TLR8 agonists remodel the tumor immune microenvironment through PF4-dependent T cell recruitment and ancillary mechanisms

Abstract

Pattern recognition receptors (PRRs) are crucial modulators of the tumor immune microenvironment (TIME); yet, their comparative efficacy remains poorly characterized. Platelet factor 4 (PF4/CXCL4) exerts paradoxical effects in cancer, and its regulation by PRR signaling is unclear. Here, we systematically screened PRR agonists in murine tumor models and identified TLR8 agonists as the most potent inducers of a pro-immunogenic TIME, superior to agonists targeting TLR3/7/9 or NOD1/2. TLR8 activation drove CD45⁺ leukocyte infiltration, potentiated conventional dendritic cell (cDC) and macrophage phagocytosis and migration, amplified recruitment of CD8⁺ and conventional CD4⁺ (cCD4⁺) T cells, lowered Treg proportions, elicited pro-inflammatory and tumoricidal phenotypes in innate immune cells and CD8⁺ T cells. Notably, TLR8 agonism suppressed tumor growth in both immunocompetent and T cell-deficient mice, indicating the involvement of both innate and adaptive immunity. Mechanistically, TLR8 agonists upregulated PF4 expression in macrophages, cDC2, and CD8⁺ T cells via the NF-κB pathway. PF4 in turn recruited cCD4⁺ T cells via CXCR3, and its local overexpression mimicked the antitumor effect of TLR8 activation. Beyond PF4, TLR8 signaling mediated PF4-independent effects, including Treg suppression via IFN-γ and enhanced macrophage phagocytosis. Combination of a TLR8 agonist with anti-PD-1 therapy markedly and synergistically improved survival of tumor-bearing mice. Thus, TLR8 agonists optimally remodel the TIME through PF4-dependent T cell recruitment and PF4-independent ancillary mechanisms. Our finding that the antitumor activity of locally induced PF4 contrasts with its reported protumor effects when expressed systemically clarifies the context-dependent duality of PF4 in cancer. These results position TLR8 agonists as promising candidates for combination immunotherapy.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00262-026-04329-8.

Introduction

Pattern recognition receptors (PRRs) are a class of innate immune sensors that play a critical role in detecting pathogen-associated molecular patterns (PAMPs) and damage-associated molecular patterns (DAMPs) [1] and have recently emerged as key regulators of tumor immunity and the tumor immune microenvironment (TIME) [2]. PRRs, such as Toll-like receptors (TLRs) and NOD-like receptors (NLRs), are expressed on both immune cells and tumor cells. Their activation can initiate signaling cascades that modulate immune responses, inflammation, and cell survival, thereby influencing tumor progression and antitumor immunity [3].

Intriguingly, PRRs can exert dual roles, either promoting or suppressing tumor growth, depending on the receptor type, cellular context, and tumor stage [2, 3]. On one hand, PRR activation can directly stimulate the antitumor activity of innate immune cells and promote adaptive immunity indirectly through regulation of dendritic cell (DC) function or directly through modulation of lymphocyte function [4]. These effects can lead to the suppression of tumor growth and the establishment of an immunogenic TIME. On the other hand, chronic or dysregulated PRR signaling may contribute to tumorigenesis by fostering an immunosuppressive TIME, characterized by the recruitment of regulatory T (Treg) cells and myeloid-derived suppressor cells (MDSCs) [5–7]. This paradoxical role of PRRs highlights their complexity in cancer biology and underscores the need for a deeper understanding of their mechanisms.

Despite significant advances in elucidating the functions of PRRs in tumor immunity, several critical gaps remain in our knowledge. First, the collective impact of PRRs on the TIME and the precise regulatory mechanisms are not fully understood. Second, the comparative efficacy of different PRR agonists in modulating the TIME and antitumor immunity has not been systematically investigated. For instance, TLR agonists such as CpG oligonucleotides and poly(I:C) have shown promise in preclinical and clinical studies, but how their effects compare with those of other TLR or NLR agonists is unclear [8–10]. Such comparisons are essential for identifying the most effective therapeutic strategies targeting PRRs.

Platelet factor 4 (PF4), also known as chemokine (C-X-C motif) ligand 4 (CXCL4), is secreted by platelets and many cell types including monocytes, DCs, natural killer (NK) cells, activated T cells, and cancer cells [11, 12]. Besides its physiological role in hematopoiesis, cell proliferation, differentiation, and activation, it is involved in multiple diseases such as vaccine-induced immune thrombotic thrombocytopenia (VITT) and cancer. Intriguingly, PF4 has been shown to exhibit an antitumor activity through its angiostatic effect or interference with protumor chemotaxis [13, 14], as well as a protumor function by promoting platelet production, infiltration of MDSC, and regulatory T cell (Treg) differentiation and proliferation[15–18]. Thus, its potential as a therapeutic target in cancer remains elusive.

Here, we evaluated the relative efficacy of various PRR agonists in shaping the TIME and identified TLR8 agonist as the most efficient inducer of a favorable tumor immune microenvironment. The therapeutic effect of TLR8 agonists was further validated in different tumor models, and PF4 was identified as a crucial downstream effector.

Results

TLR8 agonist outperforms a panel of PRR agonists in remodeling the TIME

We selected a panel of PRRs whose activation was reported to have antitumor potentials and compared their roles in shaping the TIME in a subcutaneous model (Fig. 1A). Agonists for these PRRs were divided into two groups and tested in two separate experiments. They were administered intratumorally to subcutaneous AKR tumors and TIMEs were analyzed by flow cytometry with a focus on adaptive immunity (Fig. S1). Only TLR8 and NOD2 agonists significantly increased the proportion of CD45⁺ leukocytes among total cells (Fig. 1B). While TLR8 and NOD1 agonists raised the percentage of conventional DCs (cDCs) within the CD45⁺ compartment, almost all agonists—except the TLR7 agonist—promoted the uptake of tumor-expressed ZsGreen proteins by cDCs (Fig. 1B, C), suggesting comparable enhancement of DC phagocytosis of tumor antigens. Consequently, ZsGreen⁺ DCs were increased by these agonists in the tumor-draining lymph nodes (LNs), indicating an active migration of DCs from the tumor site, although the ratio of cDCs in the LNs was only upregulated by TLR8-506. (Fig. S2A, B). Agonists for TLR8 and NOD2 increased the proportions of macrophages in both tumors and LNs (Figs. 1E & S2C). All agonists except the NOD1 agonist enhanced tumor-antigen phagocytosis by tumor-infiltrating macrophages (Fig. 1F), but only TLR8-506 significantly promoted the migration of antigen-loaded macrophages to LNs (Fig. S2D). Notably, TLR8-506 was the sole agonist that upregulated both CD8⁺ T and conventional CD4⁺ (cCD4⁺) T cells (Fig. 1G, H), whereas only the TLR7 agonist upregulated the Treg cell proportion (Fig. 1I). Accordingly, TLR8 agonists raised the ratios of effector T cells to Treg cells (Fig. 1J, K).

Fig. 1.

Dynamics of immune cell subsets in response to agonist stimulation. A Schematic of the experimental workflow. Immune profiling by flow cytometry was performed 48 h after a single intratumoral injection of PRR agonists. B Percentage of CD45⁺ leukocytes among live cells following intratumoral administration of different agonists: TLR3 agonist poly IC (25 µg/tumor), TLR7 agonist imiquimod hydrochloride (25 µg/tumor), TLR8 agonist TL8-506 (10 µg/tumor), TLR9 agonist CpG ODN 2395 (50 µg/tumor), STING agonist ADU-S100 ammonium salt (25 µg/tumor), NOD1 agonist Tri-DAP (10 µg/tumor), and NOD2 agonist murabutide (5 µg/tumor). C Proportion of cDCs within the CD45⁺ population. D Percentage of ZsGreen⁺ cells among DCs. E Proportion of macrophages within the CD45⁺ population. F Percentage of ZsGreen⁺ cells among macrophages. G Proportion of CD8⁺ T cells within the CD45⁺ compartment. H Proportion of cCD4⁺ T cells within the CD45⁺ compartment. I Proportion of Tregs within the CD45⁺ compartment, pooled from two independent experiments. J Ratio of cCD4⁺ T cells to Tregs. (K) Ratio of CD8⁺ T cells to Tregs. Data represent the mean ± SEM (n = 8 mice per group). Statistical comparisons were performed using one-way ANOVA. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

In summary, our screen identified the TLR8 agonist as the most comprehensive inducer of a pro-immunogenic TIME. Unlike other agonists, it uniquely enhanced both CD8⁺ and cCD4⁺ T cell infiltration, thereby markedly improving effector-to-suppressor ratios. Given this coordinated immunomodulatory profile—together with the relative under-characterization of TLR8 in tumor immunity—we selected TLR8 for mechanistic exploration.

TLR8 agonist suppresses tumor growth through induction of innate an adaptive immunity

To confirm that the immunomodulatory effects on the TIME were specific to TLR8 signaling and not compound-dependent, we evaluated two distinct agonists: TL8-506 and the clinical-stage agonist motolimod [19–23]. Inclusion of motolimod, which has entered clinical evaluation in oncology, strengthens the translational relevance of our findings. At their IC₅₀ concentrations, both agonists induced comparable gene upregulation in vitro (Fig. S3A). Although murine TLR8 was historically considered non-functional [24, 25], Tlr8 knockdown completely abolished agonist-induced Ifng and Tnf mRNA expression in mouse macrophages (Fig. S3C–E), which, together with accumulating evidence [26–28], confirms the functionality of murine TLR8 in our system. To achieve comparable activations in vivo, doses of the two agonists were adjusted proportionally to their IC₅₀ values. Consistent with the AKR model data, both TLR8 agonists elevated the frequency of CD45⁺ cells in MC38 tumors (Fig. 2A). The proportion of cDCs among CD45⁺ cells and their uptake of ZsGreen were enhanced (Fig. 2B). Similarly, macrophages and their phagocytotic activity were enhanced (Fig. 2C). The percentages of CD8⁺ T and cCD4⁺ T cells increased, while that of Treg cells decreased, leading to higher effector-to-Treg ratios (Fig. 2D, E). Together, these results point to a potent induction of adaptive antitumor immunity and alleviation of immunosuppression by TLR8 agonism.

Fig. 2.

Immune modulation and antitumor effects of TLR8 agonists. A Flow cytometric analysis of CD45⁺ leukocytes was performed 48 h after three intratumoral injections of TL8-506 (10 µg/tumor) or motolimod (50 µg/tumor) (n = 6). B Percentage of cDCs within the CD45⁺ population and proportion of ZsGreen⁺ cells among cDCs (n = 6). C Percentage of macrophages within the CD45⁺ population and proportion of ZsGreen⁺ cells among macrophages (n = 6). D Frequency of cCD4⁺ T cells, CD8⁺ T cells, and Tregs within the CD45⁺ compartment (n = 6). E Ratios of cCD4⁺ T cells to Tregs and CD8 T cells to Tregs (n = 6). F, G Tumor growth curves and terminal tumor weights of AKR and MC38 tumors in C57BL/6 mice administered TL8-506 or motolimod every 3 days (n = 5). H, I Tumor growth curves and terminal tumor weights of AKR and MC38 tumors in nude mice administered motolimod every three days (n = 5). J, K Macrophage phagocytosis of CFSE-labeled tumor cells following motolimod stimulation (n = 5). Data represent the mean ± SEM. Data in panels A–G were analyzed using one-way ANOVA, and data in panels H–K were analyzed using Student’s t-tests. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

We next evaluated the impact of TLR8 agonists on tumor growth. While the agonists did not affect tumor cell proliferation in vitro (Fig. S3E, F), they strongly suppressed both AKR and MC38 tumor growth in vivo (Fig. 2G, H), without altering mouse body weight (Fig. S3G, H). Interestingly, TLR8 agonists retained their antitumor activity in nude mice lacking T cells, although the effect was less pronounced than in the immunocompetent mice (Fig. 2H, I). Consistent with enhanced macrophage phagocytosis in vivo (Fig. 2C), motolimod potently promoted the phagocytosis of CFSE-labeled tumor cells by both mouse macrophages and human THP-1-derived macrophages in vitro (Fig. 2J, K). These results indicate that TLR8 agonism inhibits tumor growth via both adaptive and innate immune mechanisms.

TLR8 agonist activates PF4-CXCR3 pathway to shape the TIME

To elucidate the mechanisms underlying TLR8-mediated TIME remodeling, we performed scRNA-seq on immune infiltrates from AKR tumors (Figs. 3A–C and S4A, B). Reinforcing our flow cytometry data (Fig. 2), scRNA-seq confirmed key immunomodulatory trends: an increase in cDC and cCD4⁺ T cell frequencies alongside a decrease in Treg proportion (Fig. S4B). Notably, the cDC2 subset, which represented the predominant cDC population, was significantly elevated by motolimod treatment (Fig. S4A, B). Tlr8 mRNA expression was mainly found in CD8⁺ T cells, cDC2, macrophages, and neutrophils (Fig. 3D, E), identifying these populations as primary responders to the agonist. In human tumor-infiltrated immune cells, TLR8 was also highly expressed in cDC2 and macrophages/monocytes, though not in CD8⁺ T cells (Fig. S4C, D). As a consequence of direct TLR8 stimulation, cDC2, macrophages, neutrophils, and CD8⁺ T cells exhibited more activated phenotypes, with upregulated expression of pro-inflammatory cytokines (e.g., Il1b, Tnf, and Ifng), tumoricidal molecules (e.g., Nos2, Camp, and Lyz2), and cytotoxic proteins (e.g., Gzmb), alongside downregulation of immunosuppressive factors such as Il10, Tgfb1, and Pdcd1 (Fig. S4E–H). These data indicate a clear shift of the TIME toward an antitumor state.

Fig. 3.

TLR8 agonist activates PF4-CXCR3 pathway to shape the TIME. A t-SNE plot showing the distribution of all cells from the control and agonist-treated groups combined (n = 15,745 cells). B t-SNE plot depicting the distribution of distinct immune cell types. C Dot plot displaying the marker genes used to define each immune cell population. D Density plot showing the distribution of Tlr8-expressing cells across the t-SNE map. E Violin plot illustrating Tlr8 expression levels across different immune cell types. F Circle plot depicting the strength of Pf4–Cxcr3 receptor–ligand interactions between distinct cell types in the control and motolimod-treated groups. G Differential gene expression profiles of neutrophils, macrophages, cDC2, and CD8⁺ T cells. H Bubble plot comparing changes in specific receptor–ligand interactions between the control and motolimod-treated groups, focusing on macrophage-to-cCD4⁺ T signaling, cDC2-to-cCD4⁺ T signaling, and CD8⁺ T-to-cCD4⁺ T signaling. I Violin plot showing differential expression of Cxcr3 in cCD4⁺ T and Tregs, analyzed using the Wilcoxon rank-sum test. ns, not significant; **P < 0.01

CellChat analysis revealed a significant impact of all three Tlr8-expressing populations on conventional CD4⁺ T cells (Fig. 3F). Notably, multiple chemotactic pathways were upregulated by motolimod to control cCD4⁺ T cell infiltration, among which the Pf4/Cxcl4–Cxcr3 pathway was strongly activated downstream of macrophages, cDC2, and CD8⁺ T cells (Fig. 3G). Consistently, Pf4 mRNA was significantly upregulated in these cells (Fig. 3H), while Cxcr3-expressing cCD4⁺ T cells were increased accordingly (Fig. 3I), indicating Pf4/Cxcl4–Cxcr3 chemotaxis as a key pathway reshaping the TIME.

TLR8 agonist induces PF4 expression through NF-κB signaling pathway

To assess the impact of TLR8 activation on PF4 expression, peritoneal mouse macrophages and THP-1-derived human macrophages were treated with the TLR8 agonist motolimod. This stimulation robustly upregulated both mouse Pf4 and human PF4 expression (Fig. 4A, B). We next examined whether tumor cells contribute to this pathway. MC38 and AKR cells expressed negligible levels of Tlr8 or Pf4 mRNA (Fig. S5A, B). Moreover, motolimod treatment did not induce Pf4 transcription or enhance PF4 protein secretion from tumor cells (Fig. S5B–D), ruling out a functional tumor-intrinsic TLR8–PF4 axis. Consistent with TLR8 signaling, the NF-κB signaling—a key downstream pathway—was significantly activated (Fig. 4C, D) [29]. To directly test whether PF4 expression is regulated by the TLR8-activated NF-κB, we knocked down Rela/RELA in mouse primary macrophages and human THP-1-derived macrophages (Fig. 4E–H). The knockdown abolished motolimod-induced PF4 expression at both mRNA and protein levels (Fig. 4I–K). These results establish Pf4/PF4 as a downstream target of the TLR8-NF-κB signaling pathway.

Fig. 4.

TLR8 agonist induces PF4 expression through NF-κB signaling pathway. A, B qRT-PCR analysis of Pf4 expression in mouse and human macrophages in the motolimod-treated and control groups (n = 3). C, D Luminescence assay showing NF-κB pathway activation in mouse and human macrophages following motolimod stimulation (n = 3). E, F qRT-PCR analysis of Rela expression in Rela-knockdown and control mouse and human macrophages with or without motolimod treatment (n = 3). G, H Western blotting showing p65 protein levels in RELA-knockdown and control mouse and human macrophages with or without motolimod treatment. I, J qRT-PCR analysis of Pf4/PF4 expression in Rela/RELA-knockdown and control mouse and human macrophages with or without motolimod treatment (n = 3). K ELISA quantification of Pf4 levels in the culture supernatant from Rela-knockdown and control mouse macrophages with or without motolimod treatment (n = 3). Data represent the mean ± SEM. Statistical comparisons were performed using Student’s t-tests. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

PF4 remodels the TIME and suppresses tumor growth

We next sought to elucidate the role of PF4 downstream of TLR8 activation in regulating tumor growth and the TIME. Tumor cell lines overexpressing Pf4 were used to establish subcutaneous tumors, mimicking localized Pf4 upregulation within the TIME (Fig. S6A). In both AKR and MC38 models, Pf4 overexpression significantly inhibited tumor growth and enhanced the antitumor effect of motolimod (Fig. 5A, B). Flow cytometric analysis at the experimental endpoint revealed that Pf4 overexpression increased tumor-infiltrating CD45⁺ cells, cCD4⁺ T cells, and CD8⁺ T cells, as well as the ratios of cCD4⁺ T/Treg and CD8⁺ T/Treg ratios — effects consistent with the those of motolimod, although Pf4 moderately increased Tregs whereas motolimod did not (Fig. 5C–H). Increased intratumoral CD8⁺ T cell infiltration in the MC38 model was further confirmed by immunofluorescence staining and quantitative image analysis (Fig. 5I, J). These findings suggest that TLR8 agonists remodel the TIME at least partially through Pf4 and may have Pf4-independent mechanisms for limiting Treg cell accumulation, such as induction of IFN-γ (Figs. S3C and S4F).

Fig. 5.

PF4 remodels the TIME and suppresses tumor growth. A, B Tumor growth curves and terminal tumor weights in C57BL/6 mice bearing AKR or MC38 tumors with or without Pf4 overexpression, treated with or without motolimod administered every three days (n = 6). C Percentage of CD45⁺ cells among live cells in AKR and MC38 tumor models (n = 6). D Percentage of cCD4⁺ T cells within the CD45⁺ population in AKR and MC38 tumor models (n = 6). E Percentage of CD8⁺ T cells within the CD45⁺ population in AKR and MC38 tumor models (n = 6). F Percentage of Tregs within the CD45⁺ population in AKR and MC38 tumor models (n = 6). G Ratios of cCD4⁺ T cells to Tregs in AKR and MC38 tumor models (n = 6). H Ratios of CD8⁺ T cells to Tregs in AKR and MC38 tumor models (n = 6). I Representative immunofluorescence staining of CD8α (red) and DAPI (blue) in MC38-Vector or MC38-Pf4-OE tumor sections with or without motolimod treatment. Scale bar: 50 μm. J Quantification of CD8α⁺ positive cell ratio (%) relative to total DAPI⁺ nucleated cells per section (n = 6). Data represent the mean ± SEM. Statistical comparisons were performed using Student’s t-tests. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

To demonstrate a direct effect of Pf4 on T cell migration, we performed a transwell assay. Motolimod-treated macrophages significantly enhanced T cell migration, an effect blocked by anti-Pf4 antibody (Fig. S6B–D), consistent with in vivo observations and indicating that motolimod-induced Pf4 directly recruits T cells to the tumor site. Moreover, bioinformatics analysis revealed that higher PF4 expression correlates with better survival in patients with breast, colon, and liver cancer (Fig. S6E), supporting its antitumor role. Together, these data demonstrate that PF4 shapes the TIME and contributes to the tumor-suppressive effect of TLR8 agonists.

TLR8 agonist enhances the efficacy of anti-PD-1 therapy

We subsequently investigated the synergistic effect of a TLR8 agonist in combination with anti-PD-1 therapy. In both AKR and MC38 models, motolimod significantly potentiated the antitumor activity of the PD-1 inhibitor and prolonged the survival of tumor-bearing mice (Fig. 6A–F). These findings confirm that TLR8 agonists can enhance the efficacy of other immunotherapeutic agents.

Fig. 6.

TLR8 agonist enhances the efficacy of anti-PD-1 therapy. A Experimental workflow showing treatment schedule for AKR tumor-bearing C57BL/6 mice receiving intratumoral motolimod and intraperitoneal anti-PD-1 (αPD-1) antibody, alone or in combination. B, C Tumor growth curves and survival analysis of AKR tumor-bearing mice treated with the indicated regimens (n = 7 mice per group). D Experimental workflow showing treatment schedule and details for MC38 tumors in C57BL/6 mice receiving intratumoral motolimod and intraperitoneal αPD-1 antibody, alone or in combination. E, F Tumor growth curves and survival analysis of MC38 tumor-bearing mice treated with the indicated regimens (n = 7 mice per group). Data represent the mean ± SEM. Tumor growth curves were analyzed using one-way ANOVA, and survival curves were analyzed using log-rank testing. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

Discussion

Our study systematically compares the efficacy of diverse PRR agonists in remodeling the TIME and identifies TLR8 agonists as the most potent inducers of a pro-immunogenic TIME. This finding contributes to the evolving landscape of PRR-based therapies, which have transitioned from traditional vaccine adjuvants to direct antitumor agents. Historically, PRR agonists such as TLR9 agonists (CpG oligonucleotides) and TLR3 agonists (poly(I:C)) have been valued as vaccine adjuvants for their ability to enhance antigen presentation by DCs and amplify adaptive immune responses [30]. However, their clinical translation as standalone cancer therapies may be limited by incomplete TIME modulation. Our data reveal that while most tested agonists improve phagocytosis and LN migration of DCs and macrophages, only TLR8 agonists concurrently enhance the infiltration of CD8⁺ and cCD4⁺ T cells, and increase effector-to-Treg ratios (Figs. 1 and S2). This comprehensive remodeling of adaptive immunity, coupled with improvements in innate immune cell function (e.g., macrophage phagocytosis), positions TLR8 agonists as uniquely capable of bridging adjuvant activity and direct antitumor efficacy, offering a more holistic approach to reprogram TIME.

Notably, TLR7 and TLR8 are structurally similar TLRs that recognize single-stranded RNA [31] yet exhibit striking functional divergence in our models: TLR8 agonists augmented effector T cells and reduced Treg ratio, whereas TLR7 agonists specifically increased Treg infiltration (Figs. 1 and 2). This discrepancy aligns with their activation of distinct signaling cascades reported in previous studies [32, 33]. For instance, during RNA virus infection, TLR7 stimulation has been shown to inhibit Th1-type cytokine expression in monocytes and type I IFN response in CD4⁺ T cells, while TLR8 triggering promotes Th1 differentiation [32]. Further supporting ligand-dependent specialization, Forsbach et al. demonstrated that TLR8 recognizes RNA motifs distinct from those sensed by TLR7 [25]. These differences in signaling output likely underpin their contrasting effects on TIME.

The dual role of PF4/CXCL4 in tumor progression—observed here as an antitumor mediator but reported elsewhere as protumor [15–18]—highlights the critical influence of contextual factors, including local versus systemic expression and model-specific microenvironments. In our study, TLR8 agonists induced PF4 locally within the TIME, where it acts as a key chemoattractant for cCD4⁺ T cells via the CXCR3 pathway. In contrast, systemic PF4 overexpression has been linked to protumor effects by promoting platelet-tumor cell interactions [15, 34]. Although a protumor role of non-platelet-derived PF4 in the TIME was proposed either due to its effect on Treg polarization or MDSC recruitment, those studies often relied on genetic Pf4 deletion or systemic administration of PF4 neutralizing antibodies, which may not specifically target the TIME [16–18]. Additionally, our observation that local PF4 overexpression moderately increases Treg infiltration (Fig. 5G) but still enhances antitumor activity suggests that, in the local TIME, the pro-inflammatory effects of PF4 (e.g., effector T cell recruitment) outweigh its weak Treg-promoting activity—an equilibrium that may shift in systemic contexts.

Beyond PF4, our findings reveal that TLR8 agonists exert antitumor effects through PF4-independent mechanisms, underscoring their ability to engage both innate and adaptive immunity. This is most evident in T cell-deficient nude mice, where TLR8 agonists retain significant antitumor activity (albeit reduced compared to immunocompetent mice; Fig. 2H, I), pointing to innate immune-mediated effects. Consistent with this, TLR8 agonists strongly enhance macrophage phagocytosis of tumor cells in vitro and in vivo (Fig. 2C, J, K), as well as upregulate the tumoricidal/antimicrobial phenotypes of macrophages and neutrophils (Fig. S4D, E). Moreover, scRNA-seq analysis revealed predominant Tlr8 expression in tumor-associated macrophages, while in vitro assays ruled out direct tumor cell killing by motolimod. Together, these data strongly suggest that macrophages are primary innate immune effectors responsible for the TLR8-driven tumor control observed in T cell-deficient hosts. Additionally, TLR8 agonists may directly activate NK cells [19, 35], further contributing to tumor suppression in the absence of T cells. Regarding Treg regulation, the discrepancy between PF4 overexpression (modestly increasing Tregs) and TLR8 agonists (reducing Tregs; Fig. 5G) suggests additional pathways: TLR8-induced IFN-γ, for example, is known to inhibit Treg differentiation, providing a plausible PF4-independent mechanism to limit immunosuppression (Figs. S3C and S4F) [36, 37]. Alternatively, a direct reversal of human Treg function by TLR8 agonism has been reported [38, 39], although this mechanism is less relevant in mice as TLR8 is barely expressed in mouse Treg cells (Fig. 3E). These combined effects—activation of innate effectors (macrophages, NK cells) and modulation of adaptive immunity—position TLR8 agonists as versatile therapeutics capable of functioning across diverse immune contexts.

The clinical translation of TLR8 agonists hinges on strategies to specifically target them to tumors, minimizing systemic toxicity. Our use of intratumoral administration leverages the enriched expression of TLR8 in mouse tumor-infiltrating immune cells including macrophages, cDC2, and CD8⁺ T cells (Fig. 3D, E), and human tumor-infiltrating cDC2, macrophages, and monocytes (Fig. S4A, B), ensuring localized activation. To further enhance specificity, future formulations could employ tumor microenvironment-responsive delivery systems, such as pH-sensitive nanoparticles or matrix metalloproteinase (MMP)-cleavable carriers, which release agonists only in the acidic, protease-rich tumor milieu [40, 41]. This would reduce off-target activation of TLR8 in healthy tissues (e.g., peripheral blood monocytes), which can trigger systemic inflammation. Additionally, our data showing synergism between TLR8 agonists and anti-PD-1 therapy (Fig. 6) highlight the potential of combinatorial strategies: TLR8 agonists remodel TIME to increase effector T cell infiltration, rendering tumors more susceptible to immune checkpoint blockade. Optimizing dosing sequences—for example, priming with TLR8 agonists to activate TIME before initiating anti-PD-1 therapy—may further enhance efficacy.

In conclusion, our study establishes TLR8 agonists as superior modulators of TIME through both PF4-dependent (effector T cell recruitment) and PF4-independent (innate immune activation and Treg suppression) mechanisms. Their ability to bridge innate and adaptive immunity, coupled with their synergism with immune checkpoint inhibitors, underscores their therapeutic potential. Dissecting the context-dependent roles of PRRs and PF4 provides a framework for developing targeted therapies that harness the full power of the immune system against cancer. Future work should focus on optimizing tumor-specific delivery and exploring combinatorial regimens to maximize clinical benefit.

Materials and methods

Cell lines

AKR (mouse esophageal cancer cell line; Qingqi Biotechnology, Shanghai, China) and MC38 (mouse colon adenocarcinoma cell line; Nanjing Cobioer Biosciences, Nanjing, China) cells were cultured in DMEM (Gibco) supplemented with 10% FBS (Anlite) and 1% penicillin/streptomycin. THP-1 cells (Pricella Biotechnology, Wuhan, China) were cultured in RPMI-1640 (Gibco) with 10% FBS. For differentiation, THP-1 cells were treated with 50 ng/mL PMA (MedChemExpress, HY-19065) for 48 h, followed by M1 polarization with LPS (Sigma, #L4391, 100 ng/mL), IFN-γ (PeproTech, #315–05, 20 ng/mL), and IL-1β (PeproTech, #211-11B, 10 ng/mL). MC38 and AKR cells stably expressing ZsGreen, Pf4, or empty vector were generated by lentiviral transduction.

Mouse tumor models and treatments

Female C57BL/6 and nude mice (6–8 weeks; GemPharmatech, Nanjing, China) were housed under SPF conditions. MC38-ZsGreen or AKR-ZsGreen cells (0.75 × 10⁶) were injected subcutaneously. Tumor volumes were calculated as V = L × W²/2. When tumors reached approximately 200 mm³, PRR agonists were administered via intratumoral injection at the following doses: TLR3 agonist poly(I:C) (Sigma-Aldrich, #42,424–50-0; 25 µg/tumor); TLR7 agonist imiquimod hydrochloride (MedChemExpress, #HY-B0180A; 25 µg/tumor); TLR8 agonist TL8-506 (InvivoGen, #tlrl-tl8506; 10 µg/tumor); TLR8 agonist motolimod (MedChemExpress, #HY-13773; 50 µg/tumor); TLR9 agonist CpG ODN 2395 (Class C) (InvivoGen, #tlrl-2395–1; 50 µg/tumor); STING agonist ADU-S100 ammonium salt (MedChemExpress, #HY-12885B; 25 µg/tumor); NOD1 agonist Tri-DAP (InvivoGen, #tlrl-tdap; 10 µg/tumor); and NOD2 agonist murabutide (InvivoGen, #tlrl-mbt; 5 µg/tumor). Anti-PD-1 antibody (clone RMP1-14, Bio X Cell) was administered intraperitoneally at a dose of 100 µg per mouse. At the experimental endpoint, mice were euthanized by carbon dioxide (CO₂) inhalation. All animal procedures were approved by the Institutional Animal Care and Use Committee of the School of Biomedical Engineering, Shanghai Jiao Tong University (Approval No. 2023030). Experiments were performed in strict accordance with the National Research Council's Guide for the Care and Use of Laboratory Animals (8th edition) and reported following the Animal Research: Reporting of In Vivo Experiments (ARRIVE) 2.0 guidelines.

Flow cytometry

For the initial screen comparing different PRR agonists (Fig. 1), AKR tumors were harvested 48 h after a single intratumoral injection for flow cytometric analysis. In subsequent experiments focused on TLR8 agonists (Figs. 2), MC38tumor-bearing animals received three intratumoral injections every 3 days, and tumors were processed for analysis 48 h after the final dose. Tumors were dissociated with a Tumor Dissociation Kit (Miltenyi, #130–096-730). Cells were stained with fluorochrome-conjugated antibodies (eBioscience, BioLegend, BD Biosciences, listed in Supplementary Table S1) and a Zombie viability kit (BioLegend, #423,108). For detection of the transcription factor FoxP3, cells were surface-stained, followed by fixation and nuclear permeabilization using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific, #00–5523-00), according to the manufacturer’s protocol, and subsequently stained with anti-Foxp3 antibody. Data were acquired on BD LSRFortessa and analyzed with FlowJo.

Isolation of peritoneal macrophages and splenic T cells

Peritoneal macrophages were elicited by intraperitoneal injection of 3% thioglycollate (Sigma, #T9032) and collected by lavage. Purity (≥ 70%) was confirmed by flow cytometry using anti-F4/80 (BioLegend, BN8) and anti-CD11b (BD Biosciences, M1/70). Splenic T cells were obtained by mechanical dissociation, RBC lysis (BioLegend, #420,301), and stimulated with Dynabeads™ Mouse T-Activator CD3/CD28 (Thermo Fisher, #11456D) plus recombinant IL-2 (PeproTech, #212–12).

Macrophage phagocytosis assay

Macrophages were pretreated with motolimod (MedChemExpress, #HY-13773) and co-cultured with CFSE-labeled tumor cells (Thermo Fisher, #C34554). Phagocytosis was analyzed by flow cytometry (BD LSRFortessa) using FlowJo software.

Lentiviral RNA knockdown

shRNA sequences against mouse Tlr8, Rela, and human RELA were cloned into a lentivector (Supplementary Table S2). IC-enhancer lentiviruses were produced by Genewiz (Suzhou, China) and used to infect immune cells in the presence of 6 μg/ml polybrene (Sigma-Aldrich).

Quantitative PCR (qRT-PCR)

RNA was extracted with an RNeasy Mini Kit (Qiagen, #74,104), reverse-transcribed with a PrimeScript RT Kit (Takara, #RR037A), and amplified using TB Green Premix Ex Taq II (Takara, #RR820A) on an ABI 7900HT. GAPDH served as the internal control.

Single-cell RNA sequencing (scRNA-seq)

When tumors reached ~ 200 mm³, mice began receiving intratumoral injections of motolimod (50 µg/mouse) or vehicle every 3 days for a total of three doses. Tumors were harvested 48 h after the final injection. CD45⁺ cells were enriched with anti-CD45-biotin (BioLegend, #103,104) and magnetic beads (Miltenyi, #130–048-101). Libraries were prepared with the 10 × Genomics Chromium Single Cell 5′ Kit and sequenced on Illumina NovaSeq 6000. The final dataset comprised 10,037 high-quality cells from motolimod-treated tumors (Mean Reads per Cell = 48,813) and 8,099 cells from control tumors (Mean Reads per Cell = 63,714). Raw sequencing data were aligned to the mouse reference genome mm10 and quantified using the 10 × Genomics CellRanger pipeline (version 7.1.0) with default parameters. Subsequent analyses were performed in R (version 4.2.0) utilizing the Seurat package (version 4.1.3). Quality control metrics were applied to filter out low-quality cells, retaining those with between 200 and 6,000 detected genes, > 1,000 unique molecular identifiers (UMIs), and < 20% mitochondrial gene content. The Harmony algorithm was implemented to correct for technical batch effects across samples, followed by data normalization using the LogNormalize method with a scale factor of 10,000. Principal component analysis was conducted using the top 40 principal components, which informed both the shared nearest neighbor graph construction (k = 10 neighbors) for unsupervised clustering using the Louvain algorithm (resolution = 0.8) and subsequent t-SNE visualization. Cell type annotation was performed through a combination of canonical marker gene expression patterns (detailed in Fig. 3C) and reference-based approaches. Differential gene expression analysis was carried out using the FindMarkers function with thresholds set at absolute log2 fold change ≥ 1.5 and Bonferroni-adjusted p-value < 0.05. Cell–cell communication networks were interrogated using CellChat (version 1.6.1) to identify significant ligand-receptor interactions. All data visualizations, including violin plots, feature plots, and t-SNE projections, were generated using Seurat's visualization toolkit and ggplot2 (version 3.4.1).

Transwell migration assay

Macrophages were pretreated with motolimod (100 nM), anti-PF4 antibody (R&D Systems, #AF595), or control IgG (R&D Systems, #AB-108-C) and co-cultured with activated T cells in Transwell inserts (Corning, #3422). Migrated cells were stained with crystal violet (Yeasen, #60506ES20) and counted using ImageJ.

Immunofluorescence

Tumors were harvested at the experimental endpoint and bisected longitudinally. One half of each tumor was fixed in formalin and embedded in paraffin. For each experimental group, six independent tumors were analyzed. One representative paraffin-embedded section per tumor was prepared for immunofluorescence staining with anti-CD8α antibody (Cell Signaling Technology, #98,941) and counterstained with DAPI (Sigma, #D8417) to visualize cell nuclei. Immunofluorescence images were acquired using a Leica fluorescence microscope under identical imaging conditions for all samples.

Quantification was performed at the single-cell level using the image-analysis software Saiviewer (Servicebio; Wuhan, China). Nuclei were identified based on the DAPI channel, followed by automated cell segmentation and CD8α-positivity assignment. Results are presented as the percentage of CD8α⁺ cells among total DAPI⁺ nucleated cells per section. All automated counts were manually reviewed before data aggregation and statistical analysis.

Dual-luciferase reporter assay

Macrophages were co-transfected with NF-κB reporter (in pGL4) and Renilla control (pRL-CMV) plasmids using jetOPTIMUS (Polyplus, #117–15). Firefly and Renilla luciferase activities were measured with the Dual-Luciferase Reporter Assay System (Promega, #E1910) according to the manufacturer’s instructions.

Enzyme-linked immunosorbent assay (ELISA)

Pf4 levels in culture supernatants were quantified using Mouse PF4 ELISA Kit (Elabscience, #E-EL-M3080) according to the manufacturer’s instructions.

Western blotting

Proteins were extracted with RIPA buffer (Thermo Fisher, #89,900) plus inhibitors (Roche, #04693159001). Equal amounts of protein were separated by SDS-PAGE, transferred to PVDF membranes (Merck, #IPVH00010), and probed with antibodies against NF-κB p65 (Cell Signaling Technology, #8242) and β-actin (Proteintech, #60,008–1-Ig).

Cell proliferation assay

Cell proliferation was assessed using CCK-8 kit (Dojindo, #CK04) according to the manufacturer’s instructions.

Statistical analysis

GraphPad Prism (v9.0) and R (v4.1.3) were used. Group differences were analyzed by Student’s t-test, ANOVA, or Wilcoxon test as appropriate. Kaplan–Meier survival curves were generated using KM-Plotter (https://kmplot.com) [42]. P < 0.05 was considered statistically significant.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Q. Z., Z. W. and B. M. conceived and designed the study, analyzed data, and wrote the manuscript. Q. Z. and Z. W. performed most of the experiments. Y. C., Z. Z., W. L., Z. F. and W. L. assisted in conducting some of the experiments. W.-Q. G. and B. M. provided critical guidance and supervision.

Funding

This work was supported by the National Key R&D Program from Ministry of Science and Technology of the People’s Republic of China (2023YFC3404101), National Natural Science Foundation of China (W2431055), Science and Technology Commission of Shanghai Municipality (25ZR1401313), Natural Science Foundation of Chongqing Municipality, China (CSTB2024NSCQ-KJFZMSX0044), the National 111 Project of China (B21024), Shanghai Jiading District Health Commission (2025-KY-ZD-05), and Peak Disciplines (Type IV) of Institutions of Higher Learning in Shanghai.

Data availability

All data relevant to the study are included in the article or uploaded as Supplementary Materials. The raw scRNA-seq data reported in the present paper have been deposited in the GEO database under accession no. GSE305506. All data supporting the findings of this study are available within the paper or from the corresponding authors upon reasonable request.

Declarations

Conflict of interest

The authors declare no competing interests.