Modeling the Specific CD4⁺ T Cell Response against a Tumor Neoantigen

The genetic instability of tumors gives rise to multiple neoantigens that can potentially generate an immune response. The appearance of tumors is therefore surprising in immunocompetent individuals. It was proposed that tumors are initially ignored by the immune system (1). Only big tumors would be seen at a stage where the tumor mass would overwhelm the immune response leading to deletion, suppression of or an inadequate class (Th2 versus Th1 for instance). Accordingly, the frequency and titers of antitumor Abs found in the serums of cancer patients is proportional to the size of the tumors (2). Another school of thought based on the study of chemically induced tumor models proposes that tumor cells bearing neoantigens are rejected very early on during tumor development in a process called tumor editing (3, 4). This would be followed by an equilibrium phase during which the immune system keeps in check the growing tumors as suggested by the good prognosis significance of CD8⁺ T cell infiltrates in colon tumors (5). Tumor development would be related to the occurrence of Ag loss variants escaping the immune response.

Neoantigens are ideal targets for therapeutic purposes as there is no risk of autoimmunity if an efficient antitumor response is generated (6). Most emphasis so far has been put on CD8⁺ T cells because they can be directly cytotoxic to tumor cells and appear to be the main effectors in many models. Because most tumor cells do not express MHC class II (MHC-II) molecules, CD4⁺ T cell implication in antitumor responses has only recently been recognized. Indeed, CD4⁺ T cells can be more efficient than CD8⁺ T cells at rejecting solid tumors (7, 8). In humans, transfer of autologous-expanded tumor-specific CD4⁺ T cells induced complete remissions (9, 10). We recently evidenced that chronically activated effector CD4⁺ T cells are increased in the blood of cancer patients, and this expansion is correlated with tumor regression during neoadjuvant chemotherapy of breast cancer (11). Moreover, the presence of a similar CXCL13 secreting CD4⁺ T cell subset in breast (12) and colon (13) tumors is correlated with a better prognosis.

Because human studies are only correlative, animal models are used for mechanistic studies and to infer causal relationships. The specific antitumor T cell responses have mostly been studied using transplantable tumor cell lines that can express a model or tumor Ag for which TCR transgenic T cells are available (1, 7, 8, 14–18). Tumor growth is usually fast and kills the animal in a few weeks. Moreover, many cells die at the time of surgical implantation. This leads to the release of the nominal Ag, which is captured by dendritic cells (DCs) in an inflammatory context. The Ag-loaded DCs migrate to the tumor draining lymph node (TdLN) and may stimulate an immune response. Thus, transplantable tumor models do not reproduce the slow and noninflammatory growth of human tumors. In these models, it is difficult to determine whether a neoantigen exclusively expressed by growing tumors at “steady state” would generate an immune response and, if so, what would be the nature of this response.

One way to avoid this caveat is to wait a few days after tumor cell seeding before transferring naive tumor Ag-specific T cells. Nonetheless, although this might allow the study of the CD8⁺ T cell response as the MHC-I:peptide complexes are short lived, this is not sufficient for CD4⁺ T cells because the lifespan of MHC-II:peptide complexes can be above 3 wk (19, 20) as we will also show in this study in a tumor context. To overcome this difficulty, a cell line expressing a MHC-I epitope fused to GFP at low level could be induced to high expression upon in vitro Cre-LoxP–mediated recombination (21). The low Ag-expressing clone did not stimulate an immune response, which was only observed for the high Ag-expressing cells in a process requiring MHC-matched stromal cells. In vivo induction of Ag expression in a similar system led to anti-GFP Abs of the IgG isotype that also depended on the CD4⁺ T cells (22). To our knowledge, there is no detailed analysis of the antitumor CD4⁺ T cell response in a transplanted tumor model in which Ag expression is induced in established tumors.

In this paper, using a transplantable tumor constitutively expressing the MHC-II–restricted DBY model Ag, we show that the way the cells die at the time of tumor seeding strongly impacts the Ag lifespan. To overcome this caveat, which prevented us to study the relationship between the immune system and tumors at steady state, we generated and validated a tumor cell line expressing the DBY Ag in an inducible manner. We compared the intensity and quality of the immune response of naive CD4⁺ T cells against tumor cells expressing the Ag at the time of or 7 d after implantation of the tumor. We show that the tumors are not ignored, and we did not find any evidence for tolerance or changes in the class of the immune response.

CD45.1⁺Rag2^−/− TCR transgenic Marilyn female mice (23) were used as a source of monoclonal-specific CD4⁺ T cells recognizing the I-A^b–restricted male DBY peptide (NAGFNSNRANSSRSS) with a Vα1.1/Vβ6 TCR. Female CD45.2⁺Rag2^+/+ or Rag2^−/− C57BL/6 (B6) mice were used as recipients. Spleen cells from male CD3ε^−/− B6 mice were used as a positive control in Fig. 5, whereas the cells from female mice were used as a source of APC during ex vivo stimulation in Fig. 7. Marilyn and Rag2^−/− B6 mice were bred at the Centre d’Exploration et de Recherche Fonctionnelle Expérimentale (CERFE-Genopole, Evry, France). Wt B6 mice were purchased from Charles River Laboratories (L’Abresles, France). CD3ε^−/− mice were bred in-house at Institut Curie. Live animal experiments were carried out in accordance with the guidelines of the French Veterinary Department.

The methylcholanthrene-induced fibrosarcoma MCA101 and the stable transfectant MCA-DBY cell lines were described previously (24, 25). Cells were cultured in DMEM-GlutaMAX medium (Life Technologies) supplemented with 10% heat-inactivated FCS, 100 U/ml penicillin, 100 pg/ml streptomycin, 10 mM HEPES, 1 mM sodium pyruvate, and 50 μM 2-ME. For MCA-DBY cells, complete medium was supplemented with 0.5 mg/ml geneticin (Life Technologies).

The MCA101-LSL^DBY clone was obtained by transfection of MCA101 cells with the pcDNA3.1-LSL^DBY vector followed by double cloning by limiting dilution. The outline of the plasmids is shown in Supplemental Fig. 1. To generate pcDNA3.1-LSL^DBY, the neomycin resistance gene flanked by two LoxP sites (LoxP-stop-Neo^R-LoxP) was inserted into the DBY cDNA sequence by overlapping PCR. A stop codon was inserted just after the first LoxP sequence upstream of the Neo^R cassette to prevent translation of the DBY epitope. MCA101-LSL^DBY cells were then infected with lentiviral particles encoding the tet-inducible Cre-ER^T2. Starting with the pLVUT-tTR-KRAB-GFP plasmid (Addgene) (26), the GFP coding sequence was replaced by Cre-ER^T2 to generate pLVUT-tTR-KRAB-Cre-ER^T2. Transduced cells (MCA101-LSL^DBY-Tet-On: Cre-ER^T2) were cloned to obtain the H4B1 clone. H4B1 cells were cultured using Tet System Approved FBS (Takara Clontech). To induce DBY expression in vitro, H4B1 cells were treated with 10 μg/ml doxycycline (Dox; Takara Clontech) and 1 mM 4-hydroxytamoxifen (4-OHTam; Sigma-Aldrich) for 4 d. Recombination was detected by PCR using DBY-specific primer (5′-tgcagattcgctggaggacttctta-3′; 5′-tgcgttatcgattcaattgccccaccagtcaac-3′) (Eurogentec).

T cells were obtained from peripheral lymph nodes (LNs) of Marilyn mice. After labeling with CFSE (Life Technologies) at 5 μM for 8 min at 37°C, 5 × 10⁵ naive T cells were injected into the tail vein in 100 μl PBS 0.5% BSA.

MCA-DBY cells were treated with different cell death inducers: 600 μM mitomycin C (Sanofi Aventis) for 2 h, 30 μM doxorubicin for 24 h, gamma rays (100 Gy), UV light (150 mJ/cm²; UVITec Crosslinker), or six freeze/thaw cycles. The percentage of apoptotic cells was assessed using Annexin V-FITC Apoptosis Detection kit (BD Biosciences).

Tumor cells (2 × 10⁵ unless specified otherwise in 100 μl PBS 0.5% BSA) were injected s.c. into the flank of wild-type (wt) or Rag2^−/− CD45.2⁺ mice. Tumor size was assessed twice a week using the formula length × width × width/2. H4B1 cells were previously cultured or not in presence of Dox/4-OHTam, and injected s.c. into B6 mice. Mice with in vivo and in vitro Ag induction received the same treatment (Dox in drinking water and tamoxifen (Tam) from Sigma-Aldrich, i.p.) to control for a direct effect of the treatment on the tumor or host T cells. In vivo induction protocol is described in Supplemental Fig. 3C. Mice were sacrificed at day 7 or 15 posttransfer.

Tumors were mechanically disrupted and resuspended in CO₂ Independent Medium (Life Technologies) containing 10% FCS with 1 mg/ml collagenase D and 250 μg/ml DNase I (Roche) for 90 min at 37°C. After filtration through a 40-μm cell strainer, tumor-infiltrating lymphocytes (TILs) were isolated on a lympholyte M (Cedarlane, Ontario) gradient.

Cell viability was assessed using DAPI (Roche) or Live/Dead Fixable Violet or Aqua Dead Cell Stain Kit (Life Technologies). Cells were preincubated with anti-FcR Ab (2.4G2 hybridoma supernatant) before staining. The following Abs from eBioscience, BioLegend, or BD Biosciences were used: PerCP-eFluor710-anti-LAG3 (C9B7W); PerCP-Cy5.5-anti-CD160 (7H1); PE-anti-Vβ6 (RR4-7); PE-anti-ICOS (15F9); PE-anti–LAMP-1 (1D4B); PE-anti–CTLA-4 (UC10-4B9); PE-Cy5-anti-CD45.1 (A20); PE-Cy5-anti-βTCR (H57-597); biotin-anti-2B4 (m2B4); PE-Cy7-anti-PD1 (29F.1A12); allophycocyanin-anti–Tim-3; Alexa-Fluor647-anti-BTLA (6A6); PE-Cy7-streptavidin; Alexa Fluor 700-anti-CD44 (IM7); allophycocyanin-Cy7-anti-CD45.2 (104); BV421-anti-βTCR (H57-597); BV570-anti-CD45.2 (104); BV605-anti-CD62L (MEL-14); and BV605-anti-CD45.1 (A20); BV785-anti-CD4 (RM4-5). The cells were incubated for 20 min at 4°C and washed with PBS 0.5% BSA. Intracellular staining was performed using eF450-anti-Foxp3 (FJK-16), PE-Cy7- or BV711-anti–IFN-γ (XMG1.2) (XMG1.2), Alexa Fluor 700-anti–IL-17 (TC11-18H10.1), Alexa Fluor 647-anti-granzyme B (GB11), and eFluor450-anti–TNF-α (MP-6-XT22) using the Cytofix/Cytoperm kit (BD Biosciences). FACS data were acquired using a Fortessa or LSR II cytometer (BD Biosciences) and analyzed using FlowJo (Tree Star).

Cells from the TdLN were restimulated at 1.5 × 10⁶/ml for 5 h at 37°C in 5% CO₂ in complete RPMI 1640 medium with 10 ng/ml PMA and 250 ng/ml ionomycin (Sigma-Aldrich) and GolgiStop (BD Biosciences) for the last 4 h. Alternatively, TdLN, spleen, and TILs were resuspended at 2 × 10⁶ cells/200 μl and restimulated for 16 h with DBY peptide (30 μM) or anti-CD3 + anti-CD28 beads (Miltenyi Biotec) and an anti–LAMP-1 Ab as indicated. GolgiStop was added during the last 4 h.

Cytokine secretion was also measured after restimulation with DBY peptide (30 μM) for 48 h followed by a Cytometric Bead Array (BD Biosciences) of the supernatant.

Mice were injected i.p. twice a day with 1 mg BrdU for 5 d (Sigma-Aldrich). Cells were stained for extracellular markers before fixation overnight with Cytofix/Cytoperm Buffer (BD Biosciences). Cells were then incubated for 30 min in PBS 1% PFA and 0.5% Tween 20, washed with Perm/Wash Buffer (BD Biosciences), and refixed for 5 min. Incorporated BrdU was exposed by treatment of cells with 500 U/ml DNase I (Roche) for 1 h at 37°C. Intracellular staining was performed with an anti-BrdU allophycocyanin-conjugated Ab (Bu20A; eBioscience).

All quantitative data were analyzed on Prism software using unpaired or paired nonparametric tests (Mann–Whitney U or Wilcoxon signed rank) where indicated: *p < 0.05, **p < 0.01, and ***p < 0.001.

We set out to study the CD4⁺ T cell response against a growing tumor expressing a nominal MHC-II–restricted Ag. To avoid the artificial priming by the Ag released by the cells that unavoidably die at the time of implantation, we considered delaying the adoptive transfer of Ag-specific T cells after tumor seeding. We therefore measured how long last Ag presentation after implantation of dying cells, using the proliferation of Ag-specific CD4⁺ T cells transferred after tumor cell challenge as readout. For that purpose we used the MCA101 (methylcholanthrene-induced) fibrosarcoma cell line (24), transfected or not with a plasmid encoding the MHC-II–restricted DBY male Ag (MCA-DBY) (25). These cells do not express MHC-II molecules in vitro with or without IFN-γ treatment (Ref. 25 and data not shown) or ex vivo (25). Untreated or mitomycin-treated MCA-DBY cells were injected s.c. into B6 mice. Naive CFSE-labeled CD45.1⁺Marilyn CD4⁺ T cells that recognize the DBY epitope were adoptively transferred at different time points and analyzed 5 d later in the dLN (Fig. 1A). The mitomycin-treated cells became apoptotic and died in a few days (Supplemental Fig. 2). The DBY⁺ tumors grew like the parental cell line, whereas the mitomycin-treated cells did not generate any tumor (Fig. 1B). Marilyn T cells transferred 8 d after tumor injection proliferated in mice challenged with living MCA-DBY cells (Fig. 1C, 1D). Surprisingly, injection of mitomycin-treated MCA-DBY cells also induced proliferation of Marilyn cells, although to a lesser extent than living cells, over an extended period on time (Fig. 1C, 1D).

To determine whether this was not specific of mitomycin-induced cell death, we treated MCA-DBY cells with several chemical or physical cell-death inducers: doxorubicin, gamma rays, UV light, and freeze/thaw cycles. None of the treated cells grew in vitro (Supplemental Fig. 2) or in vivo (data not shown). However, T cell proliferation was constantly observed (Fig. 1E) but was somewhat variable from one experiment to another. A trend toward higher and long lasting proliferation was observed with doxorubicin- and gamma rays–treated cells, in line with the immunogenic apoptotic cell death induced by doxorubicin (27). Treatment that induced necrotic cell death, such as freeze/thaw cycles, did not generate a strong and prolonged Ag presentation in contrast with what has been previously proposed (28). The rapid loss of membrane integrity induced by these treatments probably causes a degradation or disappearance of intracellular proteins such as DBY.

This continuous presentation of DBY tumor Ag probably reflects the long-term persistence of Ag-loaded APCs because the proliferation of newly transferred Marilyn T cells decreases over time. Altogether, these data show that the Ag released from dying tumor cells can be efficiently presented to naive CD4⁺ T cells for an extended time. Moreover, the way the cells die impacts the intensity and duration of Ag presentation.

The long lifespan of MHC-II–restricted Ag presentation generated at the time of tumor implantation prevents the study of the immune response against a tumor Ag in a setting mimicking the human situation. To overcome this problem we generated a cell line in which Ag expression can be induced several days/weeks after tumor seeding. A LoxP-flanked stop-Neo^R cassette (LSL) was inserted into the cDNA encoding the DBY protein before the sequence encoding the DBY epitope (Fig. 2A, Supplemental Fig. 1). To obtain inducible Ag expression, we transduced the MCA101 tumor cell line with this construction and a Cre-ER^T2 recombinase encoded by Tet-on controlled single vector allowing a double layer of control by Dox and Tam (26). The H4B1 clone was obtained and further validated in vitro and in vivo.

In vitro, in the absence of Dox and 4-OHTam, the LSL cassette remained unrecombined as shown by the presence of the 1.8-kb band and the absence of the 600-bp band (Fig. 2B). DNA recombination became apparent in the presence of Dox and 4-OHTam. To assess DBY expression after in vitro induction, we measured the proliferation of Marilyn T cells transferred into mice that had been injected with untreated or in vitro–induced H4B1 cells. Marilyn T cells proliferated in mice bearing induced tumors but did not in mice bearing uninduced tumors (Fig. 2C). Thus, DBY expression can be efficiently induced in H4B1 cells, without expression by uninduced cells at the time of tumor challenge.

We then verified that in vivo treatment with Dox/Tam induced DNA recombination in growing tumors. We injected H4B1 cells into Rag2^−/− animals to avoid any rejection after Ag induction and treated the animals with the inductors for a week (Fig. 2D). PCR analysis of the tumor DNA demonstrated an efficient recombination with the appearance of a low m.w. band and a decreased of the high m.w. band (Fig. 2E). The amount of Ag expressed by the growing tumors into wt B6 host was then assessed by excising the in vivo– or in vitro–induced tumors that were then digested (Supplemental Fig. 3 A). Serial dilutions of the tumor cell suspensions were s.c. injected into B6 before transferring CFSE-labeled Marilyn T cells. T cell proliferation was assessed at day 6. The shift along the x-axis of the titration curves indicates that the amount of Ag was ∼4-fold lower after in vivo induction in comparison with in vitro induction. Furthermore, PCR analysis of the tumor DNA demonstrated incomplete recombination (Fig. 2E). We therefore optimized the conditions of in vivo induction to obtain higher recombination rate in established tumors (Supplemental Fig. 3B; data not shown). The final protocol is detailed in Supplemental Fig. 3C. We also verified that the dose of Dox and Tam used in our experiments did not modify the growth of the tumors or the immune response (data not shown).

Using the H4B1 clone, we studied whether the appearance of an Ag in growing tumors would be seen by the immune system in the absence of Ag released after implantation. We transferred Marilyn T cells into B6 mice bearing either tumors in which Ag expression had been induced in vitro (some DBY Ag is released and captured by APCs at the time of implantation) or tumors in which the Ag was subsequently induced in vivo (the DBY Ag appears as a neoantigen in established tumors) (Fig. 3A). Ag induction was initiated in palpable tumors (100 mm³) 7 d after implantation. After 10 more days, we transferred CFSE-labeled Marilyn cells and analyzed their phenotype in the TdLN either 7 or 15 d later (Fig. 3A). We chose to perform the adoptive transfer 10 d after Ag induction to allow a fair comparison of the in vitro and in vivo models with the same length of potential contact between the Ag and the T cells in a setting mimicking the encounter of a tumor neoantigen by naive CD4 T cells. In vivo treatment of immunocompetent mice with Dox/Tam efficiently induced DNA recombination (Fig. 3B). We first measured tumor growth in each group. In the uninduced group (DBY^neg), tumor growth was not affected in mice immunized with male splenocytes on the day of Marilyn cell transfer (Supplemental Fig. 3D). In vitro– and in vivo Ag–induced tumors were not rejected, in agreement with our previous reports, showing that Marilyn cells cannot reject MCA-DBY tumors in the absence of strong priming (25, 29).

We next assessed the ability of both tumor types to activate specific CD4⁺ T cells (Fig. 3C). DBY expressed as a neoantigen (in vivo induction) induced proliferation of Marilyn T cells. The high number of intermediary divided (1–4) cells (Fig. 3C) could represent continuous recruitment of naive T cells or, alternatively, arrest of T cell proliferation by T-T interactions as described previously (19). To distinguish these two hypotheses, we injected BrdU for 5 d before harvesting the TdLN 15 d after T cell transfer (Fig. 3D). In both groups, more than two-thirds of the divided cells were BrdU⁺ indicating at least one cell division during the last 5 d. The proportion of BrdU⁺ Marilyn T cells at one to two cell divisions was lower in the in vitro Ag–induced group indicating that a larger proportion of Marilyn T cells stopped dividing. This could reflect a stronger host response because of a longer contact with the Ag in this group. Nonetheless, the high proportion of BrdU⁺ Marilyn T cells in both groups indicates that naive Marilyn T cells were actively recruited into the immune response 15 d after transfer. Thus, in the TdLN a MHC-II–restricted tumor neoantigen was not ignored.

The absence of Ag presentation in an inflammatory context may lead to differences in the quality of the immune response generated in the inducible Ag model. We therefore characterized the expression of several activation and homing molecules on the responding Marilyn T cells in the TdLN 15 d after adoptive transfer (Fig. 4). CD44 expression was rapidly upregulated to reach high levels on fully divided T cells in both experimental groups (Figs. 3C, 4A). CD62L expression rapidly decreased in both groups (Fig. 4B). Moreover, a small proliferation of the Marilyn T cells indicating some Ag leakiness was observed at day 15 in the Ag-uninduced tumors providing a group with low amount of Ag. The changes in surface marker expression were very similar following in vivo or in vitro Ag induction, with only small differences in the percentage of positive cells, indicating that T cells were similarly activated and able to acquire homing molecules to recirculate and reach tissues. We did not find any Foxp3 expression by Marilyn T cells (Supplemental Fig. 4).

Activation and progressive differentiation to effector T cells are associated with upregulation of costimulatory and inhibitory receptors (iRs). Expression of iRs has also been linked to tumor associated immune regulation (30), and iR expression profiles may differ according to the differentiation status. We therefore investigated the surface expression of several iRs and one costimulatory receptor on Marilyn T cells from the TdLN (Fig. 4C–E). The two iRs, PD-1 and BTLA, together with the costimulatory receptor ICOS were highly and uniformly expressed on Marilyn T cells from the TdLN 15 d after transfer in both groups (Fig. 4C–E). In contrast, we found no expression of other iRs such as CTLA-4, TIM-3, LAG-3, CD160, and 2B4 (data not shown). Notably, PD-1 expression was slightly lower in the two groups in which the Ag appears in established tumors in comparison with the in vitro Ag–induced group. This result suggests that the initial Ag stimulation in an inflammatory context has a long-lasting impact on the immune response even on naive CD4 T cells stimulated in the TdLN 2 wk later.

We then studied the effector activities and differentiation of Marilyn T cells after challenge by tumors with in vitro or in vivo Ag–induced expression. We first measured the number of Marilyn T cells producing IFN-γ in the TdLN, 7 and 15 d after transfer, upon a short (5 h) ex vivo restimulation with PMA and ionomycin (Fig. 5A). Mice bearing Ag-uninduced tumors immunized with splenocytes from CD3ε^−/− male mice were used as positive control. Cytokine production was not measurable in mice bearing uninduced tumors, whereas more than five cell division Marilyn T cells produced IFN-γ after male cell immunization. Both in vitro and in vivo Ag induced groups exhibited the same CFSE/IFN-γ profile. The proportion of IFN-γ–secreting cells increased between days 7 and 15 in both groups suggesting that no regulatory negative feedback loop was set up.

To confirm that the Marilyn T cells had been efficiently primed after in vivo Ag induction, we measured cytokine production after TCR engagement (DBY peptide for 16 h) (Fig. 5B–F). In the TdLN, the proportion of Marilyn T cells producing IFN-γ and TNF-α was higher after in vitro induction (Fig. 5B). Some of the Marilyn T cells were IFN-γ⁺TNF-α⁺. Yet, in the spleen, the percentage of cytokine producing cells was similar in the two groups and was higher than in the TdLN (Fig. 5C). This is consistent with a progressive differentiation model in which the T cells that exit from the priming LN to other organs are those that divided the most and consequently include a higher proportion of effector cells (31).

In addition, in both groups 85 and 60–90% of the CD44^hi Marilyn T cells in the TdLN and spleen, respectively, expressed also LAMP-1, suggesting cytotoxic capacity (Fig. 5D) (32). For the in vivo Ag–induced group, the proportion of Lamp-1⁺ Marilyn T cells was slightly lower in the spleen than in the TdLN. Finally, the number of CD44^hi Marilyn CD4 T cells in the spleen was higher in the in vitro than in the in vivo Ag–induced group (Fig. 5E). Altogether, these results suggest a lower priming in the TdLN by the in vivo Ag–induced tumors that would generate a lower number of recirculating effector cells.

Functionally, the expression of the PD-1 and BTLA iRs by the Marilyn CD4 T cells (Fig. 4C, 4D) might indicate an “exhausted” phenotype (30). In the TdLN, most of the IFN-γ–secreting CD44⁺ T cells coexpressed PD-1 and BTLA, whereas BTLA was not much expressed by the IFN-γ–secreting T cells in the spleen (Fig. 5F). This could be related to the absence of Ag in the spleen and the transient expression of BTLA after Ag stimulation (33). Thus, in both the TdLN and spleen, PD-1⁺ Marilyn T cells exhibited a higher ability to produce IFN-γ, probably because PD-1 is more expressed by highly differentiated subsets, as shown for human CD8⁺ T cells (34). Altogether, these data suggest that the two modes of Ag induction generated effector Marilyn T cells able to recirculate and secrete effector mediators after in vitro restimulation.

To better assess the functionality of the anti-DBY CD4⁺ T effector cells generated after in vivo Ag induction, we measured a panel of lymphokines secreted after in vitro restimulation with the DBY peptide for 2 d. We focused on the in vivo Ag–induced group because most of the cytokine-secreting cells are the transferred Marilyn T cells in this group, whereas implantation of in vitro Ag–induced tumors primes host DBY–specific T cells (data not shown). In the TdLN, cytokine secretion was observed without peptide addition, probably because of Ag-loaded APCs or to the nonspecific immune response induced by growing tumors as this background secretion was also observed in the noninduced Ag group (Fig. 6A). Although the cytokine secretion was variable from one mouse to another, DBY specific Th0 (IL-2), Th1 (IFN-γ and TNF-α), and IL-10 responses were observed in the TdLN only when the Ag had been induced. In the spleen (Fig. 6B), we observed DBY-specific IFN-γ and TNF-α secretion but very little IL-2 and no IL-10 or IL-17 secretion (data not shown). These results indicate that Ag expression in established tumors stimulates a polyfunctional CD4⁺ T cell response in the TdLN, whereas the recirculating effector cells found in the spleen are essentially Th1.

We then studied the functionality of the Marilyn T cells found in the tumors. All the Marilyn T cells were CD44^hi and had divided at least six times (CFSE^neg) (data not shown). The proportion of Marilyn T cells among the TILs was variable from one mouse to another but similar in the two groups (Fig. 7A). Notably, 70–90% of the Marilyn CD4 T cells infiltrating the tumor produced IFN-γ (Fig. 7B). This suggests either high antigenic stimulation at the tumor site and/or the cells that infiltrated the tumors were those able to produce IFN-γ.

We then measured the secretion of several cytokines after in vitro peptide restimulation of Ficoll-isolated cells from pools of six to eight tumors of the uninduced or in vivo Ag–induced groups (Fig. 7C). The absence of IL-2 may be related to its consumption by activated cells or regulatory T cells during the in vitro culture. Interestingly, more TNF-α, IL-10, and IL-17 secretion with or without peptide was found in the in vivo Ag–induced group in comparison with the Ag-uninduced group. The absence of peptide-specific lymphokine secretion probably is due to the presence of Ag-loaded APCs leading to TCR triggering without exogenous peptide. To confirm these results in an independent experiment, we also analyzed the lymphokine secretion by Ficoll-isolated cells from individual tumors of the in vivo Ag–induced group. Although IL-2 and IL-13 were undetectable, we also observed high levels of TNF-α, IL-10, and IL-17 in many tumors but little IFN-γ (Fig. 7D). This experiment further confirms that the induction of Ag in established tumors leads to the infiltration of the tumors by polyfunctional effector cells.

In this study, we show that dying tumor cells may release Ag at the time of tumor implantation and generate long-lasting MHC-II–restricted immune response. Cell death induced by anthracycline was strongly immunogenic in line with previous reports (27, 35). The induction of Ag expression in established tumors was not ignored by the immune system because it efficiently primed CD4⁺ T cells in the TdLN and generated polyfunctional effector T cells able to recirculate back to the tumor. The recirculating T cells were the ones able to secrete IFN-γ. Like for human tumors, we did not observe any signs of tumor rejection despite the increased frequency of Ag-specific T cells.

The long lifespan of MHC-II–restricted Ag presentation we observed is consistent with our previous data using peptide-loaded LPS-matured DCs that induced proliferation of naive CD4⁺ T cells for more than 3 wk (19). A high stability of the MHC-II peptide complexes is indeed a common feature of immune-dominant MHC-II epitopes (20). The long-term Ag presentation after tumor implantation may changes the quality of the subsequent immune response. This prevents the study of the interaction between the immune system and the tumor in a situation mimicking the slow growth in non-inflammatory conditions of early-stage human tumors. This caveat is particularly important for MHC-II–restricted response and can be resolved by the use of models in which the Ag can be induced in established tumors. In the GFP-inducible tumor models described by H. Schreiber and group, the CD4⁺ T cell response was barely studied (22, 23). Thus, to our knowledge, our model is the first detailed study of the antitumor CD4⁺ T cell response in the absence of priming by Ag released at the time of tumor implantation.

Thanks to the double layer of control (transcription and cytoplasm/nucleus translocation), our system was not leaky at the time of tumor implantation and during the following week: no T cell proliferation was observed 7 d after T cell transfer (Figs. 2C, 3C), but a small proliferation was found at day 15 (Fig. 4C–E). Interestingly, this provides another in vivo Ag–induced tumor group with less Ag. In our model, the priming in the TdLN should be provided by APCs that have captured the Ag in the tumor and migrated to the dLN as previously described for a CD8⁺ response (21).

The immune response was qualitatively similar in the inducible Ag groups with a lower PD-1 expression on the proliferating Marilyn T cells in comparison with the in vitro Ag–induced group. Besides this difference, no qualitative changes in the immune response were observed between the two models, although the immune response was slightly weaker in the inducible Ag model in comparison with the in vitro Ag–induced tumors. This weaker response is probably related to a lower amount of Ag in the in vivo–induced model (Supplemental Fig. 3A). Yet, effector CD4⁺ T cells secreting IFN-γ were generated in both models. TNF-α, IL-10, and IL-17 secretion was also found and most of the CD4⁺ T cells expressed LAMP-1 after in vitro restimulation suggesting cytotoxicity features. This is consistent with our finding of cytotoxic CD4⁺ T cells in the blood of cancer patients (11), whereas tumor-specific CD4⁺ T cells developed into cytotoxic effectors able to clear established tumors in mice (8). Although activated Marilyn T cells can clear target cells loaded with the specific peptide (7), we did not yet explore this cytotoxic ability in the current model.

The pattern of PD-1 and BTLA expression in the different organs is consistent with the stable expression on memory cells of the former and the transient expression of the latter (33). The expression of these inhibitory molecules does not seem to prevent the generation of IFN-γ–secreting cells and the recirculation of effector cells to the spleen and tumor and therefore does not seem to indicate an exhaustion phenotype. Other mechanisms should be sought to explain the absence of tumor rejection. The selection of Ag loss variants (or of cells that had never recombined) is unlikely as the recombination rate was very high at the end of the experiment (Fig. 3B).

In this paper, we did not study the effector phase of the response as no tumor rejection was observed. However, we found intratumor CD4⁺ T cells able to secrete IFN-γ after in vitro restimulation, indicating that not only the priming was efficient but also functional effectors able to migrate back to the tumor were generated. The proportion of effector CD4⁺ T cells we found in the tumors (Fig. 6C) reaches the one found by Corthay et al. (36), which was sufficient to reject a myeloma tumor. Similarly to this model, the MCA101 cell line does not express MHC-II molecule (25), but DBY is a cytoplasmic protein, which is not secreted contrary to the myeloma Ig or OVA used in other reports (36–39). The intracytoplasmic location of the DBY protein may explain the absence of tumor rejection as secretion of the Ag may induce tumor rejection (14). Moreover, the Ag released by the dying cells of the in vitro induced model is probably in a very different form (apoptotic bodies, cell debris, and so on) in comparison with these soluble proteins, and this may prevent the generation of a strong antitumor reaction. Nonetheless, because nonsecreted proteins are more abundant than secreted ones, it is highly probable that most neoantigens appearing in human tumors are linked to intracellular proteins. In our model that mimics this situation, the low number of effectors generated would be related to the absence of inflammation/costimulation and/or insufficient Ag load in the TdLN as these parameters are essential to generate the high number of effectors necessary to reject tumors (25).

Because our model is still a transplantable tumor system, the genetic heterogeneity between the cell line and the host may be a source of potential immunogenic uncontrolled epitopes. The host response toward these epitopes might interfere with the CD4⁺ response against a MHC-II restricted neoantigen that we modeled by an adoptive transfer of naive T cells after Ag induction in established tumors. Moreover, the rapid growth of the transplanted tumors may preclude the observation of deletion mechanisms. Indeed, assuming that the half life of naive CD4⁺ T cells in an LN is ∼8–12 h (40), the flux of CD4⁺ T cells reaching the TdLN during the span of the experiment (2–3 wk after T cell transfer) may not be sufficient to observe a significant decrease in cell numbers. Autochthonous tumor models induced by lentiviral delivery of both the Ag and the oncogenic events in genetically modified animals (41) would be necessary to address this issue. However, the CD4⁺ T cells have not been yet studied in these models, and the lentiviral infection of hematopoietic cells strongly primes the host before tumor development (Ref. 42; H. Flament, R. Alonso Ramirez, and O. Lantz, unpublished observations)

Importantly, in comparison with these autochthonous tumor models, our inducible Ag–transplanted tumor system allows bigger synchronous tumor cohorts, which are necessary for therapeutic development. Indeed, the antitumor neoantigen MHC-II–restricted immune response is able to clear experimental (7, 8, 36) or human tumors (9, 10) and overcome any loss in the MHC-I presentation pathway by the tumor cells. Our model will be very useful to better understand these antitumor CD4⁺ T cell responses to allow their amplification or de novo generation for therapeutic purpose.

We thank E. Piaggio and C. Sedlik for comments and suggestions on the manuscript and I. Grandjean and C. Daviaud, as well as the personnel of the Curie Animal Facility for Animal Husbandry, for help with experiments.

The authors have no financial conflicts of interest.