Differential proteasomal processing of hydrophobic and hydrophilic protein regions: Contribution to cytotoxic T lymphocyte epitope clustering in HIV-1-Nef

Abstract

HIV proteins contain a multitude of naturally processed cytotoxic T lymphocyte (CTL) epitopes that concentrate in clusters. The molecular basis of epitope clustering is of interest for understanding HIV immunogenicity and for vaccine design. We show that the CTL epitope clusters of HIV proteins predominantly coincide with hydrophobic regions, whereas the noncluster regions are predominantly hydrophilic. Analysis of the proteasomal degradation products of full-length HIV-Nef revealed a differential sensitivity of cluster and noncluster regions to proteasomal processing. Compared with the epitope-scarce noncluster regions, cluster regions are digested by proteasomes more intensively and with greater preference for hydrophobic P1 residues, resulting in substantially greater numbers of fragments with the sizes and COOH termini typical of epitopes and their precursors. Indeed, many of these fragments correspond to endogenously processed Nef epitopes and/or their potential precursors. The results suggest that differential proteasomal processing contributes importantly to the clustering of CTL epitopes in hydrophobic regions.

Comprehensive analyses of several HIV antigens including Nef have revealed an enormous diversity of epitopes recognized by cytotoxic T lymphocytes (CTLs) of HIV⁺ patients (see the HIV Molecular Immunology Database, http://hiv-web.lanl.gov/content/immunology). Moreover, HIV CTL epitopes extensively overlap in so-called epitope clusters, whereas other regions of the HIV proteins are almost completely devoid of CTL epitopes (refs. 1–6 and HIV Molecular Immunology Database). Evidence for CTL epitope clusters exists also for non-HIV antigens (7). Although CTL epitope clusters often coincide with rather conserved regions of the HIV proteins (1, 4, 6), it has been postulated that the highly nonuniform distribution of HIV CTL epitopes may be related to antigen processing and presentation (2, 4, 6).

Proteasomes seem to participate in the generation of many if not most CTL epitopes. In the classical class I antigen-processing pathway proteasomes generate epitopes as well as precursors that are trimmed by cytosolic and endoplasmic reticulum-resident aminopeptidases. Trimming by carboxypeptidases seems to be highly unusual. Cytosolic endoproteases other than proteasomes may be involved in the production of certain epitopes, but thus far none have been unequivocally identified (for several excellent reviews, see ref. 8 and all articles in the same issue). Important information on the role of proteasomes in antigen processing has been obtained by in vitro digestion of proteins and protein fragments with isolated proteasomes. However, thus far analyses of CTL epitope-containing full-length antigens (ovalbumin and β-galactosidase) (9, 10) and most antigen fragments focused on a single epitope and/or related epitope precursors. In nearly all experiments with CTL epitope-containing substrates, 20S proteasomes were used, and excellent correlations with epitope recognition on antigen-presenting cells were often reported (11–13).

The HIV protein Nef is a small regulatory protein and critical for the development of the acquired immunodeficiency syndrome. Nef is located in the cytoplasm and temporarily attached to the cell membrane via myristoylation (14, 15). It is a potent CTL antigen and reveals a typical pattern of high-density CTL epitope clusters (refs. 1 and 16 and HIV Molecular Immunology Database). Nef-specific CTLs may be particularly important in the control of HIV infection, because Nef is the predominant transcript early after infection and, perhaps as the only HIV protein, is expressed even before proviral integration (17, 18). In addition, primary HIV-infected T cells are lysed efficiently by Nef-specific CTLs (19), whereas abrogation of recognition by Gag- and polymerase-specific CTLs has been reported due to the Nef-mediated down-regulation of HLA-A and HLA-B molecules (20).

We recently reported that proteasome inhibitors abrogated CTL recognition of several Nef epitopes on Nef-expressing cells, suggesting proteasome involvement in the processing of HIV-Nef (21). Now we have digested recombinant full-length Nef protein with isolated proteasomes, comprehensively analyzed the cleavage products, and compared them with database epitope maps and a panel of naturally processed class I-binding Nef peptides determined by us. Our data suggest that the pronounced sensitivity of hydrophobic protein regions to proteasomal processing contributes importantly to the generation of an enormous multiplicity of extensively overlapping CTL epitope peptides and to their clustering in hydrophobic protein regions.

Methods

Digestion of Recombinant Nef with Purified Proteasomes. The nef gene of the HIV-1 LAI isolate was inserted into a variant (pRP261, kindly provided by B. Wollscheid, Max-Planck-Institut für Immunbiologie, Freiburg, Germany) of the pGEX-3X-vectors (Amersham Pharmacia) to generate a Nef-glutathione S-transferase fusion protein. Purification used glutathione-Sepharose 4B, and the Nef protein was cleaved off by incubation with thrombin and purified further by RP-HPLC. 20S proteasomes were purified from the human lymphoblastoid cell line T1 as described (21). Recombinant Nef (60 μg) was digested with isolated proteasomes (2 μg) at 37°C in a total volume of 300 μl of buffer (20 mM Hepes, pH 7.3/1 mM EGTA/5 mM MgCl₂/0.5 mM 2-mercaptoethanol). The reaction mixture was fractionated by RP-HPLC on a Smart system (Amersham Pharmacia) equipped with a Sephasil C18 SC2.1/10 column with 0.1% trifluoroacetic acid (eluent A) and 80% acetonitrile containing 0.081% trifluoroacetic acid (eluent B). Degradation products were separated on a gradient of 10–55% B in 50 min and eluted between 16/17% and 52% of eluent B. Aliquots of the HPLC fractions were analyzed by matrix-associated laser desorption ionization/time-of-flight MS with a Voyager STR spectrometer coupled with delayed extraction (Applied Biosystems; lower molecular mass limit, 400–450 Da). In some cases, sequences were confirmed by tandem-MS analysis on a Micromass Q-TOF instrument. Aliquots of reconstituted fractions were also tested for recognition by Nef-peptide-specific CTL in a ⁵¹Cr-release assay on P815 murine mastocytoma cells transfected with HLA-A2, HLA-B7, or HLA-A3.

Identification of Naturally Processed Nef Peptides and in Vitro Induction of Primary Nef-Specific CTL Lines. Naturally processed Nef peptides were identified essentially as described (21). Briefly, peptides acid-eluted from Nef-transfected cells (HLA-A2.1⁺ T1, HLA-B7.2⁺ Jurkat, or C1R cells transfected with HLA-A3) were fractionated by RP-HPLC with shallow gradients specifically designed to resolve mixtures of related peptides. CTL epitopes were determined by ⁵¹Cr-release assay with 17 polyclonal Nef-peptide-specific CTL lines induced in vitro with synthetic peptides as described (21).

Results

Pronounced Sensitivity of CTL Epitope Clusters to Proteasomal Digestion. The vast majority of endogenously processed Nef-CTL epitopes is concentrated in the central portion of the protein (residues 68–146) and at the COOH terminus (residues 180–206) (Fig. 1). Regions 68–100 and 113–146 in the central portion have a particularly high epitope density, thus some authors consider these as two separate epitope clusters (1). We digested the full-length recombinant HIV-1 LAI Nef protein with 20S proteasomes purified from T1 lymphoma cells containing both housekeeping proteasomes and immunoproteasomes as revealed by Western blot (data not shown). The LAI (formerly Bru) sequence was chosen because most Nef-CTL epitopes recorded in the literature are defined for LAI (HIV Molecular Immunology Database). Cleavage products were identified by MS. We comprehensively analyzed the peptide mixtures at ≈40%, 75%, and 95% of substrate consumption. Essentially all cleavage products were found at all three time points. Fig. 1 shows the peptides (n = 143) produced when ≈40% of the substrate molecules had been consumed. To simplify analysis, regions 68–146 and 180–206 were taken as epitope clusters, and regions 1–67 and 147–179 were taken as noncluster regions. Although cluster and noncluster regions each cover approximately half (106/100 aa) of the Nef sequence, we detected three times as many proteasomal cleavage products in the former (107 vs. 36), corresponding to twice the density of proteasomal cleavage sites in the epitope-cluster regions compared with the noncluster regions (1/1.5 vs. 1/3.1 aa).

Fig. 1.

Map of the HIV-1-Nef protein showing CTL epitope peptides and proteasomal cleavage products. CTL epitope peptides are shown above the sequence. Black lines are those recognized by CTLs of HIV-infected subjects as recorded in the HIV Molecular Immunology Database. Red lines are the naturally processed peptides acid-eluted from Nef-transfected cells (see also Table 1). The 20S proteasomal cleavage products of recombinant HIV-1-Nef (LAI) protein are indicated as lines below the sequence. Fragments corresponding to acid-eluted HLA class I ligands are highlighted in red; fragments corresponding to their potential precursor peptides are highlighted in green. *, Epitopes identified by immunization of HLA-A2 transgenic mice with HIV-1-Nef (HXB3), with the sequences ALTSSNTA and MTYKAALDL instead of AITSSNTA and MTYKAAVDL (LAI); **, a 20-aa sequence (Nef_161–180) where a CTL epitope of unknown HLA restriction was located; ***, subdominant epitope with the sequence DPEKEVLQWK (LAI sequence, DPEREVLEWR).

The sizes of the proteasomal cleavage products are given in Fig. 2. Very few were shorter than 8 aa. For all peptides of 8 aa and longer, i.e., sizes potentially giving rise to MHC class I ligands, 2.6 times as many (92 vs. 35) were found in the cluster as in the noncluster regions. A substantial proportion of cleavage products from the epitope clusters (63 of 107) fell into the approximate size range of MHC class I peptide ligands, i.e., 8–11 aa. Only 9 of the 36 cleavage products from the noncluster regions had this length, with the majority being longer.

Fig. 2.

Length distribution of 20S proteasomal cleavage fragments from HIV-1-Nef protein. Recombinant HIV-1-Nef (LAI) protein was digested with purified 20S proteasomes, and the cleavage fragments were determined by MS (see Fig. 1). The preferred size range of MHC class I peptide ligands is highlighted in pink.

Enhanced Preference for Hydrophobic P1 Residues in CTL Epitope-Cluster Regions. At the COOH termini of the Nef proteasomal cleavage products we frequently found large hydrophobic amino acids, certain charged amino acids, as well as the small amino acid alanine, with a hydrophobic side chain (Fig. 3). At the NH₂ termini of cleavage products preferences were not so pronounced, but small and certain charged amino acids were often found. Except for the acidic amino acids (which are not frequent at the COOH termini of mammalian MHC class I ligands), particularly glutamic acid, the amino acid frequencies at the COOH termini of proteasomal cleavage fragments are in good agreement with that at the COOH termini of mammalian MHC class I ligands (see also ref. 22).

Fig. 3.

Frequencies of amino acids at the COOH and NH₂ termini of the 143 Nef proteasomal cleavage fragments shown in Fig. 1. The COOH-terminal amino acids are shown as gray bars, and the NH₂-terminal amino acids are shown as white bars. For the COOH-terminal amino acids, the proportion of peptides from the CTL epitope clusters is indicated by the hatched parts of the bars. The frequencies (%) of individual amino acids in the Nef protein are given below the x axis.

Unexpectedly, we found that the vast majority of cleavage fragments with COOH termini suitable for MHC class I binding was derived from epitope-cluster regions (Fig. 3, hatched portions of the gray bars). Conversely, many fragments with COOH termini rarely found in MHC class I ligands stem from noncluster regions (Fig. 3). These differences between fragments liberated from cluster and noncluster regions reflect pronounced differences in cleavage-site preferences. In cluster regions proteasomes cleaved after 87% (39/45) of the hydrophobic amino acids but only after 37% (12/32) in noncluster regions. For example, cleavages were observed after 83% (10/12) of the leucine residues and after 71% (5/7) of the valine residues in cluster regions and only after 20% (1/5) of leucine and 37% (3/8) of valine residues in noncluster regions. We therefore analyzed whether the sequence environment of a cleavage site (P1-P1′) had an influence on the cleavage preferences. Positions P8 to P4′ of cleavage sites with a hydrophobic P1 residue contained a higher frequency of hydrophobic amino acids in cluster regions (24–45%; mean 33%, >40% in positions P7, P5, P3, and P2′) than in noncluster regions (0–30%; mean 13%). Leucine residues in noncluster regions not used for cleavage showed a total lack of large hydrophobic amino acids in positions P8 to P3 and P2′. Together, these data reveal a more-pronounced preference of proteasomes for hydrophobic amino acids in P1 in cluster regions than in noncluster regions, which depends on a hydrophobic sequence environment of the cleavage site.

When both the sizes and COOH termini are considered, the numbers of proteasomal cleavage fragments suitable as CTL epitopes and epitope precursors differ considerably between cluster and noncluster regions. Of the 92 cleavage fragments of ≥8 aa from the epitope clusters (see Fig. 2), 75 have a COOH-terminal amino acid frequently found in mammalian MHC class I ligands; of these, 67 have a hydrophobic and 8 have a basic COOH-terminal amino acid. In contrast, of the 35 peptides ≥8 aa from the noncluster regions, only 16 have a hydrophobic or basic COOH-terminal amino acid.

Hydrophobicity and Conservation of Cluster and Noncluster Regions. As shown in Fig. 4A, the regions that correspond to the Nef-CTL epitope clusters and that are processed intensively by proteasomes are on average more hydrophobic than the less-intensively processed noncluster regions, which are predominantly hydrophilic. To investigate whether the relationship between hydrophobicity and epitope density was unique to Nef or more general, we performed a similar analysis of Gag p17. This well studied protein contains two CTL epitope clusters (11–44 and 71–101), and a third one begins at the boundary to Gag p24 (5, 6). When the epitope clusters of Gag p17 were plotted against the hydrophobicity/hydrophilicity profile (Fig. 4B), a striking relationship between epitope clusters and hydrophobicity was observed, similar to that found for Nef.

Fig. 4.

Hydrophobicity/hydrophilicity plots and conservation of the primary sequence of HIV-1-Nef (A) and HIV-1-Gag p17 (B). Hydrophobicity/hydrophilicity plots were generated with the algorithm HPLC/retention pH 7.4; similar plots were obtained with the algorithm of Abraham and Leo (ExPASy Molecular Biology Server, www.expasy.org/cgi-bin/protscale.pl). Plots created with window sizes between 15 and 21 looked almost identical. CTL epitope clusters and single CTL epitopes are represented by boxes at the top. For Nef-CTL epitopes, see Fig. 1; for Gag CTL epitopes, see the HIV Molecular Immunology Database. The variability plots (below the hydrophobicity plots) show the number of different amino acids (dots) at each position of the concensus sequences of subtypes A–D. Sequence alignment was performed by using align (www2.igh.cnrs.fr/bin/alignguess.cgi). *, a 20-aa sequence (Nef_161–180) where an HLA class I-restricted epitope (of unknown type restriction) was mapped.

It has been shown by others that CTL epitope clusters often correspond to more-conserved regions of the HIV proteins (1, 3, 4, 6, 23). In fact, the central hydrophobic portion of Nef (region 68–146) encompassing the two central CTL epitope clusters is quite well conserved among the HIV-1 subtypes. However, the hydrophobic COOH-terminal cluster (region 180–206) is not as well conserved. Conversely, the Nef protein contains conserved sequence stretches that are hydrophilic and do not correspond to epitope-cluster regions (Fig. 4A). These stretches often correspond to protein–protein interaction sites responsible for pathogenic effects of Nef (14). For Gag p17, both hydrophobic epitope clusters show a higher degree of conservation than the noncluster regions (Fig. 4B). Because hydrophobic regions are important for protein folding, they may generally be quite conserved. Some correlation of epitope clustering to sequence conservation is therefore expected.

Relationships of Proteasomal Fragments to Naturally Processed Class I-Binding Nef Peptides. To compare proteasomal cleavage fragments with naturally processed MHC-presented peptides, we determined class I-binding Nef peptides acid-eluted from Nef-transfected cells. Previously, only two naturally processed, class I-binding HIV peptides had been identified, one from Gag and one from reverse transcriptase (24). Of the 17 eluted Nef peptides listed in Table 1, 12 were identified in the present study and 5 were identified previously (21). The peptides stem from the two central Nef-CTL epitope clusters and are highlighted in red in Fig. 1. Due to cross recognition, CTL lines against the 17 Nef-HLA class I ligands distinguished a minimum of nine CTL epitopes. Six of the nine epitopes occurred in several naturally processed length variants (four in two, and two in three length variants) including the examples presented previously (21). Recognition of the longer variants by CTL did not depend on extracellular trimming by carboxypeptidases present in FCS (21). The length variants were predominantly but not exclusively COOH-terminal variants. These results suggest that the presentation of COOH-terminal epitope length variants is a frequent phenomenon. The nine epitopes are described in the literature as being recognized by HIV-infected patients and/or to bind to HLA class I molecules (Table 1).

Table 1. Relationships between naturally processed Nef peptides and full-length Nef proteasomal degradation products.

Naturally processed peptide*	Eluted from HLA*	HLA-restriction of CTLs in HIV+ subjects or HLA-binding†	Peptide found in proteasomal Nef-digest:		Nt extended precursors found by MS in proteasomal Nef-digest
			CTLs	MS‡
71TPQVPLRPM79	B7	B7/B35§	+	+	PVTPQVPLRPM
71TPQVPLRPMTY81	B7	B7/B35§	+	+	PVTPQVPLRPMTY
					EVGF PVTPQVPLRPMTY
					EEVGF PVTPQVPLRPMTY
73QVPLRPMTY81	A3	B35¶	+	-	PQVPLRPMTY
74VPLRPMTY81	A3	A3/B35	+	+	TPQVPLRPMTY∥
					PVTPQVPLRPMTY**
					EVGF PVTPQVPLRPMTY
					EEVGF PVTPQVPLRPMTY
73QVPLRPMTYK82	A3	A2/A3/A11/B35/B27/A30††	-	-	PVTPQVPLRPMTYK‡‡
74VPLRPMTYKA81	A3	VPLRPMTYK binds to HLA-A11	+	+	EEVGF PVTPQVPLRPMTYKA
75PLRPMTYK82	A3	A11	-	-	PVTPQVPLRPMTYK‡‡
75PLRPMTYKAA84	A3		+	+	PVTPQVPLRPMTYKAA
75PLRPMTYKA83	A3	binds to HLA-A2§	+	+	EEVGF PVTPQVPLRPMTYKA
77RPMTYKAA84	B7		+	-	PLRPMTYKAA∥
77RPMTYKAAV85	B7	B7/binds to HLA-B35§	+	-	-
77RPMTYKAAVDL87	B7		+	-	-
128TPGPGVRY135§§	B7	binds to B7§,§§	+	+	YTPGPGVRY
					NYTPGPGVRY∥
					WQNYTPGPGVRY
					FPDWQNYTPGPGVRY
128TPGPGVRYPL137§§	B7	B7/B35/B42/B81	+	+	-
135YPLTFGWCY143§§	B7	B7/B18/B35§/B49/B53	+	+	VRYPLTFGWCY‡‡
136PLTFGWCYKL145§§	A2	A2	+	+	-
136PLTFGWCYKLV146§§	A2	binds to A2§§	+	+	-

^*Eluted from Nef-transfected cells and grouped according to CTL cross recognition

^†As recorded in the HIV Molecular Immunology Database

^‡Additional cleavage products found by MS corresponding to other epitopes described in the literature are ¹³WPTVRERM²⁰ (HLA-B8), ⁷⁹MTYKAAVDL⁸⁷ and ¹⁸⁰VLEWRFDSRL¹⁸⁹ (HLA-A2), and ¹⁰⁶RQDILDLWIY¹¹⁵ (HLA-B7); we also found ¹⁰⁶RQDILDLWI¹¹⁴ and ¹⁰⁶RQDILDLW¹¹³, which may also bind to HLA-B7, ¹⁰⁵RRQDILDLWI¹¹⁴ (HLA-B27), ¹⁰⁵RRQDILDLW¹¹³ and ¹⁰⁵RRQDILDLWIY¹¹⁵, which may also bind to HLA-B27, ¹²⁶NYTPGPGVRY¹³⁵ (HLA-A24), ¹³⁶PLTFGWCYK¹⁴⁴ (HLA-A3), and ¹⁸⁶DSRLAFHH¹⁹³ (HLA-B35) and ¹⁸⁶DSRLAFHHV¹⁹⁴ (HLA-B51)

^§As recorded in ref. 16

^¶As recorded in ref. 1

^∥Also an epitope/HLA ligand; TPQVPLRPMTY (HLA-B35 epitope/B7 ligand) and potential precursor of QVPLRPMTY and VPLRPMTY (HLA-A3 ligands/B35 epitopes); PLRPMTYKAA (HLA-A3 ligand and potential precursor of the HLA-B7 ligand RPMTYKAA); NYTPGPGVRY (HLA-A24 epitope and precursor of the B7 ligand TPGPGVRY)

^**Also excised from a 35-mer Nef peptide by archaebacterial proteasomes (ref. 16)

^††As recorded in ref. 25

^‡‡Detected only in very small amounts

^§§Data are from ref. 21

For the majority of the naturally processed Nef peptides the exactly corresponding peptides were found among the proteasomal cleavage products of full-length Nef (Table 1 and highlighted in red in Fig. 1). Potential NH₂-terminally extended precursors were detected for the majority of but not all ligands (Table 1 and highlighted in green in Fig. 1). We also found a number of proteasomal cleavage products that correspond to CTL epitopes mapped by others (HIV Molecular Immunology Database), which may correspond to HLA ligands (see ‡ footnote in Table 1). For most of these epitopes, NH₂-terminally extended precursor peptides were also among the proteasomal cleavage products. Some cleavage products correspond to both an HLA ligand and a potential precursor (see g∥ footnote in Table 1). Taken together, the data presented suggest that nearly all of the peptides excised by proteasomes from the CTL epitope-cluster regions are related to HLA class I ligands (see also Fig. 1).

Discussion

HIV antigens contain large numbers of endogenously processed CTL epitopes that tend to concentrate in clusters. We found that the CTL epitope clusters correspond to the hydrophobic regions of the HIV proteins and that proteasomes excise large numbers of extensively overlapping peptides with the sizes and COOH termini typical of epitope and epitope-precursor peptides, particularly from hydrophobic regions of the Nef protein. Many of these cleavage products indeed correspond to endogenously processed Nef-CTL epitope peptides and their potential precursors. We propose that this is due to four fundamental properties of proteasomes: (i) the broad cleavage-site specificity; (ii) the enhanced preference for hydrophobic P1 residues in a hydrophobic sequence context; (iii) the excision of large numbers of peptides ≥8 aa; and (iv) the versatility of proteasomal processing resulting in overlapping peptide patterns. Together, our data suggest that proteasomal proteolysis contributes importantly to the multiplicity of endogenously processed epitope peptides and to their clustering in hydrophobic regions.

Systematic epitope mappings with CTLs of HIV⁺ subjects thus far include only an incomplete set of HLA alleles (HIV Molecular Immunology Database). Hence, the number of endogenously processed HIV CTL epitope peptides may still increase, even though some HLA alleles share similar peptide-binding motifs. CTL responses to multiple epitopes are often observed, i.e., in different studies individual HIV⁺ subjects recognized 1–13 (25), 2–17 (26), and 0–13 (27) epitopes. These are minimal estimates because of the incomplete knowledge of CTL epitopes and the use of reference strain-based peptides. In addition, different subjects sharing the same HLA allele may respond to different epitopes presentable by this allele, and the same subject may recognize different epitopes at different stages of infection (25–30).

HIV may not be the only virus associated with a multitude of endogenously processed CTL epitopes as well as a pronounced heterogeneity of CTL responses. The hepatitis B virus surface antigen harbors at least eight endogenously processed HLA-A2-binding CTL epitopes, and patients develop CTL responses against multiple epitopes in the hepatitis B virus antigens (31). The few hepatitis C virus proteins contain >20 different HLA-A2-binding epitopes, many of which have been shown to be endogenously processed (32), and a response against multiple epitopes seems to be characteristic of hepatitis C virus-infected individuals, particularly during acute infection (33). Tumor antigens such as the melanoma antigen gp100, Her-2/neu, the PRAME protein, and the MAGE antigens contain multiple epitopes recognized by CTLs of patients; gp100 contains at least nine different endogenously processed HLA-A2-binding epitopes (see the Peptide Database of T Cell-Defined Tumor Antigens, Ludwig Institute for Cancer Research, Brussels, at www.cancerimmunity.org/peptidedatabase/Tcellepitopes.htm).

CTL epitope clusters may not be unique to the HIV proteins either. A pronounced CTL epitope cluster has been described for the major outer membrane protein of Chlamydia trachomatis (7). In hepatitis B virus/hepatitis C virus as well as tumor antigens several of the epitopes identified thus far (mostly HLA-A2-binding epitopes) overlap with or are close to others (refs. 31 and 32 and Peptide Database of T Cell-Defined Tumor Antigens). The underlying basis of CTL epitope clustering in HIV proteins has been of considerable interest. On the one hand, CTL epitope clusters often coincide with conserved regions of the HIV proteins, possibly due in part to the use of reference strain-based peptides (4, 6). On the other hand, mechanisms based on evolutionary host/virus adaptation have been considered. HLA class I-binding motifs might have been eliminated in the more variable and therefore functionally less-important regions due to immunoselection, a putative strategy of HIV to subvert the host defense (3). Conversely, the preferential presentation of more-conserved peptides might be a strategy of the host to combat the immune escape attempts of the virus (23). Differences in antigen processing and presentation have also been considered (2, 4, 6, 34). Our data suggest that such differences indeed contribute significantly to CTL epitope clustering. We demonstrate a striking correlation between three elements of a protein sequence: the degree of hydrophobicity, the intensity of proteasomal processing, and the density of CTL epitopes. The preference of proteasomes for hydrophobic protein stretches is in accord with the cleavage preferences of the proteasome for hydrophobic P1 residues (12, 22, 35) (see Fig. 3). Moreover, our data show that hydrophobic amino acids are used more frequently for cleavage in the hydrophobic environment of epitope clusters than in the hydrophilic environment of noncluster regions. Hydrophobic interactions with hydrophobic inner surfaces of the proteasome are probably important for translocation of substrates and for their maintenance in an unfolded and easily degradable state (36, 37).

The number of proteasomal fragments potentially giving rise to CTL epitopes from the Nef clusters is 4.7-fold greater than that from the noncluster regions (75 vs. 16). Surprisingly, the difference in numbers of known CTL epitopes is even greater, namely >50 CTL epitope peptides in cluster regions compared with 2 or 3 in noncluster regions (see Fig. 1), which may be because of the more-difficult detection of epitopes in the less-conserved noncluster regions. In addition, fragments from noncluster regions that possess a suitable size and COOH terminus may lack hydrophobic internal or auxiliary anchor residues for MHC class I binding.

Analyses of 42 individual self-MHC class I ligands from as many different source proteins revealed that these were more hydrophobic than the rest of their source proteins and were derived from evolutionarily more-conserved regions (38). Although this study did not address epitope clusters, these results suggest that our conclusions may not only be relevant for HIV proteins. Hydrophobic protein regions may tend to be highly conserved because of their importance for correct protein folding. The Nef-CTL epitope clusters cover virtually the entire folded core domain of the Nef protein (see Fig. 1 and 4A; for structural information, see ref. 14). However, proteolytic enzymes cannot recognize the conservation of a protein sequence, thus it is unlikely to be causally related to proteasomal processing.

It is likely that Nef is predominantly processed by the classical proteasomal pathway, in line with our finding that proteasome inhibitors abrogated CTL recognition of several Nef epitopes on Nef-expressing cells (21). However, because in certain cases HLA class I assembly is only partially sensitive to proteasome inhibitors, a few other known endoproteases may contribute to the generation of CTL epitopes. Tripeptidyl peptidase II might contribute to the generation of CTL epitopes with basic COOH-terminal amino acids, particularly with COOH-terminal lysine residues (39), whereas proteasomes usually do not cleave efficiently after lysine (see Fig. 3 and ‡‡ footnote in Table 1). Calpains have not been shown to be involved in antigen processing. The endoproteolytic thimet oligopeptidase seems to destroy proteasomal cleavage products, including epitopes, in the cytosol (10). The endoplasmic reticulum-signal peptidase and furin in the trans-Golgi operate in nonclassical processing pathways (40, 41). Because of their more-restricted substrate and cleavage specificity, nonproteasomal endoproteases likely generate only limited subsets of class I-binding peptides. Taken together, these findings and our present data suggest that proteasomes perhaps are not the only but the major endoprotease in charge of processing Nef. The data therefore suggest an important contribution of proteasomes to epitope clustering in Nef.

CTL responses may play an important role in the control of HIV infection (2, 6). Effective HIV vaccines therefore should contain CTL epitopes restricted by many different HLA alleles including those of Nef. Because hydrophobic protein regions often seem to be quite well conserved, they may harbor epitopes particularly effective for protective host immunity. On the other hand, the generation of a large variety of epitopes, many of which overlap and therefore display pronounced structural similarity, may result in a CTL response of pronounced heterogeneity and low average affinity and therefore may be part of the immune evasion strategy of HIV. The clustering of CTL epitopes in hydrophobic regions contrasts with Nef pathogenicity, which is based on protein–protein interaction mostly involving hydrophilic parts of the protein (14, 15). The preferential association of pathogenicity and immunogenicity with hydrophilic and hydrophobic regions, respectively, of Nef seems particularly promising for the design of vaccines based on truncated versions of Nef.