Research ArticleCancer Vaccines

Simultaneous Targeting of Toll- and Nod-Like Receptors Induces Effective Tumor-Specific Immune Responses

See allHide authors and affiliations

Science Translational Medicine  08 Feb 2012:
Vol. 4, Issue 120, pp. 120ra16
DOI: 10.1126/scitranslmed.3002868


Toll-like receptor (TLR) ligands are increasingly being used as adjuvants in cancer vaccine trials to harness innate immunity and prime effective antitumor immune responses. Despite some success, enhancing tumor antigen presentation, promoting a protective antitumor response, and overcoming the immunosuppressive tumor microenvironment pose considerable challenges that necessitate further improvements in vaccine design. Here, we show that expression of the TLR ligand flagellin within tumor cells constitutes an effective antitumor vaccination strategy that relies on simultaneous engagement of TLR5 and the Nod-like receptors (NLRs) NLRC4/NAIP5 (neuronal apoptosis inhibitory protein 5) by flagellin along with associative recognition of tumor antigen for optimal antigen presentation to T cells. Although TLR5 signaling was critical for mediating rapid macrophage-dependent clearance of flagellin-expressing tumor cells in vivo, TLR5 and NLRC4/NAIP5 were equally important for priming antitumor CD4+ and CD8+ T cells and suppressing tumor growth. Vaccination with irradiated flagellin-expressing tumor cells prevented tumor development, and disrupting flagellin recognition by TLR5 or NLRC4/NAIP5 impaired protective immunization against an existing or subsequent tumor. Our findings delineate a new strategy to induce anticancer immune responses consisting of introducing microbial structures with dual TLR and NLR stimulatory activity into tumor cells. This ensures recognition of tumor-derived antigen within the inflammatory context of microbial recognition and additionally activates both the phagocytic and the cytosolic pathways of innate immune defense against the tumor.


The very nature of cancer development from normal cells presents several challenges to the immune system (1). Transformed cells lack the molecular structures characteristic of microbial cells [collectively known as pathogen-associated molecular patterns (PAMPs)] (2) and thus cannot engage host defense pathways that control protective adaptive immunity (3). Also, shared protein expression profiles between normal and transformed cells entail immune tolerance to transformed cells. Furthermore, immunity is not induced to tumor-specific proteins in the absence of signals that induce inflammation and T cell costimulatory molecule expression (46). To overcome these challenges, cancer vaccination protocols have harnessed the highly efficient ability of dendritic cells (DCs) to internalize antigen, migrate to secondary lymphoid organs, and present antigen-derived peptides on major histocompatibility complex (MHC) molecules for recognition by T cells (79). Genetically modified DCs engineered to express defined tumor antigens or immunomodulatory molecules, or pulsed with tumor-derived peptides or antigens, tumor cell lysates, or apoptotic or necrotic cells, have all been widely used in preclinical and clinical trials for immunotherapy of various types of cancer (9). Vaccination protocols have also sought to activate DCs through incorporation of PAMPs that engage Toll-like receptors (TLRs) (3, 10). These approaches have unveiled new hurdles (1, 11, 12). Some of these include inefficacy of TLR ligands as single therapeutic agents, inability to overcome the suppressive tumor microenvironment that inhibits effector CD8+ T cell functions, and lack of CD4+ T cell help as a result of defective MHC class II (MHC-II) presentation of tumor-derived peptides (1315).

Our recent studies have highlighted the complexity of TLR-mediated control of DC function in antigen presentation (16, 17). DCs establish the microbial origin of a given protein via its physical association with a PAMP. Physical association, either through direct fusion or via coexpression within one phagocytic cargo, ensures delivery of the protein with the PAMP into the same phagosome and its associative recognition with the PAMP (3, 1618). We have shown that this leads to preferential presentation of peptides derived from that protein by MHC-II (17). PAMP engagement of TLRs enhances MHC-II (17) and MHC-I (19, 20) presentation by mediating cleavage of the invariant chain, which prevents binding of peptides to MHC-II, specifically within MHC-II+ LAMP-2+ phagosomes carrying TLR ligands (17), and by recruitment of the transporter associated with antigen processing (TAP), essential for loading peptides onto MHC-I, to early endosomes carrying exogenous antigens (19). On the basis of these findings, we hypothesized that physical introduction of a PAMP within tumor cells would allow associative recognition of tumor-derived antigens with the PAMP, leading to enhanced presentation of tumor-derived peptides by MHC-I and MHC-II within the context of inflammatory cytokines and costimulatory molecules.

In choosing a PAMP to introduce into tumor cells, we picked flagellin, a protein constituent of bacterial flagellar protofilaments. In its monomeric form, flagellin is recognized by TLR5 via direct interaction with the flagellin D1 domain (21), which induces inflammatory signaling dependent on the adaptor MyD88 (21, 22). Flagellin is also recognized by two members of the Nod-like receptor (NLR) family, NLRC4 (23, 24) and NAIP5 (neuronal apoptosis inhibitory protein 5) (25, 26), which hetero-oligomerize to mediate assembly of a cytosolic caspase-1–activating multiprotein complex called the inflammasome (27). Although NLRC4 can directly recruit procaspase-1 through its caspase-activating and recruitment domain (CARD), NAIP5 lacks a CARD and associates with NLRC4 to recruit procaspase-1 (27). Subsequent activation of caspase-1 is required for processing and release of several inflammatory cytokines (27). Thus, we reasoned that expression of flagellin within tumor cells would simultaneously lead to (i) associative recognition of flagellin with tumor-derived antigens, (ii) engagement of TLR5 and subsequent enhancement of the presentation of tumor-derived antigens within an inflammatory and costimulatory context, and (iii) engagement of the inflammatory NLR pathway of host defense.

Here, we show that immunization of mice with flagellin-modified tumor cells protected against growth of parental tumors, irrespective of their expression of foreign antigens, and induced robust tumor-specific CD4+ and CD8+ T cell responses when compared to immunization with a standard regimen of co-administrating flagellin with tumor cells. Although TLR5 signaling was critical for T cell priming, NLRC4 and NAIP5 were equally important not only in the subsequent adaptive immune response against the tumor, but also in mediating tumor rejection and establishing protective antitumor immune memory. Our results reveal an unappreciated role for TLR5 cooperation with NLRC4/NAIP5 in controlling adaptive immunity and have important implications for designing new and effective immunotherapeutic strategies that rely on a combination of TLR and NLR ligands.


Introduction of flagellin within tumor cells triggers the host innate immune response in a MyD88-dependent manner

To ensure associative recognition of a tumor-derived antigen with flagellin, we used a recombinant retroviral vector encoding a fusion protein between the model antigen ovalbumin (OVA) and flagellin from Salmonella typhimurium (StfOVA) (17). We generated EL4 thymoma cell lines expressing either OVA alone or StfOVA. Transduced EL4 cells were sorted by FACS (fluorescence-activated cell sorting) for high expression of retroviral-encoded green fluorescent protein (GFP) and subsequently expressed similar amounts of OVA (fig. S1, A and B). Flagellin expression was detected only in EL4-StfOVA cells (fig. S1, C and D). TLR5 was not detected in the tumor cell lines used herein (fig. S1E).

We assessed the innate immune response to flagellin expressed by EL4-StfOVA by injecting EL4-OVA or EL4-StfOVA cells into the peritoneal cavity of syngeneic mice, where macrophages predominate at steady state [Fig. 1, A and B, Macrophages (F4/80high) panel, phosphate-buffered saline (PBS) groups]. Unlike EL4-OVA cells, which persisted and constituted 15 to 30% of the cells within the peritoneal cavity [Fig. 1, A, B (tumor cells), and D, and fig. S2A], CD45+GFP+ EL4-StfOVA cells were rapidly eliminated within 12 hours, a time frame that excluded a role for adaptive immunity in clearing the tumor (Fig. 1, A and B; 3 hours in fig. S2A). Elimination of EL4-StfOVA was associated with significant increases in the numbers of CD11b+Ly6G+ neutrophils, CD11bintF4/80int macrophages [Fig. 1, A, wild-type (Wt) panels, and B], and NK1.1+CD49b+ natural killer cells, which were Ly49C/I/F/H+ Ly49A (Fig. 1B and fig. S2B). Differences in the absolute numbers of CD11bhighF4/80high cells were not significant (Fig. 1B), and no CD11c+ DCs were detected (fig. S2B). Recruitment of CD11b+Ly6G+ neutrophils and CD11bintF4/80int macrophages in response to EL4-StfOVA cells was severely impaired in Myd88−/− mice, which further correlated with persistence of EL4-StfOVA cells (Fig. 1A and fig. S2C), demonstrating that clearance of flagellin-expressing tumor cells from the peritoneal cavity was dependent on MyD88 signaling. Co-injection of recombinant flagellin (RecFLA) with EL4-OVA cells into the peritoneal cavity also promoted MyD88-dependent eradication of tumor cells (Fig. 1A, OVA + RecFLA panels). Therefore, the flagellin expressed by EL4 tumor cells was functional, triggering rapid MyD88-dependent innate immune responses. This response could mediate clearance of flagellin-expressing EL4 tumor cells from the peritoneal cavity regardless of whether flagellin was expressed in tumor cells or co-administered with tumor cells.

Fig. 1

The innate immune system targets flagellin-expressing tumor cells. EL4-OVA, EL4-StfOVA, or EL4-OVA cells (3 × 106) co-injected with 2 ng of RecFLA as indicated were injected into the peritoneal cavity of syngeneic mice. (A) FACS analysis of cells from the peritoneal cavity of Myd88−/− or wild-type (Wt) mice 12 hours after injection. One representative of four experiments is shown. Percentages of indicated gated populations are shown. (B and C) Absolute cell numbers of designated populations in the peritoneal cavity of Wt mice after injection of PBS, OVA, or StfOVA as indicated on the x axes. Data are pooled from four independent experiments and presented as means ± SEM. *P < 0.05; **P < 0.005. n.s., not significant. (D) FACS analysis of cells from the peritoneal cavity of Wt mice left untreated or injected with clodronate-containing liposomes 12 hours before injection of EL4 cells. One representative of four experiments is shown at 12 hours. (E) Wt mice were injected with control IgG or anti–Gr-1–neutralizing antibody (clone RB6-8C5) 12 hours before injection of EL4 cells. One representative of four experiments is shown at 16 hours after injection. Percentages of indicated gated populations are shown.

Consistent with our previous findings that MyD88-dependent TLR signaling enhances phagocytosis by macrophages in vitro (16), we observed a statistically significant increase in the absolute number of tumor-carrying GFP+F4/80+ or GFP+CD11b+ macrophages in vivo after injection of EL4-StfOVA compared to EL4-OVA cells (Fig. 1C and fig. S2D). GFP+Ly6G+ neutrophils were also detected, although the differences in their numbers between the EL4-StfOVA and the EL4-OVA groups were not significant (Fig. 1C and fig. S2D). Introduction of flagellin within EL4 cells did not affect cell surface expression of CD47, which has been shown to inhibit phagocytosis (28), and these cells did not express pro-phagocytic surface calreticulin unless they were induced to undergo apoptosis (fig. S1F). These data suggested that efficient clearance relied on flagellin expression, which appeared to override the inhibitory signals from CD47.

Injection of clodronate-containing liposomes before EL4 tumor cells successfully depleted CD11b+F4/80+ cells without affecting Ly6G+ cells (Fig. 1D and fig. S2E, PBS versus clodronate/liposome panels) and led to the persistence of EL4-StfOVA cells (Fig. 1D, compare GFP+CD45+ cells in PBS versus clodronate/liposome StfOVA groups). Administration of anti–Gr-1 antibody did not affect the CD11bhighF4/80high populations, but it successfully depleted Ly6G+ neutrophils and decreased recruitment of CD11bintF4/80int cells (Fig. 1E and fig. S2E). However, anti–Gr-1 antibody had no effect on clearance of EL4-StfOVA cells (Fig. 1E and fig. S2E). Concomitant treatment of recipient mice with anti–Gr-1 and clodronate-containing liposomes depleted both neutrophils and macrophages but showed no further additive effects over clodronate/liposomes alone (fig. S2F). Collectively, these results showed that flagellin-expressing tumor cells rapidly recruited significantly greater numbers of neutrophils and macrophages into the peritoneal cavity and that macrophages were primarily responsible for clearance of flagellin-expressing tumor cells.

Flagellin-expressing tumor cells fail to establish tumor in vivo

We next determined the ability of EL4 and flagellin-expressing EL4 cells to form established tumors over time after subcutaneous injection. When monitored over the course of 24 days, EL4-StfOVA cells failed to establish subcutaneous tumors (Fig. 2A), and this was observed with two independently generated clones of EL4-StfOVA cells (Fig. 2B). In contrast, injection of EL4-OVA resulted in development of palpable tumors, which was not altered by co-injection of RecFLA or by frequent administration of RecFLA over the course of tumor development (Fig. 2A). The concentration of flagellin expressed by the number of injected EL4-StfOVA (1 × 105 to 2 × 105) was lower than the concentration of co-administered RecFLA (2 ng) (fig. S1C). Rejection of EL4-StfOVA cells was dependent on TLR5 signaling because the absence of TLR5 or MyD88 in recipient mice restored the capacity of EL4-StfOVA tumor cells to form subcutaneous tumors (Fig. 2C). This also demonstrated that the inability of EL4-StfOVA cells to form tumors in vivo was not an intrinsic property of the clones generated but was due to selective immune pressure through TLR5 signaling (29).

Fig. 2

Expression of flagellin within tumor cells induces better tumor rejection and antitumor adaptive immune responses than co-injection of RecFLA with tumor cells. (A to C) Tumor volumes in syngeneic Wt mice after subcutaneous injection of 1 × 105 EL4-OVA, EL4-StfOVA, and EL4-OVA with 2 ng of RecFLA (EL4-OVA + RecFLA), or EL4-OVA–injected mice subsequently injected twice a week with 5 ng of RecFLA [EL4-OVA + RecFLA (twice/week)]. Tumor volumes at day 20 in (B) Wt mice after subcutaneous injection with different clones cl.1 and cl.2 of EL4-StfOVA compared to EL4-OVA and (C) Wt, Tlr5−/−, or Myd88−/− mice. Each dot represents one mouse. (D) Tumor foci in lungs of Wt mice 14 days after intravenous injection of 2 × 105 B16-Eα, Stf.Eα, or B16-Eα co-injected with RecFLA (left panel). Pictures of four lungs representative of each group at day 14 (right panel). All experiments are repeated two to three times to achieve n = 8 to 15 mice per group. Data are presented as means ± SEM. *P < 0.05; **P < 0.005; ***P < 0.0005. n.s., not significant.

Eradication of StfOVA-expressing tumor cells was not limited to EL4. Poorly immunogenic B16-F10 melanoma cells expressing a fusion protein between S. typhimurium flagellin and the α chain of the MHC-II molecule I-E (Stf.Eα) were markedly impaired in colonizing the lung at 14 days (Fig. 2D) and 28 days (fig. S3) after intravenous injection. In contrast, RecFLA had no effect on B16-Eα lung metastases (day 14 in Fig. 2D and day 28 in fig. S3), further emphasizing that, for mediating tumor rejection, flagellin within tumor cells is superior to its co-injection with tumor cells.

Associative recognition of tumor antigens with flagellin effectively primes antigen-specific CD4+ and CD8+ T cells in a TLR5/MyD88-dependent manner

Injection of EL4-StfOVA cells into mice treated with a neutralizing antibody against interferon-γ (IFN-γ), a key cytokine for T cell–mediated antitumor immune responses (30), restored the ability of these cells to form tumors (Fig. 3A). These data suggested a role for the adaptive immune system in mediating rejection of flagellin-expressing tumor cells. We thus injected EL4-StfOVA cells into Rag1−/− recipient mice, which lack mature T and B cells (31). In Rag1−/− mice, EL4-StfOVA cells formed subcutaneous tumor at the injection site with similar volume and kinetics as EL4-OVA cells, formally demonstrating an important role for the adaptive immune response in protection from tumor growth (Fig. 3B). EL4-StfOVA cells formed tumor in wild-type mice treated with a neutralizing antibody against CD8, which demonstrated an important role for CD8+ T cell–mediated responses (Fig. 3C).

Fig. 3

Expression of flagellin within tumor cells induces both CD4+ and CD8+ T antitumor adaptive immune responses. (A to C) Tumor volume in mice of indicated genotypes injected with 1 × 105 EL4-OVA or EL4-StfOVA. In some experiments, mice were treated with neutralizing antibody to IFN-γ (A), or CD8 (C), or control IgG. Experiments were repeated twice (n = 10 per group). Data are means ± SEM. (D and E) FACS analysis of proliferation of OVA-specific CFSE-labeled CD8+ OT-I T cells in tumor-draining lymph nodes of mice with indicated genotypes at day 3 after injection of indicated EL4 tumor cells. (F and G) IFN-γ (F) and granzyme B (GrzB) (G) expression by these OT-I T cells analyzed by intracellular staining and FACS analysis. Data are pooled from four independent experiments (n = 4; means ± SEM). **P < 0.005. (H to L) FACS analysis assessing proliferation of CFSE-labeled T cells in tumor-draining lymph nodes of mice after injection of 1 × 105 tumor cells unless indicated otherwise. (H) OT-I (upper panels) and OVA-specific CD4+ OT-II (lower panels) on days 3 and 5, respectively. [Stf+EL4-OVA (not fused)] denotes EL4 cells expressing OVA not fused to but coexpressed with flagellin. (I) gp100-specific CD8+ Pmel T cells at day 3. (Stf+B16-Eα) denotes B16-Eα expressing flagellin not fused to Eα. (J) OT-II T cells in Wt or Tlr5−/− mice at day 5. OT-I T cells on day 3 (K) and OT-II T cells on day 5 (L) in Wt or CD11c-DTR transgenic mice treated or not with diphtheria toxin (DT) and injected with 2 × 105 EL4-StfOVA cells. Percentages of proliferating cells are shown. All data are representative of three to five independent experiments.

To assess in vivo presentation of the model tumor antigen OVA to CD8+ T cells, we evaluated the proliferation of adoptively transferred carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled OVA-specific T cell receptor (TCR) transgenic OT-I CD8+ T cells in tumor-draining lymph nodes. Neither EL4-OVA cells nor co-injection of EL4-OVA cells with RecFLA induced appreciable OT-I T cell proliferation (Fig. 3, D and E; means and SEM in fig. S4). In contrast, EL4-StfOVA cells induced proliferation of 60 to 75% of OT-I T cells [Fig. 3, D and E, StfOVA Wt histograms]. TLR5-MyD88 signaling was required because OT-I T cell proliferation was impaired when recipient mice were deficient for MyD88 (Fig. 3D) or TLR5 (Fig. 3E). Increased OT-I T cell proliferation in response to EL4-StfOVA was associated with increased production of IFN-γ (Fig. 3F) and expression of granzyme B (Fig. 3G and fig. S5A). Therefore, although co-administration of RecFLA with EL4-OVA tumor cells or injection of EL4-StfOVA tumor cells triggered clearance from the peritoneal cavity (Fig. 1A), OVA-specific CD8+ T cell priming was observed here when OVA and flagellin were expressed as a fusion protein by EL4-StfOVA cells.

We next assessed whether OVA needed to be fused to flagellin or whether co-delivery of OVA with flagellin as part of the same phagocytosed tumor cell could also prime OVA-specific CD8+ T cells. We generated EL4-OVA cells expressing flagellin that was not fused to OVA (Stf+EL4-OVA). These cells were eliminated as efficiently as EL4-StfOVA cells when injected into the peritoneal cavity of wild-type mice (fig. S5B). Stf+EL4-OVA cells also strongly induced OT-I T cell proliferation (Fig. 3H). In addition, subcutaneous injection of Stf+B16-Eα melanoma cells, where flagellin was expressed in B16 melanoma cells but not fused to a model antigen, induced proliferation of adoptively transferred CFSE+ Pmel-1 CD8+ T cells specific to the endogenous melanoma antigen gp100 (Fig. 3I) (32). Collectively, data in Fig. 3, H and I, demonstrate that fusion between flagellin and tumor-derived antigens, either xenoantigen or a bona fide tumor antigen, was dispensable for flagellin-induced cross-priming of tumor-specific CD8+ T cells. These results also suggest that other endogenous tumor antigens could also be processed and presented by MHC-I molecules through their associative immune recognition with flagellin within the same phagosome.

Subcutaneous injection of EL4-StfOVA induced proliferation of adoptively transferred OVA-specific OT-II CD4+ T cells in tumor-draining lymph nodes (Fig. 3J, StfOVA groups), whereas no such proliferation was induced upon the injection of EL4-OVA alone or with RecFLA (Fig. 3J, OVA or OVA + RecFLA groups, respectively). Moreover, similar to OT-I T cells, priming of OT-II T cells did not require fusion of the model antigen OVA with flagellin when expressed in the same tumor cell (Fig. 3H). Therefore, although co-administration of RecFLA was sufficient to induce clearance of tumor cells in the peritoneal cavity (Fig. 1A), associative recognition of tumor-derived antigen with flagellin was, similar to the priming of CD8+ T cells, also necessary for priming tumor-specific CD4+ T cells. These results argue that delivery of tumor antigen with PAMP into the same phagosome can, as we had shown in vitro (17), lead to efficient MHC-II presentation of a tumor antigen in vivo.

Priming of both adoptively transferred CD8+ and CD4+ T cells in response to EL4-StfOVA cells (Fig. 3, D and J, respectively) suggested a role for antigen-presenting cells, which was further supported by abrogation of OT-I and OT-II T cell proliferation in tumor-draining lymph nodes of mice treated with clodronate/liposomes (fig. S5C). When cocultured with apoptotic EL4 thymoma or A20 B cell lymphoma cells, splenic CD11c+ DCs up-regulated CD40 (fig. S5D) and secreted interleukin-12 (IL-12) in a MyD88-dependent manner specifically in response to flagellin-expressing tumor cells (fig. S5E), showing that DCs can be activated in response to flagellin-expressing tumor cells. Because of the superior ability of CD11c+ DCs to prime naïve T cells, we used CD11c–diphtheria toxin receptor (DTR)/GFP (CD11c-DTR) transgenic mice to transiently deplete CD11c+ cells after treatment with DT (33). Depletion of CD11c+ cells abolished OT-I and OT-II T cell priming in the tumor-draining lymph nodes after subcutaneous injection of EL4-StfOVA cells (Fig. 3, K and L, respectively, compare DT-treated Wt versus CD11c-DTR groups). These data suggested that DCs likely acquired tumor-derived antigen in the peripheral tissue to prime tumor-specific CD8+ T cells in the lymph node. Indeed, purified splenic DCs preferentially presented tumor-derived OVA after phagocytosis of tumor cells that expressed flagellin as shown by higher levels of proliferation and IL-2 production by cocultured CFSE-labeled OT-I T cells (fig. S5, F and G, respectively).

The NLRC4/NAIP5 inflammasome is required for subcutaneous elimination of flagellin-expressing tumor cells

Mutation of three leucines (L470, L472, and L473), conserved among various bacterial species and located within the C-terminal 35 amino acids (D0 domain) of flagellin, demonstrated their key role in NAIP5-dependent activation of the NLRC4 inflammasome [Fig. 4A and (25)]. To determine whether NLRs are activated in response to the flagellin expressed by tumor cells, we generated EL4 cells expressing different variants of StfOVA, where the same leucines shown to be critical for NLRC4/NAIP5 activation (25) were mutated to alanines (Fig. 4A; similar OVA/flagellin expression in fig. S6, A and B). Single mutation of L470 (StfOVA-ΔNLR.A) or double mutation of L472/L473 (StfOVA-ΔNLR.B) was previously shown to severely impair the ability of Legionella pneumophila to trigger macrophage caspase-1–dependent cell death (pyroptosis) (25). We found that the L470A and L472A/L473A mutations restored in part the ability of the corresponding flagellin-expressing EL4 cells to form tumors in vivo (Fig. 4B, ΔNLR.A and ΔNLR.B, respectively). Consistent with abrogation of NLRC4-NAIP5 activation (25), mutation of all three leucines to alanine (EL4-L470A/L472A/L473A: EL4-StfOVA.AAA) completely reversed flagellin-mediated tumor rejection, highlighting the importance of the NLR-activating capacity of flagellin in this process (fig. S6C).

Fig. 4

Adaptive immune responses triggered by flagellin-expressing tumor cells rely on both TLR5 and NLRC4/NAIP5. (A) Schematic representation of Wt and mutated forms of StfOVA fusion proteins. C-terminal amino acid sequence of flagellin is depicted. Specific residues important for NLRC4/NAIP5 (L470, L472, and L473) and for TLR5 (I411) recognition are highlighted in bold. The names of constructs used and the corresponding mutations into alanine are shown. (B) Tumor volume in syngeneic Wt mice at day 20 after transplantation of EL4 cells expressing different forms of StfOVA fusion proteins. Data are representative of two experiments (n = 10 per group) presented as means ± SEM. ***P < 0.0005. Each dot represents one mouse. (C and D) FACS analysis assessing proliferation of CFSE-labeled T cells in tumor-draining lymph nodes of mice with indicated genotypes after subcutaneous injection of EL4 cells expressing different forms of StfOVA. Data shown are representative of at least three experiments. Percentages of proliferating cells are shown. (C) OT-II T cells at day 5. (D) OT-I T cells at day 3.

To confirm that TLR5 signaling was also required for elimination of EL4-StfOVA cells, we generated EL4 cells expressing a variant of flagellin (StfOVA-ΔTLR5), where isoleucine 411 (I411) located in the conserved D1 domain was mutated to alanine. This I411A mutation impairs but does not abrogate flagellin recognition by TLR5 (29). I411A relieved TLR5-mediated suppression of tumor development (Fig. 4B, EL4-StfOVA-ΔTLR5), although not to the same extent as that seen with genetic deletion of TLR5 (Fig. 2C), corroborating the role of TLR5 in mediating rejection of StfOVA-expressing tumor cells. When both the TLR5-activating residue I411 and the NLRC4/NAIP5-activating residue L470 were mutated, corresponding ΔTLR5ΔNLR.A EL4-StfOVA grew with kinetics similar to that of EL4-OVA cells that did not express flagellin (Fig. 4B). These results showed that both TLR5 and the NLRC4/NAIP5 inflammasome effectively suppressed flagellin-expressing tumor cells from establishing tumor in vivo.

The NLRC4/NAIP5 inflammasome is required for T cell priming in response to associative recognition of tumor antigen with flagellin

We next investigated how loss of the NLRC4/NAIP5-activating residues of flagellin affected priming of T cells specific to tumor-derived OVA. Surprisingly, proliferation of both CD8+ OT-I and CD4+ OT-II T cells was severely impaired in tumor-draining lymph nodes of mice injected subcutaneously with either EL4-StfOVA-ΔNLR.A or EL4-StfOVA-ΔNLR.B cells compared to their proliferation in response to EL4-StfOVA cells (Fig. 4, C and D, for OT-II and OT-I, respectively). In addition, EL4-StfOVA.AAA cells bearing a mutation in the leucines at positions 470, 472, and 473 of flagellin that abrogate NLRC4-NAIP5 activation also failed to induce OT-I T cell proliferation (fig. S6D). Collectively, these results suggest that despite retaining the ability to activate TLR5, loss of NLRC4/NAIP5-activating residues in flagellin is sufficient to severely impair priming of T cells specific to tumor-derived OVA. In agreement with the critical role of TLR5 for T cell priming in response to flagellin-expressing cells (Fig. 3, E and J), OT-I and OT-II T cell proliferation was reduced in response to EL4-StfOVA-ΔTLR5, further confirming a role for recognition by TLR5 in initiating adaptive immune responses to flagellin-expressing tumor cells (Fig. 4, C and D, for OT-II and OT-I proliferation, respectively).

The NLRC4/NAIP5 inflammasome and not TLR5 is dispensable for elimination of flagellin-expressing tumor cells by innate immune cells

In contrast to the effects of NLRC4/NAIP5 activation on T cell priming, mutations of the NLRC4/NAIP5-activating residues L470 or L472/L473 did not impair the ability of flagellin-expressing tumor cells to recruit inflammatory neutrophils and macrophages to the peritoneal cavity (Fig. 5A, compare StfOVA-ΔNLR.A and ΔNLR.B to StfOVA). Significant reduction in intraperitoneal levels of IL-1β in response to injection of EL4-StfOVA-ΔNLR.A or EL4-StfOVA-ΔNLR.B cells (Fig. 5B) confirmed that inflammasome function was indeed impaired by these mutations. In sharp contrast, clearance of flagellin-expressing tumor cells from the peritoneal cavity was impaired in Tlr5−/− mice, showing a crucial role for TLR5 (Fig. 5C). Similarly, deficiency in TLR5 also impaired clearance of EL4-StfOVA-ΔNLR.A or EL4-StfOVA-ΔNLR.B cells, confirming the critical role of TLR5 and not NLRC4/NAIP5 in mediating tumor clearance in this case (fig. S7). Thus, here and as shown in Fig. 1A, MyD88 signaling downstream of TLR5 is sufficient to trigger clearance of flagellin-expressing tumor cells (Figs. 1A and 5C, Myd88−/− versus Wt groups).

Fig. 5

Distinct roles of TLR5 and the NLRC4/NAIP5 inflammasome in the elimination of flagellin-expressing tumor cells. (A and C) FACS analysis of cells from the peritoneal cavity of Wt mice (A) or Wt or Tlr5−/− mice (C) 12 hours after intraperitoneal injection of indicated EL4 cells. One representative of three experiments is shown. Percentages of cells in indicated gates are shown. (B) IL-1β cytokine levels measured by enzyme-linked immunosorbent assay (ELISA) from peritoneal lavages of Wt mice (n = 6 per group) 6 hours after injection of EL4 cells expressing the indicated form of StfOVA fusion protein. Three experiments are pooled. Data are presented as means ± SEM. **P < 0.005; ***P < 0.0005. n.s., not significant.

Immunization with flagellin-expressing tumor cells protects mice from tumor development via NLRC4/NAIP5 and TLR5

We next evaluated the ability of flagellin-expressing tumor cells to protect against the establishment of lung metastases by preexisting B16-Eα melanoma cells. We followed intravenous inoculation of B16-Eα cells with a subcutaneous injection of either irradiated B16-Eα or Stf+B16-Eα cells (Fig. 6A) (34, 35). Although immunization with B16-Eα or B16-Eα co-administered with RecFLA did not prevent lung metastases, no tumor foci were detectable in the lungs of mice immunized with Stf+B16-Eα (Fig. 6B), demonstrating the superiority of immunization with irradiated flagellin-expressing tumor cells.

Fig. 6

Immunization with flagellin-expressing tumor cells induces protection against growth of parental tumor cells and induces TLR5/NLRC4/NAIP5-dependent CD8+ T cell memory. (A and C) Schematics showing experimental strategies for (B) and (D) to (L), respectively. (B) Numbers of B16-Eα tumor foci in lungs of Wt mice treated with indicated irradiated B16 cells. Data are pooled from two experiments (n = 5 to 10 mice per group). One lung representative of each group is shown (right panel). (D to H) EL4-Wt (parental tumor) (D and E) or EL4-OVA (F to H) tumor volumes after challenge of mice of the indicated genotypes with live tumor cells 20 days after immunization with 2 × 105 of indicated irradiated EL4 cells (n = 5 to 15 mice per group). In (F), one group was immunized with EL4-OVA cells co-injected with recombinant flagellin (OVA + RecFLA), and another group was composed of mice that had survived injection with live EL4-StfOVA cells [StfOVA (live)]. (I and K) FACS analysis, on day 3 after challenge, of CD8+ T cells positively stained for SIINFEKL–H2-Kb tetramer in tumor-draining lymph nodes of Wt mice (I) or indicated genotypes (K) immunized with indicated irradiated OVA-expressing EL4 cells, and challenged with live Wt EL4-OVA on day 21 after immunization. One representative of three independent experiments is shown. Percentages of cells in indicated gates are shown. (J and L) Absolute cell numbers of CD8+ T cells positively stained for SIINFEKL–H2-Kb tetramer from (I) and (K), respectively. Experiments were repeated two to three times. Each dot represents one mouse (B, D, and E to H) and means ± SEM are shown (B, D to H, J, and L). *P < 0.05; **P < 0.005; ***P < 0.0005. n.s., not significant.

We next determined whether the adaptive immune response mobilized by the associative recognition of tumor antigen with flagellin could protect mice against development of parent tumors. We subcutaneously immunized mice with irradiated EL4 cells that expressed either a flagellin-OVA fusion protein, OVA alone, or no OVA, and with (Stf+) or without flagellin expression. Mice were challenged 21 days later with live EL4 wild-type cells that did not express OVA (Fig. 6C). Although immunization with irradiated EL4-OVA cells had no significant effect on subsequent growth of EL4 wild-type cells, immunization with irradiated Stf+EL4-OVA (Fig. 6D), and notably Stf+EL4 cells without OVA (Fig. 6E), offered statistically significant protection against parent EL4 tumor development when compared to control groups. Similarly, mice were protected from challenge with EL4-OVA cells when immunized with irradiated or live EL4-StfOVA, but not EL4-OVA (Fig. 6F). Protection from tumor growth was observed only by associative immune recognition of tumor-derived OVA with flagellin, which induced robust T cell proliferation (Fig. 3, D, E, and J), and not co-injection of RecFLA with EL4-OVA (Fig. 6F, compare StfOVA groups with OVA + RecFLA).

To determine the role of NLRC4/NAIP5 in the protective antitumor response after immunization with flagellin-expressing tumor cells, we immunized mice with irradiated EL4-StfOVA-ΔNLR.A or EL4-StfOVA-ΔNLR.B cells, where the NLR stimulatory activity of flagellin was impaired. Each of these mutations significantly abrogated the protection afforded by immunization with flagellin-expressing tumor cells (Fig. 6G). Protective immunization with EL4-StfOVA also relied on intact TLR5 signaling (Fig. 6H).

Because CD8+ T cell responses are important for antitumor immunity (36), we next examined the expansion of tumor-derived OVA-specific CD8+ T cells within the endogenous T cell repertoire of immunized mice. Using SIINFEKL–H2-Kb tetramers that bind to the TCR of OVA-specific CD8+ T cells, we detected a population of 0.48% CD8+CD44+tetramer+ T cells within tumor-draining lymph nodes of mice immunized with irradiated EL4-StfOVA, but not those immunized with EL4-OVA, EL4-OVA co-injected with RecFLA, or ΔNLR.A and ΔNLR.B EL4-StfOVA cells (Fig. 6I). These endogenous CD8+ T cells expressed IFN-γ and granzyme B (fig. S8), and their absolute numbers were significantly increased within the tumor-draining lymph nodes of EL4-StfOVA–immunized mice compared to the other immunized groups (Fig. 6J). TLR5 and MyD88 deficiency strongly decreased the presence of CD8+CD44+tetramer+ T cells, demonstrating that both NLRC4-NAIP5 and TLR5 signaling were equally required for activation of tumor-specific CD8+ T cells upon vaccination with flagellin-expressing tumor cells (Fig. 6, K and L). These results strongly argue that immunization with tumor cells modified to express flagellin can generate an endogenous antigen-specific CD8+ T cell memory response that protects against tumor development.


The promise of TLR ligands in cancer immunotherapy is best exemplified in Food and Drug Administration approval of the TLR7 agonist imiquimod (Aldara) for basal cell carcinoma, actinic keratosis, and external genital warts [reviewed in (12, 37)]. Mycobacterium bovis Bacille Calmette-Guérin (BCG), which stimulates TLR2 and TLR4 (38, 39), has long been used for intravesical treatment of bladder cancer (40). Various forms of TLR9 agonist CpG oligodeoxynucleotides have been used as adjuvants in monotherapies, combination therapies, and phase 1/2/3 vaccine trials in patients with non–small cell lung cancer [reviewed in (41)]. Despite setbacks in a phase 3 clinical trial of CpG in combination with chemotherapy (11, 41), CpG still holds promise and new TLR9 agonists are under development and evaluation (12, 37). Several other TLR-based vaccine formulations are being evaluated as well, including synthetic TLR2 and TLR4 agonists, the TLR5 agonist flagellin, and double-stranded RNA mimics as TLR3 agonists (12, 37).

Here, we tested the efficacy of the TLR5 ligand flagellin as an adjuvant for tumor immunotherapy by forcing its expression within tumor cells, thereby ensuring associative recognition of tumor antigens with flagellin, a property we had previously shown that enhances antigen presentation (17). This strategy was effective in synergistically activating TLR5 and NLRC4/NAIP5 to induce protection from tumor growth and to efficiently prime both tumor-specific CD8+ and CD4+ T cell responses (schematic summary in fig. S9). Fusion between flagellin and a model antigen was previously shown to be important for CD8+ T cell priming, limiting the use of flagellin as an adjuvant to diseases where target antigens are well defined (4246). Our antitumor vaccination studies demonstrated the adjuvanticity of flagellin regardless of its fusion with tumor-derived antigen, allowing presentation of not only xenoantigens such as OVA or Eα but also endogenous tumor antigens such as the melanoma gp100 antigen. Furthermore, introduction of flagellin into tumor cells without artificial foreign antigens induced protection against subsequent challenge with live parental tumor. Enhanced antigen presentation relied primarily on the physical presence of flagellin within tumor cells because co-administration of soluble RecFLA with tumor cells did not achieve the same effect. Collectively, our findings suggest that in addition to physically linking flagellin to known tumor antigens to optimize their presentation, introducing flagellin within tumor cells could enable associative recognition with flagellin of various other known and unknown tumor-specific antigens.

Consistent with our previous studies in vitro (17), associative recognition of flagellin and tumor antigens led to priming of tumor-specific CD4+ T cells, an important step because of the difficulty in inducing helper T cell responses in tumor immunotherapy (4749). Help from CD4+ T cells is critical for differentiation of cytotoxic CD8+ T lymphocytes (CTLs) (14, 15, 5052), and peptide-based vaccinations consisting of long peptides containing both helper and CTL epitopes, administered with TLR ligands, have served as efficacious cancer vaccine formulations (4749). Co-administration of RecFLA with tumor cells, which mimics many current forms of TLR-based tumor immunotherapy, was not successful in priming antitumor CD4+ T cells. Therefore, by ensuring concomitant activation of both CD8+ and CD4+ T cells, associative immune recognition of tumor antigens with PAMPs constitutes an important mechanism to consider when developing new immunotherapies.

Recent studies have proposed an important role for NLRP3, NLRC4, and NLRP6 in preventing tumor development in models of colitis-associated cancer (5356). Our study now shows that NLRC4/NAIP5 can be exploited in tumor immunotherapy by forcing flagellin expression within tumor cells. NLRC4/NAIP5 activation, which controls caspase-1–mediated responses, is essential for innate immune responses to flagellin after infection with intracellular pathogens (24, 25), but whether these NLRs are involved in promoting adaptive immunity has been unclear. Our data provide direct evidence for a critical role of NLRC4/NAIP5 in priming CD8+ T cell responses and preventing growth of flagellin-expressing tumor cells. NLRP3 was critical in inducing an adaptive immune response to an inflammatory tumor cell death induced by chemotherapy, where NLRP3-dependent release of IL-1β was essential for priming tumor-specific CD8+ T cells (57). Our data demonstrate that NLRC4/NAIP5 activation is required for IL-1β secretion in response to peritoneal injection of flagellin-expressing tumor cells, and phagocytosis of flagellin-expressing tumor cells by splenic DCs led to IL-12 secretion, known to promote T helper 1 (TH1) differentiation. Therefore, it is possible that recognition of flagellin by NLRC4/NAIP5 supports TH1-mediated immunity required for host defense against intracellular pathogens, a property that can be exploited for antitumor immunity as we demonstrate here (58). Collectively, these data support an immediate therapeutic application of manipulating NLRs to stimulate protective immunity.

Flagellin-expressing cells triggered antitumor host immunity via the concerted actions of TLR5 and the NLRC4/NAIP5 inflammasome. However, we found a dichotomy between the necessity of NLR cooperation with TLR5 for T cell priming and the dispensability of NLR for mobilizing TLR5-dependent tumor clearance. Disrupting flagellin recognition by NLRs did not affect macrophage clearance of flagellin-expressing tumor cells or the recruitment of inflammatory cells into the peritoneal cavity. These data suggested that TLR5 signaling was indeed sufficient for triggering early innate immune cell activation, consistent with our previous studies (16, 18). Although the use of TLR ligands in DC-based immunotherapies is sufficient for activating DCs (9), our data here argue that the ability of activated DCs to mount a protective antitumor adaptive response could be optimized by the inclusion of NLR ligands.

Our present study further urges a reconsideration of the manner in which PAMPs are delivered as adjuvants in cancer immunotherapy by demonstrating the superiority of introducing flagellin within tumor cells in inducing protective tumor immunity compared to the exogenous co-administration of flagellin with tumor cells. Immunization with irradiated flagellin-expressing tumor cells could prevent preexisting tumor cells from developing solid tumors at metastatic sites. Immunization with flagellin-expressing tumor cells in the presence, and more notably in the absence, of xenoantigens induced statistically significant decreases in growth of parental tumors. These results have translational potential in the human clinical setting, especially in the ability to successfully immunize with flagellin-modified tumor cells and without the need for incorporation of tumor-specific or foreign antigens. Such approaches would certainly have to be tailored to the nature and size of the tumor burden, and the amount of flagellin introduced into tumor cells must carefully be tuned to achieve appreciable clinical outcome. Flagellin provides the additional advantage of triggering both TLR and NLR signaling pathways, thereby mobilizing two important arms of the innate immune system. New clinical applications, especially immunotherapeutic applications geared toward preventing tumor relapse after tumor ablation or resection therapies, can therefore be designed on the basis of the key ingredients described in this study.

Materials and Methods

Full experimental procedures are available in the Supplementary Material.

Mouse strains

All knockout and transgenic mice used in this study were either male or female on a C57BL/6J background. C57BL/6J, C57BL/6J-Thy1.1, C57BL/6J-CD45.1, OT-I, Rag1−/−, and CD11c-DTR/GFP mice were purchased from the Jackson Laboratory. OT-I mice were crossed with C57BL/6J-Thy1.1 to obtain OT-I/Thy1.1 transgenic mice. Tlr5−/− mice were provided by R. Flavell, OT-II by R. Medzhitov, and Myd88−/− mice by S. Akira. Pmel-1 mice were generated in the laboratory of N. P. Restifo (National Cancer Institute, Bethesda, MD) and are commercially available from the Jackson Laboratory (40). We used 8- to 10-week-old animals for all experiments. All experiments were approved by the institutional animal care and use committee and performed in agreement with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health publication 86-23, revised 1985).

Tumor development and immunizations

For subcutaneous tumor development, tumor cells were washed twice and resuspended at 1 × 106 cells/ml in PBS. Cells (1 × 105 to 2 × 105) were injected subcutaneously with an insulin syringe (Becton-Dickinson) into age- and sex-matched recipient mice. Tumor volume was measured every 3 days and calculated with the following formula: tumor length × (tumor width)2 × 1/2. For immunization experiments, tumor cells were γ-irradiated at 30 Gy, and 2.5 × 105 cells were injected subcutaneously into one flank. Twenty-one days later (14 days for tetramer staining experiments), mice were subcutaneously injected in the opposite flank with 1 × 105 live control tumor cells, and tumor progression was evaluated as described above. When live cells were used for immunization, 1 × 105 live cells were injected, and mice were challenged 2 months later. When depleting antibody for IFN-γ was used, mice received two intraperitoneal injections of 250 ng of XMG-1.2 antibody (eBioscience) at days 1 and 5 after tumor inoculation. For CD8+ T cell depletion, mice were injected intraperitoneally with 500 μg of anti-CD8 antibody (clone 2.43) on days 3, 2, and 1 before tumor transplantation. Control mice were injected with an equivalent amount of isotype-matched control immunoglobulin G (IgG). For B16 tumor cell colonization of the lungs, 1 × 105 cells were intravenously injected into the tail vein of recipient mice. Twenty-one or 28 days later, mice were killed and lungs were analyzed under a tissue microscope for enumeration of tumor foci. For treatment experiments, mice were injected intravenously with 2 × 105 B16-Eα cells. Mice were then injected subcutaneously on days 3 and 10 with 5 × 106 irradiated B16-Eα cells expressing or not expressing flagellin. In some groups, B16-Eα cells were co-injected with 5 ng of recombinant flagellin (RecFLA). On day 21, mice were killed and lungs were analyzed as described above for analysis.

Assessing immune responses to peritoneal colonization by flagellin-expressing tumor cells

Recipient mice were injected intraperitoneally with 200 μl of PBS containing 3 × 106 live tumor cells. Sixteen hours later, the peritoneal cavity was washed with cold PBS, and peritoneal cells were stained for subsequent FACS after gating out false-positive “doublets.” For neutrophil depletion, 12 hours before the injection of tumor cells, mice were injected intraperitoneally with 250 ng of anti–Gr-1 monoclonal antibody (clone RB6-8C5) or an isotype-matched control. For depletion of macrophages, mice were injected as indicated with 100 μl of a standard suspension of clodronate-loaded liposomes (59) 12 hours before tumor cell injection.

In vivo T cell proliferation assay, intracellular staining, and tetramer staining

OVA-specific CD8+ and CD4+ T cells were isolated from spleen and lymph nodes of OT-I, OT-I/Thy1.1+, or OT-II transgenic mice with magnetic beads (Miltenyi Biotec) and injected into Thy1.1+ or Thy1.2+. CFSE labeling was performed as described (60), and 1 × 106 to 1.5 × 106 CFSE+ cells were injected into the tail vein of recipient mice 6 to 12 hours before tumor cell inoculation. Mice were then inoculated subcutaneously with 1 × 105 live tumor cells. At day 3 (for OT-I CD8+ T cells) or day 5 (for OT-II CD4+ T cells), cells from tumor-draining lymph nodes were stained for subsequent flow cytometry analysis with an LSRII cytometer (BD Biosciences) and FlowJo software. For intracellular staining, cells were surface-stained for 15 min on ice, fixed, and permeabilized with a staining buffer set (eBioscience). Cells were then stained for intracellular cytokines for 30 min on ice, washed three times, and subjected to FACS. For tetramer staining, a single-cell suspension from tumor-draining lymph nodes was obtained, stained for CD8, CD4, CD3, B220, CD44, and phycoerythrin–H2-Kb (SIINFEKL) tetramer for 30 min on ice, and analyzed by FACS. In some experiments, CD11c-DTR chimeras were used 8 weeks after bone marrow reconstitution. To induce CD11c+ cell depletion, we injected mice intraperitoneally with 250 ng of DT in PBS every 2 days.

Statistical analyses

Statistical significance was determined by an unpaired two-tailed Student’s t test or by a two-way analysis of variance (ANOVA) grouped test. The results are given as confidence interval (P) and are considered as significant when P < 0.05.

Supplementary Material

Materials and Methods

Fig. S1. Characterization of flagellin-expressing tumor cells.

Fig. S2. Macrophage-dependent elimination of flagellin-expressing tumor cells in the peritoneum.

Fig. S3. Flagellin expression in B16 melanoma cells prevents tumor formation in wild-type C57BL/6 mice.

Fig. S4. Flagellin-OVA (StfOVA)–expressing tumor cells induce priming of CD4+ and CD8+ T cells specific to tumor-derived antigen.

Fig. S5. Flagellin-OVA (StfOVA)–expressing tumor cells induce cross-priming of CD8+ T cells specific to tumor-derived antigen.

Fig. S6. Characterization of EL4 cells expressing StfOVA constructs mutated for NLRC4-NAIP5–activating residues.

Fig. S7. Peritoneal clearance of flagellin-expressing tumor cells is dependent on TLR5 and not NLRC4-NAIP5 activation.

Fig. S8. StfOVA-expressing tumor cells induce cross-priming of endogenous CD8+ T cells specific to tumor-derived antigen.

Fig. S9. Schematic representation of the therapeutic strategy delineated within this study.


References and Notes

  1. Acknowledgments: We thank E. B. Kopp for the flagellin-OVA construct. Pmel-1 mice were generated by N. P. Restifo and provided by Y. Saenger and M. Merad. Tlr5−/− mice were provided by R. Flavell, OT-II by R. Medzhitov, and Myd88−/− mice by S. Akira. The SIINFEKL–H2-Kb–specific 25D1.16 hybridoma was a gift from R. Germain. We are grateful to L. E. Sander, M. Villalba, and D. Amsen for critical reading of the manuscript. Funding: Supported by a “Fondation pour la Recherche Médicale” fellowship to J.G. and an American Cancer Society basic research scholar award 117254-RSG-09-191-01-LIB to J.M.B. Author contributions: J.G. and J.M.B. designed the experiments, directed the study, and wrote the manuscript. J.G. and A.K. performed the experiments. N.v.R. provided the clodronate/liposomes and related advice. J.M.B. conceived the study. Competing interests: J.M.B. and J.G. have filed a provisional patent application serial no. 61/302,052 titled “Use of flagellin-expressing tumor cells in the immunotherapies of cancer.” The other authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article