Harnessing anti-cytomegalovirus immunity for local immunotherapy against solid tumors

Inflationary and Noninflationary MCMV–Specific CD8⁺ T Cells Infiltrate Solid Tumors.

HCMV infection induces polyfunctional CD8⁺ T cell responses that increase with age and remain active in cancer patients (20). MCMV infection faithfully mimics CMV infection in humans characterized by broad T cell reactivity and memory inflation (22). First, we used MCMV-infected C57BL/6 mice to confirm the inflationary and noninflationary nature of CD8⁺ T cell responses against major histocompatibility complex-I (MHC-I) epitopes derived from the IE3 and m45 MCMV proteins (22). MHC-I tetramer staining analysis on gated CD8⁺ T cells confirmed strong responses against both inflationary IE3 and noninflationary m45 epitopes (Fig. 1 A and B). Indeed, IE3-specific CD8⁺ T cells increased between 1 and 5 mo after infection, whereas the m45-specific CD8⁺ T cells remained stable in blood (Fig. 1B). Inflationary IE3-specific CD8⁺ T cells displayed a terminally differentiated phenotype characterized by low expression levels of CD127 and CD62L molecules (23). In contrast, noninflationary m45-specific CD8⁺ T cells displayed a mixed effector-memory/central memory phenotype characterized by high expression of CD127 and a bimodal expression pattern of CD62L (23) (Fig. 1A). Restimulation of blood samples with a mixture of seven MHC-I MCMV epitopes (IE3, m38, m45, m57, m139, m141, and m164) resulted in a notable production of IFN-γ (mean positive, 11%), TNF-α to a lesser extent (mean positive, 5.5%), and no IL-2 (Fig. 1C) by CD44⁺CD8⁺ T cells, confirming the high magnitude of the responses against MCMV. Despite the striking differences in the memory phenotype of IE3- and m45-specific CD8⁺ T cells, both inflationary and noninflationary CD8⁺ T cells produced significant amounts of IFN-γ (Fig. 1D and SI Appendix, Fig. S1). Cytokine production was also assessed by MCMV-specific CD44⁺CD4⁺ T cells restimulated in vitro with m09 and m139 MHC-II MCMV-derived peptides and found to be one order of magnitude lower compared with the CD8⁺ T cell responses (Fig. 1E).

Next, we assessed the tissue distribution of inflationary and noninflationary MCMV-specific T cells in latently infected mice challenged with subcutaneous tumors. C57BL/6 mice were inoculated subcutaneously 6 mo after MCMV infection, with TC-1 tumor cells that express human papillomavirus type 16 (HPV16) E6 and E7 oncogenes and mutated H-Ras (Fig. 2A). By 12 d, established tumor masses were heavily infiltrated with both inflationary (IE3) and noninflationary (m45) MCMV-specific CD8⁺ T cells without i.t. injection of MCMV peptide epitopes. Surprisingly, the infiltration of tumors by MCMV-specific CD8⁺ T cells was comparable to infiltration in the salivary glands, a major tissue reservoir for CMV (24) (Fig. 2B). Indeed, tumors were more infiltrated by MCMV-specific CD8⁺ T cells than were secondary lymphoid organs, suggesting that MCMV-specific CD8⁺ T cells were poised to traffic into a fast-growing tumor. We asked whether these tumor-infiltrating, MCMV-specific CD8⁺ T cells were bona fide resident cells by assessing in tetramer⁺ CD8⁺ T cells the expression of CD69 and CD103, markers of tissue residence and intraepithelial lymphocytes, respectively (Fig. 2B). Unlike spleen and lymph node, CD69 was expressed by the majority (55%) of the tumor-infiltrating, MCMV-specific CD8⁺ T cells, but CD103 was not expressed (Fig. 2 B and C). These data suggest that a substantial fraction of the tumor-infiltrating, MCMV-specific CD8⁺ T cells had differentiated into resident cells. In contrast to tumors, salivary gland–infiltrating, MCMV-specific CD8⁺ T cells did express CD103. This difference might be due to the absence of detectable viral replication in the tumors, as no viral DNA or mRNA was found in tumor samples, but both were found in salivary gland tissues (SI Appendix, Fig. S2). Alternatively, TC-1 tumors might already be too dedifferentiated to provide tissue residency signals typical of a normal epithelium, as bulk CD8⁺ T cells did not express CD103 (24).

Intratumoral Administration of MCMV Epitopes Induces Broad Cellular Infiltration and Local Immune Activation of the Tumor Microenvironment.

We next sought to determine whether tumor-infiltrating, MCMV-specific T cells could be activated in situ with their cognate peptide epitopes with or without pI:C and whether in situ activation of MCMV-specific T cells could alter the tumor immune microenvironment. We chose pI:C as a molecular adjuvant because it can induce strong secretion of inflammatory cytokines, in particular, type 1 IFN, by TC-1 cells in vitro after transfection (SI Appendix, Fig. S3). Latently MCMV-infected C57BL/6 mice were inoculated with TC-1 tumor cells subcutaneously. Once TC-1 tumor volume reached 100 mm³, they were injected i.t. with IE3, m38 and m45 MCMV MHC-I peptides (1 μg each) or m139 MCMV MHC-II peptide (3 μg) with or without pI:C (50 μg). Tumors were injected three times (Fig. 3A) and 2 d after the third injection, we analyzed by microscopy the consequences of i.t. injection of MCMV-derived peptide epitopes on the tumor tissues. First, MCMV-derived MHC-I–restricted epitopes with or without pI:C induced wide areas of necrotic tissue (Fig. 3B), supporting the notion that MCMV-specific CD8⁺ T cells exerted cytotoxic functions in the tumor. Lymphocytic infiltration was also noted and confirmed by immunofluorescence analysis of CD8⁺ and CD4⁺ T cell infiltrates. In contrast, tissue sections from tumors treated with MCMV-derived MHC-II peptide plus pI:C displayed both CD4⁺ and CD8⁺ lymphocytic infiltration but no necrotic tissue (Fig. 3B).

Next, using the Nanostring mouse nCounter PanCancer Immune Profiling assay (Nanostring), we characterized the tumor cellular infiltration after the injection of MCMV peptide epitopes, summarized as a heat map of z-scores for each indicated cell type (Fig. 3C). Intratumoral injection of MHC-I peptide with or without pI:C and MHC-II peptide plus pI:C induced an increase in CD8, cytotoxic, and Th1 cells that was also associated with cell types that were not initially targeted by the MHC-I– and MHC-II–restricted peptides (Fig. 3C). Bystander natural killer (NK) cells, B cells, neutrophils, and macrophages were also increased upon i.t. injection of MHC-I–restricted peptide with or without pI:C and MHC-II–restricted peptide plus pI:C (Fig. 3C and Dataset S1). Notably, the dendritic cell score was only increased in the groups treated with pI:C alone or in combination with MCMV-derived epitopes (Fig. 3C). Together, these data indicate that i.t. injection of MCMV-derived epitopes causes profound changes in the immune activation status and cellular composition of the tumor microenvironment.

Using the same Nanostring dataset as previously described, we further analyzed the modification of the tumor immune microenvironment after i.t. injections (Fig. 4A). After normalization, differential expression of individual genes in each group was analyzed relative to saline-treated samples as a reference (Fig. 4B and Dataset S2). Intratumoral injection with pI:C induced minimal changes in the number of genes up- or down-regulated (Fig. 4B), and injection of MHC-II epitope alone had no effect. However, the combination of MHC-II epitope with pI:C led to the massive modification of the tumor gene expression profile with the up- and down-regulation of 230 and 3 genes, respectively (Fig. 4B). Interestingly, the i.t. injection of MHC-I–restricted MCMV epitopes without pI:C induced the up- and down-regulation of 359 and 43 genes, respectively (Fig. 4B). When combined with pI:C, the MHC-I epitopes induced changes in the expression of genes whose numbers were similar to MHC-I peptide alone, with the up- and down-regulation of 309 and 49 genes, respectively (Fig. 4B).

Heat-map visualization and hierarchical clustering of gene expression of all samples indicated that samples from tumors injected with MHC-I peptide with or without pI:C and MHC-II peptide plus pI:C clustered together (SI Appendix, Fig. S4A). However, the magnitude of gene regulation was less pronounced in groups treated with MHC-II peptide plus pI:C than in groups treated with MHC-I peptide with or without pI:C. Specifically, single-gene analysis showed pronounced up-regulation of genes associated with T cell function and activation, such as T-bet, granzyme k, CXCR3, and PD1, in tumor samples injected with MHC-I restricted with or without pI:C or MHC-II plus pI:C (Fig. 4C). H2-D gene expression was increased in all tumor samples injected with MHC-I–restricted epitopes with or without pI:C, MHC-II–restricted epitopes plus pI:C, and, to a lesser extent, with pI:C alone (Fig. 4C). Expression of the CD40Lg gene was increased only in the samples from tumors injected with MHC-II–restricted epitopes plus pI:C, and CD11c was increased in groups injected with MHC-II–restricted epitopes plus pI:C and, to a lesser extent, with pI:C alone (Fig. 4C). The i.t. treatment with MHC-I–restricted epitopes induced the up-regulation of Arg1, S100A8, and CD36 genes, which are involved in response to tissue damage and/or sterile inflammation (Fig. 4C and SI Appendix, Fig. S4B). Together, these results suggest that local immune activation with tissue-damage signatures were associated with MHC-I peptide i.t. injection, whereas a local immune activation signature alone was associated MHC-II peptide plus pI:C injection.

Intratumoral Injection of MHC-I–Restricted MCMV Epitopes Delays Tumor Growth.

Next, we investigated whether the local immune activation and tissue-damage response caused by MCMV peptide i.t. injection could affect tumor growth and survival in the TC-1 subcutaneous model. In latently MCMV-infected mice, TC-1 subcutaneous tumors were injected six times with a mixture of inflationary (IE3), noninflationary (m45), and intermediate (m38) MHC-I minimal epitopes with or without pI:C (Fig. 5A). Compared with i.t. saline injection, i.t. injection of 1 μg of each of MHC-I–restricted peptide mixture with or without pI:C caused significant delay in tumor growth and improved survival (Fig. 5 B and C). In addition, pI:C alone caused a significant, but less pronounced, delay in tumor growth.

However, we observed occasional acute toxicity toward the end of the series of i.t. injections of MHC-I–restricted MCMV epitope in combination with pI:C, including death before the control animals succumbed to their tumors (Fig. 5C). To determine whether decreasing the dose of injected peptide would remain effective while reducing the acute toxicity, we assessed i.t. injection with lower doses of MHC-I–restricted peptide with or without pI:C. The peptides alone at reduced doses of 0.1 μg or 0.01 μg were equally able to significantly delay tumor growth without toxicity (Fig. 5D). When combined with pI:C, reduced doses of MHC-I–restricted peptides were devoid of immune-related toxicity and still able to induce long-term survival after the end of the treatment (Fig. 5 D and E). The observation that a low dose of 0.01 μg of peptide retained antitumor effect without toxicity suggests potential for scaling up in a context of clinical evaluation.

Sequential I.T. Administration of MHC-II– and MHC-I–Restricted MCMV Epitopes Promotes Long-Term Tumor Control and Antitumor Immunity.

The Nanostring analysis of the tumor microenvironment indicated that the consequences of the injection of MHC-I– and MHC-II–restricted peptide were substantially overlapping in terms of gene expression. However, while injection of MHC-I–restricted peptides induced a T cell/IFN-γ signature together with a strong tissue-damage response and sterile inflammation, MHC-II–restricted peptide injection was not associated with a tissue-damage response and induced a T cell/IFN-γ signature together with increased antigen-presentation functions. We reasoned that sequential injection of the MHC-I– and MHC-II–restricted peptides might help to harness both local immune activation and tumor killing that would favor a strong antitumor T cell response and long-term control. To evaluate this hypothesis, TC-1 subcutaneous tumors in MCMV-infected mice were injected six times with a mixture of MHC-I (IE3, m38, m45) or MHC-II (m139) MCMV epitopes plus pI:C or, sequentially, MHC-I plus pI:C followed by MHC-II plus pI:C (n = 3 times for each peptide) or conversely with MHC-II plus pI:C followed with MHC-I plus pI:C (Fig. 6A). Tumor growth was significantly delayed in all groups injected with MCMV peptides (Fig. 6B). Interestingly, complete regression and long-term cure were only observed in the groups that received MHC-II peptide alone or in combination with MHC-I peptides (Fig. 6C). Of note, i.t. injection of MHC-I peptides led to the expansion of m45-specific (noninflationary), but not IE3-specific (inflationary), CD8⁺ T cells (SI Appendix, Fig. S5 B and C).

To assess the induction of tumor antigen–specific immunity, we measured the presence of circulating CD8⁺ T cells in blood against the HPV16 E7 oncoprotein by MHC tetramer staining (Fig. 6D). Interestingly, high levels of E7-specific CD8⁺ T cells were detected in all groups that received i.t. injection of the MHC-II peptide (Fig. 6D). More remarkably, 50% of mice in the group that received MHC-II peptide followed by MHC-I peptide developed a high level of E7-specific CD8⁺ T cells (1% to 10% of total CD8⁺ T cells). The induction of high-level tetramer⁺CD8⁺ T cells correlated with complete cure and long-term survival, compared with partial responders (Fig. 6D, circled data point, and SI Appendix, Fig. S5C).

Next, we sought to determine whether the induction of CD8⁺ T cell responses against E7 oncogenes or unidentified tumor antigens was associated to long-term immunity. Four months after treatment, tumor-free mice from the first challenge were rechallenged subcutaneously with 50,000 TC-1 cells (Fig. 6E). Age-matched naïve mice were used as a tumor-challenge positive control. While all naïve control mice developed palpable subcutaneous tumors within 12 wk, we did not detect any palpable tumor in the long-term survivors, suggesting that long-term antitumor immunity had been established upon treatment with MHC-II peptide plus pI:C during the primary tumor challenge (Fig. 6E).

Tumor Regression Upon I.T. Injection with MCMV MHC-I–Restricted Epitopes Is Preferentially Conferred by Noninflationary Epitopes.

We next asked whether inflationary (IE3 or m38) or noninflationary (m45) MCMV-specific CD8⁺ T cells preferentially contributed to tumor regression upon intratumoral recall. TC-1 tumor–bearing mice were injected i.t. with IE3, m38, or m45 alone, or as pool together with pI:C and the MHC-II (m139) peptide (Fig. 7A). Groups injected with the combination containing only the noninflationary epitope (m45) showed tumor control comparable with the combination of the mixture of MCMV peptides (IE3, m38, and m45) but with no treatment-related toxicity. The tumor control was even superior to that of the groups injected with MHC-II peptide (m139) alone or combined with the inflationary peptides, IE3 or m38 (Fig. 7 B and C). Interestingly, production of IFN-γ and TNF-α was readily observed after ex vivo peptide stimulation by noninflationary (m45) or inflationary (IE3 or m38) MCMV-specific CD8⁺ T cells, but m45-specific CD8⁺ T cells displayed a slightly higher functional avidity (SI Appendix, Fig. S6).

To track MCMV-specific CD8⁺ and CD4⁺ T cells in tumors and tumor-draining lymph nodes, we performed tetramer staining after three i.t. injections of MHC-I (IE3, m38, and m45) and MHC-II (m25) peptides for which IA^b tetramers were available. Noninflationary m45-specific CD8⁺ T cells were specifically expanded after i.t. injection with the corresponding peptide alone, whereas their inflationary IE3-specific counterparts remained unchanged in the tumors and, to a lesser extent, in the draining lymph nodes (SI Appendix, Fig. S7 A and B). Together, these data are consistent with the notion that inflationary T cells display a more terminally differentiated phenotype, whereas the noninflationary cells display a memory phenotype. We assessed the proliferative response of IE3- and m45-specific CD8⁺ T cells by measuring the expression of the Ki67 nuclear antigen, a proliferation marker. Surprisingly, Ki67 was equally induced in inflationary and noninflationary CD8⁺ T cells in tumor-draining lymph nodes (SI Appendix, Fig. S7C) and in tumors (SI Appendix, Fig. S7D) after i.t. peptide injection. Therefore, the lack of IE3-specific CD8⁺ T cell expansion in blood (SI Appendix, Fig. S5B) or in the tumor (SI Appendix, Fig. S7B) upon i.t. injection might result from decreased survival capacity rather than impaired proliferative potential (25).

To track the recall of CD4⁺ T cell responses in vivo upon i.t. injection with a CMV m25 peptide (SI Appendix, Fig. S7 A and B), we used H2-A^b tetramer harboring the MHC-II m25 MCMV epitope. We observed the expansion of m25-specific CD4⁺ T cells in both draining lymph nodes and tumor tissue, and up-regulation of Ki67 (SI Appendix, Fig. S7 C and D). These findings suggest that the proliferation and expansion of the CD4⁺ T cells might contribute to the antitumor effect, as for m45-specific CD8⁺ T cells.

Intratumoral Injection of MCMV-Derived Epitopes in Poorly Immunogenic B16-F10 Melanoma Delays Tumor Growth and Induces T Cell Infiltration and Immune Activation.

We next assessed the potential of i.t. injection of MCMV-derived peptide epitopes in an immunologically “cold” tumor model. We used the well-established melanoma model B16-F10, which does not respond to immune checkpoint blockade and is poorly immunogenic despite a high mutation burden. MCMV-infected, B16-F10–bearing mice were treated by repeated i.t. injection of combinations of MCMV-derived MHC-I– and MHC-II–restricted epitopes with pI:C (Fig. 8A). Treatment with MHC-I (m45) and MHC-II (m139) peptides significantly delayed tumor growth and improved survival compared with saline-treated groups (Fig. 8 B and C), but treatment with pI:C alone did not delay B16-F10 tumor growth (SI Appendix, Fig. S8). Flow cytometry analysis of the tumor immune infiltrate of B16-F10 tumors after i.t. treatment with MCMV-derived peptides showed a pronounced increase in CD8⁺ T cells (CD3⁺CD8⁺) and m45-specific CD8⁺ T cells, and a modest increase in NK cells (CD3⁻NK1.1⁺). The treatment led to a decrease in infiltrating B cells (CD19⁺), granulocytic cells (CD11b⁺LyG⁺), and CD4⁺ T cells (CD3⁺CD4⁺). Infiltrating monocytic cells (CD11b⁺Ly6C⁺Ly6G⁻) and dendritic cells (MHC-II⁺CD11c⁺) remained unchanged (Fig. 8D and SI Appendix, Fig. S9). To further characterize the immune modulation in the B16-F10 tumor tissue, we used the multiplex cytokine/chemokine measurement kit (Biolegend, CRS Legendplex). We found that i.t. treatment with MCMV-derived peptide led to the increase in tumor-tissue concentration of IFN-γ, CXCL9, and CXCL10 (Fig. 8E), which have been mostly associated with antitumor response, and increase of CCL2 and CCL4 (SI Appendix, Fig. S10A) that can exert pro- or antitumor effects depending on the immune status of the tumor (26). Interestingly, the i.t. injection in B16-F10 tumors led to an increase of plasma cytokines and chemokine (Fig. 8F and SI Appendix, Fig. S10B). Surprisingly, although TNF-α was not significantly induced in the treated tumor, it was induced in the plasma of treated mice (SI Appendix, Fig. S10). Finally, injection of pI:C alone did not significantly alter the tumor or plasma concentrations of any of the analyzed cytokines and chemokines (SI Appendix, Fig. S10). Together, these results show that MCMV-specific T cells remain functional in an immunologically cold tumor environment and that the MCMV-derived peptides administered i.t. can cause profound activation of the tumor immune microenvironment and potent antitumor activity.

Finally, we investigated whether i.t. injection of MCMV epitopes could lead to a more pronounced antitumor effect, using a spontaneous colon adenocarcinoma model, MC-38, which harbors a high number of mutations and responds to immune checkpoint blockade (27). As in the previously assessed tumor models, i.t. injection of a combination of MCMV-derived peptides with pI:C provoked, in some mice, a sustained reduction in subcutaneous MC-38 tumor size, leading to its complete clearance that appeared more pronounced than in the B16-F10 model (SI Appendix, Fig. S11). However, we did not observe increased CD8⁺ T cell reactivity to the previously reported MC-38 neoepitopes derived from Adpgk- and Resp1-mutated antigens (SI Appendix, Fig. S11), which represent a fraction of the neoepitope repertoire that remains to be identified in MC-38 (28).

Cell Lines and Viruses.

TC-1 cells were obtained from Dr. Tzyy-Choou Wu (The Johns Hopkins University) (35) and maintained in Roswell Park Memorial Institute (RPMI; GIBCO) containing 10% (volume per volume) fetal bovine serum (FBS; SIGMA), l-glutamine (GIBCO), and 100 μg/mL G418 (InvivoGen). B16-F10 cells were acquired from ATCC and maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% FBS. MC-38 cells (60) were obtained from Dr. James Hodge (National Cancer Institute, NIH) and maintained in DMEM (GIBCO) containing 10% FBS, l-glutamine, nonessential amino acids (GIBCO), sodium pyruvate (GIBCO) and HEPES (GIBCO). Mouse M2-10B4 fibroblasts were purchased from ATCC and maintained in RPMI-Glutamax medium (Life Technologies) containing 10% FBS. Cell lines were negative for mycoplasma and maintained in culture at 37 °C, 5% CO₂. Highly virulent salivary gland–derived MCMV stocks were obtained from salivary gland homogenates of 6-wk-old infected BALB/c mice after serial passages (SI Appendix, Materials and Methods).

Mice and Experimental Procedures.

BALB/c and C57BL/6 female mice were purchased from the Jackson Laboratory and housed under specific pathogen-free conditions at the animal care facilities of the National Cancer Institute. All animal protocols used in this study were approved by the National Cancer Institute Animal Care and Use Committee. Mice aged 8 to 10 wk were infected intraperitoneally with MCMV (5 × 10³ pfu) diluted in 200 μL of PBS.

Tumor cells were injected subcutaneously (2.5 × 10⁵ TC-1 and 5 × 10⁵ MC-38) or intradermally (5 × 10⁵ B16-F10) into the right flank of MCMV-infected mice at least 4 mo after MCMV infection. At this time, MCMV is latent with no detectable replication (61). Tumors were measured twice per week with digital calipers, and tumor size was calculated according to the formula W² × L/2 (where L = length and W = width). For survival experiments, humane end point was based on tumor volume (>2,000 mm³) or average diameter (>20 mm) and overall condition of the animal.

When tumors reached a volume between 50 and 150 mm³, mice were randomized into different experimental treatment groups and injected i.t. (day 0) with the following formulations: MCMV minimal peptide (Genscript; SI Appendix, Table S1) epitopes to recall MCMV-specific CD8⁺ and CD4⁺ T cells referred to as MHC-I–restricted (MHC-I peptides) or MHC-II–restricted (MHC-II peptides), as described in SI Appendix, Table S1. MHC-I or MHC-II peptides admixed with 50 μg pI:C) (low molecular weight; InvivoGen), a TLR 3 and RIG-I–like receptor agonist that potentiates tumor antigen cross-presentation (62). Mice injected with saline (InvivoGen) or pI:C alone were used as controls. For each injection, the formulation was diluted with saline to 50 μL. Tumors were injected three times per week for 1 wk (n = 3 doses) or 2 wk (n = 6 doses).

Preparation of Cell Suspensions for Flow Cytometry and Elispot.

Tumors, lymph nodes, and spleens were finely minced and incubated at 37 °C at 250 rpm in RPMI supplemented with 2% FBS, 0.5 mg/mL Collagenase A (Roche), and 0.1 mg/mL DNase1 (Roche) for 30 min (tumor) or 15 min (lymph nodes and spleen). After incubation, single suspensions were filtered through a 70-μm mesh. Remaining erythrocytes were lysed in ammonium–chloride–potassium buffer (Life Technologies). Circulating leukocytes were obtained from ammonium–chloride–potassium–treated whole-blood samples. Single-cell suspensions were stained with fluorochrome-conjugated antibodies and MHC tetramers for flow cytometry analysis of T cell responses and for the characterization of the tumor leukocyte infiltrate (SI Appendix, Materials and Methods). Single-cell suspensions were incubated with the indicated peptide epitopes (SI Appendix, Table S1) for in vitro, antigen-specific T cell stimulation (SI Appendix, Materials and Methods) for intracellular cytokine staining or ELISPOT. For details, see SI Appendix, Materials and Methods.

Tissue Processing for RNA and Protein-Content Quantification and Microscopy Analyses.

For gene expression analysis, tumor tissues were collected 48 h after the last of three i.t. injections and flash-frozen in liquid nitrogen. RNA was isolated using TRIzol and stored at −80 °C until analysis with the Pancancer immune profiling panel (nCounter, Nanostring). For protein content analysis, ethylenediaminetetraacetic acid–potassium–treated plasma and tumor-tissue lysates were obtained 24 h to 36 h after the last i.t. treatment and stored at −80 °C until analysis with the cytokine-release syndrome panel (Legendplex, Biolegend). For microscopy analysis, fresh tumor tissues were fixed in paraformaldehyde, incubated in sucrose, and then embedded in optimal cutting temperature compound (Tissue-Tek). Frozen tissue blocks were stored frozen at −80 °C until histology and immunofluorescence analyses. For details, see SI Appendix, Materials and Methods.

Statistical Analyses.

Statistical analyses were performed using Prism software (GraphPad Software, Inc.). Paired or unpaired Mann–Whitney tests were used to analyze statistical differences between two groups. For multiple comparisons, Dunn’s test was used. The Mantel–Cox test was used for survival analysis. Treatments were compared using the extra sum-of-squares F test. Values of P < 0.05 were considered significant.