TLR5 Signaling Causes Dendritic Cell Dysfunction and Orchestrates Failure of Immune Checkpoint Therapy against Ovarian Cancer

Abstract

Ovarian cancer accounts for more deaths than any other cancer of the female reproductive system. Patients who have ovarian tumors infiltrated with high frequencies of T cells are associated with a greater survival probability. However, therapies to revitalize tumor-associated T cells, such as PD-L1/PD-1 or CTLA4 blockade, are ineffective for the treatment of ovarian cancer. In this study, we demonstrate that for ovarian cancer, Toll-like receptor 5 (TLR5) signaling, for which the only known ligand is bacterial flagellin, governed failure of PD-L1 and CTLA4 blockade. Mechanistically, chronic TLR5 signaling on CD11c+ cells in vivo and in vitro impaired the differentiation of functional IL-12-producing XCR1+CD103+ conventional type 1 dendritic cells, biasing CD11c+ precursor cells toward myeloid subsets expressing high levels of PD-L1. This culminated in impaired activation of CD8+ T cells, reducing CD8+ T-cell function and ability to persist within the ovarian tumor microenvironment. Expansion of XCR1+CD103+ conventional type 1 dendritic cells in situ using Flt3L-Ig in combination with PD-L1 blockade achieved significant survival benefit in TLR5 knockout mice bearing ovarian tumors, whereas no benefit was observed in the presence of TLR5 signaling. Thus, we have identified a host-intrinsic mechanism leading to the failure of PD-L1 blockade for ovarian cancer, demonstrating that chronic TLR5 signaling on CD11c+ cells is a barrier limiting the efficacy of checkpoint therapy.

Figure 1.. TLR5 signaling impairs the efficacy of immune therapy for ovarian cancer.

Tumors were initiated in wild-type (WT), TLR5 KO or TLR4 KO mice as indicated. After 10 days of tumor progression, mice were given either anti-PD-L1 or anti-CTLA4 every 3–4 days for a total of 4 injections. Survival of mice bearing ID8-Defb29/Vegf-A (A), UPK10 (B), or PPNM (C) ovarian tumors after treatment with anti-PD-L1. (D) Survival of TLR4 KO or wild type controls bearing ID8-Defb29/Vegf-A ovarian tumors after treatment with anti-PD-L1. (E) Survival of mice bearing ID8-Defb29/Vegf-A ovarian tumors after treatment with anti-CTLA4. A log-rank test was used to compare survival proportions. Numbers in parentheses are number of mice/group with a total of at least 2 independent experiments. (* p < .05, ** p < .01, *** p < .001, **** p < .0001).

Figure 2.. TLR5 expression on myeloid and dendritic cells, but not T cells, corresponds with tumor-associated cellular changes within the ovarian tumor microenvironment.

(A) Peritoneal wash exudates from TdTomato TLR5 reporter mice bearing ID8-Defb29/Vegf-A ovarian tumors were analyzed by flow at day 25 for TLR5-expressing (TdTomato⁺) cells in the TME. Percentages based upon frequency of TLR5⁺ or TLR5⁻ cells out of total CD45⁺. N=14 total mice. (B-C) Analysis of TLR5 expressing cell subsets in the peritoneum of naïve mice (no tumor) and in the TME of mice with ovarian tumors at 10-, 20-, and 30-days post-tumor. N=5 mice per group. Percentages are based upon total TLR5⁺ or TLR5⁻ cell proportions expressing each set of surface markers defined in the legend and depicted as stacked bar graphs (B) or pie graphs (C). Phenotypic analysis of TLR5⁺ (D and F) or TLR5⁻ (E and G) MHCII hi and CD11c⁺ DC cell subsets of TdTomato TLR5 reporter mice. (D and E) proportions of each phenotypic DC subset from TLR5⁺ or TLR5⁻ cells. (F and G) MFI of each marker from cells in D and E. N=5 per group. For all p values, Mann-Whitney unpaired t-test (* p < .05, ** p < .01, *** p < .001, **** p < .0001) was used to calculate significance. Error bars represent mean ± SEM. All graphs are representative of at least 2 individual experiments.

Figure 3.. TLR5 signaling reduces the frequency and functionality of cDC1s in the ovarian tumor microenvironment.

A-D: Dendritic cells in the peritoneal environment of wild-type and TLR5 KO mice were analyzed in the absence of a tumor (non-tumor-bearing) or 7- and 15-days post initiation of ID8-Defb29/Vegf-A ovarian tumors. N = 4 mice/group. Plots representative of at least 3 individual experiments. (A) Total cDC1 and (B) IL-12+ cDC1 in the peritoneal cavity. (C) MFI (D) and representative histograms of intracellular IL-12 and surface CD80 levels in cDC1s from the ovarian TME 15 days post ID8-Defb29/Vegf-A tumors. (E) Schematic depicting strategy to neutralize TLR5 signaling. Peritoneal wash samples and tumor nodules were analyzed on day 30 post-tumor. N = 4 mice/group. Plots are representative of at least 3 individual experiments. (F) t-SNE map of pre-gated CD11c+ MHCII+ cells. Red circle depicts cDC1 populations based upon CD103 and XCR1 expression, highlighted by the arrow in G and H. (G) Population frequencies and (H) heat map of phenotypic marker expression by DCs from the clusters in F. (I) Quantitation of cDC1 numbers and (J) MFI of PD-L1 levels with representative histograms. Unpaired non-parametric t-tests with Mann-Whitney correction was used to calculate statistical significance for all plots (* p < .05, ** p < .01, *** p < .001, **** p < .0001). Error bars represent mean ± SEM.

Figure 4.. TLR5 signaling in vivo attenuates the accumulation and functionality of tumor-associated DC subsets.

(A) Schema for mixed bone marrow chimera and treatment regimen. Briefly, wild-type CD45.1 congenic mice were lethally irradiated, followed by transfer of a 50:50 mixture of wild-type (CD45.1) and TLR5 KO (CD45.2) bone marrow. After 10 weeks, ID8-Defb29/Vegf-A tumors were initiated, followed by the initiation of PD-L1 blockade. After 15 days post-tumor, peritoneal wash exudates/tumor nodules were analyzed using flow cytometry. (B) Frequency of wild-type (CD45.1) or TLR5 KO (CD45.2) XCR1⁺ CD103⁺ (cDC1), (C) CD11b⁺ CD103⁺ (mucosal DC), (D) CD11b⁺ CD103⁻ (myeloid DC), (E) PD-L1hi cDC1 (XCR1⁺ CD103⁺) with or without anti-PD-L1 therapy. (F) Frequency of wild-type (CD45.1) or TLR5 KO (CD45.2) IL-12⁺ cDC1s. Unpaired Mann-Whitney (between mice) or paired t-tests (within mice) were used to calculate significance (* p < .05, ** p < .01, *** p < .001, **** p < .0001). Dots represent individual animals. Plots are representative of three independent experiments, with 5–10 mice/group.

Figure 5.. Chronic TLR5 signaling by flagellin promotes differentiation and expansion of PD-L1 myeloid subsets in the absence of a tumor.

Bone marrow from wild-type or TLR5 KO mice were cultured for 8 days with FLT3L and purified flagellin (Salmonella typhimurium) for 2 days (2D - acute) or 8 days (8D - chronic) of culture. (A) Total numbers of differentiated myeloid and dendritic cell subsets: cDC1s (CD11c⁺, MHCII^hi, XCR1⁺, SIRPα⁻), cDC2s (CD11c⁺, MHCII^hi, XCR1⁻, SIRPα⁺), Ly6C (CD11c⁻, CD11b⁺, Ly6C⁺, Ly6G⁻), Ly6G (CD11c⁻, CD11b⁺, Ly6C^low, Ly6G⁺), and other undefined myeloid cells (CD11c⁻, CD11b⁺, Ly6C⁻, Ly6G⁻). (B) Total numbers of PD-L1 high expressing subsets from A. (C-D) Bulk DC/myeloid cultures were pulsed with SIINFEKL peptide followed by incubation with OT-1 transgenic T cells, followed by analysis of T cell proliferation. (C) Proliferation index and (D) representative histograms of cell trace violet-labeled proliferating OT-1 CD8 T cells. Unpaired t-test with Mann-Whitney correction was used to calculate significance. (* p < .05, ** p < .01, *** p < .001, **** p < .0001). Each point represents cultures from an individual animal. Plots represent all data points combined from two experiments. (E-G) CITE-seq was performed on FLT3L bone marrow cultures with or without chronic flagellin stimulation for 8 days. (E) Gene set enrichment analysis between cultures expanded in the presence of chronic flagellin stimulation versus cultures expanded in the presence of FLT3L only. (F) Volcano plot of differentially expressed genes for cDC1 cells (pseudobulk analysis was performed on cells in cluster 10, identified cDC1s, grouped according to their origin). (G) Pathway analysis of differentially expressed genes identified in cluster 10 by grouping cells to their origin.

Figure 6.. CD8 T cells are necessary for survival of TLR5 KO ovarian tumor-bearing mice treated with anti-PD-L1.

(A) Treatment schema with IP administration of anti-CD8a (αCD8a) and anti-PD-L1. Briefly, wild-type or TLR5 KO mice were treated with αCD8a prior to and after tumor initiation with ID8-Defb29/Vegf-A tumors and PD-L1 blockade. (B) Survival proportions of TLR5 KO and wild-type mice bearing ID8-Defb29/Vegf-A tumors. Log-rank test for survival compared to wild-type (* p < .05, ** p < .01, *** p < .001, **** p < .0001 ) N=5 per group and is representative of two independent experiments. (C) Treatment schema for rechallenge of TLR5 KO mice surviving ID8-Defb29/Vegf-A or UPK10 tumors after PD-L1 blockade (from 1A and B). After 100 days, surviving or naïve TLR5 KO mice were injected with the same ovarian tumor cell line as established in surviving mice on day 0 with survival being monitored in the absence of PD-L1 blockade. Survival of rechallenged TLR5 KO mice with ID8-Defb29/Vegf-A tumors (D) or UPK10 tumors (E). Log-rank test was used to compare survival proportions. Numbers in parentheses indicate total mice within each group combining at least 2 independent experiments. (* p < .05, ** p < .01, *** p < .001, **** p < .0001). (F-H) (F) Treatment schema to investigate T cell function during acute (7 days post-tumor, +7D) or 7 days after tumor rechallenge of TLR5 KO mice surviving a primary or secondary challenge with ID8-Defb29/Vegf-A tumors. Secondary challenge indicates surviving TLR5 KO mice that received a re-injection of ID8-Defb29/Vegf-A tumors 100 days after surviving primary tumor challenge. Primary challenge indicates surviving TLR5 KO mice after primary tumor challenge and treatment with PD-L1 blockade. All mice were assessed 7 days following ID8-Defb29/Vegf-A tumor administration. (G) Representative gating of CD8 T cells expressing CD107a and IFNγ. (H) Quantification of T cells within the ovarian TME. Unpaired t-test with Mann-Whitney Correction (* p < .05, ** p < .01, *** p < .001, **** p < .0001 ). Error bars represent mean ± SEM. Each dot represents an individual animal. Plots are representative of two experiments with an 4–5 mice/group.

Figure 7.. Dendritic cell-specific deletion of TLR5 results in significantly increased survival during anti-PD-L1 therapy, corresponding with enhanced cDC1 frequencies and increased CD8 T cell function.

A-I TLR5fl/fl × CD11c cre negative (CD11c.TLR5^wt.cre^neg), CD11c Cre positive (CD11c.TLR5^wt.cre), and TLR5fl/fl × CD11c Cre positive (CD11c.TLR5^ko.cre) mice were given ID8-Defb29/Vegf-A ovarian tumors, followed by administration of PD-L1 blockade 10 days post-tumor-initiation. (A) Survival proportions of CD11c.TLR5^wt.cre^neg, CD11c.TLR5^wt.cre, and CD11c.TLR5^ko.cre mice. Log-rank test for survival compared to CD11c.TLR5^wt.cre^neg and CD11c.TLR5^wt.cre strains (* p < .05, ** p < .01, *** p < .001, **** p < .0001). Numbers in parenthesis indicate total number of mice and encompass two independent experiments. (B-I) 25 days post-initiation of ID8-Defb29/Vegf-A tumors in CD11c.TLR5^wt (cre negative) and CD11c.TLR5^ko mice (cre positive), peritoneal wash exudates and tumor nodules were evaluated using flow cytometry. (B) Total numbers of cDC1s, (C) PD-L1hi cDC1s, (D) MFI of PD-L1 and (E) representative histograms of PD-L1 on CD11c⁺ MHCII⁺ cells. (F) Total PD-1⁺ Lag3⁺ CD8 T cells in the TME. (G) Representative gating of CD3⁺ CD44hi CD107⁺ or IFNγ⁺ T cells. (H) Total CD3⁺ CD107a⁺ and (I) CD3⁺ IFNγ⁺. Unpaired t-test with Mann-Whitney correction was used to calculate significance (* p < .05, ** p < .01, *** p < .001, **** p < .0001). Error bars represent mean ± SEM. Dot plots are representative of at least three experiments, with each dot representing an individual animal. (J-K) Survival proportions of TLR5 KO and wild-type mice bearing ID8-Defb29/Vegf-A tumors (J) or PPNM tumors (K) and treated with FLT3L and/or anti-PD-L1. Log-rank test for survival compared to wild type (* p < .05, ** p < .01, *** p < .001, **** p < .0001). N=5 per group, results representative of at least two independent experiments.