Phagosome-associated autophagosomes containing antigens and proteasomes drive TAP-independent cross-presentation

Abstract

Activation of naïve CD8-positive T lymphocytes is mediated by dendritic cells that cross-present MHC class I (MHC-I)-associated peptides derived from exogenous antigens. The most accepted mechanism involves the translocation of antigens from phagosomes or endolysosomes into the cytosol, where antigenic peptides generated by cytosolic proteasomes are delivered by the Transporter associated with Antigen Processing (TAP) to the endoplasmic reticulum, or an endocytic antigen-loading compartment, where binding to MHC-I occurs. We have described an alternative pathway where cross-presentation is independent of TAP but remains dependent on proteasomes. We provided evidence that active proteasomes found within the lumen of phagosomes and endolysosomal vesicles locally generate antigenic peptides that can be directly loaded onto trafficking MHC-I molecules. However, the mechanism of active proteasome delivery to the endocytic compartments remained unknown. Here, we demonstrate that Phagosome Associated LC3A/B structures (PhALs) deliver proteasomes into subcellular compartments containing exogenous antigens, and autophagy drives TAP-independent, proteasome-dependent cross-presentation.

Introduction

Virally infected cells or tumor cells express MHC-I molecules bound to peptides derived from endogenously synthesized viral or tumor antigens that are processed by cytosolic proteasomes and delivered by the Transporter associated with Antigen Processing (TAP) to the endoplasmic reticulum (ER), where peptide binding occurs(1). Activated cytotoxic T lymphocytes (CTL) recognize peptide-MHC-I complexes to mediate their effector functions. Naïve CTL precursors must be activated by dendritic cells (DCs) bearing MHC-I-peptide complexes identical to those on the ultimate CTL target cells(2). Rather than synthesizing the antigens DCs acquire them by endocytosis or phagocytosis: MHC-I binding of peptides derived from the internalized antigens and recognition of the complexes by CD8⁺ T cells is called cross-presentation(2). The cellular mechanisms responsible for cross-presentation have been extensively studied but aspects remain controversial.

A generally accepted mechanism involves translocation of internalized antigens into the cytosol(3–5). These antigens then follow the same processing pathway as endogenous antigens and depend on the peptide loading complex (PLC), which includes TAP(3, 6–8). A proposed second, TAP-independent, pathway, in which endocytic or phagosomal proteolysis generates the peptides within the endosome or phagosome(9–11), suffers from the problem that endolysosomal proteases may not generate the same peptides as the proteasome. This objection was overcome by our finding that active proteasomes are present within phagosomes and endolysosomes (12). We suggested that these proteasomes generate peptides from internalized antigens that bind to trafficking MHC-I molecules and argued that this would limit competition for MHC-I binding by peptides derived from cytosolic proteins. How active proteasomes enter the phagosome or endolysosome remains unknown. Here, we examine the hypothesis that autophagy is responsible.

Autophagosomes deliver cytosolic components, including mitochondria, to endolysosomal compartments to facilitate their degradation. However, autophagy plays a broader role in cellular processes, including unconventional protein secretion and membrane repair(13). Autophagosomes are defined as LC3/ATG8-associated double membrane-bound intracellular organelles. There are seven isoforms of LC3, including two splice variants of LC3A, LC3B, LC3C, GABARAP, GABARAPL1 and GABARAPL2. An essential step of autophagosome biogenesis involves the covalent conjugation of phosphatidylethanolamine (PE) to the C-terminal glycine residue of the LC3 variants. PE-conjugated LC3 is recruited to the autophagosomal membrane and is required for autophagy(14). Here, we show that inhibition of autophagy by enzymatically reversing PE-LC3 formation inhibits TAP-independent antigen cross-presentation, as does pharmacological inhibition of upstream components. We also show that LC3A and LC3B localize to membrane-bound structures containing proteasomes as well as antigens that have exited the phagosome and suggest a novel mechanism that explains how access of proteasomes to phagosomal antigens drives TAP-independent cross-presentation.

Materials and methods

Mice

C57BL/6 and TAP1^−/− breeding pairs were purchased from Jackson Laboratories. The mice were housed and bred at Yale Animal Resource center according to institutional guidelines.

Reagents and antibodies

Abs used were: anti-hβ₂m (BBM.1, Thermo Fischer Scientific, MA1–26040), anti-LC3A (Proteintech, 18722-AP), anti-LC3B (Novus, NB100–2220), anti-LC3C (Proteintech, 18726-AP), α-GABARAP (Proteintech, 18723–1-AP), α-GABARAPL1 (Proteintech, 11010–1-AP), anti-GABARAPL2 (Proteintech, 18724–1-AP) and anti-Proteasome 20S α1, 2, 3, 5, 6 and 7(MCP231, Enzo, BML-PW8195–0100). Secondary Abs were F(ab’)2-Goat anti-Mouse IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor^™ 488 (Thermo Fischer Scientific, A-11017) and F(ab’)2-Goat anti-Mouse IgG (H+L), Alexa Fluor^™ 568 (Thermo Fischer Scientific, A-11019. mAbs against H2-Kb conjugated to Alexa-647 (Y3)(15) and FITC-mouse CD11c antibody (Invitrogen, 11-0114-85) were used for flow cytometry.

Plasmid Constructs

Generation of pRetroX-Tight-GFP, encoding GFP under the control of TRE, has been described previously(12). Multicistronic pRetroX-Tight-GFP-CMV-hβ₂m plasmid, was engineered by cloning DNA encoding hβ2m along with an upstream CMV promoter, generated by overlap primer extension, into the BamHI and EcoRI sites of pRetroX-Tight-GFP. pRetroX-GFP-RavZ-CMV-hβ₂m was generated by cloning RavZ ORF into BamHI and NotI site of pRetroX-Tight-GFP-CMV-hβ₂m. rtTA expressing blasticidin resistant pRetroX-TetOn Advanced-BSD was generated by replacing the Neo-resistant gene of pRetroX-TetOn Advanced with the BSD gene using the BstXI and XhoI sites.

BMDC generation and transduction.

BMDCs were differentiated from bone marrow of wild type and TAP1^−/− C57BL/6 mice and retroviruses generated from 293T cells and BMDCs were co-transduced with viruses to deliver the expression construct and rtTA as previously described, (12). Transduced cells were selected by culturing the cells in puromycin(3μg/ml) and blasticidin(10μg/ml) 24hr post transduction. After 72hr transgene expression was induced with doxycycline(1μg/ml) for 16hr.

Cross-presentation assay

Transduced BMDC were incubated with varying numbers of Ovalbumin coated latex beads for 5hr, fixed, washed and incubated with B3Z cells for 16hr as previously described, (12). Cross-presentation efficiency was assessed by measuring IL-2 secretion using ELISA (OptEIA Mouse IL-2 ELISA kit; Thermo Fischer, 555148).

ULK1 inhibition or activation assay

For ULK1 inhibition, wild type BMDC or TAP^−/− transduced hβ2M BMDC were incubated with Ovalbumin coated latex beads (30 beads/cell) in the presence of either DMSO or 5 μM ULK-101 (inhibitor, Selleckchem, S8793) or 5 μM LYN-1604 (agonist, Selleckchem, S8597) for 5hr, followed by fixation, washing, and co-culture with B3Z as described above using ELISA MAXTM for IL2 (BioLegend, 431001).

Flow Cytometry

Staining for surface H2-K^b using Alexa 647-conjugated Y3 mAb and FITC-conjugated CD11c mAb (to label the BMDCs) was previously described (12). Data were acquired on an Accuri flow cytometer (BD Biosciences) and analyzed with FlowJo software. After gating on single cells, H2-K^b staining was assessed on CD11c⁺ cells

Immunofluorescence microscopy.

BMDC were mixed with latex beads coated with Alexa647-conjugated Ovalbumin at a density of 10 beads per cell, pelleted onto a glass coverslip, cultured for 4hr, fixed with 4% paraformaldehyde and permeabilized with 0.5% saponin as described (12). The cells were stained with anti-LC3A (0.5μg/ml), anti-LC3B (100μg/ml), anti-LC3C (55μg/ml), anti-GABARAP (90μg/ml), anti-GABARAPL1 (50μg/ml), anti-GABARAPL2 (50μg/ml) and anti-Proteasome 20S α1, 2, 3, 5, 6 and 7(1:50). Alexa-488- or Alexa568-conjugated secondary antibodies were used. The percentage of phagosomes associated with PhALs was determined using Fiji software.

Immunoelectron microscopy

6×10⁶ BMDCs were incubated with OVA-coated beads for 1h at 37C, layered onto 2ml of FBS and spun down for 5min at 150g. The BMDCs were transferred to 6-well plates at 2×10⁶/well, incubated at 37C for 3h, and fixed as described, (12). After rinsing in PBS and resuspension in 10% gelatin, blocks were trimmed and rotated in 2.3M sucrose overnight at 4°C, transferred to aluminum pins and frozen rapidly in liquid nitrogen. The frozen blocks were cut and placed on carbon/formvar coated grids and floated in a dish of PBS as previously described, (12). For immunolabeling, grids were incubated with anti-MCP231 at 1:200, which required a rabbit anti-mouse bridge (Jackson ImmunoResearch), and rabbit anti-LC3A diluted 1:30. 10nm and 5nm protein A gold particles (Utrecht Medical Center) were used. Grids were rinsed in PBS, fixed using 1% glutaraldehyde for 5mins, rinsed, and transferred to a UA/methylcellulose drop before being dried and viewed using FEI Tecnai Biotwin TEM at 80Kv. Images taken using Morada CCD and iTEM (Olympus) software.

Result and Discussion

Inhibition of autophagy inhibits TAP-independent antigen cross-presentation.

We showed that BMDCs from TAP1^−/− mice fail to cross-present phagocytosed antigens, but that presentation is restored by expressing human β₂-microglobulin (hβ₂m)(12), which stabilizes mouse MHC-I molecules lacking bound high-affinity peptides to allow their export from the ER(15). This form of cross-presentation involves entry of active proteasomes into endolysosmal compartments(12). To investigate the mechanism, we generated a retroviral construct, TRE-GFP-CMV-hβ₂m, which expresses GFP under the control of a doxycycline-inducible Tet responsive element (TRE) together with hβ₂m driven by the constitutive CMV promoter (Fig 1A). TAP1^−/− BMDCs were transduced with TRE-GFP-CMV-hβ₂m and cross-presentation of the model antigen ovalbumin (OVA) non-covalently associated with latex beads was assessed as described(12). Cross-presentation was observed in TAP1^−/− BMDC transduced with TRE-GFP-CMV-hβ₂m but not in BMDCs transduced with a control retrovirus lacking hβ₂m. This confirmed our previous finding and validated the use of retroviruses encoding tetracycline-inducible genes and constitutive hβ₂m as tools to investigate TAP-independent cross-presentation(12).

Figure 1. Inhibition of autophagy reduces cross-presentation by TAP1^−/− BMDCs expressing hβ₂m while activation enhances it.

A. Schematic representation of the expression systems used to express hβ₂m with GFP or GFP- tagged RavZ.

B. Cross-presentation by TAP1^−/− BMDCs co-expressing GFP and hβ₂m was compared to TAP1^−/− BMDC co-expressing hβ₂m and GFP-RavZ. IL-2 secretion by the K^b-SIINFEKL restricted CD8+ hybridoma B3Z cultured with BMDCs fixed 6hr post phagocytosis of a varying number of OVA coated latex beads was determined by ELISA.

C. Cross-presentation of OVA by wild type BMDCs co-expressing GFP and hβ₂m compared to wild type BMDCs expressing hβ₂m and GFP-RavZ.

D. Flow cytometric analysis of surface hβ₂m on transduced BMDCs expressing GFP only, GFP and hβ₂m, or GFP-RavZ and hβ₂m.

E,F. Cross-presentation of OVA and surface expression of H2-K^b on WT and TAP1^−/−+ hβ₂m BMDCs in the presence of ULK-101 (E) or Lyn-1604 (F) compared to DMSO treated controls. MFI are normalized to the average MFI of the DMSO controls. A representative experiment of three independent experiments is shown for (B,C,E,F), with the mean (±SD) of assays performed in triplicate; **** indicates p<0.005, ** indicates p< 0.01 (ANOVA test).

We hypothesized that autophagy might deliver proteasomes to the phagocytic/endocytic compartments. RavZ is a Legionella effector protein that cleaves PE from LC3 to inhibit autophagosome formation(16), providing a tool to examine this question. We engineered TRE-GFP-RavZ-CMV-hβ₂m, a retroviral plasmid that allows the expression of GFP-tagged RavZ under the control of TRE along with constitutively expressed hβ₂m. Upon doxycycline addition, the cross-presentation efficiency of phagocytosed OVA by TAP1^−/− BMDC co-expressing hβ₂m and inducible GFP-RavZ was lower than TAP1^−/− BMDC expressing hβ₂m alone (Fig 1B), consistent with the involvement of the autophagic machinery. To eliminate possible off-target effects we transduced the same retroviral constructs into WT BMDCs and found that RavZ had no significant impact on cross-presentation (Fig 1C). We used flow cytometry to ensure that expression of hβ₂m was equivalent in TAP1^−/−BMDCs transduced with TRE-GFP-CMV-hβ₂m and TRE-GFP-RavZ-CMV-hβ₂m (Fig 1D). TAP1^−/− BMDCs transduced with control vector were used to establish background staining.

We also used small molecule inhibitors and activators of processes upstream of LC3 lipidation by targeting ULK1 (unc-51 like autophagy initiating kinase 1), a serine/threonine kinase that is the ortholog of yeast ATG1(17, 18). ULK-101 selectively inhibits ULK1 and suppresses autophagy(18), while the activator Lyn-1604 upregulates it(19). Treatment of TAP1^−/− BMDCs transduced with hβ₂m with ULK-101 reduced cross-presentation compared to DMSO controls (Fig 1E, left), with no change in surface H2-K^b (Fig 1 E, right). We also found that treating the BMDCs with the ULK1 activator Lyn-1604 (19) enhanced cross-presentation (Fig. 1F, left) without affecting surface H2-K^b expression (Fig. 1F, right). Thus TAP-independent cross-presentation relies on ULK1 function as well as LC3 lipidation suggesting that it is driven by canonical autophagy. Non-canonical autophagy occurs when the double-walled autophagosome forms without the involvement of the accessory proteins normally involved in autophagy.

In contrast to the results obtained with RavZ (Fig 1C), treatment with ULK-101 decreased cross-presentation in WT BMDCs (Fig 1E) with no impact on surface H2-Kb levels. While experimental differences may account for the effect, previous studies have shown that pharmacological inhibition of autophagy reduces cross-presentation by BMDCs when evaluated a few hours after antigen uptake (19). Also Atg7- or Atg5-deficient DCs initially retain cross-presentation capacity (20, 21) but when evaluated >24h after antigen uptake it is enhanced (19). Our findings are consistent with these data. With the activator Lyn-1604 there was a slight increase in cross-presentation in WT BMDCs (Fig 1F), but it did not reach statistical significance.

Autophagic flux is regulated by multiple extracellular signals(22), and ULK1 is immediately downstream of both the energy-sensing pathway mediated by AMP-activated protein kinase (AMPK) and the nutrient-sensing pathway mediated by mechanistic target of rapamycin complex 1 (mTORC1) that initiate production of autophagic membranes. Since ULK1-dependent autophagy is regulated by the extracellular environment, endocytic proteasome delivery may be similarly dependent. No published studies examine the effect of enhanced autophagy on cross-presentation, but rapamycin has been shown to increase CD8+ T cell responses to vaccines(23, 24). DCs may shift cross-presentation from TAP-dependence to TAP-independence depending on external stressors that affect autophagy.

Antigen derived from phagosomes localizes to LC3A and LC3B positive structures.

We analyzed by confocal microscopy the localization of 6 different isoforms of LC3 in BMDCs 4hr after phagocytosis of latex beads coated with Alexa-647-OVA (Fig 2). We observed large LC3A-positive punctate structures juxtaposed to phagosomes (Fig 2A, C). These structures were also positive for OVA, likely acquired from the phagocytosed beads (Fig 2B, C). While we observed punctate structures positive for other LC3 isoforms (Fig 2D, G, J, M, P), LC3A demonstrated the most prominent staining in proximity to phagosomes. LC3B structures were also positive for OVA(Fig 2E, F, arrowhead), while those positive for the other LC3 isoform puncta were not. Our findings suggest that there is an exchange of contents between LC3A- and LC3B- positive organelles and phagosomes. We designated these LC3A/B puncta Phagosome Associated LC3A/B structures (PhALs).

Figure 2. Phagocytosed OVA localizes to LC3A- and LC3B-positive autophagosomes.

BMDCs were fixed and stained for LC3A(A-C), LC3B(D-F), LC3C(G-I), GABARAP1(J-L), GABARAPL1(M-O), GABARAPL2(P-R) 4hr post-uptake of Alexa647-OVA coated latex beads. Single representative optical sections of confocal images are shown (Scale bar- 10μm). The data is representative of 5 independent experiments, with >10 fields imaged in each experiment for LC3A, and 2 independent experiments for LC3B and 3 independent experiments for LC3C, with at least 5 fields imaged for each experiment.

Only 33 percent of OVA-containing phagosomes are proximal to PhALs, based on analysis of confocal images of 44 cells that phagocytosed OVA-647 coated beads and were stained with LC3A. This suggests that a localized mechanism may promote phagosomal association and antigen delivery. A prosaic explanation is that the interaction occurs during a transient phagosomal maturation stage, such that at a given instant only a subset of phagosomes is in that state. A more interesting possibility, because autophagy plays a role in endolysosomal membrane repair, is that leakage of phagocytic cargo into the cytosol acts as a signal for the initiation of a membrane repair pathway involving the recruitment of autophagosomes(25). Studies from our laboratory and others have shown that phagocytosed antigens can enter the cytosol of cross-presenting DCs(5). Rather than involving a specific protein transporter entry may result from local disruption of the membrane(26). Indeed, recent work suggests a role for perforin-2 in this process (27). It has been argued that phagosomal membrane disruption results in inflammasome activation and cell death(28): coupling localized disruption to autophagy-mediated membrane repair could prevent this(29). The precise signals that drive recruitment of autophagosomes to the cross-presenting phagosome remain to be defined.

Proteasomes localize to the PhAL lumen.

Because autophagy delivers cytosolic components into endolysosomal compartments we hypothesized that the PhALs deliver proteasomes into BMDC phagosomes (12). To test this, BMDCs were incubated for 4hr with latex beads, coated with Alexa-647-labeled OVA, to allow phagocytosis, and then stained for proteasomes, using an anti-pan-α-subunit antibody, and LC3A. Confocal microscopy clearly showed proteasomes intermixed with the PhAL structures (Fig 3A). To determine their localization more precisely we used immuno-electron microscopy (IEM), labeling sections with anti-LC3A and the anti-pan-α-subunit antibody. We observed double membrane bound structures containing both LC3A (Fig 3B and Supplemental Fig 1, black arrowhead) and proteasomes (white arrowhead) proximal to OVA-bead containing phagosomes (Fig 3B). The proteasome-containing double membrane-bound organelles are consistent with the role of canonical autophagy in PhAL generation.

Figure 3. PhAL structures contain OVA and proteasomes.

A. BMDCs were fixed and stained for LC3A and proteasome α-subunits 4hr post-uptake of Alexa-647-OVA-coated latex beads. A single optical section is shown of a representative cell from 2 independent experiments with >10 fields imaged in each. An enlarged image of PhAL as marked by white box is shown.

B. EM micrograph of immunogold-labeled BMDCs fixed 4hr post latex bead uptake using proteasome α-subunit and LC3A Abs. α-proteasome subunits are labeled with 5nm gold particles (white arrows) and LC3A with 10nm gold particles (black arrow) (Scale bar = 200nm). Data is representative of two experiments in which more than 10fields were imaged.

Overall, the data show that both proteasomes and antigen are within the lumen of PhALs. Exchange of cytosolic components (proteasomes) and phagosomal contents (antigen) at a repair interface between autophagosomes and phagosomes would be consistent with these findings. Based on this, and the involvement of autophagy in TAP-independent cross-presentation that we define here, we postulate that an exchange of cargo between phagosomes and autophagosomes, schematically presented in Fig 3C, allows the intermixing of antigens and proteasomes within a membrane-enclosed organelle that results in the generation of antigenic peptides that are loaded onto phagosomal MHC-I, likely recycling from the cell surface.

Key points:

• Canonical autophagy drives TAP independent antigen cross-presentation.

• LC3A/B positive vesicles associate with phagosome.

• Proteasomes localize within the lumen of LC3A vesicles associated with phagosomes.