Purpose: Tumor-specific antigens of 3-methylcholanthrene (MCA)-induced sarcomas were defined by the narrow immune responses they elicited, which uniquely rejected the homologous tumor, with no cross-reactions between independently derived syngeneic MCA-induced tumors. This study examines whether an autophagosome-enriched vaccine derived from bortezomib-treated sarcomas can elicit an immune response that cross-reacts with other unique sarcomas.
Experimental Design: Mice were vaccinated with either MCA-induced sarcomas or autophagosomes derived from those tumors and later challenged with either homologous or nonhomologous sarcomas. In addition, 293 cells expressing a model antigen were used to understand the necessity of short-lived proteins (SLiP) in this novel vaccine. These findings were then tested in the sarcoma model. Autophagosomes were characterized by Western blotting and fluorescent microscopy, and their ability to generate immune responses was assessed in vitro by carboxyfluorescein succinimidyl ester dilution of antigen-specific T cells and in vivo by monitoring tumor growth.
Results: In contrast to a whole-cell tumor vaccine, autophagosomes isolated from MCA-induced sarcomas treated with a proteasome inhibitor prime T cells that cross-react with different sarcomas and protect a significant proportion of vaccinated hosts from a nonhomologous tumor challenge. Ubiquitinated SLiPs, which are stabilized by proteasome blockade and delivered to autophagosomes in a p62/sequestosome-dependent fashion, are a critical component of the autophagosome vaccine, as their depletion limits vaccine efficacy.
Conclusion: This work suggests that common short-lived tumor-specific antigens, not physiologically available for cross-presentation, can be sequestered in autophagosomes by p62 and used as a vaccine to elicit cross-protection against independently derived sarcomas. Clin Cancer Res; 17(20); 6467–81. ©2011 AACR.
This article is featured in Highlights of This Issue, p. 6367
For more than 50 years, the basic paradigm that vaccination with chemically induced tumor cells is effective at protecting mice with a challenge from that specific tumor, but not other tumors induced by the same carcinogen, has been repeatedly documented, is well accepted, and has driven numerous autologous cancer vaccine trials. The importance of the current article is that it identifies a methodology that improves vaccine efficacy sufficiently enough to prime immunity against shared antigens in models where whole tumor vaccines fail to do so. It also provides the mechanisms responsible for augmented priming and insights to further enhancements. Better vaccines will help, but generating therapeutic immune responses in patients with advanced cancer will likely require a multimodality combination approach. A clinical study of this strategy is underway.
Cross-presentation was identified as a means by which antigens can be presented by cells in which they were not synthesized, thus obviating direct presentation as the sole mechanism to prime an immune response (1). These findings were expanded by the demonstration that cross-presentation of melanoma antigens during vaccination was essential for the generation of an effective antitumor immune response (2). One component of cross-presentation that has been debated and still remains unknown is the source of antigen and the method of its delivery to professional antigen-presenting cells (APC). Although some groups have shown that the source of antigen is cellular protein, others argue that it is peptides chaperoned by HSPs (3, 4). To complicate the debate further, Johanthan Yewdell's group showed that cross-priming results from both the donation of proteasome substrates and stable cytosolic peptides in conjunction with HSP90 (5, 6). Recently, we described a pool of antigen used for cross-presentation that is dependent on macroautophagy (hereafter referred to as autophagy; refs. 7, 8).
Autophagy is a cellular process in which portions of the cytoplasm are sequestered by double-membrane vesicles termed autophagosomes that range in size from 300 to 900 nm (9). The contents of these autophagosomes are degraded in a lytic compartment, which facilitates the turnover of long-lived proteins and is critical for maintaining the pool of amino acids needed for anabolism. The formation, function, and isolation of autophagosomes are distinct from that of exosomes, 100-nm single-membrane vesicles originating from endosomes. Exosomes are actively secreted into the extracellular environment to signal to or otherwise modulate neighboring cells via a variety of stimulatory or inhibitory mechanisms (10–12). Despite their discrete biophysical properties, both dendritic cell–derived exosomes and tumor-derived autophagosomes can generate robust immune responses (12). However, unlike tumor-derived autophagosomes, tumor-derived exosomes have been shown by different groups to be immunosuppressive via a variety of mechanisms (9, 10, 13, 14).
A key marker of the induction of autophagy is the conversion of the cytosolic form of microtubule-associated protein 1 light chain 3 (LC3-I) via a series of ubiquitin-like conjugation steps to the lipidated form (LC3-PE or LC3-II) that is tightly associated with autophagosomes. Conversion to LC3-II is not only a definitive marker of autophagy but can also promote curvature and hemifusion of membranes and is therefore critical to the process (15). A recently described protein, p62/SQSTM1 (sequestosome), binds both polyubiquitin and LC3 and thus facilitates degradation of ubiquitinated proteins via autophagy (16, 17). Interaction of LC3 with p62 has added a layer of complexity to the autophagic network and suggests that this bulk degradation process may be more selective than previously appreciated.
We have shown that autophagy in tumor cells is essential for efficient cross-presentation and subsequent induction of tumor immunity in a B16 melanoma model (8). Cross-presentation, which was measured by proliferation of carboxyfluorescein succinimidyl ester (CFSE)-labeled antigen-specific T cells, was significantly inhibited when autophagy was blocked and increased during autophagy promotion. Interestingly, when cell lysates were fractionated and used as an antigen source, the fraction with the greatest cross-presentation activity also had the highest level of the specific autophagosome marker LC3. By treating cells with the proteasome inhibitor bortezomib and the lysosomotropic agent NH4Cl, which prevents fusion of autophagosomes with lysosomes, autophagosome-containing vesicles could be isolated. These isolated autophagosome-containing vesicles, termed DRibbles (7), served as a potent antigen source in cross-presentation assays and in in vivo vaccine studies. In combination with the results of 2 recent publications, which showed enhanced antigen presentation related to autophagy (18, 19), our work has further defined the function of autophagy as a means of sequestering antigen for cross-presentation.
To understand better the function of autophagy in cross-presentation, we developed a model that incorporates the DRiP hypothesis (20). A significant proportion of MHC class I binding peptides originate from defective ribosomal products (DRiPs), including misfolded and truncated polypeptides, which are degraded by the proteasome shortly after their translation and loaded onto MHC class I molecules (5). Because DRiPs and other short-lived proteins (SLiP) are stabilized by proteasome inhibition, we hypothesized that autophagosome-containing vesicles isolated from bortezomib-treated cells would contain DRiPs and SLiPs and thereby provide a unique spectrum of potential tumor rejection antigens. We further hypothesized that using these vesicles to prime an immune response will generate a broader T-cell response.
Prehn and Main established the unique specificity of chemically induced 3-methylcholanthrene (MCA) sarcomas, whereby sarcomas generated in genetically identical mice with similar morphology and growth characteristics would only protect vaccinated mice from a challenge with the immunizing tumor but not other syngeneic sarcomas. Although there has been a paucity of antigens associated with the unique specificity of this tumor model (21), genetic analysis of an MCA-induced sarcoma after cytolytic T lymphocytes (CTL) immunoselection revealed a deletion in a region rich with oncogenes and tumor suppressor genes (22). Even though a unique immunodominant antigen results from each MCA treatment, these data show that specific loci or chromosomal regions are more susceptible to the mutating effects of MCA. Moreover, using CTL immunoselection, a secondary tumor antigen shared by an independent sarcoma cell line was uncovered, showing that the unique rejection antigen is only part of the tumor antigen profile (23). Others have shown cross-reactivity among heterogenic clones of the MCA-106 sarcoma using effector cells primed with the parental MCA-106 line but no cross-reactivity with an antigenically distinct MCA-205 sarcoma (24). There are, therefore, limited examples of common antigens among the MCA-induced sarcomas in the few publications reported.
In this article, we examine the role of autophagy in tumor immunity by focusing on autophagosomes as the source of antigen for cross-presentation. We find that vaccination with antigens derived from autophagosomes can broaden the T-cell response beyond that seen following whole-cell vaccination. Studies using MCA sarcomas and human embryonic kidney (HEK) 293T cells stably expressing a short-lived model antigen both indicate that SLiPs are necessary for this unique autophagosome-mediated immune response. Furthermore, we show that the ubiquitin/LC3-binding protein p62 (sequestosome) has a key role as a regulator of selective autophagy, as it associates with both ubiquitinated antigen and LC3 and is needed for the sequestration of SLiPs into the autophagosomes. On the basis of these findings, we propose that the broad array of tumor antigens contained in autophagosomes is dependent upon ubiquitinated SLiPs incorporated by the sequestosome and that cross-presentation of these autophagosomes primes a unique cross-protective immune response.
Materials and Methods
Mice and cell lines
Female C57BL/6J (B6) mice were purchased from Charles River. OT-1 breeders transgenic for the T-cell receptor (TCR) that recognizes chicken ovalbumin (peptide sequence SIINFEKL) in the context of H-2Kb were purchased from Jackson Laboratory. All mice were maintained and used in accordance with the Earl A. Chiles Research Institute Animal Care and Use Committee and recognized principles of laboratory animal care were followed (Guide for the Care and Use of Laboratory Animals, National Research Council, 1996).
HEK 293T cells were cultured in complete medium, which consisted of Dulbecco's Modified Eagle Medium (Lonza) supplemented with 10% FBS (Life Technologies). Cell lines were maintained in T-75 or T-150 culture flasks in a 5% CO2 incubator at 37°C. Stable expression of both the stable OVA (mOVA) and short-lived OVA (rOVA) in HEK 293T cells was achieved with lentiviral transduction as described earlier (8). For the generation of recombinant lentiviruses, HEK 293T cells were transiently transfected with vector plasmid pWPT or pGIPz, virus packaging plasmid pPAX2, envelope plasmid VSV-G MD2, and helper plasmid pAdv. Viral supernatant was used to infect HEK 293T cells, and transduced cells were sorted on the basis of green fluorescent protein (GFP) expression by flow cytometry.
MCA-induced sarcomas were previously generated in our laboratory (25, 26). All tumor aliquots used had been in vivo passaged less than 6 times. Unique MCA sarcomas were passaged in vivo by excising a tumor from a female C57BL/6 mouse followed by triple-enzyme digestion of the tumor (incubation at room temperature with hyaluronidase, collagenase, and DNAse for 1–3 hours with agitation) and subcutaneous injection into a naive mouse. Uniqueness of each sarcoma was confirmed by vaccine challenge studies summarized in Fig. 2G.
B16BL/6J-D5 (D5), a poorly immunogenic subclone of the spontaneously arising B16BL6 melanoma, was kindly provided by Dr. S. Shu (Earle A. Chiles Research Institute, Portland, OR). A stock of D5 cells, screened to be free of Mycoplasma, was banked and assayed for immunogenicity and expression of melanoma-associated antigens, gp100 and TRP2. D5 cells from this stock were used within 4 to 6 weeks of culture. D5 cells were cultured in complete medium, which consisted of RPMI 1640 (BioWhittaker) supplemented with 10% FBS (Life Technologies), 50 μmol/L 2-mercaptoethanol (Aldrich), 0.1 mmol/L nonessential amino acids, 1 mmol/L sodium pyruvate, 2 mmol/L l-glutamine, and 50 μg/mL gentamicin sulfate. Cell lines were maintained in T-75 or T-150 culture flasks in a 5% CO2 incubator at 37°C.
APCs were generated in the spleens of C57BL6 mice via hydrodynamic gene transfer (27). Female C57BL/6J (B6) mice were sequentially injected with 2 mL intravenous dose of 2 μg of plasmid DNA encoding murine Flt3 ligand or granulocyte-macrophage colony-stimulating factor. The spleens are typically enlarged and enriched for CD8+/CD11c+ cells (8, 28, 29). Spleens were harvested and processed into a single-cell suspension and frozen.
Tumor vaccine and challenge
Female C57BL/6J (B6) mice were vaccinated in the lower right flank with a subcutaneous injection of 5 × 106 irradiated tumor cells from a freshly digested sarcoma or with 3 × 106 APCs pulsed with autophagosomes (3 × 106 cell equivalents) for 6 to 8 hours. To control for the use of dendritic cells in the autophagosome vaccine, 3 × 106 APCs pulsed with 5 × 106 irradiated tumor cells for 6 hours, were washed and used as a vaccine as described earlier. For the cyclohexamide experiments, 5 × 106 frozen tumor cells (previously treated with cyclohexamide) were thawed, irradiated, and used as vaccine. Mice were challenged 2 weeks after vaccination by subcutaneous injection in the lower left (opposite) flank with 2 × 104 to 3 × 104 viable digested MCA tumor cells. Tumor growth was assessed by measuring the perpendicular diameters of the sarcoma. Mice were sacrificed when the area of the tumor, determined by the product of the perpendicular diameters, reached 150 mm2 or greater.
Isolation of autophagosome-containing vesicles
Autophagosome-containing vesicles (DRibbles) were harvested from HEK 293 cells or tumor cells after treating these cells with 100 nmol/L bortezomib (Velcade) and 10 mmol/L of NH4Cl in complete medium for 18 to 24 hours in a 5% CO2 incubator at 37°C. The cells and the supernatant were harvested and spun at 480× g, as described by Stromhaug and colleagues (30). The supernatant was then spun at 12,000× g to harvest the autophagosome-containing pellet. On the basis of the amount of cells originally seeded, the autophagosome pellet was resuspended at 108 cell equivalents per mL (typically in the range of 0.5–1 mg/mL total protein by bicinchoninic acid).
DNA construction and transfection
Plasmid DNA vector cloning was described earlier (8). Briefly, Ub-X-GFP–expressing plasmids were fused with an OVA antigen by PCR with Vent polymerase and cloned into the lentiviral vector pWPT (kindly provided by Dr. D Trono, Department of Microbiology, Geneva School of Medicine, Geneva, Switzerland). The LC3 fusion plasmid pCMV-GFP-LC3 and the p62 fusion plasmid pCMV-tdTomato-p62 were kindly provided Dr. T Johansen (Biochemistry Department, Institute of Medical Biology, University of Tromso, Tromso, Norway). Both the LC3 and p62 constructs were subcloned into the pWPT vector by PCR with Vent polymerase. Transient transfections of HEK 293FT cells with LC3, p62, or OVA expressing vectors were done using METAFECTENE PRO (Biontex Laboratories GmbH).
Measurement of cytokine production by primed LN T cells from vaccinated mice
To measure the priming of naive lymphocytes, mice were vaccinated in both the fore and hind flanks by subcutaneous injection of 1 × 106 irradiated sarcoma cells or 2 × 106 dendritic cells pulsed with 3 × 106 cell equivalents of autophagosomes. Draining lymph nodes were harvested after 11 days and cultured with soluble anti-CD3 (5 μg/mL) for 48 hours and then expanded with interleukin (IL)-2 (60 IU/mL) for 72 hours. After this in vitro activation and expansion, effector T cells were washed, resuspended in complete medium and IL-2 (60 IU/mL), and seeded at 2 × 106/2 mL/well in a 24-well plate. The cells were either cultured without further stimulation or stimulated with 2 × 105 D5, primary (triple-enzyme digested) MCA-304, MCA-309, MCA-310, MCA-311 tumor cells, primary kidney cells, or immobilized anti-CD3 (positive control). Supernatants were harvested after 24 hours and assayed for IFN-γ by ELISA using commercially available reagents (IFN-γ; BD Biosciences Pharmingen). The concentration of cytokines in the supernatant was determined by regression analysis.
Short interfering RNA knockdown of sequestosome (p62)
siGENOME SMARTpool M-010230-00-0005 against human SQSTM1 (p62) was purchased from Dharmacon (Thermo Scientific). Nonspecific control short interfering RNA (siRNA) sc-36869 was purchased from Santa Cruz Biotechnology. A 50 nmol/L solution of either siRNA was transfected into HEK 293 T cells, using Invitrogen's Lipofectamine 2000.
Radioimmunoprecipitation assay (RIPA) buffer with a protease inhibitor cocktail (Roche) was used to lyse autophagosomes isolated from treated 293rGFP-OVA or parental HEK 293T cells. The lysate (50 μg) was initially incubated with 2 μg of goat anti-human IgG (Jackson ImmunoResearch). The precleared lysate (∼25 μg) was then incubated overnight with 1 μg of a goat anti-GFP antibody (Rockland). Protein A/G agarose was added to the anti-GFP/lysate mixture and incubated for 4 hours. Sixty microliters of 2× NuPAGE LDS buffer was added directly to beads. The beads were subjected to SDS-PAGE and Western blotting as described later.
For Western blots, HEK 293T cells (1 × 106) or autophagosomes (10 × 106 cell equivalents) were lysed in 100 μL of RIPA buffer with a protease inhibitor cocktail (Roche). The lysates were mixed with 4× NuPAGE LDS sample buffer and samples were resolved by 4% to 20% SDS-PAGE (Invitrogen). Proteins were transferred to a nitrocellulose membrane, incubated with primary antibodies, diluted in blocking buffer (5% dry milk) overnight, and then incubated with horseradish peroxidase (HRP)–conjugated secondary antibodies for 1 hour. Protein bands were revealed by using chemiluminescent reagents (Pierce). The primary antibodies included rabbit anti-actin (1:2,000; Sigma), goat anti-GFP (1:1,000; Rockland), rabbit anti-ubiquitin (1:1,000; Upstate), rabbit anti-LC3 (1:1,000; MBL), and goat anti-sequestosome (1:700; Santa Cruz Biotechnology). The secondary antibodies were goat antirabbit HRP (1:10,000; Jackson ImmunoResearch), and donkey anti-goat HRP (1:10,000; Jackson ImmunoResearch).
CFSE proliferation assay
CFSE-labeled naive T cells from ovalbumin-specific OT-1 TCR transgenic mice were added to APCs pulsed with autophagosomes, SIINFEKL peptide (10 μg), or whole soluble ovalbumin (50 μg). Cross-presentation of antigens to labeled OT-I T cells was assessed by measuring the dilution of CFSE by flow cytometry. Splenocytes from OT-I mice were labeled with 5 μmol/L of CFSE according to the manufacturer's protocol (Invitrogen). T-cell proliferation was measured as loss of CFSE intensity after 3 or 4 days of APC and T-cell coincubation.
The images were acquired at the Advanced Light Microscopy Core at The Jungers Center of Oregon Health and Science University on a high-resolution wide-field Core DV system (Applied Precision). This system is an Olympus IX71 inverted microscope with a proprietary XYZ stage enclosed in a controlled environment chamber; differential interference contrast transmitted light and a short arc 250-W xenon lamp for fluorescence. The camera used was a Nikon Coolsnap ES2 HQ. Each image was acquired as Z-stacks in a 1,024 × 1,024 format with a 60 × 1.42 NA Plan Apo N objective in 3 colors: green, red, and blue. The pixel size was 0.107 × 0.107 × 0.5 μm. The images were deconvolved with the appropriate optical transfer function, using an iterative algorithm of 10 iterations. The histogram was optimized for the most positive image and applied to all the other images for consistency before saving the images as 24-bit merged TIFF. A reference differential interference contrast image was acquired from the middle of the Z-stack.
For the time-to-death endpoint, the survival distribution was estimated by the Kaplan–Meier method. The log-rank test was used to compare the hazard rates of the vaccines. For ELISA data, the log10-transformed IFN-γ values were analyzed by a mixed-effects model that appropriately accounted for different sources of variability introduced by the experimental design. This model had fixed effects for the vaccine group, stimulator, and their interaction, and random effects for the experiment (nested within vaccine group), the stimulator–pool interaction, and the assay replicate. Different variances were allowed for each vaccine group. Contrasts using least-squares means for each stimulator–vaccine group combination were used to test for mean differences. A significance level of α = 0.05 between different vaccine groups was limited to individual stimulators. The significance level was α = 0.05. All analyses were done using SAS v9.2, using either PROC LIFETEST or PROC MIXED.
Vaccination with autophagosome-containing vesicles derived from 3-MCA–induced sarcomas generates cross-reactive T cells
Multiple MCA sarcomas were irradiated and used in an attempt to protect mice from independently generated sarcomas. Similar to what has been described (31), irradiated MCA-310 cells protected mice against an MCA-310 tumor challenge but not an MCA-311 tumor challenge. Conversely, irradiated MCA-311 cells protected mice against an MCA-311 tumor challenge but not against MCA-310 (Fig. 1A). Previous works in our laboratory and others have described how pre-effector T cells found in the lymph node draining a tumor could be sequentially activated with anti-CD3 and IL-2 and retain tumor specificity (32, 33). MCA-304 and MCA-310 tumor-draining lymph nodes were harvested, activated in vitro with anti-CD3, and expanded with IL-2. The resultant effector T cells were assayed for IFN-γ secretion after stimulation with different tumor targets. Both MCA-304 and MCA-310 primed effector T cells secreted tumor-specific IFN-γ when stimulated with the immunizing tumor (Fig. 1B, black bars). However, both groups of effector T cells failed to secrete appreciable IFN-γ in response to all other targets, including the nonvaccinating MCA-induced sarcomas. On the basis of our hypothesis that SLiPs or DRiPs incorporated into autophagosomes contain unique antigens that will broaden the T-cell response, autophagosome-containing vesicles were isolated from bortezomib-treated, intact sarcomas, and used as a subcutaneous vaccine. Surprisingly, MCA 304-derived autophagosomes pulsed onto APCs primed an immune response to not only MCA-304 but also other independently derived sarcomas (Fig. 1B, white bars). Similar results were observed with an autophagosome vaccine derived from the MCA-310 sarcoma. These responses were specific because the autophagosome vaccine did not prime T cells to react with the syngeneic, but unrelated, B16 melanoma subclone (D5) or a primary kidney cell line (Fig. 1B and data not shown). This suggests that an autophagosome vaccine may prime a more diverse repertoire of T cells to sarcoma-related antigens.
Autophagosome vaccination protects against multiple independently derived sarcomas
To determine whether the autophagosome vaccine could not only prime a broader T-cell response to tumor or tumor-associated antigens but could also protect a vaccinated host from a nonhomologous (nonimmunizing) sarcoma challenge, groups of mice were vaccinated with irradiated MCA-induced tumors or autophagosomes derived from these tumors and subsequently challenged with tumor. Vaccination with irradiated whole tumor cells provided complete protection against the homologous tumor. Although the irradiated whole-cell vaccine protected more mice than an autophagosome vaccine derived from the same tumor, the autophagosome vaccine provided immunity against not only the homologous sarcoma but also other independently derived sarcomas (Fig. 2A–F). These results coupled with the previous experiment in Fig. 1B show that autophagosome-containing vesicles derived from MCA-induced sarcomas contain epitopes that can be cross-presented and prime a T-cell population capable of protecting the host from a challenge with sarcomas other than the one used during vaccination.
Earlier experiments showed that autophagosomes pulsed onto APCs were more effective than autophagosomes alone at generating a productive vaccine (data not shown). Therefore, to control for the possibility that the addition of APCs potentiates the cross-protection observed with the autophagosomal vaccination, APCs were pulsed with irradiated whole cells and used as a vaccine. Similar to what was observed in experiments without APCs, whole tumor cell vaccination with APCs could protect vaccinated mice from a challenge with the immunizing tumor (Supplementary Fig. S1). However, the inclusion of APCs with the whole-cell vaccination could not cross-protect mice from a challenge with a closely related but independently derived sarcoma in any of the vaccines tested (Supplementary Fig. S1). This autophagosome-mediated cross-protection showed in both Fig. 2 and Supplementary Fig. S1 was evident in 8 of the 9 combinations tested, whereas the whole-cell vaccine provided cross-protection in 0 of 9 combinations tested (Fig. 2G).
Autophagosomes contain SLiPs that can induce proliferation of naive antigen-specific T cells
To understand the mechanism of cross-protection with the autophagosome vaccine, we used a model antigen system to test the hypothesis that SLiPs were necessary for autophagosome-induced immunity. HEK 293T cells were transduced to stably express the model antigen ovalbumin (OVA) with a moderate half-life of approximately 8 to 10 hours or a shorter half-life of approximately 20 minutes (8). The half-life of the model proteins was controlled by inserting either a stabilizing methionine residue (mOVA) or a destabilizing arginine residue (rOVA) at the N-terminus of the protein as determined by the N-end rule (34). In addition, to detect the model antigen more easily, the OVA proteins were fused to an enhanced green fluorescent protein (eGFP; ref. 8). The accumulation of fusion proteins in the HEK 293 T cells was measured by flow cytometry for the presence of eGFP (Fig. 3A). The amount of the stable model protein (mGFP-OVA) remained nearly constant when cultured overnight with or without proteasome inhibition. In contrast, the turnover of the short-lived model antigen (rGFP-OVA) was proteasome-dependent; blockade of the proteasome with bortezomib increased the expression of the rGFP-OVA protein.
Autophagosomes (3 × 106 cell equivalents) were isolated from 293rGFP-OVA cells treated with or without the proteasome inhibitor bortezomib and pulsed onto APCs for 6 hours, which were used to stimulate naive OVA-specific, CFSE-labeled T cells. Proliferation of the stimulated T cells was measured by dilution of CFSE. Although autophagosomes isolated from untreated 293rGFP-OVA cells caused a significant increase in T-cell proliferation, autophagosomes isolated from 293rGFP-OVA cells treated with bortezomib induced substantially more proliferation (Fig. 3B). In addition, autophagosomes isolated from bortezomib-treated 293rGFP-OVA cells were superior at stimulating T-cell proliferation when compared with the cell lysate (Fig. 3C).
Inhibition of protein synthesis before proteasome blockade prevents accumulation of SLiPs
To further define the relationship of SLiPs and autophagosomes, cells were pretreated with the protein synthesis inhibitor cyclohexamide before the addition of bortezomib to inhibit the production of SLiPs and prevent their inclusion into autophagosomes. To determine a dose of cyclohexamide capable of preventing protein synthesis, HEK 293T cells were depleted of all methionine and then treated with cyclohexamide at 10 or 50 μg/mL, or with vehicle alone for 4 hours. Concomitantly, the cells were pulsed with nonradioactive labeled methionine (l-azidohomoalanine) and visualized using a streptavidin-HRP enzymatic reaction. The HRP detection system can react with endogenous biotin, such as the biotin carboxylase subunits of acetyl-CoA (39-kDa band in Fig. 4A). In 4 hours, 10 μg/mL cyclohexamide prevented synthesis of all proteins (Fig. 4A).
We hypothesized that because mOVA is a stable protein, its expression would be detected after a 4-hour pretreatment with cyclohexamide before either a 4-hour or overnight culture with bortezomib whereas there would be diminished expression of rOVA because of its short half-life. Expression of GFP in HEK 293mGFP-OVA cells was essentially unchanged after 4 hours regardless of whether they were treated with bortezomib or cyclohexamide (Fig. 4B, left). However, the expression of GFP in HEK 293rGFP-OVA cells was diminished in the absence of proteasome inhibition and absent following cyclohexamide pretreatment (Fig. 4B, right). The reduced expression of the model SLiP was not due to decreased cell viability or a secondary effect of bortezomib or cyclohexamide, as both 293rGFP-OVA cells and the 293mGFP-OVA cells treated with cyclohexamide and bortezomib or bortezomib alone had similar morphology (slightly more rounded cells with cyclohexamide), growth patterns, and were able to exclude trypan blue even after an overnight culture (data not shown).
The expression of both mGFP-OVA and rGFP-OVA in HEK 293T cell lysates was also examined by Western blotting after a 24-hour treatment. Similar to what was observed by flow cytometry, bortezomib caused a modest increase in the more stable mOVA protein. However, HEK 293mGFP-OVA cells cultured overnight with cyclohexamide and bortezomib exhibited a substantial reduction in the levels of the 64-kDa fusion protein (Fig. 4C). Bortezomib increased the half-life of the rGFP-OVA fusion protein in HEK 293rGFP-OVA cells only in the absence of cyclohexamide.
Autophagosomes derived from cells that lack SLiPs have diminished capacity to stimulate antigen-specific T cells
Autophagosomes derived from cells expressing either the short-lived rGFP-OVA protein or the stable mGFP-OVA protein were isolated, used to pulse APCs, and drive naive antigen-specific OT-1 T-cell proliferation. Autophagosomes from cells expressing the stable mGFP-OVA protein caused OT-1 T cells to proliferate at nearly the same level whether they were generated in the presence or absence of bortezomib, with or without cyclohexamide (Fig. 4D). These data indicate that cyclohexamide pretreatment does not prevent stable proteins from being sequestered in autophagosomes, nor does it prevent the formation of functional autophagosomes. However, autophagosomes from cyclohexamide-treated HEK 293rGFP-OVA cells were unable to induce antigen-specific T-cell proliferation (Fig. 4E). The diminished proliferation was more apparent when higher doses of autophagosomes were used to stimulate the naive T cells. These experiments show that stable proteins, but not SLiPs, are incorporated into autophagosomes when protein synthesis is inhibited.
Autophagosomes that lack SLiPs fail to protect hosts from MCA tumor challenge
On the basis of data from the HEK 293 model antigen system, which suggest that SLiPs are incorporated into autophagosomes and can be necessary for their efficacy as a vaccine, MCA-induced sarcomas were treated with cyclohexamide to inhibit protein synthesis before the addition of bortezomib. Autophagosomes isolated from MCA-311 cells treated with or without cyclohexamide were used to vaccinate mice, which were then challenged with MCA-304 to assess the ability of MCA-311 autophagosomes to cross-protect (Fig. 5A). Vaccination with autophagosomes isolated from bortezomib-treated cells (complete autophagosomes) protected 85% (17 of 20) of the mice. In contrast, vaccination with autophagosomes isolated from cells that were pretreated with cyclohexamide before the addition of bortezomib (cyclohexamide autophagosomes) protected only 47% (9 of 19 mice) of the mice, which were not significantly different from nonvaccinated mice (8 of 20 mice protected) or mice vaccinated with nonhomologous MCA-311 whole cells (4 of 10 mice protected). These results suggest that the presence of SLiPs in the autophagosome vaccine was necessary to generate cross-protection to an independently derived sarcoma.
To understand further the role that SLiPs may play in generating a protective immune response against a homologous tumor challenge, autophagosomes isolated from cyclohexamide-treated MCA-304 cells were pulsed onto APCs and used to vaccinate groups of mice. Whereas only 20% (4 of 20) of mice vaccinated with an MCA-304 whole cells and 32% (8 of 25) with MCA-304 autophagosome vaccine grew tumor, 64% (16 of 25) of mice vaccinated with cyclohexamide-treated MCA-304 autophagosome grew tumor (Fig. 5B). This incidence of tumor growth was nearly the same as the nonvaccinated mice (17 of 25).
Although cyclohexamide can have dramatic biological effects, MCA-304 cells treated overnight with bortezomib and with or without cyclohexamide displayed similar viability and the autophagosomes isolated from cyclohexamide-treated tumors had similar protein concentrations (1.97 and 2.15 mg/mL, respectively). To assess the potential negative effects of cyclohexamide on autophagy, which could limit more than just the incorporation of SLiPs and DRiPs into autophagosomes, LC3, a critical component necessary for the induction of autophagy and thus autophagosome formation, was measured by Western blotting. The ratio of LC3 to total protein was the same or even greater with cyclohexamide pretreatment in both HEK293mGFP-OVA (Fig. 6A) and MCA-304 autophagosomes (Fig. 5C).
Autophagosomes contain ubiquitinated proteins and the sequestosome/p62, both of which colocalize with LC3
To examine the effects of cyclohexamide on protein incorporation into autophagosomes, expression of mGFP-OVA and rGFP-OVA in isolated autophagosomes was determined by Western blotting. Autophagosomes isolated from HEK 293mGFP-OVA cells had nearly the same expression of mGFP-OVA regardless of treatment with bortezomib or cyclohexamide (Fig. 6A). Interestingly, the autophagosomal marker LC3 was overexpressed in the cyclohexamide-treated cells relative to the ± bortezomib groups (Fig. 6A). This increase in LC3 has been shown to occur when cells undergo stress, such as inhibition of protein synthesis (9). It is also of note that while the cellular expression of mGFP-OVA pretreated with cyclohexamide was reduced (Fig. 4C), its expression in the autophagosome-containing vesicles was not (Fig. 6A). This suggests that mOVA-containing autophagosomes blebbed from the cells, reducing the cellular expression of this fusion protein while maintaining its expression in the autophagosomes. The model SLiP rGFP-OVA was detected only if the proteasome was inhibited in cells with normal protein synthesis; it was absent in autophagosomes produced from cells treated with cyclohexamide. Therefore, by inhibiting protein synthesis before blocking the proteasome, isolated autophagosomes contain minimal amounts of SLiP.
Because SLiPs were a critical component of the autophagosome vaccine (Figs. 4E and 5), we were interested in how they were delivered to the autophagosomal pathway. Our hypothesis predicts and others have shown (8) that treating cells with bortezomib increases the pool of ubiquitinated proteins, which may provide a means of delivery to the autophagosomal pathway. Interestingly, higher molecular weight bands were detected with anti-GFP in HEK 293rGFP-OVA autophagosomes isolated from bortezomib-treated cells (Fig. 6A) but not in the cells used to produce the autophagosomes (Fig. 4B). Although these high-weight bands were also present in autophagosomes isolated from cells expressing the more stable mGFP-OVA protein, they were less abundant. The N-end rule predicts that the model SLiP will be quickly degraded in a ubiquitin-proteasome–dependent fashion (34). We, therefore, hypothesized that these larger proteins were a ubiquitinated species of the fusion protein that escaped degradation by the proteasome and were enriched in the autophagosomes isolated from the cells. To test this hypothesis, autophagosomes from HEK 293rGFP-OVA cells were immunoprecipitated with anti-GFP and Western blotted with anti-ubiquitin. The anti-ubiquitin antibody bound only to the higher-molecular-weight proteins found in autophagosomes isolated from cells treated with bortezomib (± cyclohexamide; Fig. 6B) but not in autophagosomes from untreated 293rGFP-OVA cells or bortezomib-treated parental HEK 293T cells. This observation shows that ubiquitinated SLiPs were present in autophagosomes.
Recent work has shown that the ubiquitin-binding protein sequestosome/p62 can interact with both ubiquitinated proteins and autophagy-related initiator LC3, potentially delivering these proteins to the autophagy pathway (16, 17, 35). Western blots showed clearly that p62 was present in the autophagosomes (Fig. 6C). Colocalization of LC3 and p62 in HEK 293 T cells was seen with confocal microscopy using GFP-tagged LC3 and tomato-tagged p62. The SLiP rGFP-OVA also colocalized with tomato-tagged p62 in transfected HEK 293 T cells but not with a control GFP vector (Fig. 6D). These results show that ubiquitinated proteins are packaged into autophagosomes with ubiquitin/LC3 binding protein p62, especially when cells are treated with bortezomib.
p62 is necessary for the delivery of ubiquitinated proteins to autophagosomes
To understand how the sequestosome may influence trafficking of specific proteins to autophagosomes, p62 expression was knocked down in HEK 293T cells. A titration of p62 siRNA showed that protein expression was greatly reduced with 50 nmol/L of the p62 smart pool (Fig. 7A). HEK 293T cells were then transfected simultaneously with both p62 siRNA and a plasmid that expressed either rGFP-OVA or mGFP-OVA. After 24 hours, bortezomib was added and the cells were cultured for an additional 24 hours before isolating the autophagosome-containing vesicles. A portion of the autophagosomes isolated from the transfected cells was used to detect the fusion protein by Western blotting (Fig. 7B). The knockdown of p62 in both rGFP-OVA- and mGFP-OVA–transfected cells greatly reduced the level of the GFP-OVA. This knockdown did not affect the levels of LC3 in the autophagosomes, suggesting that autophagosome formation was independent of p62 expression (Fig. 7B).
On the basis of the colocalization experiments and the presence of the ubiquitinated protein species in isolated autophagosomes, we hypothesized that knockdown of the sequestosome would greatly diminish the ability of autophagosomes to stimulate naive T-cell proliferation. To test this hypothesis, autophagosomes generated from the cotransfection described earlier were pulsed onto APCs and used to stimulate CFSE-labeled naive antigen-specific T cells as previously described. T cells stimulated with autophagosomes generated from 293 cells with the p62 knockdown proliferated less than those stimulated without the knockdown (Fig. 7C). Although the p62 knockdown diminished the proliferation of T cells stimulated with mGFP-OVA autophagosomes, the inhibition of proliferation was even greater with rGFP-OVA autophagosomes knocked down for p62. These results coupled with the presence of ubiquitinated proteins in the rGFP-OVA autophagosomes and the colocalization of the LC3 and p62 showed that delivery of SLiPs to autophagosomes is sequestosome dependent.
We have shown that an autophagosome vaccine can prime a repertoire of T cells with a broader tumor-specific reactivity than the intact sarcoma from which it was derived. In contrast to the irradiated whole-cell vaccine, vaccination with tumor-derived autophagosomes protected mice from a challenge with independently derived MCA sarcomas. Experiments using MCA sarcomas and the OVA model antigen system both suggest that SLiPs incorporated into autophagosomes are necessary in generating a robust immune response. In addition, these data suggest that the sequestering of SLiPs into autophagosomes is dependent on the ubiquitin-binding protein sequestosome/p62. This novel vaccine strategy has redefined the tumor rejection antigens associated with MCA-induced tumors and has created the potential for new vaccine strategies.
Although MCA-induced sarcomas can share similar histology, growth patterns, and sensitivity to chemotherapeutics and are generated in syngeneic mice using the same carcinogen, each sarcoma induces a unique tumor-specific immune response. This observation has been made by many researchers (31, 33, 36–38), and it is also evident in our studies by the failure of our MCA-induced sarcoma whole-cell vaccine to cross-protect against independently derived but related MCA-induced sarcomas (Fig. 1A). Although this failure to cross-protect has been attributed to each sarcoma's unique antigenic profile (38), it seems unlikely that a large panel of sarcomas of similar etiology will not have common antigens. A more plausible description of this unique specificity could be described by the concept of a tumor rejection antigen, which functionally relates to how well an immune response directed against a tumor antigen can limit tumor growth (39, 40). The hierarchy of a tumor rejection antigen, which ranges from poor to strong, can be influenced by the ability of a specific antigen to prime an immune response. Thus, although there may be a common pool of antigens shared between independently derived sarcomas, individual tumors can express unique antigen(s) that may dominate the immune response, effectively limiting the priming of shared epitopes. This immune dominance may relate to multiple factors. Pion and colleagues have shown that immune dominance is determined by not only the affinity of the TCR and peptide-MHC complex (p:MHC) but also the surface density of p:MHC on the APC (41), which is directly proportional to an epitope's MHC binding affinity, linking the establishment of immune dominance to antigen presentation. This observation coupled with previous findings, which show the necessity of cross-presentation during tumor immunity (2), illustrates how cross-presentation may determine immune dominance and consequently the tumor specificity seen with MCA-induced sarcomas. Thus, a possible tumor rejection antigen must not only be stable enough to be transferred to an APC it must also possess an epitope that can compete for binding of the APC's MHC molecules as well as have an affinity for the TCR. Because of their transient nature, SLiPs are poorly cross-presented and are therefore removed from the pool of possible rejection antigens.
The rules governing cross-presentation of tumor or tumor-associated antigens with a cellular vaccine may not apply to an autophagosome vaccine, as SLiPs can be stabilized by proteasome blockade, isolated in autophagosomes, and pulsed onto APCs ex vivo or even injected directly into a lymphnode (intranodal). Consequently, we have focused on the repertoire of antigen, specifically the SLiPs sequestered, as a means of altering the tumor specificity with an autophagosome vaccine. On the basis of the observation that both protection and cross-protection from an MCA sarcoma challenge were reduced with elimination of SLiPs from the autophagosome vaccine (Fig. 5), antigens common to independently derived MCA-induced sarcomas are likely to be SLiPs. These SLiPs may provide epitopes with a high affinity for MHC molecules, allowing them to compete with the dominant whole-cell rejection antigen for surface density on the APC. Alternatively, the dominant rejection antigen may not be sequestered into autophagosomes, providing subdominant, neo-, or cryptic antigens access to MHC molecules for cross-presentation.
Although the potential of a varied repertoire of antigen in autophagosomes is a novel and significant finding, further studies are needed to augment the efficacy of an autophagosome vaccine. The secretion of IFN-γ by autophagosome-primed effector T cells indicates that cross-reactive tumor immune responses are generated (Fig. 1B), yet complete protection of vaccinated mice was not achieved. The lack of a memory T-cell response during the autophagosome vaccination may explain why tumor growth curves were only shifted in a proportion of mice vaccinated with autophagosomes. A vaccine strategy that incorporates an OX-40 agonist could boost the generation of memory T cells, providing a further delay in tumor progression and lead to a larger frequency of protected mice. Alternatively, the difference between delayed tumor growth versus complete protection could relate to the initial magnitude of primed T cells. The addition of a TLR agonist during the vaccine could increase critical costimulatory molecules during cross-presentation, leading to a stronger immune response. In addition, our data suggest that increasing the concentration SLiPs or their delivery to the autophagosomes would be a means of increasing the potency of the vaccine.
The targeting of SLiPs into autophagosomes seemed to be coordinated, as cell lysates could not stimulate antigen-specific T cells as effectively as could autophagosomes, suggesting a relative concentration of SLiPs in autophagosomes. This observation was supported by the remarkable increase in the proliferation of antigen-specific T cells when simulated with isolated autophagosomes versus whole-cell lysate (Fig. 3C). Western blot analysis suggests that the short-lived OVA protein was readily detected at much lower total protein concentrations in autophagosomes than in cell lysates (Figs. 4C and 6A and data not shown). In addition, higher-molecular-weight fusion proteins were detected only in autophagosomes and not in cell lysates. These higher-molecular-weight proteins were more obvious in autophagosomes derived from bortezomib-treated cells and especially those containing the model SLiP antigen that is rapidly ubiquitinated versus the more stable antigen (Fig. 6A). The higher weight band, confirmed to be ubiquitinated by Western blot analysis after coimmunoprecipitation (Fig. 6B), was likely responsible for directing SLiPs into autophagosomes based on recent publications that established a link between autophagy and ubiquitinated proteins via the LC3 and ubiquitin binding protein p62/sequestosome (42–44). Indeed, HA-tagged p62 was readily detected in autophagosomes isolated from transfected cells and fluorescent microscopy showed colocalization of p62 with both the model SLiP and LC3 (Fig. 6C and D), all supporting the hypothesis that p62 regulates the sequestering of ubiquitinated proteins into autophagosomes. Although using siRNA to knockdown expression of p62 may destabilize or limit expression of the GFP-OVA construct, these experiments corroborate the role of sequestosome in antigen delivery to autophagosomes, as reduced expression of p62 limited the model antigen found in the autophagosomes and diminished proliferation of antigen-specific T cells stimulated with these autophagosomes. Autophagosomes isolated from HEK293rGFP-OVA cells treated with p62 siRNA had a greater impairment in their ability to stimulate T-cell proliferation than autophagosomes isolated from p62 siRNA-treated cells expressing the more stable mGFP-OVA antigen (Fig. 7C), suggesting a correlation between the stability or state of ubiquitination of the protein and the necessity of p62 to transfer the protein into the autophagosome.
Collectively, the data point to a model whereby SLiPs or other ubiquitinated proteins are sequestered into autophagosomes in a p62-dependent fashion when the cell is treated with a proteasome inhibitor. Furthermore, data from the MCA-induced sarcoma cross-protection experiments suggest that these SLiPs include antigenic determinants common to multiple independently derived sarcomas. Several possibilities may explain how a whole-cell vaccine does not afford the cross-reactivity observed with an isolated autophagosome vaccine. The most simple explanation is that tumors rapidly turnover SLiPs in a proteasome-dependant fashion, thereby preventing their accumulation. This explanation makes the assumption that MCA-induced tumors do not directly prime an immune response because rapidly degraded proteins fill a large proportion of MHC class I molecules on the cell surface (20). Alternatively, SLiPs may accumulate at low levels but are not stable enough to be efficiently cross-presented, which is supported by the observation that superior T-cell proliferation occurs with much less autophagosomes than cell lysates. If both SLiPs and whole-cell tumor rejection antigens are equally cross-presented, the rejection antigens may assert immune dominance by binding MHC class I more efficiently or, in the context of MHC class I, have a higher affinity for the TCR (41). Addressing these possibilities will provide a greater understanding of not only this novel finding but also a potential class of novel tumor antigens.
The implication of these studies from a translational point of view is that autophagosomes may be a way to uncover common antigenic determinants created during the transformation process. Evidence supporting this concept in this model comes from 2 studies. The first study identified a region of the mouse genome rich in tumor-modifier genes and associated with MCA-induced tumor antigens (22). A second study showed that mice vaccinated with an MCA-induced tumor that had lost the dominant antigen could generate CTLs, which recognized a shared antigen expressed not only by the immunizing cell line but also by independent sarcoma cell lines (23). The idea that tumor-derived autophagosomes may contain common tumor-specific or tumor-associated antigens is clinically appealing. These autophagosomes could have the potential of an off-the-shelf vaccine in an allogeneic setting, particularly if an autologous tumor is unavailable to the patient. In addition, this strategy could be useful in priming immune responses against tumor-antigen loss variants seen with a progressively growing tumor and might even serve as an alternative vaccine to treat patients coming out of remission. This strategy has the potential to augment immunity in clinical applications and positively contribute to the field of immunotherapy.
Disclosure of Potential Conflicts of Interest
H.-M. Hu and B.A. Fox have ownership interest in UbiVac.
This study was supported by NIH grants CA080964 (B.A. Fox) and CA141278 (H.-M. Hu) and support from Robert W. Franz, Wes and Nancy Lematta, The Chiles Foundation and Providence Medical Foundation.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.