Cross-dressing of CD8α⁺ Dendritic Cells with Antigens from Live Mouse Tumor Cells Is a Major Mechanism of Cross-priming

Apoptotic, necrotic, and live cells are natural sources of antigens for priming of antitumor immunity and tolerance against self-antigens. However, there is a vast asymmetry between the understanding of the immune system's interactions with apoptotic and necrotic cells compared with live cells, even though live cells constitute the overwhelming majority of cancer cells in a tumor. Although apoptotic and necrotic cells are extensively studied (1–6), very few studies have examined live cells as source of antigens (6–11). Dendritic cells (DC), but not macrophages, acquire antigens expressed by many live cell types (7). DCs can cross-present HIV antigen, ovalbumin, and other antigens from both live and apoptotic cells (9–11). In addition, DCs loaded with exosomes derived from live tumor cells induce potent CD8⁺ T-cell–dependent antitumor effects (12).

Although these studies show live cells can be immunogenic, there are major holes in our understanding of the mechanism of their immunogenicity. There has been no comparison of the immunogenicity of live cells with that of dead cells. In addition, how antigen-presenting cells (APC) uptake and present antigen from live cells is unclear. Splenic CD8α⁺ DCs selectively endocytose apoptotic but not live cells (6), whereas CD8α⁺ DCs are the predominant APC subset in spleen, which take up antigens from intravenously administered live cells (10). However, both of these studies restricted their analysis to only the CD8α⁺ and CD8α⁻ DC subsets in spleen. Although we understand the mechanisms underlying presentation of antigens derived from dead cells (1–5), little is known about the mechanism underlying acquisition and presentation of live cell–associated antigens by APCs at skin draining lymph nodes (sdLN).

Here, using mouse tumor models, we dissected the mechanism of immunogenicity of live tumor cells and made three significant observations: (i) live tumor cells were more immunogenic than irradiated (IR) or apoptotic tumor cells; (ii) antigens from live and IR tumor cells are taken up equally by CD8α⁺ DCs, whereas CD169⁺ macrophage readily takes up cell-associated antigens from IR cells but not from live cells; and (iii) cross-dressing of APCs at sdLNs is a major mechanism underlying presentation of live tumor cell antigens.

Female BALB/c, C57BL/6, B6.SJL-Ptprc^a Pepc^b/BoyJ (CD45.1 congenic strain of C57BL/6), Batf3^−/−, and H2K^bm1 mice were purchased from The Jackson Laboratory. CD45.1⁺ RAG^−/− OT-I T-cell receptor (TCR) Tg mice (OT-I) were bred and housed in barrier facilities maintained by the Center for Laboratory Animal Care. Animal work was performed with permission from the Institutional Animal Care and Use Committee of the University of Connecticut Health Center (Farmington, CT) and was in compliance with established guidelines.

Cell lines Meth A, CMS5, EL4, and EG7 were used. These lines were obtained in 1997 and have been in our cell bank since. We typically obtain a fresh-frozen vial approximately every 6 months. Cells were not authenticated in the past year. All lines were tested for Mycoplasma contamination in 2017 by the Mycoplasma PCR Elisa (Millipore Sigma). All cells cultured in RPMI 1640 (Thermo Fisher Scientific; catalog no. 11875085) with 10% FBS (Sigma; catalog no. F8067-500 ml) supplemented with sodium pyruvate (Thermo Fisher Scientific; catalog no. 11360070) and penicillin–streptomycin–glutamine (Thermo Fisher Scientific; catalog no. 10378016) were used. Additional supplements, i.e., nonessential amino acids (Thermo Fisher Scientific; catalog no. 11140050) and 2-mercaptoethanol (Thermo Fisher Scientific; catalog no. 21985023) were used for EG7 cells.

Mice were immunized intradermally on the right flank in Hank's Balanced Salt Solution (HBSS) without serum in 100 μL volume. Viability of live tumor cells used in this study for immunization and tumor challenge was >97%. The IR tumor cells used for immunization were cultured for 24 hours in complete medium, after γ-radiation (200 Gy) using a Gammacell 1000 irradiator (Atomic Energy of Canada Limited) to allow them to undergo apoptosis. In the case of immunization with freshly mitomycin C (MC)–treated cells, the tumor cells were treated with MC (Sigma-Aldrich; catalog no. M7949-2MG) diluted in RPMI medium (50 μg/mL for 1 hour), followed by thorough washing with complete RPMI medium, and were used right after that, when their viability was comparable with live tumor cells. Tumor growth was monitored with calipers for 3to 4 weeks or until the average diameter of tumors exceeded 15 mm.

These assays were performed on single-cell suspensions of lymph nodes (LN) or spleens. The LNs or spleens were transferred to different wells of a 24-well plate, containing a mixture of Collagenase D (500 μg/mL; Sigma; catalog no. 11088858001) and DNase I (500 μg/mL; Sigma; catalog no. 10104159001) in RPMI (1 mL/well). The tissues or their pieces were teased apart using 26-gauze needles and incubated in the enzyme mixture for 45 minutes at 37°C with intermittent mixing every 10 minutes. The digested tissue was further physically crushed (using the piston of a 1 mL syringe), filtered through a nylon mesh (100 μm; Falcon; catalog no. 08-771-19), and collected in 10 mL FACS buffer (PBS supplemented with 5% FBS and 5 mmol/L EDTA). CD8⁺ T cells were purified by negative immunomagnetic selection using a kit (Table 1) from pooled single-cell suspension of sdLNs and spleens of immunized mice. Isolated CD8⁺ T cells were then stimulated with live tumor cells for 12 hours. Brefeldin A (BD; BD Biosciences; catalog no. 555029 in RPMI medium) was added at a final concentration of 1 μg/mL 1 h after adding live tumor cells to the culture of CD8⁺ T cells. Cells were stained with appropriate antibodies (Table 1) for 30 minutes at 4°C in the dark before washed, acquired, and analyzed by flow cytometer (LSRII, BD Biosciences). The FACS DIVA software was used for sample acquisition, and FlowJo software (FlowJo vX.0.7) was used for analysis of flow data. Typically, 30,000 CD3⁺ CD8⁺ T cells were analyzed for each sample.

Ovalbumin-specific CD8⁺ T cells from spleen, mesenteric LNs, and sdLNs of OT-I mice were enriched as described before. Enriched OT-I CD8⁺ T cells were then labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE; Biolegend; catalog no., 422701) diluted in RPMI medium (2.5 μmol/L) and cocultured with different APCs in equal proportion (1:1). Proliferation of OT-I CD8⁺ T cells was measured 3.5 days later in terms of dilution of CFSE by flow cytometry as described below. OT-I CD8⁺ T cells were gated and identified by staining with antibodies for the congenic marker CD45.1, lineage marker CD3, and CD8.

OT-I CD8⁺ T cells were enriched from OT-I mice as described above and labeled with CFSE (5 μmol/L). The CFSE-labeled OT-I CD8⁺ T cells (10⁶) were adoptively transferred into recipient C57BL/6 mice, i.v. (through retro-orbital sinus in 100 μL volume). These mice were then injected intradermally with appropriate tumor cells, 24 hours later. Note that 3.5 days after OT-I adoptive transfer, cells were harvested from draining inguinal LNs, stained with appropriate antibodies, and analyzed by flow cytometry as described below.

The sdLNs were harvested from mice and washed with Ca⁺²- and Mg⁺²-containing HBSS in individual wells of a 24-well tissue culture plate. After thorough washing, the LNs were transferred to different wells containing a mixture of Collagenase D (500 μg/mL) and DNase I (500 μg/mL) in RPMI (1 mL/well). The LNs were minced using 26-gauze needles and incubated in the enzyme mixture for 45 minutes at 37°C with intermittent mixing of the digested tissue, every 10 minutes. The digested tissue was further physically crushed and filtered through nylon mesh (100 μm) and collected in 50 mL centrifuge tubes in 10 mL of ice-cold FACS buffer (PBS supplemented with 5%FBS and 5 mmol/L EDTA). The cell suspension was then centrifuged at 1,600 RPM for 5 minutes at 4°C, and cell pellet was suspended in 1 mL of FACS buffer and transferred to 96-well deep-well plates for antibody staining. The cells were then immediately acquired in a flow cytometer (LSR II, BD Biosciences) and analyzed later by FlowJo software (FlowJo vX.0.7).

Cells harvested from sdLNs of mice were stained with a cocktail of antibodies to label the different APC subsets as described before. The four APC subsets, i.e., CD8α⁺ DCs, CD8α⁻ DCs, plasmacytoid DCs (pDC), and CD169⁺ macrophages were identified and gated, and then sorted into FACS tubes containing 1 mL of sterile RPMI medium with 10% FBS. The sorting was done using BD FACSARIAII flow cytometer. The purity of individual-sorted populations was checked using a very small aliquot of sorted APCs and was observed to be close to 99%. Similarly, the sorting of bone marrow–derived DCs (BMDC) from a coculture of BMDCs and EG7 cells was done after labeling the BMDCs with CD11c-APC and gating them.

CMS5 and EG7 cells were labeled with CellVue Burgundy (red) using CellVue Burgundy Fluorescent Cell Labeling Kit (LI-COR, catalog no. 929-90010). The labeling was done as per the manufacturer's instructions. Briefly, 1 × 10⁷ cells/mL were stained with CellVue Burgundy dye at a final concentration of 2 μmol/L, for 5 minutes at 25°C. The labeled cells were washed thoroughly with HBSS and injected into mice. Labeled tumor cells (10⁷) were injected intradermally in 100 μL volume of HBSS, in the right flanks of individual mice. After 24 hours, sdLNs and non-draining lymph nodes (ndLN) were harvested and imaged ex vivo, using LI-COR Odyssey CLx system.

CMS5 tumor cells (10⁷ cells in 50 μL volume) were labeled with CFSE diluted in HBSS (to 5 μmol/L) and injected into the footpad of BALB/c mice. After 24 hours, popliteal LNs were surgically exposed as previously described (13). Lymphatics were visualized by using anti-Lyve1 (223322; R&D Systems) that was conjugated with Qdot655 using a Site Click Qdot 655 Antibody Labeling Kit (Life Technologies) and injected into mice intravenously (10 μg/mouse, retro-orbitally), 1 hour before imaging. A MaiTai Ti:sapphire laser (Newport/Spectra-Physics) was tuned to 920 nm for excitation of the YFP and the Qdot 655. Optical sections (512 × 512 pixels) were acquired on an Ultima IV multiphoton microscope (Prairie Technologies) to provide imaging volumes of 100 μm in depth. Imaging data were analyzed using Imaris 3D software (Bitplane, Inc.).

Femurs and tibias of 6- to 8-week-old mice were harvested aseptically. After cutting out the ends of theses bones, the bone marrow cells were flushed out with HBSS under sterile conditions. The harvested bone marrow cells were then washed thoroughly, and about 2 to 3 million cells were cultured in complete RPMI 1640 medium supplemented with 20 ng/mL recombinant murine GM-CSF (Peprotech; catalog no. 315-03) in 10 cm² bacteriological Petri dishes. Incubation was done at 37°C for 7 days to generate BMDCs. After the incubation, the nonadherent cells in the culture were harvested from the Petri dishes and used in subsequent experiments.

Exosomes were purified from culture supernatants of EG7 cells using total exosome isolation kit from Invitrogen (catalog no. 4478359), as per the manufacturer's instructions. Briefly, the EG7 culture was harvested and centrifuged. The supernatant was used for isolation of exosomes, and the EG7 cells in the pellet were counted to have an idea about the number of cells from which the isolated exosomes were derived. The exosome preparation was then subjected to appropriate dilution with sterile PBS and incubated with BMDCs for 4 hours at 37°C in a serum-free medium. Exosomes derived from approximately 1 × 10⁶ EG7 cells were incubated with 1 × 10⁶ BMDCs. After this, the BMDCs were thoroughly washed to get rid of free exosomes and used in OT-I proliferation assay.

All statistical analyses used unpaired two-tailed Student t test, and P < 0.05 was considered significant. GraphPad Prism version 6.0 was used for all statistical analysis.

The antibodies and kits used are described in Table 1.

BMDCs were washed and incubated with serum-free medium at a density of 1 × 10⁶/mL for 2 hours. EG7 tumor cells were harvested from culture, labeled with CFSE (2.5 μmol/L), and incubated with 10 μg/mL of CD47 monoclonal (B6H12) antibody (eBioscience, catalog no. 16-0479-81) or corresponding mouse isotype antibody (IgG1kappa) for 30 minutes at 37°C in complete medium. EG7 cells were then washed and incubated with serum-free medium at a density of 1 × 10⁶ cells/mL for 2 hours. Finally, 5 × 10⁴ BMDCs were cocultured with 5 × 10⁴ EG7 cells in individual wells of a 96-well tissue culture plate at 37°C. After incubation for 2 hours, the cells were washed and stained using Fc Block, CD11c, and ef-780 before being analyzed by flow cytometer.

After immunizing BALB/c mice with subtumorigenic dose of live CMS5 or Meth A tumor cells, the mice were injected with 250 μg of CD8 (Clone 2.43, Bio X Cell) or CD4 (Clone GK1.5, Bio X Cell) depletion antibodies, intraperitoneally. This administration of depletion antibodies was done 2 days before the lethal tumor challenge, and 4 and 10 days after the lethal tumor challenge, respectively.

BALB/c mice challenged with 200,000 CMS5 fibrosarcoma cells consistently displayed tumor growth in all mice; challenge with lower doses of CMS5 (100,000 cells) was consistently nontumorigenic. In order to test if mice challenged with subtumorigenic doses of CMS5 cells develop immunity to a subsequent lethal challenge, BALB/c mice were primarily challenged with 100,000, 50,000, 25,000, or 20,000 CMS5 cells. Fourteen days after the first tumor challenge, no tumors formed in these mice. All mice were now challenged second time, with a lethal dose of 200,000 CMS5 cells. Mice initially challenged with as few as 25,000 live CMS5 cells were resistant to subsequent lethal challenges with 200,000 CMS5 cells (Fig. 1A). In similar experiments with the Meth A sarcoma, as few as 15,000 live tumor cells elicited tumor protection (Supplementary Fig. S1A and S1B). In the EG7 lymphoma of C57BL/6 mice, as few as 5,000 live cells elicited tumor immunity (Supplementary Fig. S2A and S2B). These very small numbers of live cells that elicited tumor immunity in different tumor models of different haplotypes were surprising as immunization with as many as 25 × 10⁶ IR tumor cells is typically required to elicit immunity to lethal tumor challenges, as observed by us previously in Meth A and CMS5 (14), and as was observed here in the EG7 model (Supplementary Fig. S2C). We also performed experiments in parallel comparing live and IR cells in the three tumor models, and consistently observed that small numbers of live but not IR CMS5, Meth A (Supplementary Fig. S1C and S1D), or EG7 cells (Supplementary Fig. S2A–S2C) elicited antitumor immunity.

The simplest explanation of these findings is that initial challenge with subtumorigenic doses of tumor cells leads to their progressive replication in vivo, such that the initial subtumorigenic challenge dose of live cells multiplies several folds before it is eliminated by the developing antitumor immune response. Although the initial tumor inoculum is small, the effective inoculum in vivo might be much higher, closer to the 25 × 10⁶ IR cells, to get robust tumor immunity. In order to test this possibility, we monitored the multiplication of live tumor cells in vivo using in vivo imaging. BALB/c mice were injected with live CMS5 cells; 2 days prior to imaging, mice were administered with fluorescent 2-deoxy glucose (FDG) intravenously, which preferentially accumulates in rapidly dividing tumor cells (Fig. 1B, inset). FDG accumulation increased with time in mice injected with 200,000 live CMS5 cells, at the site of progressively growing tumors; however, in mice injected with 25,000 live CMS5 cells, FDG accumulated to a modest degree till day 8 after which it declined (Fig. 1B). Thus, subtumorigenic doses of tumor cells are not progressively dividing in vivo. As a control, freshly MC-treated live tumor cells were injected, which are incapable of proliferation, and failed to accumulate FDG (Fig. 1B).

Because subtumorigenic doses of live cells proliferate to a limited degree (Fig. 1B), we tested the immunogenicity of live tumor cells which were incapable of proliferation. Mice (n = 10 per group) were therefore injected with 25,000 live replicating or the same number of live freshly MC-treated nonreplicating CMS5 cells; as before, no tumors (0/10) developed in mice in either group after 14 days. All mice were now challenged with 200,000 live CMS5 cells. Surprisingly, even mice immunized with MC-treated cells showed complete resistance to a lethal tumor challenge (Fig. 1C). The influence of proliferation of tumor cells at the site of inoculation was tested further, using inoculation-site-excision studies. Excision of the initial tumor challenge at 6 hours or earlier leads to complete loss of immunity; on the other hand, excision at 24 hours did not lead to loss of tumor immunity (Fig. 1D). As seen in Fig. 1B, tumor cells underwent very little local proliferation within 24 hours; thus, division and persistence of live tumor cells at the site of inoculation do not appear to be critical for establishing antitumor immunity.

Superior immunogenicity of live tumor cells was also observed in assays of CD8⁺ T-cell activity in vivo. Live CMS5 cells (25,000) elicited a measurable and potent CD8⁺, CD44^high, IFNγ⁺ T-cell response, which was significantly higher than that elicited by equivalent dose of IR CMS5 cells (Fig. 1E). CD8 cells were gated as shown in Supplementary Fig. S2F and analyzed further for CD44^high, IFNγ⁺ as shown in Fig. 1E. Higher immunogenicity of live tumor cells was also seen using EG7 cells in an OT-I cell proliferation assay (Supplementary Fig. S2D and S2E).

The characteristics of antitumor immunity elicited by live tumor cells were dissected. By immunizing and cross-challenging BALB/c mice with Meth A and CMS5 tumor cells, immunity elicited by live tumor cells is tumor-specific (Supplementary Fig. S3A and S3B). After immunizing the BALB/c mice with single subtumorigenic dose of live CMS5 (Supplementary Fig. S3C) or live Meth A (Supplementary Fig. S3D) tumor cells and subjecting them to corresponding lethal challenge different days post immunization, antitumor immunity was long lasting (lasts greater than 90 days). The antitumor immunity elicited by live tumor cells was amenable to adoptive transfer after 5 days of immunization (Supplementary Fig. S3E and S3F). To examine the role of CD4⁺ and CD8⁺ effector T cells in this antitumor immunity elicited by live tumor cells, we depleted CD4⁺ or CD8⁺ T cells in vivo, which showed a dependency on CD8⁺ T cells (Supplementary Fig. S3G and S3H). Thus, immunity elicited by subtumorigenic doses of live tumor cells was long lasting, tumor-specific, and CD8⁺ T-cell–dependent (Supplementary Fig. S3).

PKH26, a lipophilic membrane labeling dye, was used to track uptake of tumor cell–associated antigens by APCs at sdLNs. C57BL/6 mice were injected intradermally with fluorescent (PKH26-labeled) live or IR EG7 cells; 24 to 36 hours later, different APC subsets [i.e., CD8α⁺ DCs, CD8α⁻ DCs, B220⁺ pDCs, and CD169⁺ macrophages) from sdLNs were stained using the appropriate antibodies (Table 1) and analyzed by flow cytometry for uptake of PKH26-labeled membrane components. The gating strategy for these APCs is shown in Supplementary Figs. S4 and S5. We observed that CD169⁺ macrophages, the major APC subset that takes up IR tumor cell–associated antigens (2), did not take up labeled membrane components from live tumor cells but did take up antigen from IR tumor cells (Fig. 2A). However, CD8α⁺ DCs acquired PKH26 from both live and IR tumor cells, almost equally and significantly (Fig. 2A). Other APC subsets (i.e., CD8α⁻ DCs and pDCs) showed very little uptake of PKH26 from either live or IR cells (Fig. 2A). These same APC subsets, FACS-sorted from mice immunized with live or IR EG7 cells, were also tested for their ability to stimulate proliferation of naïve OT-I cells ex vivo; consistent with the data on uptake of PKH26, only CD8α⁺ DCs stimulated OT-I cells (Fig. 2B) following immunization with live EG7 cells. In the case of IR cell immunization, both CD169⁺ macrophages and CD8α⁺ DCs were the predominant APC subsets capable of stimulating OT-I cells (Fig. 2B).

Transport of subcutaneously injected IR tumor cells to draining lymph nodes (dLN) through lymphatics has been demonstrated (2). We wanted to determine the mode of transportation of live tumor cell–associated antigens from the site of tumor challenge to the APCs in sdLNs. BALB/c mice were injected intradermally on the right flank with live or IR 10⁷ CMS5 cells (in 100 μL volume) labeled with CellVue-Burgundy (LI-COR), a lipophilic membrane labeling dye. C57BL/6 mice were similarly injected with live or IR 10⁷ EG7 cells labeled with the same dye. After 24 hours, dLNs and ndLNs were harvested and imaged ex vivo. Only dLNs were observed to take up CellVue Burgundy (Fig. 2C). In a similar experiment, CFSE-labeled live 10⁷ CMS5 tumor cells were injected in the footpad of BALB/c mice, and draining popliteal LNs were surgically exposed after 24 hours and subsequently monitored by intravital 2-photon microscopy. We observed accumulation of CFSE-labeled cells in the dLNs (Fig. 2D and Supplementary Video S1). However, accumulation of dyes or dye-labeled cells at dLNs did not necessarily indicate the presence of tumor cells in the dLN. In order to test if tumor cells were actually reaching the dLNs, we made use of CD45.1⁺ congenic C57BL/6 mice, which were injected with unlabeled EG7 tumor cells (CD45.2⁺). Twenty-four hours after injection, single-cell suspensions of dLNs were cultured in vitro (in complete medium for EG7 cells) for 6 days. Ten percent of the culture was harvested on days 0, 2, 4, and 6 and analyzed by flow cytometry for the presence of live CD45.2⁺ (EG7) cells. A time-dependent increase of live EG7 cells was observed in the cultures (Fig. 2E; Supplementary Fig. S6A). Thus, the successful recovery and propagation of live tumor cells from sdLNs of mice injected with live tumor cells suggested that intact live tumor cells find their way to the sdLNs through lymphatics. In another approach, PKH26-labeled CD45.1⁻ EG7 tumor cells were injected to CD45.1⁺ mice, and single-cell suspension of the sdLNs was analyzed by flow cytometry after 24 hours. The ndLNs had negligible live CD45.1⁻ PKH26⁺ cells, whereas the dLNs had a significantly higher proportion of these cells among the CD45.1⁻ cell population (Supplementary Fig. S6B), supporting our earlier observation regarding the arrival of intact live tumor cells at sdLNs.

The mechanism of transfer of antigen from live tumor cells to APCs was evaluated in vitro and in vivo. In vitro, EG7 cells (live or IR) were stained with either CFSE (a dye labeling cytosolic proteins) or PKH26(a dye labeling membrane). The fluorescently labeled EG7 cells were then cocultured with BMDCs from C57BL/6 mice for 12 hours, and the BMDCs were then analyzed by flow cytometry for acquisition of CFSE or PKH26 from labeled tumor cells. BMDCs cultured with live tumor cells acquired significant quantities of PKH26 but little CFSE (Fig. 3A), suggesting that membrane components but not the cytosolic components of live tumor cells were accessible to APCs. In contrast, BMDCs cultured with IR tumor cells acquired significant quantities of both dyes (Fig. 3A).

In vivo, C57BL/6 mice were injected with live or IR EG7 tumor cells, stained with CFSE or PKH26. Since our previous observations (Fig. 2A and B) showed that CD8α⁺ DCs were the only APCs which could take up antigens from both live and IR tumor cells at sdLNs, this APC subset was analyzed for the uptake of CFSE/PKH26 from the injected tumor cells. Consistent with the data in vitro, CD8α⁺ DCs readily picked up PKH26 but not CFSE from live tumor cells (Fig. 3B). However, both CFSE and PKH26 were taken up from IR tumor cells by CD8α⁺ DCs. These results suggested that live tumor cells were resistant to be phagocytosed (6, 15, 16) but may transfer membranes to APCs by other mechanisms including trogocytosis (17–20), exosomes (21–23), or tunneling nanotubes (23–25). Consistent with this interpretation, analysis of surface expression of CD47, a “don't-eat-me-signal,” revealed that live murine tumor cells overexpressed CD47 compared with normal tissues (Fig. 3C). Similar observations were made with multiple patient-derived solid tumors (26). Blocking of CD47 with a monoclonal antibody enhanced the degree of phagocytosis of live tumor cells by BMDCs (Fig. 3D) without inducing apoptosis. We attributed the modest degree of enhancement to the fact that the live tumor cells did not express “eat-me” signals such as phosphatidyl serine.

Phagocytosis of apoptotic/dead cells and cross-presentation of their cell-associated antigens are well documented (1, 2, 4, 6, 27), though the form of antigen that constitutes the source for cross-priming is under debate (3, 28). The primary physiologic form of antigen that gets transferred from live tumor cells to the APCs was examined. BALB/c BMDCs (H-2^d) were cocultured with equal numbers of live or IR EG7 cells (K^b) as antigen donor cells for 12 hours. BMDCs were then analyzed for the presence of SIINFEKL-K^b complexes, by staining with the 25-D1.16 antibody: BALB/c BMDCs incubated with live EG7 cells had significantly higher expression of SIINFEKL-K^b than those incubated with IR EG7 tumor cells (Fig. 4A), suggesting the transfer of SIINFEKL-K^b complexes from EG7 cells to the BALB/c BMDCs by cross-dressing. Even though equal numbers of live and IR cells were used in this assay, a close examination of Fig. 4A showed that the number of IR EG7 cells (CD11c⁻SIINFEKL-Kb⁺) was lower than the number of live EG7 cells; this appearance was misleading because of the exclusion of nonviable (ef 780 viability dye⁺) cells during flow cytometric analysis.

We then measured the transfer of SIINFEKL-K^b complexes through cross-dressing by culturing BALB/c BMDCs with live or IR EG7 cells, then FACS sorted the BMDCs, and used them to stimulate OT-I cells. Consistent with the observations in Fig. 4A, significantly higher expansion of OT-I cells was observed with sorted BALB/c BMDCs precultured with live EG7 cells compared with IR EG7 cells (Fig. 4B). Although these data suggested that BMDCs were being cross-dressed with SIINFEKL-K^b complexes from tumor cells, there was the possibility that OT-I proliferation was driven by a small number of contaminating EG7 cells in the FACS-sorted BMDC population. EG7 cells constituted about 0.1% of the sorted BMDC population (Fig. 4D). Using this number as a benchmark, we performed the stimulation of OT-I cells by artificial mixtures of BMDCs and 100 to 500 times more (i.e., 10%–50%) contaminating EG7 cells. We observed that only at a contamination of 500 times more EG7 cells than present in the FACS-sorted BMDC populations in Fig. 4D stimulated equivalent OT-I proliferation (Fig. 4C). The issue of contamination of EG7 cells as an explanation for apparent cross-dressing did not arise in Fig. 4A because the cross-dressed CD11c⁺ SIINFEKL-K^b+ BMDC population was exclusively gated in this experiment. As an additional control, fixing the BMDCs using paraformaldehyde before incubation with EG7 tumor cells abrogated the transfer of SINNFEKL-K^b complexes from EG7 cells by both the above assays.

Cross-dressing of APCs by live tumor cells was also evaluated in vivo. The H-2K^bm1 mouse harbors a mutation in K^b that allows the mutant K^bm1 to bind SIINFEKL but prevents the resulting SIINFEKL-K^bm1 complexes from being recognized by the TCR of CD8⁺ T cells due to an alteration in the region of K^bm1 that interacts with the TCR (29). This mouse was used with the reasoning that endogenous APCs of H-2K^bm1 mice cannot stimulate SIINFEKL-K^b–specific CD8⁺ T cells (OT-I cells), unless they acquire the SIINFEKL-K^b complexes from donor cells through cross-dressing. H-2K^bm1 mice were immunized with live or IR, SIINFEKL-pulsed EL4 (K^b) cells; 36 hours later, CD8α⁺ DCs, CD8α⁻ DCs, and CD169⁺ macrophages were sorted from pooled single-cell suspension of sdLNs and used to stimulate naïve OT-I cells. Sorted CD8α⁺ DCs from mice immunized with live cells induced significant expansion of naïve OT-I cells, whereas CD8α⁻ DCs and CD169⁺ macrophages did not (Fig. 4E; Supplementary Fig. S7). CD8α⁺ DCs from mice immunized with IR cells did not induce OT-I expansion (Fig. 4F; Supplementary Fig. S7). These results suggested that CD8α⁺ DCs were cross-dressed with SIINFEKL-K^b complexes from live tumor cells. Figure 4E and F minimizes the possibility that the OT-I cells were being stimulated directly by live cells: the live or IR cells were used to immunize mice, and 36 hours later, the various APCs were FACS-sorted into individual populations and used to stimulate OT-I cells. The contamination, if any, of live tumor cells in these FACS-sorted populations would be identical; yet, only the CD8α⁺ DC population from the live tumor cell–injected group stimulated the proliferation of OT-I cells. We concluded from these three independent and well-controlled experiments that cross-dressing of DCs was a major mechanism of presentation of live tumor cell–associated antigens.

The in vitro cross-dressing experimental system described in Fig. 4A and B was used to determine if cross-dressing of BMDC required physical contact between the BMDCs and tumor cells or it could be mediated by exosomes. BMDCs were precultured with live or IR cells, followed by sorting of BMDCs and their staining for presentation of SIINFEKL-K^b complexes or use in OT-I proliferation assay. In both assays, physical contact between BMDC and the tumor cells was required, because the separation of BMDCs from tumor cells by a Transwell membrane (1 μm pore size) abrogated the transfer of SIINFEKL-K^b (Fig. 5A and B). Exosomes or such other vesicles (which are 30–100 nm in size and hence could pass through the Transwell membrane) were not mediating the transfer of SIINFEKL-K^b complexes.

To examine the contribution of exosomes more definitively, BMDCs of BALB/c or C57BL/6 mice were precultured with exosomes derived from EG7 cells, washed thoroughly, and subsequently used to stimulate naïve OT-I cells. When purified exosomes from culture supernatants of EG7 cells were used to stimulate naïve OT-I cells directly, significant proliferation of OT-I cells was observed indicating the presence of SIINFEKL-K^b complexes on surface of exosomes (Fig. 5C, bottom left plot). Such stimulation of CD8⁺ T cells directly by exosomes has been demonstrated (30). However, BALB/c BMDCs precultured with above exosomes failed to stimulate OT-I cells (Fig. 5C, bottom middle plot), suggesting that SIINFEKL-K^b complexes present on EG7 cell–derived exosomes were not directly transferred to the surface of BALB/c BMDCs. On the contrary, and as a positive control, C57BL/6 BMDCs precultured with the same exosomes stimulated OT-I cells (Fig. 5C, bottom right plot).

The role of CD8α⁺ DCs in priming of antitumor immunity following immunization with live/IR tumor cells was further evaluated by using Batf3^−/− mice, which lack CD8α⁺ DCs in LNs and spleen (Supplementary Fig. S8A and S8B). Wild-type and Batf3^−/− mice (BALB/c origin) were injected intradermally with 25,000 live CMS5 cells or 25 × 10⁶ IR CMS5 cells (Fig. 6A). Some of the Batf3^−/− mice immunized with 25,000 live CMS5 cells developed tumors, but all other mice were tumor-free. Two weeks after immunization, all the tumor-free wild-type and Batf3^−/− mice were challenged with 200,000 CMS5 cells. Wild-type mice, but not the Batf3^−/− mice, were protected from tumor development (Fig. 6B). The failure of Batf3^−/− mice to establish antitumor immunity following immunization with live or IR tumor cells confirmed the role of CD8α⁺ DCs in antitumor immunity. This result was consistent with our previous observation on antigen uptake and presentation from live and IR tumor cells (Fig. 2A and B) and in vivo cross-dressing experiment (Fig. 4E and F; Supplementary Fig. S7).

Apoptotic and necrotic cells have been extensively studied as sources of cellular antigens for APCs. However, little is known about live cells as antigen donor cells (ADC). Here, we showed that live tumor cells were far more immunogenic than dying tumor cells. We identified two significant points of divergence between the mechanisms of ADC–APC interaction between live and apoptotic cells. Firstly, CD169⁺ macrophages, which readily acquire antigens from apoptotic cells (2), did not do so from live cells presumably because of the absence of eat-me signals (e.g., phosphatidyl serine); CD8α⁺ DCs did not show any such discrimination. These observations were consistent with the reported inability of macrophages, but not DCs, to extract cell-associated antigens from live cells (7). Secondly, phagocytosis and cross-presentation are the primary mechanisms underlying antigen presentation from apoptotic cells (2), whereas here trogocytosis and cross-dressing of MHCI-bound peptides present on living cell surface were identified as the dominant mechanisms used by APCs for acquisition and presentation of live tumor cell–associated antigens. The findings of this study also highlighted that even though the APCs took up more antigen (cytosolic and membrane) from apoptotic cells (primarily through phagocytosis), this did not translate to presentation of more antigen. The antigen uptake from live cells, mediated by trogocytosis, appeared to be the more efficient mechanism.

Our observations do not completely address the mechanisms of disparity of immunogenicity between live and IR cells, but they do offer a concrete suggestion. Apoptotic cell–associated antigens were taken up through phagocytosis by CD8α⁺ DCs and CD169⁺ macrophage alike, whereas live cell–associated antigens were taken up by trogocytosis mostly by CD8α⁺ DCs. Phagocytosis of apoptotic cells skews APCs (i.e., monocytes, macrophages, and DCs) away from inducing immunity toward the induction of immunosuppression or tolerance (16, 31, 32). Thus, much of the antigenic stimulus provided by large doses of apoptotic cells is abrogated by the tolerance elicited by their phagocytic uptake. This idea is strengthened by the previous observation that due to extensive degradation of antigen in IR or apoptotic, but not live cells, significantly more apoptotic cells than live cells are required for DCs to achieve comparable MHCI-peptide presentation (11); our data are consistent with this finding. In addition, parallel analyses of the uptake of dying and live tumor cells showed that unlike IR or apoptotic tumor cells, the membrane-associated antigens of live tumor cells were efficiently cross-dressed by CD8α⁺ DCs.

Trogocytosis, a contact-dependent mechanism of antigen transfer between interacting cells (33, 34), was an additional mechanism of transfer of antigen from live tumor cells to APCs. This was likely the dominant mechanism of antigen transfer from live tumor cells physiologically in light of syngeneic MHC I expression in most tumor cells. In an elegant study, Puaux and colleagues (17) use multiple membrane labeling lipophilic dyes and dyes labeling cytosolic proteins to study trogocytosis. They concluded that membrane markers, but not the cytosolic markers, allow for the efficient detection of trogocytosis. Accordingly, we used membrane labeling dye PKH26 and cytosolic labeling dye CFSE, which showed that there was a significant transfer of membrane but not the cytosolic components from live tumor cells to APCs. Uptake of PKH26 by itself is not a marker of donation of membranes by trogocytosis, yet it is the uptake of PKH26 in the relative absence of uptake of cytosolic materials (stained by CFSE) that indicates trogocytosis. We also observed that the transfer of membrane-associated antigens from live tumor cells to APCs required physical contact between the donor and recipient cells, which was not mediated by exosomes. Another possibility was that the transfer of antigens occurs through actin-based membranous intercellular bridges called tunneling nano tubes (TNT) which mediate transfer of organelles, plasma membrane components, and cytoplasmic molecules between the cytosol of adjacent cells. These TNTs are 50 to 1,000 nm in diameter and can extend up to 200 μm (23–25). However, the complete abrogation of antigen transfer in our transwell experiments suggested that TNTs have little role in the antigen transfer from live tumor cells to APCs. Together, these findings suggested that the antigen transfer from live tumor cells to APCs primarily involved trogocytosis. Antigen presentation by cross-dressing (that primarily involves antigen transfer through trogocytosis) merits discussion vis-à-vis other reported mechanisms of antigen presentation. MHC I–deficient (34) and MHC I disparate cells (35) can cross-prime antigen effectively, and trogocytosis is not the mechanism of antigen transfer in those systems. Other mechanisms such as heat shock protein–peptide complexes (3) and transfer of intact antigen (28) mediate cross-priming under those circumstances.

Cross-dressing has been demonstrated in vitro (36–39). In a viral system, Wakim and Bevan show cross-dressing of splenic CD11c⁺ cells with viral peptide–MHC complexes, and the ability of such cross-dressed DCs to activate memory, but not naïve, CD8⁺ T cells (30). In their study, splenic CD8α⁻ DCs are more efficient in cross-dressing of viral antigens as opposed to CD8α⁺DCs. Li and colleagues demonstrate that cross-dressed CD8α⁺/CD103⁺ DCs can effectively prime CD8⁺ T cells following vaccination with DNA or a mixture (1:1) of live and dead genetically modified mouse embryonic fibroblasts (40). However, here we validate higher immunogenicity of live tumor cells and demonstrate presentation of their cell-associated antigens by CD8α⁺ DCs at sdLN through cross-dressing, leading to CD8-mediated antitumor immunity. These findings have important implication for DC-based immunotherapy, where autologous live tumor cells can be exploited as potential source of antigen for DCs.

No potential conflicts of interest were disclosed.

A. Das Mohapatra: Formal analysis, supervision, investigation, methodology, writing–original draft, writing–review and editing. I. Tirrell: Investigation, methodology, project administration. A.P. Bénéchet: Investigation. S. Pattnayak: Investigation. K.M. Khanna: Conceptualization, supervision, investigation. P.K. Srivastava: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

The authors thank Dr. Evan R. Jellison for help with FACS sorting.

This research was funded by the Neag Cancer Immunology Translational Program (P.K. Srivastava), Northeastern Utilities Chair in Experimental Oncology (P.K. Srivastava), and the Personalized Immunotherapy Core Interest Group of the Connecticut Institute for Clinical and Translational Science (P.K. Srivastava).

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.