomatic phases as well as later in HIV infection 35,77,78, and high frequencies of CTL precursors have been found in noninfected individuals 20. A detailed knowledge of MHC class I–peptide ligands and their intracellular generation, as well as the molecular basis of the adverse effects of Nef, should be instrumental in the development of an HIV vaccine including Nef CTL epitopes.

The CTL donor H.K.'s generous participation in this study is acknowledged. We thank Dr. Ian Haidl (Max-Planck Institute of Immunobiology) for his critical reading of this manuscript. We also thank Ms. K. Dannappel for her peptide analysis (Biochemistry Service, Max-Planck Institute of Immunobiology), and E. Geier (Max-Planck Institute of Immunobiology) for her helpful discussions.

This study was supported by grants from the Deutsche Forschungsgemeinschaft to G. Niedermann (Ni 368/2-1), R. Maier, and A. Meyerhans (Me 1061/2-3), and from the European Union to K. Eichmann (Biotech 98-0242). M. Lucchiari-Hartz was funded by a fellowship from the Max-Planck Gesellschaft with the Conselho Nacional de Pesquisa-CNPq (contract 49.0303-97.8).