Prolonged antigen survival and cytosolic export in cross-presenting human γδ T cells

Delayed Endosomal Acidification and Proteolysis in γδ T-APCs.

We reported previously that γδ T-APCs outperform human moDCs in the cross-presentation of soluble protein (10). The reason for this difference is unknown but might involve differences in antigen uptake and/or processing. We first analyzed the endocytic activity of γδ T-APCs using different soluble substrates (BSA, dextran, and Lucifer yellow). Whereas immature DCs are known for their high level of endocytosis, γδ T-APCs showed diminished endocytic activity that did not differ substantially from that of human αβ T cells, B cells, and monocytes (Fig. S1A). Thus, the efficiency of cross-presentation did not correlate with antigen uptake, which agrees with recent findings in DCs (9, 19). We further studied the mechanism of endocytosis using dimethyl amiloride (DMA), a Na⁺/H⁺ channel inhibitor that selectively blocks macropinocytosis, and cytochalasin D, which inhibits phagocytosis by preventing actin polymerization. Although both inhibitors blocked endocytosis in moDCs, only DMA inhibited the uptake of BSA in γδ T-APCs, indicating that cytoskeleton rearrangement was not involved in the uptake of soluble antigen by γδ T-APCs (Fig. S1B). Intracellular processing of endocytosed antigen is a critical next step in cross-presentation (20). Whereas fast degradation was shown to prevent cross-presentation, a delay in lysosomal proteolysis increased its efficiency (20, 21). In support of this finding, initial endosomal alkalinization followed by slow acidification was reported to promote cross-presentation (22). Collectively, diminished lysosomal activity is thought to allow antigen “escape” into the cytosol for processing by the cross-presentation machinery and induction of CTL responses.

To examine antigen degradation, we pulsed γδ T-APCs and moDCs with FITC-BSA, followed by a 0.5 to 6 h chase and analysis by flow cytometry (FACS). Surprisingly, after a 6 h chase, γδ T-APCs still emitted the same level of fluorescence as after the pulse (0 h chase) (Fig. 1A). In contrast, moDCs lost >75% of their fluorescence within 3 h and >95% within 6 h. This phenomenon was observed in fresh and activated γδ T cells, whereas fluorescent signals were decreased by ≥50% in αβ T cells, B cells, and monocytes after a 6 h chase (Fig. S2A). Rapid loss of the FITC signal in moDCs could have been due to degradation/excretion and/or antigen sequestration in acidic compartments. When we pulsed moDCs with BSA coupled to the pH-insensitive fluorochrome Alexa Fluor 488, we observed full maintenance of the fluorescence over a 24 h chase period (Fig. S2B), demonstrating that rapid fluorescence quenching was the cause of the observed loss of the FITC signal. To corroborate this conclusion, we measured the pH in BSA-containing endosomes. Whereas the pH in moDCs dropped from 7 to 4.5 during the 3 h chase period, γδ T-APCs maintained a neutral pH (Fig. 1C), indicating that most of the antigen in γδ T-APCs did not access acidic compartments (pH < 5.5) within the first 6 h. In DCs, the NADPH oxidase NOX2 has been shown to control endosomal pH (22); however, using a NOX2-specific inhibitor, we found no evidence of NOX2 involvement in pH control in γδ T-APCs (Fig S3). To determine whether the slow endosomal acidification in γδ T-APCs affected lysosomal proteolysis, we assessed BSA degradation by Western blot analysis. Whereas moDCs degraded most of the BSA within 3 h, γδ T-APCs exhibited a significant delay in proteolysis (Fig. 1C). We conclude that γδ T-APCs and moDCs differ fundamentally in terms of the kinetics of endosomal acidification and protein degradation.

Traffic of Internalized Antigen in γδ T-APCs and moDCs.

The cellular compartments involved in cross-presentation in γδ T-APCs are unknown. While the importance of the cytosolic pathway is widely accepted, the presence of vesicular compartments mediating cross-presentation (e.g. ER–phagosome fusion compartment or recycling endosomes) remains controversial (2, 5, 23). We examined the fate of endocytosed protein by confocal microscopy in γδ T-APCs and moDCs after a 1 h pulse (Fig. 2) and after a 1.5 h chase (Fig. S4). As expected, in the majority of cells (60% of γδ T-APCs and 100% of moDCs), during the pulse, BSA initially colocalized with the early endosomal marker EEA-1 (Fig. S5). Within individual positive cells (>0% colocalization in Fig. S5B), only 30% of endocytosed BSA colocalized with EEA-1 in both APCs, possibly reflecting the dynamic process of endocytosis. After a subsequent chase period, this colocalization was lost in all cells analyzed, indicating that the internalized protein had left early endosomes. In >60% of both APCs, BSA was found to partially colocalize (21%–27%) with the late endosomal/lysosomal marker Lamp-1. Surprisingly, in all cells analyzed, a major fraction of BSA-containing vesicles was not associated with Lamp-1, even after the chase period; thus, we investigated whether BSA might have traveled to recycling endosomes, a compartment recently proposed to be involved in cross-presentation (5). No colocalization with rab11⁺ recycling endosomes was identified in γδ T-APCs or moDCs, however. Similarly, preliminary experiments in γδ T-APCs revealed no BSA colocalization with the ER marker calnexin. Although these experiments cannot prove that antigen does not traffic through these compartments in γδ T-APCs at any given time point, they certainly cannot account for the antigen-containing Lamp-1–negative vesicles.

Of interest, early after endocytosis we detected substantial colocalization of BSA with the insulin-regulated aminopeptidase (IRAP) in the majority (>80%) of γδ T-APCs and moDCs (colocalization of 22% and 44%, respectively). IRAP is a peptide-trimming peptidase suggested to be involved in proteasome-dependent cross-presentation and found in early endosomes (24). After the 1.5 h chase, this colocalization was lost in 40% of γδ T-APCs and in all moDCs. In conclusion, our data indicate that in both APCs, internalized soluble protein travels from early to late endosomal or lysosomal compartments, with no substantial contribution of recycling endosomes.

Cytosolic Export and Processing.

Because we found no distinct vesicular compartments in support of the endosomal/vacuolar pathway in γδ T-APCs, we turned our attention to the cytosolic pathway. The fundamental step in this model is the translocation of internalized antigen from endosomes to the cytosol for degradation by the proteasome (1–4). To directly analyze antigen translocation, we used the recently described model of cytochrome c (cyt c)-induced apoptosis (25). Because the presence of cyt c in the cytosol is highly proapoptotic, only cells capable of exporting antigen into the cytosol will undergo apoptosis on exposure to exogenous cyt c. Horse cyt c interacts with cytosolic Apaf-1, a component of the apoptosome required for caspase 9 activation, whereas yeast cyt c is inactive and was used here as a negative control. Surprisingly, γδ T-APCs incubated with increasing concentrations of horse cyt c induced substantial apoptosis, whereas yeast cyt c did not raise the level of apoptotic cells above the culture medium control (Fig. 3A and Fig. S6A). The addition of the caspase inhibitor Z-VAD-FMK fully inhibited apoptosis, confirming the proapoptotic function of horse cyt c (Fig. 3A and Fig. S6A). In clear contrast, no apoptosis induction was seen in moDCs (Fig. 3B). To extend our investigation on human DCs, we turned to physiologically more relevant DC subsets present in peripheral blood. Neither conventional DCs (lin⁻, HLA-DR⁺, CD1c⁺) nor plasmacytoid DCs (lin⁻, HLA-DR⁺, CD304⁺) demonstrated enhanced apoptosis in response to horse cyt c (Fig. 3B). In addition, no other human cell type that we analyzed (fresh and activated αβ T cells, B cells, monocytes) showed clear evidence of apoptosis induction (Fig. S6B). In control experiments, mouse CD8⁺ DCs, but not CD8⁻ DCs, were sensitive to horse cyt c treatment, in agreement with the findings of Lin et al. (25). Collectively, we report a remarkable efficiency of γδ T-APCs to translocate exogenous soluble antigen from endosomes into the cytosol, which was not seen in human in vitro or blood-derived DCs.

We next analyzed the involvement of the proteasome in antigen degradation. γδ T-APCs and moDCs were pulsed for 1 h with FITC-BSA and chased for 24 h in the presence of the proteasome inhibitor lactacystin before analysis by FACS and Western blot. The cytosolic pH is neutral, allowing the study of cytosolic FITC-BSA without interference from the acidic endosomal/lysosomal compartment, in which the FITC fluorescence is quenched. While increasing concentrations of lactacystin increased the residual fluorescence in γδ T-APCs, no difference was detected in moDCs (Fig. S7A). Correspondingly, protein degradation was significantly delayed by lactacystin in γδ T-APCs, but not in moDCs (Fig. S7B). To more accurately examine the relative contribution of the proteasome versus the lysosome to antigen degradation, we pulsed APCs with the self-quenched form of DQ-BSA, which releases its pH-independent fluorescence only after protein degradation. Confirming the delay in FITC-BSA degradation (Fig. 1C), γδ T-APCs continued to release fluorescent degradation products for up to 3 h after the pulse when DQ-BSA was washed away (Fig. 3C). In contrast, moDCs degraded most of the BSA during the pulse and the first 30 min chase period. Rapid decline of the fluorescent signal in moDCs during the chase period might be due to fluorochrome instability or excretion. While the proteasome inhibitor lactacystin did not affect antigen degradation in moDCs, it inhibited about 25% of the proteolysis in γδ T-APCs (Fig. 3D). In addition, the acidification inhibitor NH₄Cl and the protease inhibitor leupeptin reduced BSA degradation to a similar extend in both APCs (Fig. 3D). We conclude that the main contribution to proteolysis was provided by endosomal/lysosomal compartments in both APCs; however, in clear contrast to the situation in moDCs, a substantial portion of internalized antigen in γδ T-APCs was exported to the cytosol and degraded by the proteasome.

γδ T-APCs Take Up and Cross-Present Virus-Infected Cellular Debris and Virions.

Up to this point, we limited our analysis to soluble antigen. γδ T-APCs were not able to take up particulate antigen, such as bacteria, zymosan, or beads, which clearly distinguishes these APCs from phagocytes, including moDCs. However, our analysis of the uptake of fluorescently labeled cellular debris from necrotic cells found that γδ T-APCs ingested particles, as demonstrated by confocal microscopy (Fig. 4A). We next investigated whether γδ T-APCs were able to cross-present antigen derived from cellular debris. We incubated γδ T-APCs with debris generated from influenza virus–infected cells, and studied cross-presentation of the influenza matrix protein M1-derived immunodominant peptide M1p58-66 using HLA-A2–restricted, M1p58-66–specific CD8⁺ T cells as responder cells (10). The functional readout included IFN-γ production in M1p58-66 tetramer–positive responder cells after coculture with APCs. In contrast to debris from noninfected cells, γδ T-APCs incubated with influenza-infected cellular debris induced strong CD8⁺ T cell responses (Fig. 4B). Responder cells did not become directly activated by debris from influenza-infected cells. Cellular debris was UV irradiated and did not contain live virus, excluding a potential complication by virus infection in our system (see below). We next explored the ability of debris-treated γδ T-APCs to induce proliferation in primary blood CD8⁺ T cells. γδ T-APCs loaded with debris from influenza-infected cells, but not uninfected cells, led to strong proliferation of M1-specific CD8⁺ T cells (Fig. 4C).

To determine whether γδ T-APCs cross-presented free virus particles, we incubated cells with increasing concentrations of influenza virus, washed the cells, and then incubated them with M1p58-66–specific CD8⁺ T cells. Remarkably, γδ T-APCs loaded with small quantities of live virus induced robust IFN-γ responses (Fig. 5A). Similar to debris-loaded γδ T-APCs (Fig. 4), M1-specific primary CD8⁺ T cell responses were observed when γδ T-APCs were incubated with influenza virus alone (Fig. 5B). Of note, M1p58-66–specific T cell responses were inhibited by lactacystin, indicating involvement of the proteasome in the processing of influenza virus, as was seen previously with purified M1 protein (Fig. 5C) (10). In contrast to cells permissive to influenza infection, such as monocytes and moDCs, we did not detect viral protein synthesis (matrix protein M1 and nucleoprotein NP) in γδ T-APCs during incubation with live virus (Fig. S8 A and B). Moreover, UV- and heat-inactivated (56 °C) virus largely retained its ability to induce M1-specific CD8⁺ T cell responses (Fig. 5C). In contrast, the response was abolished if the virus was heat-inactivated at 65 °C, a temperature that caused the denaturation of influenza hemagglutinin, which controls fusion of the viral nucleocapsid with endosomes and subsequent viral delivery into the cytosol. It was reported previously that DCs loaded with inactivated influenza virus elicited CTL responses only if the virus retained its fusogenic activity (26). To mimic physiological conditions, γδ T-APCs were incubated with virus-infected live cells before culture with CD8⁺ responder T cells. Even though HEK293 cells are HLA-A2⁺ and thus potentially able to directly present viral antigen, CD8⁺ T cell responses to influenza-infected HEK293 cells were low. The responses were increased substantially when HLA-A2⁺ γδ T-APCs were added, however (Fig. 5D). In agreement with the requirement for HLA-A2⁺ APCs, only background responses were seen when HLA-A2^- γδ T-APCs were used. These results were confirmed in experiments using the HLA-A2⁻ cell line SK-N-MC as a virus source (Fig. 5D). Collectively, these experiments demonstrate that γδ T-APCs are fully capable of cross-presenting cellular debris and virus particles for induction of strong anti-influenza CTL responses.

Cell Isolation and Culture.

Peripheral blood mononuclear cells (PBMCs) were isolated from healthy donors using Lymphoprep (Axis-Shield), according to the local ethical guidelines on experimentation with human samples. Vγ9Vδ2 T cells (96%–99% pure) were isolated using Vδ2-PE or Vγ9-PE-Cy5 and anti-PE microbeads or, for confocal microscopy, anti-TCR-γδ-biotin. γδ T cell lines (80–90% pure) were generated from PBMCs stimulated with synthetic HMB-PP (kindly provided by H. Jomaa, Giessen, Germany) and cultured for 3 weeks with 20 U/mL of IL-2 (Proleukin; Chiron). γδ T-APCs were generated from purified γδ T cells via stimulation for 18–48 h with 20–50 nM HMB-PP, 20–50 U/mL IL-2, and 70-Gy–irradiated HLA-A2⁻ EBV-transformed B cells at a γδ:feeder cell ratio of 5:1. Occasionally, γδ T cell lines were used with similar results. αβ T cells were purified using the Miltenyi Pan T Cell Isolation Kit and stimulated for 2 days with 1 μg/mL of PHA (Sigma-Aldrich). MoDCs were derived from monocytes purified with CD14 microbeads (Miltenyi) and cultured for 6–7 days with 50 ng/mL of GM-CSF and 10 ng/mL of IL-4 (Peprotech). Blood DCs were enriched from PBMCs by depleting lin⁺ cells (CD3, CD14, CD19, and CD56). Spleen DCs from C57BL/6 mice were purified using CD11c microbeads (Miltenyi).

Endocytosis.

Cells were pulsed for 0.5–1 h with 1 mg/mL of FITC-BSA, Alexa Fluor 488-BSA, or Alexa Fluor 647-BSA or with 0.5 mg/mL of DQ-BSA (all from Invitrogen) and washed extensively before the chase period. In some experiments, cells were preincubated for 30 min with 5 μM cytochalasin D (Sigma-Aldrich), 250 μM DMA (Sigma-Aldrich), 50 mM NH₄Cl, 100 μM leupeptin, 10 μM diphenylene iodonium, and 20 μM Z-VAD-FMK or 1–10 μM lactacystin (Alexis). Inhibitors were kept in the medium throughout the chase period. The percent inhibition was calculated as the reduction in geometric MFI of inhibitor-treated cells versus untreated cells after background fluorescence (4 °C) was subtracted. Cells incubated for 6 h with cyt c (Sigma-Aldrich) were stained with annexin V (BD Biosciences). Debris was generated by repetitive freezing-thawing of virus-infected MDCK cells or the PKH26 (Sigma-Aldrich)-labeled cell line RAW264.7.

Influenza A Virus Infection.

Cells were infected for 2 h with influenza A strain A/PR8/34 (PR8) virus (MOI 0.5–3; kindly provided by A. Gallimore, Cardiff, United Kingdom) in serum-free medium and cultured for 24 h. Replication-deficient virus was generated by UV inactivation for 30 min at a distance of 5 cm or by heating for 30 min. Virus inactivation was confirmed by standard plaque-forming unit assay.

Antigen Presentation Assays.

M1-specific CD8⁺ T cells were generated by stimulating PBMCs with 0.1 nM M1p58-66 peptide (kindly provided by P. Romero, Lausanne, Switzerland). After 3 days, 40 U/mL of IL-2 was added. Cells were restimulated after 10–14 days in the presence of irradiated (40 Gy) PBMCs. Resting cells with 25–45% M1p58-66–specific cells were used for assays, as determined by tetramer staining (kindly provided by A. Sewell, Cardiff, United Kingdom). Recombinant influenza M1 protein (kindly provided by M. Luo, Birmingham, AL), virus, or debris was added to γδ T-APCs for 18 h. The APCs were washed extensively before culture with responder cells at an APC to responder cell ratio of 1:2–3:1. Brefeldin A (10 μg/mL; Sigma-Aldrich) was added after 45 min; 5 h later, intracellular IFN-γ was measured by gating on CD8 and M1p58-66 tetramer–positive cells. For primary T cell stimulation, CD8⁺ T cells were cocultured with antigen-loaded and irradiated (12 Gy) γδ T-APCs at an APC to responder cell ratio of 1:10. Twenty units per milliliter IL-2 was added on day 3, and tetramer-positive cells were analyzed on day 11.

pH Measurement.

pH was measured as described previously (39). In brief, cells were incubated for 1 h with 1 mg/mL of FITC-BSA and Alexa Fluor 647-BSA, chased, and analyzed by FACS. The FL1:FL4 ratio was calculated, and pH values were extrapolated from calibration curves obtained from the same batch of cells incubated for 4 min in buffers (pH 5–8) in the presence of 10 μM monensin and nigericin (Sigma-Aldrich).

Western Blot Analysis.

Equal numbers of cells were lysed in sample buffer, sonicated, and run on a 12.5% SDS-PAGE gel. Gels were blotted on a nitrocellulose membrane and stained with anti-FITC or β-actin antibodies. Blots were developed using ECL Western blotting substrate (Pierce) and captured on radiographic film (GE Healthcare). Densitometric analysis of scanned autoradiographs was done using ImageJ software.

Confocal Microscopy.

FITC-BSA–, Alexa Fluor 488-BSA–, or Alexa Fluor 647-BSA–loaded cells were fixed in formaldehyde, permeabilized with 0.1% saponin, and stained for EEA-1, Lamp-1, rab11, or IRAP. Cells were incubated with Alexa Fluor 488– or Alexa Fluor 549–labeled secondary antibodies (Invitrogen), followed by the nuclear stain DRAQ5 (Axxora) or Sytox Orange (Invitrogen). For simplicity, nuclear stains were false-colored in blue, endocytosed BSA was colored in green, and vesicular stains were colored in red. Before endocytosis of cellular debris, γδ T-APCs were surface-stained with PKH67 (Sigma-Aldrich) in accordance with the supplier's protocol. Cells were mounted using Prolong Gold (Invitrogen). Images were acquired by scanning in sequential mode on a Leica LSM TCS SP5 confocal microscope (1.4 NP Plan Apochromat) and processed with Imaris 6.1.3 (Bitplane). Quantitative colocalization was performed using the ImarisColoc module and the automatic threshold function.

Statistical Analysis.

Data are expressed as mean ± SEM and analyzed using the two-tailed unpaired Student t test, with differences considered significant as indicated in the figures: *P < 0.05; **P < 0.01; and ***P < 0.005.

The media, reagents, and antibodies used are described in SI Materials and Methods.