Pathogen-boosted adoptive cell transfer therapy induces endogenous antitumor immunity through antigen spreading

Abstract

Loss of target antigens in tumor cells has become one of the major hurdles limiting the efficacy of adoptive cell therapy (ACT)-based immunotherapies. The optimal approach to overcome this challenge includes broadening the immune response from the initially targeted tumor-associated antigen (TAA) to other TAAs expressed in the tumor. To induce a more broadly targeted antitumor response, we utilized our previously developed Re-energized ACT (ReACT), which capitalizes on the synergistic effect of pathogen-based immunotherapy and ACT. In this study, we showed that ReACT induced a sufficient endogenous CD8⁺ T-cell response beyond the initial target to prevent the outgrowth of antigen loss variants in a B16-F10 melanoma model. Sequentially, selective depletion experiments revealed that Batf3-driven cDC1s were essential for the activation of endogenous tumor-specific CD8⁺ T cells. In ReACT-treated mice that eradicated tumors, we observed that endogenous CD8⁺ T cells differentiated into memory cells and facilitated the rejection of local and distal tumor re-challenge. By targeting one TAA with ReACT, we provided broader TAA coverage to counter antigen escape and generate a durable memory response against local relapse and metastasis.

Introduction

Despite the unprecedented success of adoptive cell therapy (ACT), a considerable amount of patients become resistant to and relapse from the therapy after initial responses (1). Recurrence is frequently associated with the emergence of tumor cells that lose the initial ACT-targeted antigens, due to genomic instability and high mutation rates of malignant cells (2,3). As antigen escape–related relapse has repeatedly been observed following monolithic antigen-targeted therapy, it has become one of the greatest obstacles for improving the clinical outcome of patients and expanding the spectrum of cancers treated with ACT. To overcome these issues, ACT-mediated immune responses must be broadened from the initial targeted antigen to other tumor-associated antigens (TAAs)(4,5). Such a phenomenon, called antigen spreading, has been well-documented and positively correlates to the efficacy of immunotherapies, such as cancer vaccines and immune checkpoint blockade (6,7).

Antigen spreading usually occurs following initial therapy-mediated tumor destruction, which leads to the release of secondary (i.e. non-targeted) TAAs (4,5,8). Subsequently, these antigens can be taken up by professional antigen-presenting cells (APCs), such as dendritic cells (DCs), to induce T-cell responses. As a result, these broadly targeted antitumor immune responses may prevent the outgrowth of antigen escape variants and provide long-term protection against relapse. Despite these advantages, antigen spreading in ACT is rarely achieved at a meaningful level for therapeutic efficacy. This is likely associated with insufficient activation Batf3-dependent conventional type 1 DCs (cDC1s), which are required for the initial cross-priming of antitumor CD8⁺ T cells and effector T-cell recruitment into the tumor microenvironment (TME)(9). Therefore, adequate activation of this DC subset is crucial to initiate the endogenous T-cell response against newly released TAAs (10).

In addition to primary tumor control, the ultimate goal of ACT is to generate memory T cells that protect against local relapse and distant metastasis. To avoid tumor growth at the site of the primary tumor, induction of a newly discovered subset of memory cells, known as tissue-resident memory (T_RM) cells, is essential (11). T_RM cells were first discovered in an infection model (12), where they provide first-line protection at the portal of pathogen entry during a secondary local infection (13,14). A similar role of T_RM cells in cancer has also been observed (15). In support of this notion, a few studies in a variety of human cancers have shown that increased tumor infiltration by T cells with a T_RM cell–like phenotype correlates with improved overall survival in lung cancer (16,17), breast cancer (18), bladder cancer (19), and ovarian cancer (20). More convincingly, local vaccines that induce melanoma-specific T_RM cells potently suppress tumor growth in mucosal tissues compared to systemic immunizations, as observed in mouse models (16,21). In contrast, preventing cancer metastasis to distal sites through the blood or lymph system may rely more on circulating memory cells. These cells, including central memory (T_cm) or effector memory (T_em) cells, recirculate between the blood and lymphoid organs to continuously scan for previously encountered antigens (22). Consistent with this, the presence of circulating memory cells has been correlated with improved survival in melanoma patients (23). Taken together, the optimal ACT should aim to generate both circulating and tissue-resident memory CD8⁺ T cells.

In an attempt to overcome antigen escape and prevent recurrence and metastasis, we utilized a newly developed approach that combines ACT with a pathogen-based vaccine, named Re-energized ACT (ReACT)(24). As described previously (24), ReACT is a combination of ACT of dual-specific T cells that are tumor reactive CD8⁺ T cells engineered in vitro with an additional TCR that recognizes a bacterial antigen and intratumoral injection of the bacteria. In response to the bacterial infection, dual-specific CD8⁺ T cells vigorously expand in vivo and migrate to the tumor site, where they kill targeted tumor cells and robustly eradicate primary tumors (24). In this study, we aimed to test if ReACT could elicit antigen spreading to overcome antigen escape and generate sustained immunity against cancer relapse and metastasis. In our antigen escape model, we first created a mutant B16-gp100-KO cell line by CRISPR. Next, we mixed the parental B16-F10 cells with various ratios of B16-gp100-KO cells to establish tumors with escape variants when applied with ReACT. In this model, we demonstrated that ReACT sufficiently induced endogenous T cell–mediated antigen spreading to eliminate antigen escape variants. Batf3-dependent DCs were crucial for the induction of endogenous antitumor CD8⁺ T-cell responses. Endogenously activated CD8⁺ T cells could further differentiate into T_RM cells and circulating memory T cells to provide optimal protection against both local and systemic re-challenge. Overall, our study suggests that ReACT, although targeting only one tumor antigen, can mobilize endogenous T-cell responses against new TAAs to overcome antigen escape and form long-term memory protection against cancer relapse and metastasis.

Materials and Methods

Mice

C57BL/6 and C57BL/6 CD45.1/CD45.1 were purchased from NCI (Rockville, MD). Batf3^−/− mice and CD11c-DTR mice were provided by Dr. Aimin Jiang from Roswell Park Comprehensive Cancer Center (Buffalo, NY). C57BL/6 Thy1.1/Thy1.1 mice, Cd8a^−/− mice, and C57BL/6 Thy1.1/Thy1.1 Pmel mice were obtained from Jackson laboratories (Bar Harbor, ME). Pmel mice were crossed with C57BL/6 mice for at least four generations to generate Thy1.2/ Thy1.2 Pmel mice. All in vivo experiments used 8–12-week-old females, and all animal procedures were approved by the IACUC Committee of the Medical College of Wisconsin.

Bone marrow chimeras and T-cell depletion

To generate Thy1.1 endogenous CD8⁺ T-cell bone marrow chimeras, Cd8^−/− mice were irradiated with a Gammacell 40 Exactor with two doses of 500 rad, each 4 hours apart. To obtain donor bone marrow from Thy1.1⁺ mice and Cd8^−/− mice, femurs and tibiae were harvested and bone marrow was flushed out using RPMI (Corning, NY) with 10% Hyclone fetal bovine serum (Thermo Fisher Scientific, Waltham, MA). 3×10⁶ cells consisting of a 30%:70% mixture of Thy1.1⁺ and CD8^−/− bone marrow cells were intravenously transferred to the irradiated recipients and allowed to engraft for 8 weeks. The reconstitution was confirmed by flow cytometry before the initiation of experiments. To deplete Thy1.1-expressing cells, anti-mouse Thy1.1 (BioXcell, Hanover, NH) was administered intraperitoneally (i.p.) 6–9 days after tumor inoculation, which was also one day before ReACT. As control, another group of mice were injected with mouse IgG2a isotype control (BioXcell). All antibodies were administered every other day over the course of two weeks.

For CD11c^DTR/WT and CD11c^DTR/Batf3^−/− mixed chimeras, 3×10⁶ cells of a 1:1 mixture was injected into irradiated B6 mice and allowed to reconstitute for 8 weeks. For DC depletion, diphtheria toxin (Sigma-Aldrich, St. Louis, MO; 500ng/dose/mouse) was injected i.p. 5 and 7 days post-ReACT. 1 × 10⁶ P14 cells was transferred on day 6.

Tumor cell lines

All tumor cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) from Lonza (Basel, Switzerland) supplemented with 10% Hyclone fetal bovine serum (Thermo Fisher Scientific), glutamine (2 mmol/L; Corning) and penicillin/streptomycin (100 U/mL; Corning). For in vivo inoculation, tumor cells were passaged fewer than 5 times before each use. B16-F10 and E0771 were originally purchased from ATCC and obtained from Dr. Susan Kaech (Salk Institute, San Diego, CA) in 2013, both of which have not been further authenticated and tested for Mycoplasma.

CRISPR–Cas9 system was used to generate a B16-gp100-KO cell line (25). Two gRNAs have been designed (5’CTGGCTGTGGGGGCCCTAGA’3 and 3’CCACCCACACTCACCTTCTA’5) targeting the ‘ESGRNQDWL’ immunogenic peptide on gp100 using benchling (https://benchling.com/crispr) and the CRISPR tool from the Feng Zhang lab, MIT (http://crispr.mit.edu/). Electroporation with a CRISPR/Cas9 ribonucleo-protein complex (RNP) from IDT (Coralville, IA) was performed following the protocol provided by IDT CRISPR genome editing (www.idtdna.com) using Amaxa 4D Nucleofector system with program number DJ-110, specific for the B16-F10 cell line. Verification of CRISPR/Cas9 editing of gp100 in B16-F10 cell line was done by a Surveyor nuclease detection kit (IDT), according to the protocol (www.idtdna.com), using a single cell clone following serial dilution.

The deletion of gp100 was further confirmed by in vitro killing assays using activated Pmel-1 TCR transgenic CD8⁺ T cells, which recognize the MHC class I (H-2D^b)-restricted epitope of gp100.

E0771-GP33 was transduced with LCMV-GP_33–41 minigene under the control of actin promoter, which was kindly provided by Dr. Hanspeter Pircher (University of Freiburg, Germany). The expression of GP33 was verified by in vitro killing assays using activated P14 TCR transgenic CD8⁺ T cells, which recognize GP33. The B16-GP33 cell line was used because it enables monitoring of GP33 antigen-specific T-cell responses in tumor-bearing hosts using tetramer staining by flow cytometry.

In vitro killing assay

Pmel or P14 CD8⁺ T cells were activated with relevant peptides and IL2 for one day and used as effector cells. Target tumor cells (B16-F10, B16-gp100-KO, E0771-GP33, and E0771) were plated out on 96-well plates and incubated for 48 hours either with activated effector cells (5:1 E:T) in the presence of IncuCyte Caspase-3/7 Green Apoptosis Assay Reagent (Cat # 4440, Essen BioScience, Ann Arbor, MI). Addition of this reagent enables simultaneous detection of cells undergoing apoptosis. Images were taken every 2 hours and analyzed by the IncuCyte S3 Live-Cell Analysis System (Essen BioScience).

Tumor inoculation and other treatments

2 × 10⁵ B16 tumor cells in 100 μL of sterile PBS were injected subcutaneously (s.c.) in the flank of C57BL/6 mice. For mimicking antigen escape, mice were inoculated with different combinations of B16-F10 and B16-gp100-KO cells as following: i) 2 × 10⁵ B16-F10; ii) 1.9 × 10⁵ B16-F10 with 0.1 × 10⁵ B16-gp100-KO; iii) 1.8 × 10⁵ B16-F10 with 0.2 × 10⁵ B16-gp100-KO; iv) 1.6 × 10⁵ B16-F10 with 0.4 × 10⁵ B16-gp100-KO; 5) 2 × 10⁵ B16-gp100-KO. 7–10 days following injection, the ReACT treatment was initiated as previously published (24). Briefly, Pmel T cells were activated with gp100 peptide (Genscript, Piscataway, NJ ) and IL2 (10 ng/mL; Peprotech, Rocky Hill, NJ) for 24 hours, followed by transducing with the MSCV-IRES-GFP-OT-I vector (24) to generate dual specific CD8⁺ T cells. After expansion, 10⁶ cells were adoptively transferred into tumor-bearing mice, accompanied with intratumoral injection of 1 × 10⁴ colony forming units (CFU) of listeria monocytogenes expressing OVA (LM-OVA), which was developed by Dr. Hao Shen (University of Pennsylvania School of Medicine, Philadelphia, PA) and provided by Dr. Susan Kaech. Tumor sizes were measured by caliper three times per week and expressed as tumor volume, calculated as [longest dimension × (perpendicular dimension) ²]/2. The mice with complete tumor eradication were classified as responders, whereas others that succumbed to tumor growth were classified as non-responders. Once the tumor size exceeded 2500 mm³, mice were euthanized. For experiments performed in Figures 3 and 4, mice were sacrificed 8 days after ReACT for immune cell isolation.

Figure 3. ReACT elicits endogenous tumor-reactive CD8⁺ T-cell responses.

(A) C57BL/6 mice received P14 cells as surrogates for endogenous tumor-reactive T cells one day before inoculation with B16-GP33 melanoma. 7 to 10 days following, mice were treated with ACT therapy, Listeria, or ReACT, respectively. 8 days after therapy, mice were sacrificed, and tissues were analyzed via flow cytometry. (B-C) Thy1.1^–CD45.2⁺ CD8⁺ T cells were defined as P14 cells; Thy1.1⁺CD45.2⁺CD8⁺ T cells were defined as Pmel cells. (B) Representative flow cytometry plots and (C) cumulative results showing the frequency of P14 and Pmel T cells isolated from tumors. (D-E) P14 from ACT, Listeria, or ReACT-treated mice were re-stimulated with GP33. IFNγ production was assessed by intracellular cytokine staining. (D) Representative flow cytometry plots and (E) cumulative results are shown. (F) Representative flow cytometry plots and (G) cumulative results for granzyme B expression in P14 cells, as examined by flow cytometry. Data represent cumulative results from two independent experiments with n=7 for ACT, n=5 for Listeria and to n=7 for ReACT. Mean±SEM is shown. The significance was determined by t-test. *P<0.05; **P<0.01. n.s. denotes not significant.

Figure 4. Activation of endogenous tumor-reactive CD8⁺ T cells is dependent on BATF3-driven cDC1s.

B16-F10 tumor-bearing mice were treated with either ACT or ReACT 7–10 days following tumor inoculation. One week after therapy, tumor samples were harvested and analyzed by flow cytometry for the cDC1 population. (A-B) The frequency of XCR1^highCD172a^low cDC1s among Lin^–CD11c⁺CD26⁺MHCII⁺ cells are shown in (A) representative flow plots and (B) cumulative dot graphs. (C) CD11c^DTR/WT (WT) and CD11c^DTR/Batf3^−/− (Batf3^−/−) bone marrow chimeras were generated and inoculated with B16-GP33 tumors, followed by ReACT. 5 days after ReACT, diphtheria toxin was given to deplete DTR-expressing DCs. The following day, 1 × 10⁶ P14 T cells were transferred, and their accumulation within the tumor was assessed 8 days later by flow cytometry. (D-H) The frequency of Lin^–CD11c⁺CD26⁺ MHCII⁺XCR1^highCD172a^low cells is (D) shown in representative flow plots and (E) a scatter plot. The frequency of GP33⁺ P14 and GFP⁺ Pmel CD8⁺ T cells in tumors are (F) shown in representative flow plots and scatter plots for (G) Pmel and (H) P14 cells. Data represent pooled data from two independent experiments with n=8 for WT and n=9 for Batf3^-/−. Mean±SEM is shown. The significance was determined by t-test. ***P<0.01; ***P<0.001. n.s. denotes not significant.

In all recall experiments, mice that cleared primary tumors after ReACT and remained tumor-free for one or two months were used as memory mice, while age and gender-matched naïve mice were included as controls. For local re-challenge, mice were inoculated with 1 × 10⁵ E0771-GP33 tumor cells on the same flank where primary tumors (B16-F10) were eradicated. For systemic recall responses in the metastasis model, all mice were intravenously (IV) injected with 1 × 10⁵ B16-F10 cells and sacrificed 15 days later for harvesting lung tissues, which were perfused with PBS and fixed by 10% formalin (Sigma-Aldrich). Subsequently, all tumor nodes in the lung were counted, and the images were acquired using a Zeiss Lumar V12 Stereoscope (Zeiss, Oberkochen, Germany). For CD8 T cells depletion, 500 μg anti–mouse CD8 (clone 2.43, BioXcell) was administrated i.p. twice weekly for two weeks after tumor rechallenge.

Immune cell isolation and flow cytometry

The tumors were dissected from mice 8 days after ReACT and minced into small pieces using scissors, followed by incubation with collagenase XI (0.7 mg/mL; Sigma-Aldrich) and type IV bovine pancreatic DNase (30 mg/mL; Sigma-Aldrich) for one hour at 37°C. These tumor tissues were then mashed against a 70μm cell strainer to yield cell suspensions for isolating tumor-infiltrating lymphocytes (TILs) by Lymphocyte Cell Separation Medium (Cedarlane Labs, Burlington, ON, Canada). Subsequently, these single-cell suspensions were subjected to red blood cell lysis and filtration before staining for flow cytometry, as previously described (24). For Figure 5A, mice were bled from the facial vein, and immune cells were isolated by Histopaque (Sigma-Aldrich). For Figure 5B–E, skins of the tumor-cleared area and distal area were dissected from memory mice and processed as described above. The spleens were mashed against a 70μm cell strainer, followed by red blood cell lysis and filtration to generate single-cell suspensions for flow cytometry staining.

Figure 5. ReACT-induced endogenous tumor-reactive CD8⁺ T cells provide long-term protection against local recurrence.

(A) P14 chimeric mice were generated as described in Figure 3 using Thy1.1⁺ P14 cells. Mice were inoculated with B16-GP33 tumor cells, followed by ReACT using Thy1.2⁺ Pmel cells. All the mice were bled weekly following therapy to detect the frequencies of P14 cells and Pmel cells among total CD8 T cells. (B-C) Two months after ReACT-mediated tumor eradication, mice were sacrificed and assessed for the presence of Thy1.1⁺ P14 cells within the proximal and distal skin of the tumor-inoculated area. (B) Representative flow cytometry plots and (C) cumulative results. (D-E) Expression of CD62L and T_RM markers, CD103 and CD69, were examined on P14 cells and compared to CD44⁺CD8⁺ T cells from the spleen. (D) Representative histogram and (E) geometric mean fluorescence intensity (gMFI) are shown. (F) After initial tumor eradication following ReACT, mice were inoculated with E0771-GP33 and tumor rejection was assessed. Naïve mice were used as controls. The cumulative results from two independent experiments with n=6 and significance was determined by t-test. *P<0.05, **P<0.01. (F) log-rank (Mantel–Cox) test, **P<0.01 with n=7 for naïve and n=18 for ReACT.

For Granzyme B staining, TILs were stained for surface markers before fixation and permeabilization (all buffers from Biolegend, San Diego, CA). The permeabilized cells were then washed and stained with antibodies against granzyme B in permeabilization buffer. To assess the cytokine production of transferred P14 cells, 1 × 10⁶ TILs were stimulated with complete T-cell medium in the presence of GP33 peptide (1 μg/mL; Genscript), brefeldin A (Biolegend), and IL2 (10 ng/mL; Peprotech) for 6 hours. The complete T-cell medium included RPMI with 10% heat-inactivated fetal bovein serum (Hyclone), 10 mM HEPES (Corning), 1% non-essential amino acids MEM (Corning), 1% sodium pyruvate (Corning), L-glutamine (2 mmol/L; Corning), penicillin/streptomycin (100 U/mL; Corning) and ß-mercaptoethanol (Sigma-Aldrich). After incubation, surface markers were stained before fixation and permeabilization, followed by IFNγ staining. For IRF8 staining, TILs were fixed and permeabilized using the True-Nuclear™ Transcription Factor Buffer Set (Biolegend) after surface staining. All antibodies and buffers are listed in Supplementary Table S1. All flow cytometry data were acquired on an LSRII (BD Biosciences, San Jose, CA) and analyzed by FlowJo (BD Biosciences).

Enzyme-linked immunospot assay (ELISPOT)

For ELISPOT assays, an Immunospot mouse IFNγ ELISPOT Kit from CTL (Cleveland, OH) was used. 1 × 10⁶ target cells (B16-F10 or B16-F10^gp100ko)were treated with mouse IFNγ (500 U/mL; Peprotech) for 12 hours, and then irradiated by Gammacell 1000 (120 Gy). Effector cells were CD8⁺ T cells isolated from spleen, draining lymph nodes (DLNs), and LNs from ReACT-treated mice that cleared tumors one month prior. Targets cells were seeded at 25,000 cells per well. Effector cells were seeded at 10⁶ cells per well. Plates were wrapped in foil and incubated for 24 hours, then developed according to manufacturer’s protocol. Plates were scanned using a CTL-ImmunoSpot Plate Reader, and data was analyzed using CTL-ImmunoSpot Software.

Statistical Analysis

Graphs were generated and statistical analyses performed using GraphPad Prism version 5.02 (GraphPad Software, San Diego, CA). The overall tumor growth from endogenous CD8⁺ T cell depletion experiments was analyzed by two-way ANOVA, whereas the comparison of tumor-free mice after secondary challenge was determined by log-rank (Mantel–Cox) test. For all other comparisons, t-tests were used to determine the statistical significance. *p<0.05; **p<0.01.

Results

ReACT induces an endogenous antitumor CD8⁺ T-cell response

As we previously showed, Re-energized ACT (ReACT; schematic shown in Supplementary Figure S1A) sufficiently induces a strong antitumor CD8⁺ T-cell response and reverses the immunosuppressive TME, which leads to complete tumor eradication in ~70% of B16-F10 tumor-bearing mice (24). Motivated by the robust efficacy of ReACT in primary tumor regression, we attempted to further explore the memory response in the current study. To this end, tumor-free survivors were inoculated with same tumor cells. The majority of mice (75%) were protected from re-challenge, indicating long-lasting adaptive immunity (Figure 1A). Because ACT of tumor-reactive CD8⁺ T cells was part of the therapy, we reasoned that this protection was mediated by a CD8⁺ T-cell memory response. Indeed, depletion of CD8⁺ T cells in survivors before the tumor re-challenge almost completely abolished tumor rejection (Figure 1A). In an attempt to characterize the presence of memory T cells, we failed to detect any adoptively transferred CD8⁺ T cells in spleen and lymph nodes by flow cytometry (Figure 1B). Taken together, these results raised at least two of the following possibilities: (i) the adoptively transferred CD8⁺ T cells formed memory cells, but they are too few to be detected (ii) ReACT might have elicited endogenous CD8⁺ T-cell responses against tumor cells, which formed long-lasting immunity following the primary tumor eradication.

Figure 1. ReACT elicits an endogenous tumor-reactive CD8⁺ T-cell response.

(A) Two months after ReACT-mediated tumor eradication, mice were re-inoculated with B16F10 cells with (n=8) or without (n=4) CD8⁺ T cells. B16F10 tumor growth in naive mice (n=4) that had not received any previous tumor injections or immunotherapy were included as controls. (B) The transferred cells in mice from (A) were assessed for CD8⁺ T-cell expression of Thy1.1 in the spleen, draining lymph nodes (DLN), and non-draining lymph nodes (NLN). Representative plots are shown. (C) Bone marrow cells from Thy1.1 congenic B6 and Cd8^−/− mice were mixed at 1:1 ratio and transferred into irradiated recipient Cd8^−/− mice to generate CD8-specific Thy1.1 bone marrow chimeric (BMC) mice. (D) Representative flow cytometry plots showing the chimerism of CD4 and CD8 T cells. (E-G) Following reconstitution, mice were inoculated with WT B16-F10 cells. One week later, mice received Thy1.1-depleting antibody + ReACT (n=7) or isotype antibody + ReACT (n=12). After treatment, the depletion efficacy was confirmed by flow cytometry. (E) Shown are representative dot plots. (F) The frequency of Thy1.1⁺ cells among total CD8⁺ T cells is shown in scatter plots. (G) Tumor growth was monitored. Solid line: non-responders, dotted line: responders. Data were compiled from two independent experiments. (A) log-rank (Mantel–Cox) test was used to compared naïve with other two group respectively., **P<0.01. (G) Significance was determined two-way ANOVA.

To address this possibility, we first sought to evaluate the contribution of endogenous T cells in eliminating the primary tumor. To specifically deplete endogenous CD8⁺ T cells without affecting the transferred Pmel cells and other immune cells, we generated bone marrow chimera (BMC) mice. As illustrated in Figure 1C, Thy1.1⁺ C57BL/6 (Thy1.1-B6) bone marrow (BM) cells were mixed with Thy1.2⁺ Cd8a^–/– BM cells at a 30%:70% ratio and transferred into irradiated Cd8a^–/– mice. After reconstitution, we confirmed that all CD8⁺ T cells were Thy1.1⁺, whereas only 30% of all other T cells expressed Thy1.1 (Figure 1D). The monoclonal Thy1.1 antibody was used to deplete all CD8⁺ T cells without affecting the majority of other immune cells (Figure 1E–F, Supplementary Figure S1B). Thy1.2⁺ Pmel cells were used as the cellular source of ReACT; therefore, they were spared from Thy1.1 depletion. BMC mice were inoculated with B16-F10 followed by ReACT, as previously described (24). The treatment of anti-Thy1.1 or IgG2a isotype control was started one day before ReACT therapy and sustained for a week. As expected, endogenous Thy1.1⁺ CD8⁺ T cells were effectively depleted without affecting the transferred Thy1.2⁺ Pmel cells (Figure 1E–F). Consistent with previous results, ReACT induced robust tumor regression and eradication in around 70% of control isotype–treated mice, whereas depletion of endogenous CD8⁺ T cells resulted in a significantly reduced antitumor response. Although initial tumor regression was comparable between the two treatment groups until 10 days after therapy, eventually 5 out of 7 Thy1.1-depleted mice relapsed and succumbed to the disease (Figure 1G). Overall, these data suggested that ReACT mobilized the endogenous CD8⁺ T-cell response, which played a vital role in eliminating B16-F10 tumors.

ReACT prevents tumor antigen escape in vivo

ReACT induced endogenous CD8⁺ T-cell responses and prompted us to ask if we could use the ReACT strategy to overcome baseline antigen heterogeneity and antigen escape–mediated immune evasion by targeting a single tumor antigen. To test this idea on a B16-based melanoma tumor model with a controlled incidence of antigen escape, we generated a tumor variant that could not be recognized by a gp-100–specific TCR expressed by Pmel cells. Because it is well established that Pmel cells recognize the immunogenic epitope (EGSRNQDWL) on the melanoma antigen gp100 (26), we employed CRISPR/Cas9 technology to create a gp-100 null B16-F10 cell line. This new line had an insertion of one nucleotide at the beginning of this epitope, which induced a frame-shift mutation that altered the reading of subsequent codons and abrogated its expression. Based on this, the B16-gp100-KO tumor cell line was generated and sequenced to confirm the deletion (Supplementary Figure S2A). Activated Pmel CD8⁺ T cells could not elicit cytotoxic activity against B16-gp100-KO cells (Figure 2A). Taken together, we successfully generated a tumor escape variant, which evaded tumor destruction mediated by transferred Pmel cells.

Figure 2. In vivo antitumor activity of ReACT in the B16 tumor antigen escape model.

(A) B16-F10 and B16-F10^gp100ko cells were incubated with activated Pmel cells in the presence of Caspase 3/7 apoptosis green fluorescence detection reagent. The killing of tumor cell lines was measured in real-time by IncuCyte S3 Live-cell analysis system and shown in bar graph with SEM among replicates. (B) C57BL/6 mice were inoculated with WT B16-F10 or mixed (95% WT with 5% B16-gp100^KO; 90% WT with 10% B16-gp100^KO and 80% WT with 20% B16-gp100^KO). 7–10 days later, all mice were treated with ReACT, and tumor growth was monitored. Solid lines: non-responders, dotted lines: responders. Data were compiled from two independent experiments; 10 mice/group. Significance was determined as *p≤0.05 by two-way ANOVA

By inoculating mice with mixed parental B16-F10 with B16-gp100-KO cells at the indicated ratios (100%:0%, 95%:5%, 90%:10%, 80%:20% or 0%:100%), we established tumors with various degrees of antigen escape. After 7 days, mice were treated with ReACT as previously described (24). Consistent with our previous observation, more than 70% of mice implanted with parental B16-F10 tumors cleared those tumors after ReACT (Figure 2B). Complete and durable tumor regression was achieved in 60% of mice bearing tumors containing 5% of tumor escape variants (Figure 2B), which suggested that ReACT induced protection against antigen loss–mediated tumor evasion. To further explore the efficacy of this protection, we increased the proportion of tumor escape variants (B16-gp100-KO) to 10% in the inoculation mixture. ReACT still induced pronounced tumor regression in these mice and tumor eradication was achieved in 30% of mice (Figure 2B). However, this tumor eradication rate was reduced to 10% in ReACT-treated mice when the inoculation mixture contained 20% B16-gp100-KO (Figure 2B). As expected, ReACT failed to protect any mice when 100% B16-gp100-KO tumor cells were used (Figure 2B). Collectively, these results indicated that ReACT could efficiently control antigen escape to a certain degree in the B16-F10 tumor model.

ReACT elicits antigen spreading to expand the antitumor response

Next, we sought to characterize the ReACT-induced endogenous immune response, which potentially reacted to the new antigenic variants of tumor cells. For this study, we utilized B16-GP33 tumor cells, which are a derivative of the B16-F10 cell line modified to constitutively express GP_33–41, a glycoprotein epitope from lymphocytic choriomeningitis virus (LCMV). In this model, GP_33–41 serves as a surrogate TAA that permits the tracking of CD8⁺ T-cell responses that are not directly targeted by ReACT. Given that GP_33–41 tetramer⁺ cells were too few to be reliably detected in the tumor, P14 transgenic T cells, which express a TCR specific to GP_33–41, were transferred into congenic mice to facilitate the detection of tumor-specific CD8⁺ T cells (Figure 3A). These P14 chimera mice were inoculated with B16-GP33 tumors and received conventional ACT with Pmel cells, Listeria injection, or ReACT, respectively. Congenic markers were used to distinguish between P14 and Pmel cells, as well as CD8⁺ T cells from the hosts. Consistent with our previous report (24), ReACT significantly increased the accumulation and function of transferred Pmel (Thy1.1⁺CD45.2⁺) cells in contrast to conventional ACT (Figure 3B–C).

Given the critical role of tumor-reactive T cells, other than Pmel cells, in the therapeutic effect against tumor escape, we next assessed the impact of ReACT on the quality and quantity of P14 cells. One week after treatment, we found that ReACT induced substantially greater intratumoral expansion of P14 cells than ACT or Listeria alone (Figure 3B–C). Intratumoral injection of Listeria did elicit a stronger recruitment of P14 T cells than ACT, suggesting a bystander effect of bacterial infection–induced T-cell priming. Although P14 cells in Listeria-treated tumors released significantly higher IFNγ than the ACT-treated group in response to ex vivo re-stimulation with cognate GP_33–41 peptide, ReACT treatment provoked the strongest production of IFNγ among all treated groups (Figure 3D–E). The enhanced effector function of P14 cells was further confirmed by ReACT-boosted granzyme B expression (Figure 3F–G). Overall, our data suggested that ReACT induced antigen spreading, which was likely mediated by bacterial infection, and rendered the additional induction of CD8⁺ T-cell responses against tumor antigens not initially targeted by adoptively transferred T cells.

Batf3-dependent cDC1s are required for ReACT-induced antigen spreading

An important component of our therapeutic strategy is bacterial infection, which is known to induce an inflammatory environment that bolsters DC-mediated cross-priming of T cells. Having established the critical role of antigen spreading in overcoming antigen escape during ReACT, we speculated that ReACT mobilized these CD8⁺ T cells by harnessing DC subsets that specialize in cross-presentation. To address this possibility, we first examined the effect of ReACT on the accumulation and activation of cDC1s by flow cytometry. Compared to ACT, ReACT significantly enriched the frequency of the tumor-residing CD11c⁺CD26⁺MHCII⁺XCR1^highCD172a^low cDC1 population (Figure 4A–B, Supplementary Figure S3A), with heightened expression of classical cDC1 markers, such as CD103, IRF8, and CD24 (Supplementary Figure S3B) (27, 28). Their expression of DC activation markers CD80 and CD40 was also enhanced by ReACT (Supplementary Figure S3C).

Because cDC1s are essential for effector T-cell trafficking to the tumor and exhibit superior cross-presentation of cell-associated antigens, co-stimulation, and IL12 production over their counterpart DCs (10,29), we hypothesized that cDC1s were required for ReACT-mediated antigen spreading. To test this, we utilized a mixed BMC model that can temporally deplete cDC1s to allow normal development of an antitumor immune response (Supplementary Figure S4A). BMC mice were generated by mixing bone marrow cells from transgenic mice expressing the diphtheria toxin receptor (DTR) (under the control of a CD11c promoter) and Batf3-deficient mice at 1:1 ratio. The mixed bone marrow cells were then transplanted into lethally irradiated C57BL/6 mice to generate CD11c^DTR/Batf3^–/– BMC mice. Because Batf3 is known to be essential for development of cDC1s (30), the reconstituted cDC1s from these BMC mice can only be derived from CD11c^DTR mice, and are, therefore, subject to depletion by administering diphtheria toxin (DT). To create control mice that preserved cDC1s after DT depletion, lethally irradiated C57BL/6 mice were grafted with bone marrow cells from CD11c^DTR and WT mice to generate CD11c^DTR/WT chimeras.

After the successful reconstitution (Supplementary Figure S4B), both CD11cDTR/Batf3^–/– and CD11cDTR/WT BMC mice were inoculated with B16-GP33 tumors, followed by ReACT one week later. To determine the role for Batf3-lineage cDC1s in endogenous T-cell recruitment, depletion of DTR⁺ DCs was performed 5 days and 7 days after ReACT in both groups of mice (Figure 4C). One day after DT depletion, P14 CD8⁺ T cells were adoptively transferred intravenously to the mice, followed by assessment of tumor recruitment 48 hours later (Figure 4C). As expected, a significant reduction in cDC1s was observed in the tumors from Batf3^–/– reconstituted mice compared to WT reconstituted mice (Figure 4D–E). Consequently, although Pmel cells remained unchanged (Figures 4F–G, Supplementary Figure S5A–B), the recruitment of P14 cells into tumors was almost abolished in CD11c^DTR/Batf3^−/− BMC mice (Figure 4H). Collectively, these results indicated that tumor-residing cDC1s were required for ReACT-induced antigen spreading.

ReACT induces tumor-reactive CD8⁺ T cells to differentiate into TRM cells

Batf3-dependent cDC1s promote the generation of tissue-resident memory (T_RM) cells (11,31), which have been shown to provide optimal tumor protection. We previously showed that 60% of mice that cleared tumors after ReACT rejected a local re-challenge (24) and that this protection was dependent on CD8⁺ T cells (Figure 1A–B). Next, we sought to distinguish the relative contribution of CD8⁺ T_RM cells derived from transferred Pmel cells versus those recruited endogenously by ReACT. To test this, we characterized memory T-cell development in a model that tracks antigen spreading CD8⁺ T cells (P14 cells) recognizing a distinct TAA (GP33). P14 T cells declined in the blood two weeks following treatment and underwent a similar contraction as the adoptively transferred Pmel cells (Figure 5A). Consistent with data shown in Figure 1B, one month after tumor eradication, we did not readily detect Pmel cells in circulation, whereas a small portion of P14 cells developed into circulating memory cells (Figure 5A). P14 cells, but not Pmel cells, were also recovered in the lesion area of the skin where the tumor was eradicated (Figure 5B–C). Because these antigen-specific T cells were found within the skin lesion, we further examined their expression of T_RM markers. Almost all P14 cells from the local skin were CD62L⁻ with high expression of CD103 and CD69 compared to splenic CD44⁺ CD8⁺ T cells (Figure 5D–E), which is consistent with T_RM phenotype.

Based on emerging evidence that highlights the important role of T_RM cells in mediating optimal tumor protection (21,32,33), we next assessed whether P14 T_RM cells were capable of providing protection against local recurrence. ReACT-treated mice that cleared tumors were re-challenged with an irrelevant E0771 breast tumor cell line expressing GP33 (the only shared antigen with the previously eradicated B16-GP33 tumor cells). As a control group, naïve mice were inoculated with E0771-GP33 cells and successfully established tumors within two weeks. On the contrary, ReACT-treated mice not only significantly delayed tumor establishment, but also ~60% of mice achieved complete protection against re-challenge (Figure 5F), suggesting successful formation of a memory CD8⁺ T-cell response against escape variants. These data demonstrate that ReACT-induced tumor-specific P14 cells differentiated into memory T cells, including T_RM cells, which were the main contributors of long-term protection against local relapse.

ReACT offers long-term protection against systematic recall

Besides local T_RM cells, circulating memory T cells are also required for ideal antitumor immunity, especially long-term immune protection against distal metastasis (32). To investigate the impact of ReACT on memory cell formation in lymphoid organs, mice that cleared tumors after ReACT were sacrificed one month after tumor eradication. After stimulation with B16-F10 tumor cells, their splenic T cells produced significantly more IFNγ than naïve splenic T cells (Figure 6A–B). The gp100-deleted B16-F10 cell line also could stimulate IFNγ production in splenic T cells from ReACT-treated mice, suggesting that ReACT induced a broader T-cell response to multiple tumor antigens. Next, we explored whether ReACT-induced circulating memory cells could confer protection against systemic metastasis in a well-established lung metastasis model (34). To do this, ReACT-treated (after eradication) or naive mice were i.v. re-challenged with B16-F10 cells and sacrificed 15 days later for examination of the tumor foci in the lungs. As expected, mice with eradicated tumors after ReACT displayed a 30.7-fold reduction in number of lung tumor foci compared to naïve mice, as confirmed by histologic examination (Figure 6C–D). This indicated that ReACT also induced the development of effector memory CD8⁺ T cells for systemic protection against metastases. Overall, we have shown that ReACT could expand the immune response beyond the initial target to include new antigenic targets, which could provide protection against antigen escape and relapse.

Figure 6. Mice with eliminated tumors post ReACT are protected against systemic metastasis.

(A-B) One month after ReACT-mediated tumor eradication, CD8⁺ T cells were isolated from draining lymph nodes, stimulated ex vivo with irradiated B16-F10 cells, and assessed for IFNγ secretion using an ELISPOT assay. IFNγ-producing cells are shown in (A) representative images along with (B) quantification. (C-D) Naïve and ReACT memory mice (two months after tumor clearance) were intravenously injected with 1 × 10⁵ B16-F10 cells and sacrificed 15 days later. (C) Representative images of lungs and (D) quantification of tumor foci in lungs isolated 15 days after tumor challenge. Data represents cumulative results from two independent experiments with n=6/group for (A-B) and n=10/group for (C-D). The significance was determined by t-test. Mean±SEM is shown. **P<0.01, ***P<0.001.

Discussion

CAR T-cell therapy trials and other ACT studies demonstrate that antigen escape has emerged as a major issue (35–37), impacting the durability and efficacy of therapies. In this study, we showed that ReACT induced antigen spreading and expanded the antitumor response from one initial target antigen to multiple TAAs. As a result, this approach significantly prevented outgrowth of antigen escape variants. ReACT also provided optimal priming for endogenous T cells to differentiate into tissue-resident and circulating memory cells that provided long-term protection against local relapse and distant metastasis.

Antigen escape is reported in CD19 CAR T-cell trials in B-cell acute lymphoblastic leukemia (38). The spontaneous mutation and selective expansion of antigen-negative tumor cells in CD19 CAR T-cell therapy is well-documented, and the durability of clinical response is severely limited by the outgrowth of CD19-negative leukemia cells (39). Currently, clinical trials demonstrate that the loss of the target antigen CD19 accounts for 38%−75% of total relapses (35–37). The outgrowth of tumor escape variant cells has also been observed in solid tumors and contributes to acquired resistance of melanoma to vemurafenib (40). Therefore, an innovative therapy is desperately needed to induce an antitumor immune response targeting a broad range of TAAs to overcome this challenge. To combat antigen escape, transferred T cells can be equipped to simultaneously target two TAAs (41–43). Despite the initial success of bispecific ACT, high tumor mutation rate will inevitably result in the emergence of tumor escape variants that lose both targeted antigens. A more reasonable strategy to successfully tackle this problem is engaging endogenous CD8⁺ T cells to recognize and eliminate escape variants. Indeed, the endogenous immune system is capable of antigen spreading, the generation of an immune response against antigens distinct from the initial immune-targeted epitope (4,5). Such a phenomenon has also been observed in a study using a murine CAR model targeting EGFR (44). Mice that cleared EGFR⁺ tumors with the therapy later rejected EGFR⁻ tumors when re-challenged (44). This elegant model demonstrated that antigen spreading can be induced by ACT therapy, and may be important in dealing with tumor heterogeneity and preventing tumor escape via antigen loss variants (45). However, it is unclear how to enhance the incidence and effectiveness of this phenomenon to counter immune escape.

As we previously reported (24), we developed ReACT, which mediates significant initial tumor destruction, providing a source of new TAAs, which can be cross-presented by DCs, priming endogenous CD8⁺ T cells towards an effector phenotype, thus potentially allowing enhanced antigen spreading. Indeed, infectious agents have been shown to induce antigen spreading from microbial to self-epitopes in several autoimmune diseases (5,46,47), which is associated with their ability to activate DCs. Stimulation of Toll-like receptor (TLR) signaling on the cDC1 lineage induces prominent secretion of pro-inflammatory cytokines, IL12p70 (48) and chemokines such as CCL3, CCL4, and CCL5 (49). The combination of T cells causing tumor cell lysis, bacteria-induced recruitment of DCs, and the inflammatory milieu of immune cytokines creates an optimal environment for DC activation and cross-priming in the presence of tumor antigen. Using a model of mice engrafted with a mixture of B16-F10 and target-loss mutant (B16-gp100-KO) cells, we demonstrated that ReACT successfully elicited an endogenous CD8⁺ T-cell response with heightened effector function and reactivated to the surrogate TAAs not targeted by ReACT. Consequently, tumors containing escape mutants effectively regressed or were completely eradicated. Our approach also has potential to overcome baseline tumor heterogeneity. Due to the substantial diversity of tumor cells, single-target ACT often fails to achieve complete eradication of the tumor. ReACT-induced antigen spreading provides an opportunity to eliminate the tumor cells that do not express the originally targeted antigen.

Mechanistically, we also discovered that the expanded T-cell response critically relied on cDC1s, which are required to orchestrate an effective endogenous effector T-cell response (9). These migratory cDC1s depend on a transcription factor, BATF3, and very efficiently cross-present extracellular antigens, particularly cell-associated antigens, to CD8⁺ T cells (49). Therefore, many studies employ Batf3-knockout mice that are deficient in cDC1s to reveal their critical role in T-cell recruitment to the tumor site (9). cDC1s produce chemokines CXCL9 and CXCL10, the corresponding ligands for CXCR3 on effector T cells, which mediates their infiltration into the tumor (9). Our results are also supported by reports on Listeria monocytogenes, the bacteria used in the ReACT, showing that cDC1s are found to not only recruit effector memory T cells but also to reactivate them at the site of infection (50,51). Consistent with these studies using the same pathogen, we have clearly demonstrated that ReACT failed to induce effector CD8⁺ T cell migration into tumors in the absence of cDC1s. Lack of Batf3 in these mice led to a selective loss of cDC1s in the tumor site, whilst all other CD11b⁺ DC subsets remained undisturbed. When these mice were challenged with B16-GP33 tumors and treated with ReACT, a significant reduction in the intratumoral endogenous CD8⁺ T cells was seen, which is crucial for overcoming antigen escape.

Optimal tumor immunotherapy should offer long-term protection against relapse by generating a potent memory CD8⁺ T-cell response, which requires specific subsets of DCs to provide optimal priming (11). It has been shown that Batf3-driven DCs provide unique priming signals including IL12, IL15, and CD24-mediated co-stimulation for differentiation of skin T_RM cells (11). Compared to circulating memory T cells, T_RM cells respond faster to stimulation and are equipped with a more potent effector response, thus providing instant protection against local re-challenge (12,15,17). In our study, we demonstrated that ReACT generated functional endogenous T_RM cells in skin following tumor elimination, which contributes to its robust recall response, resulting in tumor rejection. We also provided evidence on the formation of circulating memory T cells in mice with ReACT-eradicated tumors, which played a vital role in preventing distal metastasis.

Despite the translational potential of ReACT, several limitations and potential side effects must also be considered and addressed. First, therapy is based on the local injection of bacterial material, therefore, the targeted tumor has to be relatively accessible and over a certain threshold of size. Consistent with other intratumoral immunotherapies (52), our preliminary study also revealed that ReACT could induce an antitumor immune response against the distant tumor, which requires further characterization in future studies. Secondly, antigen spreading can be beneficial to the host by resulting in protection against antigen-negative tumor cells, but it carries the risk of autoimmunity (21). No obvious adverse effects and health issues were observed in treated mice. Occasionally, we did observe that few mice developed mild vitiligo that was confined in the local region of tumor inoculation, without affecting their health condition. Consistently, a study has elegantly shown that skin-resident memory T-cell responses to melanoma are generated naturally as a result of vitiligo (53). Therefore, future work must be done to eliminate the possibility of immune response specificity spread to self-antigens, causing serve inflammation and tissue damage. As to the choice of pathogens, because multiple genera of bacteria, such as Clostridia and Salmonella, have been clinically verified for safety (54), future studies are needed to determine the efficacy of ReACT using other pathogens with proper tropisms to various types of cancer for broader clinical applications.

Overall, although the therapeutic efficacy of ACT could be impeded by tumor-mediated immune escape via loss of targeted antigen expression, we presented a possible solution to this challenge. ReACT, an approach that promotes endogenous CD8⁺ T-cell priming with TAAs via cDC1s effectively prevented the outgrowth of antigen escape variants. We also demonstrated that ReACT optimized T_RM development, forming long-term immunological memory to prevent cancer relapse.