HIV Protease Inhibitor–Induced Cathepsin Modulation Alters Antigen Processing and Cross-Presentation

This article is featured in In This Issue, p.3497

Human immunodeficiency virus infection has been considered a chronic illness since the availability of highly active antiretroviral therapies (ART), a combination of nucleoside reverse-transcriptase inhibitors, nonnucleoside reverse-transcriptase inhibitors, protease inhibitors (PIs), and integrase inhibitors given to HIV-infected patients to suppress HIV replication (1). HIV PIs block the HIV aspartyl protease, preventing the cleavage of HIV Gag-Pol polyproteins and the conversion of HIV particles into mature infectious virions (2). The ability of proteasomes to cleave similar bonds as HIV-1 protease (3) raised questions about interactions between HIV PIs and proteasomal catalytic sites.

Proteasomes play a central role in the generation of MHC class I (MHC-I) peptides (4). They degrade full-length proteins and defective ribosomal products into peptides that can be further shortened by cytosolic aminopeptidases and endopeptidases. Some peptides are translocated by the TAP complex into the endoplasmic reticulum, where they can be further trimmed by endoplasmic reticulum–resident aminopeptidases 1 or 2 and, if they contain appropriate anchor residues, loaded onto MHC-I and displayed at the cell surface. HIV PIs ritonavir (RTV), saquinavir (SQV), and nelfinavir (NFV) inhibit the proteolytic activities of the proteasome, a threonine protease (5, 6), causing intracellular accumulation of polyubiquitinated proteins (7). In mice infected with lymphocytic choriomeningitis virus and treated with RTV, CTL responses against lymphocytic choriomeningitis virus epitopes were changed (6). We previously showed that HIV PIs altered proteasome and aminopeptidase activities in primary human cells, modified HIV protein degradation patterns, epitope production, and presentation, and altered CTL responses (8).

Professional APCs, like dendritic cells (DCs) and macrophages (MΦs), can cross-present exogenous Ags (9, 10). Ags transiting through endosomes or phagosomes are partially degraded by cathepsins. Degradation fragments can escape into the cytosol for processing and presentation in the direct Ag-processing pathway (11) or be processed completely in the phagosomes by cathepsins before loading onto MHC-I (12, 13). The mechanisms of cross-presentation and connection with the direct presentation pathway remain incompletely understood (12–17). For instance, hen egg-white lysozyme (HEL), targeted to early endosomes, is completely processed and loaded onto MHC-I in the endosomal compartments, whereas HEL targeted to late endosomes is not cross-presented (15). Soluble OVA escapes into the cytosol for proteasomal processing before cross-presentation, whereas HEL is processed and loaded within the endosome (15). The inhibition of cathepsin activities, especially cathepsin S, altered epitope cross-presentation, showing the importance of cathepsins in this pathway (13, 15, 16). Cathepsin activities are controlled through pH regulation. Upon phagocytosis, phagosomes fuse with early and late endosomes and with lysosomes, thus acquiring progressively the acidification machinery (mainly the V-ATPase complex) and lysosomal proteases (18). Acidification results in the activation of cathepsins with optimal proteolytic activities between pH 4 and 6 (18, 19). The activation and recruitment of NADPH oxidase 2 (NOX2) at the phagosomal membrane generates reactive oxygen species (ROS), increasing pH and limiting cathepsin activities and Ag degradation in DCs (20, 21).

We showed previously that HIV PIs alter the hydrolytic activities of nonaspartyl proteases, such as proteasomes and aminopeptidases (8). However, the impact of HIV PIs on cathepsins, cysteine and aspartyl proteases involved in cross-presentation, has never been assessed. The aim of this study was to investigate the effect of five HIV PIs (RTV, SQV, NFV, the ubiquitously used lopinavir (LPV)/RTV (Kaletra), and the most recently prescribed darunavir [DRV]) on cathepsin activities in primary APCs and CD4 T cells, the main targets of HIV infection and subsets involved in clearing pathogens or priming immune responses. Our results showed that RTV reduced cathepsin activities, whereas SQV and NFV enhanced them. We identified two complementary mechanisms: a direct effect of PIs on cathepsins and an indirect effect on the regulatory pathway of lysosomal pH through reduction of kinase activities, NOX2 activation, phagolysosomal ROS production, and pH, leading to enhanced cathepsin activities. Altered cathepsin activities modified Ag-degradation patterns and epitope production in a sequence- and cell type–dependent manner and were linked to variable cross-presentation of HIV epitopes, altered CTL recognition, and partial modifications of the MHC self-peptidome.

PBMCs were isolated from buffy coats collected from anonymous blood donors approved by the Partners Human Research Committee under protocol 2005P001218 (Boston, MA). PBMCs from HLA-typed blood donors were obtained after written informed consent and approval under protocol 2010P002121 by the Partners Human Research Committee.

HIV PIs were obtained from Selleckchem (RTV, LPV, and DRV), Sigma (SQV), and Santa Cruz Biotechnology (NFV). All drugs were dissolved in 100% DMSO. Kaletra is a combination of LPV and RTV at a 5:1 ratio. Kaletra concentrations mentioned in this article correspond to LPV amount. Aliquots of stock solutions of 10 mM were kept at −20°C.

Monocytes and CD4 T cells were isolated from PBMCs by immunomagnetic enrichment (STEMCELL Technologies). Monocytes were differentiated into DCs and MΦs, as described (22). Phagosomal pH and ROS production were measured, as described (23). Briefly, beads were covalently coupled with FITC and FluoProbes 647 Dihydrorhodamine 123 (DHR; Life Technologies). Cells treated for 30 min with different HIV PIs (5 μM) or inhibitors (10 μM diphenylene iodonium [DPI] or 1 μM bafilomycin) were pulsed with beads for 20 min and washed extensively in cold PBS. Then cells were incubated at 37°C and analyzed by flow cytometry at different times.

A total of 2 nmol pure peptides was digested with 15 μg whole-cell extracts at 37°C in pH 4 degradation buffer (24). The degradation was stopped with 2.5 μl 100% formic acid, and peptide fragments were purified by 5% trichloroacetic acid precipitation. Degradation peptides were identified by in-house mass spectrometry (MS), as previously described (8, 19).

Proteolytic activity measurement was performed as we described previously (8, 19, 25). Briefly, 5 × 10⁴ DCs, MΦs, or CD4 T cells were pretreated with the indicated PIs, and omnicathepsin, cathepsin S, cathepsin D, or cathepsin E activities were measured using Z-FR-AMC–specific, Z-VVR-AMC–specific, Ac-RGFFP-AMC–specific, or MOCAc-GSPAFLAK-Dnp-R–specific fluorogenic substrates, respectively. Specific inhibitors were used to confirm the specificity of the reactions: E64 for omnicathepsin, Z-FL-COCHO for cathepsin S, and pepstatin A for cathepsin D and E. The rate of fluorescence emission, which is proportional to the proteolytic activity, was measured every 5 min at 37°C in a VICTOR3 Plate Reader (Perkin Elmer).

Protein extracts (30 μg/lane) from treated (5 μM PIs and/or 0.4 μg/ml PMA) or untreated DCs were subjected to SDS-PAGE on 4–12% gradient gel. Protein bands were visualized using a dual infrared fluorophore Odyssey Infrared Imaging System (LI-COR Biosciences), and densitometric quantification was performed as described (22). The Kinase Selectivity Profiling System: AGC-1 was used in addition to the Kinase-Glo Luminescent kit (both from Promega) to measure kinase activities.

Immature DCs were exposed to rHIV-1 p24 protein (Abcam) for 1 h at 37°C. DCs were washed thoroughly and cultured for 4 h before adding the epitope-specific CTL clones at a 4:1 E:T ratio in 96-well plates. DCs pulsed with the indicated peptide concentrations were used as controls for Ag presentation and CTL specificity. Target cell lysis was measured with a fluorescent-based assay developed in the laboratory (26).

MHC-bound peptides (including MHC-I and MHC class II [MHC-II]) from control and drug-treated live PBMCs were isolated directly from live cells and identified by high-resolution tandem MS (MS/MS) (M. Rucevic, G. Kourjian, and S. Le Gall, unpublished observations). In brief, 5 × 10⁷ cells were exposed briefly to mild acid treatment (10% acetic acid, 5 min), and the peptide pool obtained was immediately subjected to ultrafiltration (3-kDa cut-off) for MHC peptides enrichment. The pools of eluted peptides were subjected to liquid chromatography (LC)-electrospray ionization–MS/MS analysis. They were loaded onto a ChromXP-C18 5-μm cHiPLC capillary column (75 μm × 15 cm; Eksigent) interfaced with an LTQ Orbitrap XL Mass Spectrometer (Thermo Fisher) and resolved with a gradient consisting of aqueous mobile phase A (0.1% formic acid in water) and organic mobile phase B (0.1% formic acid in 100% acetonitrile). The 10 most abundant multiply charged ions were selected for high-resolution MS/MS sequencing. Data were analyzed with Proteome Discoverer 1.4 software (Thermo Fisher) and searched against the Swiss-Prot human database using MASCOT search engine. To ensure high accuracy of identified sequences, the mass tolerance on precursor was set to 5 ppm and 0.02 Da for fragment ions, with all assignments made at 1% false-discovery rate. The candidate MHC peptides identified were selected based on their presence in triplicate samples. All identified MHC peptides were blasted against the nonredundant Swiss-Prot database restricted to human entries to identify their corresponding source proteins.

Empirical data were analyzed using GraphPad Prism version 6. We compared two variables with t tests. All p values are two-sided, and p values < 0.05 were considered significant. Multiple comparisons were done using one-way ANOVA and the Dunnett posttest.

We aimed to assess the effect of five HIV PIs (RTV, SQV, NFV, Kaletra, and DRV) on omnicathepsin activities (which measure the combined activities of cathepsin S, L, and B) and cathepsin S, D, and E, which are expressed in APCs important for Ag processing in the phagosomes (8, 13, 19, 21, 27, 28). The PI concentrations used in this study correspond to plasma levels of ART-treated persons and are not toxic for primary cells (8, 29). The cleavage of a peptidase-specific fluorogenic peptide substrate was measured over time. The specificity of substrate cleavage was checked by preincubation of cells with cognate inhibitors of omnicathepsin (E64) or cathepsin S (Z-FL-COCHO) (1A). The hydrolysis kinetics was calculated as the maximum slope of fluorescence emission after background subtraction (Fig. 1A). 100% represents the maximum slope of fluorescence emission by cells preincubated with DMSO (maximum slope of 67 for control DMSO, 430 for SQV, 36 for RTV, 5 for E64) (Fig. 1B). The effects of each HIV PI on omnicathepsin and cathepsin S activities were assessed in freshly isolated PBMCs, CD4 T cells, monocyte-derived DCs, and MΦs from at least six HIV negative donors (Fig. 1C–F). Baseline cathepsin activities were lower in CD4 T cells and PBMCs than in DCs and highest in MΦs. In PBMCs, RTV reduced both omnicathepsin and cathepsin S activities by 3.3-fold. In contrast, SQV and NFV increased omnicathepsin activity by >3-fold, but only SQV increased cathepsin S activity. DRV did not change cathepsin activities (Fig. 1C). Similar alterations were seen in CD4 T cells, DCs, and MΦs, albeit with different fold changes (Fig. 1D–F).

We tested whether HIV PIs can directly affect cathepsins. The pH of PBMC extracts was reduced to 4–5.5 to specifically activate cathepsins (24). Similar PI-induced alterations in cathepsin activities were observed in PBMC extracts, albeit with a lower fold change (G. Kourjian and S. Le Gall, unpublished observations). The effect of PIs was also tested on recombinant cathepsin S and D. RTV, NFV, and Kaletra reduced cathepsin S activity, whereas SQV increased it (Fig. 1G). Cathepsin D activity was reduced by RTV and increased by SQV and NFV (Fig. 1H). Cathepsin D and E activities measured in live DCs showed similar alterations upon PI treatment as did purified enzymes (data not shown). We then tested whether HIV PIs affect the maturation of inactive procathepsins into cathepsins. Activation is triggered by a decrease in pH in endolysosomes, which induces the proteolytic removal of a prodomain blocking the catalytic site (30). Procathepsin K was treated with 5 μM of different PIs at different pHs before measuring cathepsin K activity (Fig. 1I). At pH 7, only SQV and NFV activated procathepsin K, resulting in a 7.5-fold increase in activity. At pH 5.5, SQV and NFV activated procathepsin K by 1.8-fold. Thus, HIV PIs have a direct effect on cathepsin maturation and activities. They may bind directly to the catalytic site of cathepsins and inhibit the activity or interact with noncatalytic effector sites, allosterically modulate the conformation of the enzyme (31), and decrease or increase hydrolytic activities. However, differences in fold changes between live cells and recombinant enzymes suggest that other mechanisms might alter cathepsin activities.

The activity of endolysosomal cathepsins is controlled, in part, through pH regulation (18, 19). We assessed whether HIV PI-induced alterations in cathepsin activities are due to changes in phagolysosomal pH in DCs, MΦs, and CD4 T cells. We used a flow cytometry–based measurement of phagosomal pH using latex beads coated with pH-sensitive (FITC) and pH-insensitive (FluoProbe 647) fluorescent dyes (23). The fluorescence intensity was quantified by flow cytometry at different times. The pH-insensitive dye allowed gating on cells that phagocytosed only one bead, whereas the fluorescence intensity of the pH-sensitive dye reflected the phagosomal pH (Fig. 2A). HIV PI treatment did not affect bead uptake (G. Kourjian and S. Le Gall, unpublished observations). Dextran-pHrodo was used to measure endosomal pH in CD4 T cells (Fig. 2A). Standard curves were obtained by permeabilizing and incubating cells in predetermined pH media. FITC intensity decreased and pHrodo intensity increased upon acidification (Fig. 2B). The pH values in phagosomes were determined by comparing the mean fluorescence intensity (MFI) in the cell population to the pH standard curves. As shown previously (20), the pH in DC phagosomes was stable at 6.5 during the 90-min monitoring (Fig. 2C, left panel). In contrast, the phagosomal pH in MΦs decreased more quickly and reached pH 5 (Fig. 2C, middle panel). Bafilomycin, a specific inhibitor of vacuolar H⁺ ATPase, increased the pH by 1.3 pH points in all cell types tested. SQV and NFV, but not RTV and Kaletra, reduced DC phagosomal pH by ∼1 pH point (Fig. 2C, 2D, left panel). HIV PIs did not significantly affect the already lower phagosomal pH in MΦs (Fig. 2C, 2D, middle panel). CD4 T cell endolysosomal pH was reduced ∼1 pH point in 60 min by SQV and NFV treatment. RTV and Kaletra did not change the pH significantly (Fig. 2C, 2D, right panel). Thus SQV and NFV, but not RTV and Kaletra, reduced phagosomal pH, which may contribute to the PI-induced increase in cathepsin activities in live cells.

The phagosomal pH is regulated, in part, by the consumption of protons through the generation of ROS by NOX2 (20, 32). To confirm this finding in our experimental model, we incubated DCs with DPI, a specific inhibitor of flavin-containing enzymes, such as NOX2. DPI reduced phagosomal pH by ∼1 pH point (Fig. 3A). DPI treatment increased omnicathepsin and cathepsin S activities in DCs by >2-fold, confirming that phagosomal ROS affect cathepsin activities (Fig. 3B). We hypothesized that HIV PIs might modulate ROS generation through NOX2, subsequently affecting phagosomal pH and cathepsin activities (Fig. 3C). Latex beads were coated with DHR, a dye that only emits fluorescence under oxidative conditions (23), and an oxidation-insensitive (FluoProbe 647) fluorescent dye was added to MΦs or DCs. At different times after phagocytosis, we measured ROS generation by quantifying the fluorescent intensity of DHR on cells that phagocytosed only one bead (Fig. 3D). As expected, DC phagosomes produced ≥2-fold more ROS than did MΦs, and ROS generation was strongly inhibited by DPI treatment in both cell types (20) (Fig. 3E, 3F). RTV and Kaletra did not significantly alter ROS production in DCs and MΦs. In contrast, SQV and NFV significantly reduced ROS generation in DCs (−63 and −72 RFU, respectively) and, to a lesser extent, in MΦs (−33 and −42 RFU, respectively) (Fig. 3E–H), showing that SQV and NFV inhibition of ROS generation in DCs and MΦs led to phagosomal acidification and increased cathepsin activities.

The activation of NOX2 is triggered by the phosphorylation of one of its subunits, phox-p47, at multiple locations and is regulated by different kinases, including Akt and PKC (33). To examine whether an SQV- or NFV-induced reduction in NOX2 activity is linked to changes in phox-p47, we used Western blot to analyze the phosphorylation of phox-p47 at serine 345 and 359 in DCs after PI treatment (Fig. 4A). A reduction in the phosphorylation through a point mutation at either serine reduced NOX2 activity by at least half (34). SQV and NFV reduced p47 Ser³⁵⁹ phosphorylation by 1.7-fold but did not affect p47 Ser³⁴⁵ phosphorylation (Fig. 4A, 4B). To test the significance of this inhibition, we assessed whether SQV and NFV could counteract the strong stimulatory effect of PMA on p47 phosphorylation (Fig. 4C). As previously reported (34, 35), PMA increased p47 Ser³⁵⁹ phosphorylation by 1.5-fold. SQV and NFV pretreatment blocked the stimulatory effect of PMA, showing that PMA and PIs exert opposite effects on pathways leading to p47 phosphorylation (Fig. 4D). To assess whether this reduction in phosphorylation was caused by direct inhibition of kinases of the PI3K/Akt kinase pathway involved in the upstream regulation of phox-p47, we measured the activities of Akt, PKA, PKC, and phosphoinositide-dependent kinase-1 (PDK1) in the presence or absence of different PIs. SQV and NFV, but not RTV and Kaletra, moderately reduced all four kinase activities by 1.5-fold (Fig. 4E), suggesting that SQV- and NFV-induced inhibition of PDK1/Akt reduced phox-p47 phosphorylation, leading to lower NOX2 activity and reduced ROS production in phagosomes. This mechanism is complementary to the direct enhancement of cathepsin activities induced by these two PIs.

We used our in vitro Ag-degradation assay to analyze how PI-induced changes in cathepsin activities alter Ag processing in compartments involved in cross-presentation (19, 24). We showed previously that the reduction in cellular extracts’ pH to 4 activates cathepsins, inhibits proteasome and aminopeptidase activities, and mimics degradation conditions in purified endolysosomes (19, 24). DC, MΦ, and CD4 T cell whole-cell extracts placed at acidic pH were used to degrade a synthetic 35-mer peptide in HIV-1 Gag p24 (MVHQAISPRTLNAWVKVVEEKAFSPEVIPMFAALS, aa 10–44 in Gag p24) containing the HLA-B57–restricted epitopes ISW9 and KF11 (36). Supplemental Fig. 1 shows all degradation peptides from the p24 35-mer identified by LC-MS/MS after a 60-min degradation in DC, MΦ, or CD4 T cell extracts in the absence of PIs; each bar corresponds to a peptide with a defined peak intensity. We quantified the relative amount of each fragment by measuring its contribution to the total intensity of all degradation fragments, as done previously (19). We displayed the relative amount of peptides starting at any N-terminal residue or ending at any C-terminal residue (top and bottom bars, respectively), thus showing the relative frequency of cleavage sites in the fragment (Fig. 5A). For instance, a frequent N-terminal cleavage site is observed at a valine in between boxed epitopes ISW9 and KF11. We observed differences in the frequency of cutting sites among DC, MΦ, and CD4 T cell extracts, confirming the differential Ag-processing activities by different cell types previously reported by our group (19, 22, 25) (Fig. 5A, left panel). A similar analysis was performed for degradations done in the presence of different HIV PIs. In DCs, SQV introduced new cutting sites within the ISW9 epitope (e.g., N-terminal of proline position 8 and C-terminal of arginine position 9). RTV and SQV also changed the frequency of certain cutting sites (e.g., N-terminal of valine position 17 and C-terminal of proline position 29) (Fig. 5A, right panel). We generated heat maps showing the percentage changes in the frequency of cleavages at each amino acid induced by each PI over DMSO. Fig. 5B shows that HIV PIs increased cleavages at some locations and decreased them at others. For instance, SQV increased the cleavage within the ISW9 epitope in DCs (Fig. 5B, top panel), potentially leading to the destruction of ISW9. Similar changes were observed when the full HIV p24 protein was used for degradation (Supplemental Fig. 2). To further analyze the effect of PIs on epitope production, we measured the amount of epitope-containing fragments. In DCs, SQV reduced the production of B57-ISW9–containing fragments by 1.7-fold, which was expected because of the increased cleavage within the epitope (Fig. 5B, top panel). RTV reduced the production of B57-KF11–containing fragments by 1.8-fold and B57-ISW9–containing fragments by 6-fold. In MΦs, RTV and SQV reduced the amount of B57-ISW9–containing fragment by 2-fold. The amount of B57-KF11–containing fragments was reduced 2.6-fold by RTV and increased 2-fold by SQV, 2-fold by NFV, and 1.4-fold by Kaletra. In CD4 T cells, the amount of B57-ISW9–containing fragments was reduced 6-fold by SQV. The amount of B57-KF11 containing fragments was increased 2-fold by RTV and 3-fold by SQV, NFV, and Kaletra (Fig. 5C). In addition to the PI-induced changes in the amount of ISW9- and KF11-containing fragments, HIV PI treatments changed the size distribution of the degradation products. The amount of 8–12-aa-long peptides was reduced by all PIs by an average of 2-fold in DCs and MΦs. In CD4 T cells, the amount of 13–18-aa-long peptides was reduced 2.7-fold by RTV, 1.7-fold by SQV, 1.4-fold by NFV, and 3-fold by Kaletra (Fig. 5D). Because cathepsins are involved in MHC-II epitope processing (37), we assessed the effect of PIs on MHC-II epitope production. DC whole-cell extracts placed at acidic pH were used to degrade a synthetic 24-mer peptide in HIV-1 Gag p24 (FRDYVDRFYKTLRAEQASQEVKNW, aa 161–185 in Gag p24) rich in MHC-II epitopes (36), in the presence or absence of PIs. Supplemental Fig. 3 shows that SQV and NFV increased cleavages at some locations and decreased them at others. These changes in the cutting sites patterns altered the amount of MHC-II epitope production. For instance, the amount of MHC-II DQA1*01/DQB1*05-restricted FT11 (FRDYVDRFYKT) was increased upon SQV treatment, and the amount of MHC-II DRB1*01/DRB1*04-restricted FE13 (FYKTLRAEQASQE) was decreased by NFV (Supplemental Fig. 3). Together, these results suggest that, by altering cathepsin activities, PIs changed the patterns and frequency of cutting sites, affecting the amount of MHC-I and MHC-II epitopes produced.

One third of HIV-infected individuals on ART are coinfected with other pathogens, such as Mycobacterium tuberculosis or hepatitis C virus (HCV) (38, 39). We hypothesized that, by modifying cathepsin activities, PIs might alter the processing of epitopes from coinfecting pathogens in APCs. We assessed the effect of the PIs on the processing of two Ags: M. tuberculosis Ag85 and HCV NS3, which both contain multiple MHC-I and MHC-II epitopes (40–45). DC whole-cell extracts placed at acidic pH were used to degrade M. tuberculosis Ag85 and HCV NS3 in the presence of different PIs. The frequency of cleavage sites (Fig. 6A) and the percentage changes in the frequency of each cutting site upon PI treatment (Fig. 6B) were analyzed as previously described. HIV PIs increased the cleavage at some locations, decreased it at others, and created new cleavages sites. These alterations in the degradation patterns changed the amount of epitopes produced during M. tuberculosis Ag85 and HCV NS3 degradation in the presence of different PIs. Of the seven epitopes exactly generated during M. tuberculosis Ag85 degradation, PIs changed the amounts produced of each one. For instance, all of the PIs with the exception of NFV reduced the cleavage within the RA15 (RAQDDFSGWDINTPA 47–61 aa) epitope, which, as expected, led to a 2–5-fold increase in RA15 epitope production (Fig. 6B). In contrast, all of the PIs tested increased the cleavage within the LM19 (LQANRHVKPTGSAVVGLSM 113–131 aa) epitope, leading to a 2–4-fold decrease in epitope production. In addition to alterations in epitope production, by introducing a new cleavage site within the AT15 (AMSGLLDPSQAMGPT 152–166 aa) epitope, RTV and SQV completely abolished its production (Fig. 6B). PI treatment also altered the degradation of HCV NS3 protein, causing increased or decreased cutting frequencies at various locations (data not shown). Together, these results suggest that, by altering cathepsin activities, PIs changed the patterns and frequency of cutting sites in M. tuberculosis and HCV Ags, leading to changes in epitope production. How these PI-induced variations in epitope production will modify the hierarchy or specificity immune responses against coinfecting pathogens in HIV-positive individuals remains to be determined.

To understand whether these PI-induced alterations in epitope production measured in vitro also affect endogenous processing and presentation by DCs, we analyzed the cross-presentation of the three optimally defined HLA-B57–restricted HIV epitopes originating from HIV-1 p24 protein: HLA-B57–restricted ISW9 (ISPRTLNAW, aa 15–23 in Gag p24), HLA-B57–restricted KF11 (KAFSPEVIPMF, aa 30–40 in Gag p24), and HLA-B57–restricted TW10 (TSTLQEQIGW, aa 108–117 in Gag p24) (36). Monocyte-derived DCs from HLA-B57⁺ patients loaded with HIV-1 p24 protein in the presence of different HIV PIs were challenged with epitope-specific CTLs (19). Target cell killing was monitored using a fluorescent-based assay developed in the laboratory (26). As previously seen in B cell lines (8), HIV PIs had no effect on the level of MHC-I surface expression in DCs (Fig. 7A). The three epitopes tested were restricted by HLA-B57 and shown to have similar affinities (36, 46). DCs pulsed with increasing amount of peptides were increasingly lysed after the addition of cognate CTLs, showing the similar sensitivity of these CTL clones for their cognate peptides (Fig. 7B). RTV reduced B57-KF11 cross-presentation by DCs by 1.4-fold, whereas the other PIs had no effect (Fig. 7C). B57-ISW9 cross-presentation was reduced by RTV and SQV (3.2- and 1.6-fold, respectively). In contrast, NFV increased killing by 1.35-fold (Fig. 7D). RTV reduced B57-TW10 cross-presentation by 2.4-fold, whereas SQV increased it by 1.4-fold (Fig. 7E). During endogenous processing and presentation of p24, the effect of PIs, the asynchronous timing of epitope production (19, 47), the relative amount of each epitope, and the number of B57 molecules available for loading also define the relative amount of the three epitopes presented by HLA-B57. Although we cannot quantify the relative amount of each of the three B57 epitopes endogenously processed and presented by infected cells, the changes in CTL recognition induced by HIV PIs are in accordance with changes in epitope production measured with our degradation assays. For instance, SQV enhanced cleavages within the B57-ISW9 epitope (Fig. 5B), leading to reduced production of B57-ISW9–containing fragments (Fig. 5C), which resulted in lower endogenous cross-presentation of B57-ISW9 by DCs to CTLs (Fig. 7D). These results demonstrate the link between drug-induced alteration of epitope production and subsequent changes in epitope-specific CTL responses to cross-presented epitopes.

This and our previous study (8) showed that HIV PIs affect the direct endogenous Ag-processing and cross-presentation pathways. In addition, these alterations were not limited to HIV proteins; they also affected the processing of M. tuberculosis and HCV Ags. Therefore, we wondered whether HIV PIs also affect self-derived MHC-bound peptidome (MHC-I and MHC-II) of uninfected cells. We analyzed the self-derived total MHC-bound peptidome (MHC-I and MHC-II) of healthy PBMCs that were pretreated with DMSO or SQV for 2 d. MHC-bound peptides were eluted by acid from live PBMCs, and 686 and 680 human peptides were identified in triplicate runs with high mass accuracy by in-house LC-MS/MS on Ctrl PBMCs and SQV-treated PBMCs, respectively (M. Rucevic, G. Kourjian, and S. Le Gall, unpublished observations). We performed an in-depth comparative analysis of the peptides commonly identified with the highest confidence scores among repeated MS runs (148 Ctrl PBMCs and 155 SQV-treated PBMC peptides listed in Supplemental Table I). The comparative mapping revealed that 68% of peptides were commonly presented by Ctrl- and SQV-treated PBMCs. Thirty-two percent of peptides were uniquely presented by Ctrl- or SQV-treated cells (Fig. 8A). As expected, the majority of common MHC-bound peptides were 8–12 aa long (compatible with MHC-I loading), and 40% were 13–18 aa long (compatible with MHC-II loading) (Fig. 8B). Intriguingly, >80% of peptides uniquely presented on SQV-treated PBMCs were longer than 13 aa compared with only 60% of peptides presented uniquely on Ctrl PBMCs (Fig. 8B). Because SQV treatment did not change the surface expression of MHC-I or MHC-II (G. Kourjian and S. Le Gall, unpublished observations), these results suggest that SQV may increase the diversity of MHC-II peptidome and/or lead to the presentation of longer peptides by MHC-I. The total peptidome of Ctrl- and SQV-treated PBMCs was sampled from 56 host proteins. Many of the commonly identified MHC peptides were derived from the same proteins and corresponded to nested peptides, with an average of three peptides/protein (Fig. 8C, 8D). However, peptides uniquely identified from SQV- or Ctrl-treated PBMCs were sampled from diverse proteins, with one peptide/protein (Fig. 8C).

Sixty-four percent of all proteins identified were represented in untreated and SQV-treated peptidome, whereas 18% were uniquely represented on Ctrl PBMCs, and another 18% were represented only on SQV-treated PBMCs; this suggested that SQV partially changed the origin of proteins represented on the surface (Fig. 8D). This might be explained, in part, by an SQV-induced alteration of transcription and expression. However, we also identified changes in the number and location of peptides originating from a common protein, in line with PI-induced changes in protein-degradation patterns (Fig. 8D). The group of proteins represented by treated and untreated PBMCs was divided into subgroups according to the diversity of peptide location. Peptides presented commonly or uniquely were derived from the same location in 64% of proteins represented by treated and untreated PBMCs (Fig. 8D, Similar). In 22% of the proteins from this group, SQV treatment narrowed the locations from which the peptides were derived (Fig. 8D, Narrowing), whereas in the remaining 14% of the proteins, SQV broadened the locations of the peptides (Fig. 8D, Broadening). We concluded that SQV-induced changes in peptidase activities and degradation patterns contributed, in part, to change the MHC self-peptidome.

Any perturbation of protease activities could modify protein-degradation patterns and, consequently, epitope presentation and immune recognition. In this study, we showed that several HIV PIs modified cathepsin activities through two complementary mechanisms and altered the MHC-peptidome and recognition of infected cells by epitope-specific CTLs.

Four of the five PIs tested variably affected cathepsin activities. RTV mostly reduced all cathepsin activities tested, whereas NFV reduced cathepsin S activity and enhanced cathepsin D activity, suggesting different interactions between each drug and cathepsins. It is intriguing that HIV PIs designed to inhibit the HIV aspartyl protease alter the activities of aspartyl proteases like cathepsin D and E, as well as cysteine proteases, such as cathepsin S. HIV PIs might bind directly to the catalytic site of cathepsins, interact with noncatalytic effector or allosteric sites that are known to regulate the hydrolytic activities (31), or induce the maturation of procathepsin into cathepsin. These off-target effects of HIV PIs might explain the inhibition or enhancement of cathepsin activities, although molecular modeling and structural studies are required to test this hypothesis.

An increase in cathepsin activity could also be triggered by a drug-induced maturation of inactive procathepsins into cathepsins. At neutral pH, glycoaminoglycans can loosen the binding of the propeptide on the catalytic site and accelerate its autocatalytic removal, subsequently activating cathepsins (48, 49). We discovered that SQV and NFV, but not RTV, facilitated the maturation of procathepsin K, even at neutral pH. SQV and NFV might be using the same mechanism as glycoaminoglycans to trigger cathepsin maturation at neutral pH. Cathepsin deregulation and the extracellular presence of active cathepsins are linked to several diseases, including damage to the mucosal barrier in inflammatory bowel disease (50), osteoporosis and osteopenia, cancer, rheumatoid arthritis, atherogenesis, and muscular dystrophy (30). Additional research is needed to investigate whether PI-induced cathepsin alterations contribute to chronic disease development in patients receiving long-term ART.

First-generation PIs, like RTV, SQV, and NFV, had stronger effects on cathepsin activity and its upstream regulators than did newer PIs like DRV. First-generation PIs also induced more rapid and profound adverse effects on lipid metabolism (51). The half-life of perilipin, a protein involved in regulating adipocyte lipolysis, was reduced by 6-fold in NFV-treated adipocytes, an effect reversed by NH₄Cl but not by proteasome inhibition (52). The NFV-induced increase in lysosomal cathepsin activity identified in this study might provide the missing link between NFV and perilipin lysosomal proteolysis in lipid metabolism disorders.

SQV and NFV have pleiotropic anticancer activities and are being repurposed for cancer treatment (53). The anticancer effects of some PIs observed in vitro were linked to the suppression of the Akt signaling pathway, but the actual molecular targets remain unknown (54, 55). The direct inhibition of multiple members of the protein kinase–like superfamily (PDK1, Akt, and PKA/C) by SQV and NFV lends support to a drug-induced inhibition of cellular processes vital for carcinogenesis and metastasis and provides a molecular basis to explain the broad-spectrum anticancer effect of SQV and NFV (53, 55). Akt and PKC are also involved in regulating activation of the NOX2 complex (33). NOX2 produces ROS in the phagosomes of neutrophils, eosinophils, monocytes, and MΦs. It contributes to the destruction of engulfed pathogens (32), as well as to the regulation of phagosomal pH in DCs (20). Whether PI-induced NOX2 inhibition affects the capacity of MΦs to kill phagocytosed pathogens requires further investigation.

SQV and NFV have the unique ability to activate cathepsin activities in APCs through two independent mechanisms: inhibiting NOX2 activity and increasing phagosome acidification and directly enhancing cathepsin activities. The mechanism underlying PI-induced enhanced cathepsin activity is less clear in CD4 T cells, but it involves direct activation of cathepsins and pH acidification. Enhanced cathepsin activity modified the degradation patterns of HIV, HCV, and M. tuberculosis proteins, altering the frequency of cleavage sites, size distribution of the fragments, and epitope production. It altered MHC-I and MHC-II epitope production and the self-derived MHC peptidome. The increase in 13–18-aa-long peptides presented by PBMCs upon SQV treatment suggests that SQV might broaden the MHC-II surface peptidome. It is unknown whether this HIV PI-induced modulation of the self-peptidome is sufficient to induce autoimmune symptoms. No increase in autoimmune disease in HIV patients receiving ART, including Kaletra, has been reported in the literature. Thymic selection of T cells involves negative and positive selection in medullary and cortical thymic epithelial cells. The unique expression of thymoproteasome in cortical thymic epithelial cells, but not in medullary thymic epithelial cells (56), contributes to shaping the immunocompetent repertoire of CD8 T cells (57) and determines the Ag responsiveness of CD8 T cells (58). One can speculate that the variable and broad peptide displayed in the thymus during selection (59) might cover some variations in the self-peptidome. A direct comparison of the thymus peptidome and PI-treated or untreated PBMCs is needed to address this question.

The alterations in degradation patterns were drug and sequence dependent and varied according to cell subsets, in line with our previous findings on cell subset–specific variations in peptidase activities (22, 25). Given the sequence specificity of cathepsins (30), each PI may variably increase the cleavage of certain motifs and decrease others. Interestingly, HIV PIs altered the degradation of HIV, HCV, and M. tuberculosis proteins in accordance with sequence-specific, rather than virus-specific, modifications in degradation patterns. A computational analysis of the degradation patterns, as done previously (8, 60), may allow us to predict the effect of each PI on epitope production.

PI-induced changes in cathepsin, proteasome, and aminopeptidase activities (this study) (8) modified protein degradation and the MHC-peptidome and altered CTL recognition of HIV cross-presenting or infected cells. In ART-treated persons with ongoing replication of drug-resistant viruses or coinfected with M. tuberculosis or HCV, PIs might alter the presentation of pathogen-derived epitopes by APCs or target cells. A drug-induced reduction or impairment of epitope presentation could render some pathogen-specific T cell responses useless or less efficient at recognizing target cells, or it might affect the T cell memory compartment if an epitope is never presented. Conversely, increased presentation of other peptides or presentation of new peptides may broaden the T cell compartment. The effects would be determined, in part, by the amount of MHC peptides displayed and the binding affinity of peptides for their MHC and corresponding TCRs. Although the lack of appropriate clinical samples precludes us from testing this hypothesis, the PI-induced modifications in M. tuberculosis and HCV Ag-degradation patterns and epitope production suggest that HIV PIs may modify the spectrum of the pathogen-specific T cell compartment. Mouse models of viral or bacterial infections will allow us to test this hypothesis. Shock-and-kill approaches to clear HIV reservoirs propose to pharmacologically reactivate provirus in the presence of ART to induce immune clearance (61). If ART includes PIs, such as Kaletra, it will be important to assess the impact of HIV PIs on epitope presentation by CD4 T cells and DCs when designing therapies to eliminate reactivated HIV-infected cells (62).

By acting directly on cathepsins and altering the physiology of phagocytic compartments, HIV PIs modify Ag processing, the MHC peptidome, and immune recognition. If HIV PIs allow the diversification of epitope presentation, a temporary PI treatment that would not induce the adverse effects observed in long-term highly active ART may provide a complementary approach to modify or improve immune recognition in the context of various immune diseases.

We thank Drs. F. Pereyra, D. Kavanagh, B. Walker, and S. Pillai for stimulating discussions and input on the manuscript.

The authors have no financial conflicts of interest.