Fusobacterium nucleatum-driven CX3CR1⁺ PD-L1⁺ phagocytes route to tumor tissues and reshape tumor microenvironment

ABSTRACT

The intracellular bacterium Fusobacterium nucleatum (Fn) mediates tumorigenesis and progression in colorectal cancer (CRC). However, the origin of intratumoral Fn and the role of Fn-infected immunocytes in the tumor microenvironment remain unclear. Here, we observed that Fn-infected neutrophils/macrophages (PMNs/MΦs), especially PMNs, accumulate in tumor tissues and fecal Fn abundance correlates positively with an abundance of blood PD-L1⁺ PMNs in CRC patients. Moreover, Fn accumulates in tumor tissues of tumor-bearing mice via intragingival infection and intravenous injection. Mechanistically, Fn can survive inside PMNs by reducing intracellular ROS levels and producing H₂S. Specifically, the lysozyme inhibitor Fn1792 as a novel virulence factor of Fn suppressed apoptosis of phagocytes by inducing CX3CR1 expression. Furthermore, Fn-driven CX3CR1⁺PD-L1⁺ phagocytes transfer intracellular Fn to tumor cells, which recruit PMNs/MΦs through the CXCL2/8-CXCR2 and CCL5/CCR5 axes. Consequently, CX3CR1⁺PD-L1⁺ PMNs infiltration promotes CRC metastasis and weakens the efficacy of immunotherapy. Treatment with the doxycycline eradicated intracellular Fn, thereby reducing the CX3CR1⁺PD-L1⁺ PMNs populations and slowing Fn-promoted tumor growth and metastasis in mice. These results suggest phagocytes as Fn-presenting cells use mutualistic strategies to home to tumor tissues and induce immunosuppression, and treatment with ROS-enhanced antibiotics can inhibit Fn-positive tumor progression.

GRAPHICAL ABSTRACT

Introduction

The tumor-specific microbiome, an emerging component of the complex tumor ecosystem, has been examined in a variety of cancer types, including gastrointestinal and nongastrointestinal tumors.¹ The microbiome is predominantly located in the intracellular cytoplasm of both tumor and immune cells and is considered to have multifactorial functions in the tumor, especially in controlling the local immune system.^2,3 Recently, emerging evidence has indicated that bacteria can translocate to tumor tissue⁴ and that intratumoral microbes, including bacteria and fungi, contribute to carcinogenesis and therapeutic responses.^5,6 However, the mechanism that drives the movement of these microbes into tumor tissues and whether they are retained in the tumor tissues remains unclear.

Fusobacterium nucleatum (Fn) is a common oral symbiotic bacterium. In previous studies, Fn was regarded as a periodontal pathogen that could lead to systemic infection.⁷ In recent years, numerous studies have confirmed that an overload of Fn in the intestinal tract is associated with tumor progression.⁸ Studies have also demonstrated that intratumoral Fn functions as an oncogenic bacteria implicated in the development of various tumor types, such as oral, esophageal, gastric, pancreatic, and colon.⁹ It is striking that as a prevalent oncogenic bacterium that can reshape the tumor immune microenvironment by virulence factor secretion and inducing host secretion of proinflammatory chemokines,¹⁰ Fn combats with immune cells, especially phagocytes, to counter engulfment and digestion, and how Fn localizes to tumor tissues remain largely unknown.

Phagocytes, polymorphonuclear neutrophils (PMNs), and macrophages (MΦs) compose the frontline of defense against bacterial infection. Curiously, our previous investigation found that although Fn is an obligate anaerobic bacterium, Fn can survive and replicate in THP-1 MΦs, so we have defined Fn as a facultative intracellular bacterium.¹¹ Recently, we confirmed that the virulence factor Fn-Dps (Fn1079) can facilitate the survival of Fn inside MΦs by upregulating the expression of the chemokine CCL2/7.¹² These results suggested that Fn can defend itself against reactive oxygen species (ROS) and lysozyme, which are fundamental tools for MΦs to eliminate invasive pathogens. A recent study also showed that Fn can be disseminated in oral cavity infections by PMNs.¹³ It is still unknown how Fn escapes from PMNs, which produce more ROS than MΦs after activation, and what role Fn-infected PMNs play in the tumor microenvironment (TME).

Recently, we found that Fn replicates inside tumor cells¹⁴ and that Fn-infected tumor cells escape T-cell attacks by upregulating the expression of PD-L1, thus protecting both host tumor cells and intracellular Fn.¹⁵ Although it is known that phagocytes can be co-opted in the spread of intracellular bacteria, whether Fn hijacks phagocytes to route its spread to tumor tissues and thereby acts as tumor-type-specific intracellular bacteria remains unknown. Additionally, when Fn-infected phagocytes traffic into tumors as tumor-associated phagocytes, their exact contributions to tumor pathology remain largely unknown. In particular, whether and how Fn infection reshapes tumor-associated neutrophils (TANs), which have diverse and plastic roles in tumor progression and therapy,¹⁶ remains largely unclear.

In the present study, to explore the origin and role of intratumoral Fn, we investigated the relationship between Fn and the major types of phagocytes and described the roles and mechanisms of Fn-loaded phagocytes in influencing tumor progression and immunotherapy in CRC. Moreover, we evaluated the role of ROS-based antibiotics in eradicating Fn for CRC therapy.

Results

Fn accumulates in phagocytes in CRC tissues, and overabundance of Fn correlates with the prevalence of PD-L1⁺ neutrophils in CRC patients

First, we examined Fn abundance in the DIANA database and found that Fn-DNA levels were significantly higher in tumor tissues of CRC patients (n = 22) compared to healthy subjects (HS, n = 432) (p < 0.0001) (Figure 1a). To further verify this result, another database-GutMEGA was exploited to analyze the Fn abundance and consistent results were obtained (Table S1). In addition, we also collected paired paraffin-embedded tumors and adjacent normal tissues from patients with CRC (n = 10) to detect the Fn. We defined the high abundance of Fn-DNA if ΔCt (the cycle threshold) < 10 (ΔCt = Ct(Fn)-Ct(18S) as Fn^high sample. We found that Fn-DNA levels were significantly higher in tumor tissues (p < 0.05) and strongly correlated with Fn levels in adjacent normal tissues (r = 0.902, p < 0.001) (Fig S1A-B; Table S2). Notably, Fn showed very intense staining in tumor tissues (including the nest and stroma) and adjacent normal tissues, and Fn-positive cells were mainly located in the cytoplasm in CRC tissues (Figure 1b). We then assessed the number of PMNs (CD66b⁺) and MΦs (CD68⁺) in Fn-positive (Fn⁺) CRC tissues. Interestingly, Fn^high tumor tissues showed significantly higher PMNs and MΦs than Fn^low tumor tissues (Figure 1c). Also, IF assay demonstrated colocalization of CD68-positive (green) MΦs with Fn (red) (Figure 1d). Moreover, immunohistochemistry (IHC) assays showed more CD66b⁺ cells in Fn^high tumor tissues than in Fn^low tumor tissues (p < 0.001) (Figure 1e–f). Meanwhile, most of the red fluorescence (Fn) was colocalized with green fluorescence (PMNs) (Figure 1f).

Figure 1:

Fn accumulates in phagocytes in CRC tissues, and fecal Fn abundance correlates positively with the abundance of blood PD-L1⁺ neutrophils in CRC patients.(a) DIANA database analysis of Fn abundance in tumor tissues from patients with CRC. (b) IHC staining of Fn in tumor and adjacent normal tissues from CRC patients. (c) if staining of CD68 (green) and CD66b (red) in tumor tissues from CRC patients. (d) IF staining of CD68 (green) and Fn (red) in tumor and adjacent normal tissues from CRC patients. (e) IHC staining of Fn and CD66b in tumor tissues from CRC patients. The relative positive ratio of CD66b is shown in the bottom panel. (f) IF staining of CD66b (green) and Fn (red) in tumor and adjacent normal tissues from CRC patients. Right panel: quantitative analysis of relative positive ratio of CD66b+Fn+ cells. (g) TCGA database analysis of differential expression of PD-L1 in tumor and adjacent normal tissues from CRC patients. (h–j) IHC staining of PD-L1 (h), or PD-L1 and CD66b (i), or if staining of CD66b (green) and PD-L1 (red) (j) in tumor and adjacent normal tissues from CRC patients. The samples of H and J were divided into two groups: Fn^low and Fn^high. (k) qPCR analysis of Fn relative abundance in fecal samples from CRC patients (cohort 2 = 31). (l) Flow cytometry analysis and quantification of PD-L1+ PMNs (%) in CRC patient blood. (m) Spearman’s correlation analysis between Fn abundance and the percentage of PD-L1+ PMNs (PD-L1+ PMNs (%). (n) CRC patients were grouped by the relative abundance of Fn, and PD-L1+ PMNs (%) between high abundance and low abundance of Fn were compared. Data are presented as the mean ± SEM, p values were determined by two-sided unpaired t-test (c–f, h, j and k), and Mann‒Whitney test (a, l, n), *p < 0.05, **p < 0.01, ***p < 0.001 for groups connected by horizontal lines. The samples in B-J from cohort 1 and K-N from cohort 2. Scale bars: 100 μm.

We previously found that intracellular Fn can upregulate PD-L1 expression in tumor cells.¹⁵ Then, we examined PD-L1 expression in TCGA datasets and found it significantly higher in CRC tumor tissues (n = 471) compared to normal tissues (n = 349) (p < 0.0001) (Figure 1g). Next, we measured PD-L1 expression in matched CRC and normal tissues. Notably, PD-L1 protein levels were higher in Fn^high CRC and in Fn^high normal tissues than in Fn^low tissues as indicated by IHC assay (Figure 1h). Strikingly, PMNs infiltration and PD-L1 expression were higher in Fn^high tumor tissues than in Fn^low tumor tissues and in normal tissues (Figure 1i-j). In paired peripheral blood and fecal samples of patients with CRC (n = 31), the Fn abundance and percentage of PD-L1⁺ PMNs were significantly higher in tissues with distant metastasis and late stage than in tissues without distant metastasis or with early-stage (Figure 1k-l; Figure S1G; Table S3). Moreover, there was a significant positive correlation between Fn abundance and the percentage of PD-L1⁺ PMNs in patients with CRC (r = 0.57; p < 0.001, Figure 1m). Consistently, the number of peripheral blood PD-L1⁺ PMNs was higher in Fn^high than in Fn^low CRC patients (p < 0.01; Figure 1n). Furthermore, we collected 21 cases as validation cohort (cohort 3), the results were essentially similar to those above (Figure S1C-G; Table S4). Taken together, these results suggest that high Fn abundance correlates with the prevalence of PD-L1⁺ neutrophils in CRC patients.

Intracellular survival of Fn in peripheral blood phagocytes by reducing ROS accumulation and producing H₂S

We performed 16S sequencing of fresh blood samples from three patients with CRC and 3 hS from cohort 2. Interestingly, the bacterial composition of the whole blood of patients with CRC included the beneficial microbes Lactobacillus and Bifidobacterium (Figure S2A-D). Notably, the genus Fusobacterium showed the highest enrichment in peripheral blood leukocytes (PBLs) compared to other components, such as erythrocytes and plasma (Figure 2a). Then, we injected Fn via the caudal vein and isolated PMNs and MΦs/monocytes (MCs) at 8 h post-infection (Figure 2b). Mouse PMNs and MCs were lysed and cultured in blood plates under anaerobic conditions, then clones were confirmed as Fn by Gram staining and 16S rRNA sequencing (Figure S2E; Addition file 1). No any Fn-like bacteria were isolated from PMNs/MCs of un-infected mice. In addition, PMNs and MCs were stained with anti-Ly6G and anti-CD14, respectively, and IF assays showed that Fn (red) was present inside Ly6G⁺ and CD14⁺ cells (Figure 2c).

Figure 2:

Fn survives inside peripheral blood phagocytes by driving ROS efflux and producing H₂S. (a) analysis of 16S sequencing in leukocytes, erythrocytes and serum from the peripheral blood of CRC patients (n = 3). (b) Schematic drawing of the model of Fn infection in vivo. (c) IF staining of Fn (red) and Ly6G (green), Fn (red) and CD14 (green) in leukocytes from C57BL/6 mice after Fn infection. Scale bars: 5 μm. (d and i) the cytotoxicity of infected PMNs, MCs (d) and PMs (i) treated with Fn at an MOI of 10:1 for 3, 6, and 12 h was measured by LDH assay. (e and j) Representative images of infected PMNs, MCs (e) and PMs (j) infected with Fn (MOI 10:1) for 12 h. Scale bars: 50 μm. (f and k) the apoptotic rates of Fn (MOI 10:1, 12 h)-infected PMNs, MCs (f) and PMs (k) were measured by flow cytometry. Right panel: statistics of the apoptosis rate. (g and l) numbers of viable Fn in PMNs, MCs (g) and PMs (l) were enumerated on blood agar plates. (h and m) three-dimensional view of α-tubulin and Fn in PMNs, MCs (h) and PMs (m) following infection with Fn (MOI 10:1) for 12 h. Scale bars: 25 μm, zoom in: 5 μm. (n) Flow cytometry analyses of total intracellular ROS levels. Cells were infected with Fn or E. coli strain DH5a (ec) at an MOI of 10:1. (o–q) flow cytometry analysis of the total intracellular ROS (o), superoxide anion (p) or hydroxyl radical (q). The right panel shows the statistics of the positive rate and cells were infected with Fn or ec at an MOI of 10:1 for 12 h. (r) Extracellular ROS in the cell culture supernatant of PMNs was measured by the fluorescence probe DCFH. (s) Intracellular H2S in PMNs was measured by the fluorescent probe WSP-5. PMNs were treated with Fn (MOI 10:1) and 50 μM sodium hydrogen sulfide (NaSH) for 12 h. (t) RT‒qPCR analysis of the mRNA expression of N1-type and N2-type markers in Fn (MOI 10:1, 12 h)-infected PMNs. (u) ELISA analysis of the expression of VEGFA and MMP9 in Fn (MOI 10:1, 12 h)-infected PMNs. PBS treatment was used as control (con). Data are presented as the mean ± SEM, p values were determined by one-way ANOVA (d, i, l, and o-s), two-way ANOVA (n), two-sided unpaired t test (f, k, t and u) and Kruskal-Wallis test (g and l). ns: no significant difference, *p < .05, **p < .01, ***p < .001 for groups connected by horizontal lines or versus con.

To further investigate the interaction between Fn and phagocytes, human PMNs and MCs were isolated and then infected with Fn (MOI 10:1). The two types of phagocytes showed no or only a slight reduction in cell viability at 12 h post-infection by LDH assay (Figure 2d). Similar results were also observed in DMSO-differentiated dHL60 cells and RAW264.7 monocytes (Figure S2F-G). In addition, there was no obvious difference in the apoptosis rate of these phagocytes based on whether they were infected with Fn or not (Figure 2e-f; Figure S2H-I). Moreover, the amount of intracellular Fn showed no substantial changes for the first 3 to 12 h based on manual counting of bacterial colonies (Figure 2g; Figure S2J). The intracellular localization of Fn in these phagocytes was observed by confocal assays (Figure 2h; Figure S2K-L). Consistently, we isolated mouse peritoneal macrophages (PMs), and similar results were observed in Fn-infected PMs (Figure 2i-m). Additionally, mouse dendritic cells DC2.4 exhibited phagocytic activity and sustained survival of intracellular Fn for at least 6 h (Figure S2M-O).

Given that PMNs are much more numerous than MCs in human blood, we detected the production of intracellular ROS in PMNs by flow cytometry. Using the extracellular Gram-negative bacteria E. coli strain DH5α (Ec) as a control, the levels of intracellular ROS, superoxide anion (O2•−) and hydroxyl radical (OH•) production were significantly lower (Figure 2n-q), while the release of extracellular ROS was higher in the Fn-treated PMNs than in the Ec-treated PMNs (Figure 2r). Moreover, Fn is an H₂S-producing species, and H₂S can quench ROS.¹⁷ We verified that the intracellular level of H₂S was increased in PMNs after Fn treatment (Figure 2s). Next, we confirmed that Fn infection induces an increased release of N2-associated proteins (VEGFA and MMP9) and a decreased release of N1-associated proteins (CCL2) in PMNs (Figure 2t-u).¹⁸ The results revealed that intracellular Fn resistance to killing of PMNs by reducing ROS accumulation and producing H₂S, induces N2-type polarization of PMNs.

Fn-infected phagocytes migrate into CRC tissues by the CXCL2/8/CXCR2 and CCL5/CCR5 axes

To observe how Fn enters tumor tissues, the DID-labeled Fn (DID-Fn) was injected intravenously into MC38 tumor-bearing mice. After 7 days of infection, still fluorescence signals retained at the tumor site of Fn-infected mice (Figure 3a). Moreover, the fluorescence accumulated in tumor tissues and in organs which host high proportion of resident phagocytes, including the liver, lung and spleen (Figure 3b). Similar results were observed in intragingival injection group (Figure 3a-b).

Figure 3:

Fn promotes the infiltration of phagocytes into CRC tissues via the CXCL2/8/CXCR2 and CCL5/CCR5 axes. (a) schematic representation of in vivo fluorescence imaging of tumor-bearing mice after intravenous injection and intragingival injection of did-labeled PBS, Fn and K-Fn. (b) Fluorescence imaging of isolated mouse organs. (c) Schematic drawing of experiments with cell-to-cell transmission. (d) Numbers of viable Fn in HCT116 and RKO cells were enumerated on blood agar plate. (e) IF staining of α-tubulin (green) and Fn (red) in HCT116 and RKO cells (f) TEM analysis of Fn inside HCT116 cells. Scale bars: 2 μm. (g and h) RT–qPCR and ELISA analysis of CXCL2 (g) or CXCL8 (h) expression in HCT116 and RKO cells. Scale bars: 10 μm. (i) Western blot analysis of CXCR2 expression in Fn (MOI 10:1, 12 h)-infected PMNs. (j–k) schematic drawing of the transwell assay (j), representative images and statistical analysis (k) (n = 5). (l) RT‒qPCR and ELISA analysis of CCL5 expression in HCT116 and RKO cells. (m) Western blot analysis of CCR5 expression in Fn-infected MCs. (n–o) schematic drawing of the transwell chemotaxis assay (N), representative images and statistical analysis (o) (n = 5). PBS treatment served as control (con). Data are presented as the mean ± SEM, p values were determined by one-way ANOVA (d, k and o), and two-sided unpaired t-test (g, h and l). ns: no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 for groups versus con.

Next, we used a reductionist in vitro coculture approach to understand the source of Fn in tumor tissues. We constructed a coculture model using Fn-infected PMNs/MCs in the upper chamber and uninfected tumor cells in the lower chamber of a Transwell plate (Figure 3c). The data reported in Figure 3d-e demonstrate that Fn exhibited intracellular proliferation inside both RKO and HCT116 cells. Fn was also found to localize in the cytoplasm in CRC cells (Figure 3e). Transmission electron microscope analysis revealed the presence of Fn inside tumor cells (Figure 3f). This finding suggests that PMNs can carry Fn into the TME for delivery to tumor cells.

Canonically, the chemokines CXCL2 and CXCL8, which induce PMNs chemotaxis through CXCR2 receptors, are important for PMNs recruitment in many cancer types.^19,20 GEPIA analysis revealed that CXCL2/8 expression significantly increased in CRC tumor tissues (T) compared with normal tissues (N). (Figure S3A). RT‒qPCR and ELISA analysis indicated that CXCL2/8 expression levels were significantly increased after Fn infection in five CRC cells (HCT116, RKO, SW480, CT26 and MC38) (Figure 3g-h; Figure S3B). CXCR2 protein levels showed no difference in Fn infection groups compared with control groups in PMNs and dHL60 cells (Figure 3i; Figure S3C).

To determine the effect of Fn-infected CRC cells on phagocytes, we performed coculture experiments with PMNs in the top chamber and CRC cells in the bottom chamber of Transwell plates (Figure 3j). The results showed that Fn-infected HCT116 and RKO cells significantly prompted the migration of PMNs and dHL60 cells compared to uninfected CRC cells (Figure 3k; Figure S3D). Moreover, treatment with a CXCR2 neutralizing antibody (CXCR2 nAb) notably diminished Fn-induced PMNs migration (Figure 3l; Figure S3D).

Additionally, CCL5 expression was significantly higher in CRC tissues than in normal tissues in GEPIA datasets (Figure S3E). RT‒qPCR and ELISA analysis indicated that CCL5 levels were significantly elevated after Fn infection in CRC cells (Figure 3m). Consistently, the chemotaxis role of CCL5 is in keeping with those seen in CCR5-positive MCs (Figure 3m-o). Collectively, these results suggest that phagocytes can carry and transfer Fn to CRC tissues and that infected tumor cells increase CXCL2/8 and CCL5 expression and actively induces the recruitment of neutrophilic/monocytic through the CXCL2/8-CXCR2 and CCL5-CCR5 axes.

The lysozyme inhibitor Fn1792 suppressed apoptosis of host phagocytes by inducing CX3CR1 expression

In our previous study, Fn1792 was found to be a highly abundant soluble protein under stressful conditions.¹² Structure analysis and functional prediction showed that Fn1792 belongs to the MliC family (Figure S4A-D). We constructed a pET28a-Fn1792 plasmid to prepare recombinant protein, and SDS‒PAGE analysis showed that purified Fn1792 exists in dimeric forms (Figure 4a; Figure S4E). Then, the Fn1792 dimer interacting partner proteins were pulled down from the mouse leukocyte lysates supernatant and validated as mouse lysozyme by nano-LC‒MS/MS analysis (Figure S4F-G). Next, Fn was treated with egg lysozyme and/or Fn1792 for 72 h. Fn1792 significantly suppressed the growth inhibition of Fn by lysozyme (Figure 4b). Similar roles were also observed in the other four lysozyme-treated gram-positive bacteria (Figure S4H). These results confirmed that the Fn1792 dimer functions as a lysozyme inhibitor.

Figure 4:

The lysozyme inhibitor Fn1792 suppressed apoptosis of host phagocytes by inducing CX3CR1 expression. (a) pull-down assay of the interaction of Fn1792 with mouse leukocyte lysate. (b) Fn growth curves were determined in the presence of lysozyme (20.0 μg/μl) and Fn1792 (0.1 μg/μl). (c and l) the cytotoxicity of PMNs, MCs (c) and PMs (l) treated with Fn1792 at different concentrations (0.01, 0.05 and 0.1 μg/ml) for 12 h was measured by LDH assay. (d and m) Representative images of PMNs, MCs (d) and PMs (m) treated with Fn1792 (0.01 μg/ml) for 12 h. Scale bars: 50 μm. (e and n) the apoptotic rates of Fn1792-treated PMNs, MCs (e) and PMs (n) were measured by flow cytometry. Right panel: statistics of the apoptosis rate. (f) Heatmap of differentially upregulated genes in mouse leukocytes and Fn1792 treated leukocytes. (g and o) RT–qPCR analysis of CX3CR1 mRNA expression in Fn1792-treated-PMNs, MCs (g) and PMs (o). (h and p) Western blot analysis of CX3CR1 protein expression in PMNs, MCs (h) and PMs (p). Cells were treated with Fn1792 (0.01 μg/ml) or Fn (MOI 10:1) for 12 h. (i) Flow cytometry analyses of the apoptosis rates of dHL60 and RAW264.7 monocytes after transfection with siNC or siCX3CR1 or treatment with Fn1792 (0.01 μg/ml) for 12 h. Right panel: the statistics of apoptosis rate. (j) Numbers of viable Fn were enumerated in dHL60 and RAW264.7 monocytes by the gradient dilution coating method. (k) IF staining of Fn (red) and α-Tubulin (green) in Fn-infected RAW264.7 monocytes after transfection with siNC or siCX3CR1 or treatment with Fn1792 (0.01 μg/ml) for 12 h. (q) Schematic drawing of the Transwell assay. (r) Representative images and statistical analysis (right panel) of (r) are shown (n = 5). (s) RT‒qPCR analysis of the mRNA expression of N1-type and N2-type markers in Fn1792 (0.01 μg/ml, 12 h) treated PMNs. (t) A flow diagram of MC38 tumor-bearing C57BL/6 mice and administration of Fn1792 and JMS-17-2. (u) Representative images of the harvested tumors. (v) Mean tumor volumes/ weights ± SEM. (w) Representative images of H&E staining of lung and quantitative analysis. (x-y) flow cytometry analysis and quantification of CX3CR1+ PMNs in Ly6G+ PMNs (%) (x) or CX3CR1+ mΦs/F4/80+ mΦs (%) (y)in tumors. PBS treatment was used as control (con). Data are presented as the mean ± SEM, p values were determined by one-way ANOVA (C, I, J, L, R and V-y), two-way ANOVA (b and the right panel of v), and two-sided unpaired t test (e, g, n, o and s). ns: no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 for groups connected by horizontal lines or versus con.

Next, we investigated the cytotoxic effects of Fn1792 on phagocytes. As shown in Figure 4c-d, Fn1792 treatment did not cause significant cytotoxic or morphological changes in human PMNs and MCs. Similar results were observed in Fn1792-treated dHL60 and RAW264.7 monocytes (Figure S4I-J). Surprisingly, flow cytometry analysis indicated that Fn1792 treatment dramatically suppressed spontaneous apoptosis of PMNs (52.97% vs. 32.19%, p < 0.01) and MCs (10.13% vs. 5.71%, p < 0.05) (Figure 4e). Fn1792 treatment also suppressed spontaneous apoptosis of dHL60 and H₂O₂-induced apoptosis of RAW264.7 monocytes (Figure S4K-L).

Then, we performed transcriptome profiling experiments in Fn1792-treated mouse leukocytes to explore the mechanism by which Fn1792 attenuates apoptosis. A total of 16 genes were found to be significantly upregulated in leukocytes after Fn1792 treatment (logFC >1.5; p < 0.05) (Figure 4f). Among these genes, excluding unknown function genes, CX3CR1 was the most significantly upregulated gene, with an average of 14.8-fold higher expression than in control leukocytes (Figure 4f). Then, upregulated CX3CR1 expression was verified in both Fn1792-treated and Fn-infected human PMNs and MCs by RT‒qPCR and Western blot analysis (Figure 4g-h; Figure S4M-N). Also, the mRNA expressions of the other upregulated genes were detected by RT-qPCR analysis, and these upregulation extents were not as obvious as the CX3CR1 genes (Figure S4O). To further understand the functions of CX3CR1 in phagocyte apoptosis, we silenced CX3CR1 in dHL60 and RAW264.7 monocytes using specific siRNAs (Figure S4P). Compared to the control cells (siNC), knockdown of CX3CR1 (siCX3CR1) promoted apoptosis of dHL60 and RAW264.7 monocytes, but Fn1792 treatment did not suppress the apoptosis of siCX3CR1 cells (Figure 4i). In addition, the counts of intracellular bacteria and IF analysis showed that dramatic decreases in CX3CR1-knockdown dHL60 and RAW264.7 monocytes compared to those in control cells (Figure 4j-k).

We also observed that Fn1792 treatment suppressed spontaneous apoptosis of PMNs (10.16% vs. 4.10%, p < 0.05) and induced the expression of CX3CR1 (Figure 4l-p). CX3CR1 is required for the adhesion and migration of leukocytes.²¹ To determine whether Fn-induced CX3CR1 expression is critical for phagocyte transmigration across the endothelium, and given that PMNs are the most abundant leukocytes in circulation, PMNs were placed onto HUVEC endothelial monolayers in the upper compartment of a Transwell plate to establish PMN/HUVEC cocultures (Figure 4q). Our results showed that the CX3CR1 inhibitor JMS-17-2 significantly inhibited both Fn- and Fn1792-stimulated PMNs migration to endothelial cells (Figure 4r; Figure S4Q). Moreover, we found that Fn1792 treatment increased the release of N2-associated proteins VEGFA and MMP9 and decreased the release of N1-associated proteins CCL2 in PMNs (Figure 4s).

Furthermore, we injected Fn1792 into MC38 tumor-bearing mice by tail vein to evaluate Fn1792 activity in vivo (Figure 4t). Fn1792 treatment significantly increased the tumor size and the number of lung metastatic foci compared with the control group (p < 0.001) (Figure 4u-w). The CX3CR1 inhibitor JMS-17-2 attenuated the stimulating effects of Fn1792 on tumor progression (Figure 4u-w). Moreover, the Fn1792-injected group presented a significantly increased infiltration of CX3CR1⁺ PMNs and CX3CR1⁺ MΦs in tumor tissues by flow cytometry analysis (Figure 4x-y; Figure S4R-S).

Taken together, these findings suggest that Fn1792 can resist lysosomal killing and induce CX3CR1 expression in phagocytes, which inhibits host cell apoptosis, promotes PMNs migration, and promotes tumor progression in vivo.

Intracellular Fn induces PD-L1 expression in phagocytes, and PD-L1⁺ neutrophils exhibit immunosuppressive functions

Intriguingly, Fn infection increased PD-L1 protein expression in PMNs and dHL60 cells in a time-dependent manner (Figure 5a-c; S5A-C). Notably, only live Fn treatment induced the expression of PD-L1, while treatment with K-Fn or Ec did not affect PD-L1 expression (Figure 5d-e; Figure S5D). Moreover, intracellular Fn and upregulated PD-L1 expression were observed in Fn-infected PMNs and dHL60 cells by IF staining (Figure 5f; Figure S5E). Similar results were obtained in Fn-infected PMs and RAW264.7 cells (Figure S5F-G). Moreover, after infection with Fn from phagocytes, intracellular Fn also induced PD-L1 expression in CRC host cells (Figure S5H).

Figure 5:

Fn induces PD-L1 expression in phagocytes, and PD-L1⁺ neutrophils exhibit immunosuppressive functions. (a-c) flow cytometry analysis (a, b) and western blot analysis (c) of PD-L1 expression in PMNs. (d and e) flow cytometry analysis (d) and western blot analysis (e) of PD-L1 protein expression in PMNs. (f) if staining of Fn (red) and PD-L1 (green) in Fn (MOI 10:1, 12 h)-infected PMNs. Right panel: quantification of PD-L1 expression. Scale bars: 25 μm. (g) Western blot analysis of protein expression in PMNs infected with Fn (MOI 10:1) for 15, 30, 60, and 120 min. (h and i) western blot analysis of protein expression in PMNs. Cells were pretreated with 30 nM TPCA-1 (h) or 100 μM NSC74859 (i) and infected with Fn for 1 h. (j) Schematic diagram showing that CD3+ T-cells were cocultured with human peripheral blood PMNs (1:1) or with an anti-PD-L1 antibody (20 μg/ml) for 48 h. (k and l) Representative flow cytometry and statistical analysis of T-cell- proliferation (k) and iFn-γ production (l) are shown (n = 3). (m) Schematic drawing of T-cell/crc cell coculture system. (n-p) flow cytometry assay of apoptosis rates (n) and the statistical analysis (o) or CCK-8 assay of the cell viability rate of HCT116 and RKO cells (p). (q) Numbers of viable Fn were enumerated in PMNs by the gradient dilution coating method. PBS treatment was set as control (con). Data are presented as the mean ± SEM, p values were determined by one-way ANOVA (a, b, d, k, l, and o-q), and two-sided unpaired t-test (f). ns: no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 for groups connected by horizontal lines or versus con.

We next investigated the underlying mechanisms of PD-L1 expression in Fn-infected phagocytes. Western blot analysis identified that the phosphorylation levels of STAT3 and NF-κB subunits p65 and p50 were up-regulated, while NFκB inhibitor IκB-α was down-regulated by Fn treatment in PMNs and dHL60 cells (Figure 5g; Figure S5I), suggesting activation of the NF-κB/STAT3 pathway following Fn infection in PMNs. Using the NF-κB/STAT3 signaling inhibitor TPCA-1 or NSC74859, we observed that Fn failed to upregulate the expression of PD-L1 in PMNs and dHL60 cells, highlighting the importance of the NF-κB pathway in the upregulation of PD-L1 during Fn infection in PMNs (Figure 5h-i; Figure S5J-K). These results suggest that intracellular Fn survival induces the expression of PD-L1 through the activation of the NF-κB/STAT3 signaling pathway in PMNs.

To understand the role of PD-L1⁺ PMNs, purified CD3⁺ T lymphocytes from human peripheral blood were cocultured with Fn-infected (Fn⁺) and uninfected PMNs (Figure 5j; Figure S5L-M). Interestingly, Fn⁺ PMNs significantly suppressed T-cell proliferation and IFN-γ production, and this inhibition was reversed by an anti-PD-L1 antibody (Figure 5k-l). Moreover, purified CD8⁺ T-cells were cocultured with Fn⁺ PMNs and CRC cells in vitro to observe crosstalk within the TME. The effect of PMN tumor modulation on cytotoxic T-cell function was also determined. The involvement of Fn⁺ PMNs in the system significantly suppressed the cytotoxic T-cell- lytic function of CRC tumor cells as indicated by cell apoptosis and viability analysis (Figure 5m-p). Then, we examined Fn survival in PMNs by bacterial counting in BHI plates. Figure 5q shows that, compared to control cells, the amount of Fn decreased significantly in cells treated with the anti-PD-L1 antibody. These results indicate that PMNs are activated by intracellular Fn and acquire the ability to suppress T-cell function through the expression of PD-L1, which protects both host PMNs and intracellular Fn bacteria.

Fn-infected CX3CR1⁺PD-L1⁺ neutrophils diminish the efficacy of αPD-L1 therapy

To investigate the effect of immunotherapy under Fn-infected condition, MC38 tumor-bearing mice were pre-infected with Fn by tail vein injection and then underwent mouse αPD-L1 treatment (Figure 6a). The Fn-infected mice showed increased tumor size and lung metastasis compared to noninfected group. Moreover, Fn infection diminished therapeutic effects in mice undergoing αPD-L1 treatment (Figure 6b-e). In addition, the infiltration of CX3CR1⁺PD-L1⁺ PMNs/MΦs in tumors of Fn-infected mice with αPD-L1 treatment were increased, while CD4⁺ T-cells and CD8⁺ T-cells were decreased compared to noninfected mice (Figure 6f-h; Figure S6A-C), and there was a significantly higher abundance of intratumor Fn-DNA in the Fn-infected group with αPD-L1 treatment compared to αPD-L1 treatment alone by qPCR analysis (Figure 6i).

Figure 6:

Fn-infected CX3CR1⁺PD-L1⁺ neutrophils diminish the efficacy of αPD-L1 therapy. (a) a flow diagram of MC38 tumor-bearing C57BL/6 mice and administration of Fn and αPD-L1. (b) Representative images of the harvested tumors. (c and d) mean tumor volumes/ weights ± SEM. (e) Representative images of H&E staining of lung and quantitative analysis. (f-h) flow cytometry analysis and quantification of CX3CR1⁺ PD-L1⁺ PMNs in Ly6G+ PMNs (%) (f), CX3CR1⁺ PD-L1⁺ mΦs/F4/80+ mΦs (%) (g) or CD4+ CD3+ cells and CD8a+ CD3+ cells in CD3+ T-cells in tumors (h). (i) qPCR analysis of Fn relative abundance in tumors. (j) A flow diagram of HCT116 tumor-bearing NCG mice. (k) Representative images of the harvested tumors. (l and m) mean tumor volumes/ weights ± SEM. (n) Representative images of H&E staining of the lung tissues and quantitative analysis. (o) qPCR analysis of Fn relative abundance in tumors and lungs. (p-r) flow cytometry analysis and quantification of PD-L1+ PMNs in PMNs (%) from peripheral blood/tumor tissues (p-q) or PD-L1+ mΦs (%) in tumors (r). (s and t) FACS quantification of the number of CD3+ T-cells in tumors (s) and the percentages of CD4+ CD3+ cells and CD8a+ CD3+ cells in CD3+ T-cells from tumors (s) or spleens (t). (u) ELISA analysis of CXCL2, CXCL8 and CCL5 levels in serum. (v) Western blot analysis of PD-L1 protein expression in tumors. (W) IHC staining of Ly6G, Fn and PD-L1 in tumors. (x and y) if staining and quantitative analysis of Ly6G (green) and Fn (red) (x) or Fn (red) and PD-L1 (green) (y) in tumors. Scale bars: 100 μm. Data are presented as the mean ± SEM, p values were determined by one-way ANOVA (d-i, m, n, o-u and x-y) and two-way ANOVA (c and l). ns: no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 for groups connected by horizontal lines.

To test the suppressive effect of Fn⁺ phagocytes on T-cell immunity in vivo, we treated PMNs or Fn⁺ PMNs with human αPD-L1 or control IgG and then injected them together with autologous T-cells into NSG mice bearing HCT116-derived CRC (Figure 6j). As expected, compared to T-cell plus PMNs transfusions, mice treated with T-cells plus αPD-L1-treated PMNs showed reduced tumor size and lung metastasis, but this effect was reversed in mice treated with T-cells plus αPD-L1-treated Fn⁺ PMNs (Figure 6k-n). Moreover, qPCR assays showed that Fn abundance in the mice treated with T-cells plus Fn⁺ PMNs was higher than that in the mice treated with T-cells plus αPD-L1-treated Fn⁺ PMNs (Figure 6o). In addition, mice treated with T-cells plus αPD-L1-treated PMNs exhibited decreased numbers of PD-L1⁺ PMNs and PD-L1⁺ MΦs and increased CD4⁺ T-cell and CD8⁺ T-cell- infiltration in the peripheral blood, tumor and spleen, but this effect was not observed in mice treated with T-cells plus αPD-L1-treated Fn⁺ PMNs (Figure 6p-t; Figure S6D-K). Moreover, the serum levels of CXCL2, CXCL8 and CCL5 were determined by ELISA analysis (Figure 6u). The result was similar to that obtained via Western blot, IHC and IF analysis (Figure 6v-y). Collectively, these data indicate that Fn-infected peripheral blood PMNs can traffic into tumor tissues and promote the accumulation of CX3CR1⁺PD-L1⁺ PMNs in tumor tissues, that enhance CRC metastasis and diminish immunotherapy efficacy.

Doxycycline effectively inhibited Fn-promoted tumor progression by reduced the size of the CX3CR1⁺ PD-L1⁺ PMNs/MΦs populations

We developed a tool by testing different antibiotics to selectively eliminate tumor-resident bacteria. An in vitro intracellular antibiotic susceptibility test showed that only doxycycline (DOX) was able to almost eliminate intracellular and extracellular Fn, while metronidazole (MTZ) and moxifloxacin (MOX) only partly killed intracellular Fn. In addition, gentamicin (GM) eliminated extracellular but not intracellular Fn, while azithromycin (ATZ) could not eliminate either extracellular or intracellular Fn (Figure S7A).

Next, MC38 tumor-bearing mice were infected with Fn by intravenous injection (i.v.) and then administered different antibiotics through drinking water and gavage (i.g.) (Figure 7a). We found that DOX or MTZ treatment showed an inhibitory effect, while GM treatment caused a slight increase in tumor size, lung metastasis and intratumoral Fn abundance in mice inoculated with Fn (Figure 7b-f). This suggests that the presence of intracellular Fn but not extracellular Fn promotes CRC tumor growth.

Figure 7:

Doxycycline effectively inhibited Fn-promoted tumor growth and metastasis by reducing the size of the CX3CR1⁺ PD-L1⁺ PMNs/mΦs populations. (a) a flow diagram of MC38 tumor-bearing C57BL/6 mice. (b) Representative images of the harvested tumors. (c and d) mean tumor volumes/ weights ± SEM. (e) Representative images of H&E staining of lung and quantitative analysis. (f) qPCR analysis of Fn relative abundance in tumors and lung. (g-i) flow cytometry analysis and quantification of PD-L1+ PMNs in PMNs (%) from peripheral blood/tumor tissues (g-h) or PD-L1+ mΦs (%) in tumors (i). (j and k) FACS quantification of the number of CD3+ T-cells in tumors (j) and the percentages of CD4+ CD3+ cells and CD8a+ CD3+ cells in CD3+ T-cells from tumors (j) or spleens (k). (l) ELISA analysis of CXCL2, CXCL8 and CCL5 levels in serum. (m) Western blot analysis of PD-L1, VEGFA and MMP9 protein expression in tumors. (n) IHC staining of Ly6G, Fn, CX3CR1 and PD-L1 in tumors. (o and p) if staining and quantitative analysis of Ly6G (green) and Fn (red) (o) or Fn (red) and PD-L1 (green) (p) in tumors. Scale bars: 100 μm. Data are presented as the mean ± SEM, p values were determined by one-way ANOVA (d-l, o and p), and two-way ANOVA (C). ns: no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 for groups connected by horizontal lines.

Therefore, to determine whether immune cell accumulation might be affected by tumor-intracellular Fn, we used flow cytometry to compare the levels of PD-L1⁺ PMNs/MΦs, CD4⁺ T-cells and CD8⁺ T-cells between tumors from antibiotic-treated mice inoculated with Fn. Compared to the MTZ treatment group, DOX significantly reduced the size of the Fn-driven PD-L1⁺ PMNs and PD-L1⁺ MΦs populations in the peripheral blood and tumor regions (Figure 7g-i; Figure S7B-F). In addition, DOX treatment significantly abrogated the effect of Fn infection on reducing the CD4⁺ and CD8⁺ T-cell- populations in the tumor and spleen regions (Figure 7j-k; Figure S7G-I). Similarly, DOX treatment abrogated Fn-induced increases in the protein expression of CX3CR1, PD-L1, MMP9 and VEGFA in tumor tissues, serum levels of CXCL8, CXCL2, and CCL5, and PMNs infiltration in tumor tissues (Figure 7l-p). These results suggest that CRC tumor progression due to the presence of intracellular Fn might be mediated by CX3CR1⁺PD-L1⁺ PMNs/MΦs and that DOX is a more effective antibiotic than MTZ or MOX for eradicating intracellular Fn and inhibiting Fn-stimulated tumor progression.

Discussion

Here, we demonstrated that Fn is located in the cytoplasm, including that of tumor cells and tumor-associated phagocytes, in CRC tissues. We also observed the presence of Fn in the tumor and adjacent normal tissues. This is consistent with recent research that has found Fn presence in myeloid cells in tumors using single-cell sequencing.²² It is currently accepted that some bacteria present within the bloodstream might travel to the tumor. Given that blood MCs and PMNs can be mobilized to sites of infection or inflammation by chemotaxis, our finding of Fn inside PMNs and MΦs in tumor and paracarcinoma tissues indicates that some intratumoral bacteria may also exploit phagocytes as a port of entry to colonize the tumor.

Recently, the circulating microbiome has been identified in noninfectious disease states.²³ We detected a high abundance of traditional beneficial microbes Lactobacillus and Bifidobacterium, in the peripheral blood of patients with CRC. A study showed a higher abundance of Bifidobacterium adolescentis in the blood of patients with CRC than in that of healthy controls.²⁴ These findings suggest that gut microbial entry into the bloodstream might be prevalent and nonselective. A study showed that the blood microbiota was less diverse and that the abundance of Fn in blood was increased after surgery in patients with CRC.²⁵ Taken together with our investigation in which the leukocyte microbiota was dominated by the genus Fusobacterium, this indicates that the enrichment and translocation of Fn might be selectively controlled by leukocytes in patients with CRC. Especially, since the high serum IgG level of anti-Fn is prevalent in CRC patients,²⁶ it is possible that Fn had to escape into leukocytes to shield against the stress of humoral antibodies in peripheral blood.

To date, besides Fn, very few intracellular obligate anaerobic bacteria have been reported. Indeed, to exert bactericidal activity inside phagocytes, intracellular oxygen restriction is usually observed during infection. Interestingly, intracellular hypoxic conditions provide a favorable environment for the survival of obligate anaerobes. These observations suggest that more undiscovered obligate anaerobic intestinal bacteria might exploit this “bug” to parasitize or shuttle freely in host cells. Moreover, although PMNs produce high amounts of ROS to kill phagocytized pathogens, some bacteria, such as Helicobacter pylori, are also able to defend against oxidative stress by inducing ROS efflux to the extracellular space.²⁷ We verified that Fn-infected PMNs produce more extracellular and less intracellular ROS than noninvasive E. coli-infected PMNs, suggesting that Fn-infected phagocytes might use a similar strategy to reduce the overburden of intracellular ROS. Furthermore, Fn can produce large amounts of H₂S. H₂S acts as a bacterial defense mechanism against killing by immune cells.²⁸ It would be plausible that Fn can also utilize H₂S to manage ROS. These might be extraordinarily important mechanisms for Fn to escape excess ROS and survive inside PMNs.

In addition to ROS production, phagocytes can eliminate pathogens via the production of enzymes such as lysozyme. MliC is a periplasmic lysozyme inhibitor that can strongly and specifically inhibit C-type lysozymes.²⁹ We confirmed that the dimer of Fn1792 functions as a lysozyme inhibitor, although Fn1792 has no homology with MliC of bacteria other than those of the genus Fusobacterium. Fn1792 may help Fn settle in the intestinal mucosa by countering intestinal lysozyme. In addition, a lysozyme inhibitor of Yersinia pestis has been reported to be important for its survival against human PMNs,³⁰ implying that Fn1792 can help Fn resist lysozyme from phagosomes. These results show that MliC might be necessary for the pathogenicity of Fn, and Fn has evolved sophisticated strategies, including phagosome escape by releasing lysozyme inhibitors, to evade innate immune killing.

Astonishingly, Fn undergoes limited intracellular replication in the cytoplasm of human PMNs, even though PMNs have only scant cytoplasm with granules. Fn survival inside PMNs was also observed in a recent study,¹³ even though Fn induces PMNs extracellular traps.³¹ Interestingly, both Fn and Fn1792 induced the expression of CX3CR1 in PMNs. Usually, CX3CR1 is expressed in lymphocytes, NK cells and MCs and is involved in the adhesion and migration of leukocytes. CX3CR1 expression was also observed in CXCL8-induced PMNs.³² Importantly, some immune cells, such as blood MCs and tumor-associated macrophages, require CX3CR1 expression for survival.³³ We demonstrated that Fn-induced expression of CX3CR1 provides a favorable environment for both the host and bacteria. This is very different from infections caused by other facultatively intracellular bacteria. For example, impaired Listeria clearance and higher susceptibility to Salmonella infection were observed in CX3CR1-deficient animals.³⁴ Our results uncovered a symbiotic relationship between Fn and host phagocytes, suggesting that PMNs serve as reservoirs for Fn employing Fn-induced CX3CR1 expression.

TANs exhibit plasticity between the antitumorigenic N1 and tumor-promoting N2 phenotypes, which is determined by signals from surrounding tissues.³⁵ In CRC, accumulating evidence has shown that PMN infiltration plays an important role in tumor progression.³⁶ We identified Fn-infected PMNs as CX3CR1^hi subtypes that produce MMP-9 and VEGF and displayed the tumor-promoting N2 phenotype. These results confirm that the recruited PMNs infected by Fn can aggravate the progression of tumors.

The PD-1/PD-L1 axis mediates immune tolerance and promotes tumor progression via the inhibition of antitumor immunity. PD-L1 expression was reported to be inducible on PMNs to suppress host immunity during polymicrobial sepsis.³⁷ A multicenter study showed that PD-L1 expression is negatively correlated with the apoptosis rates of PMNs in patients with sepsis, and increased PD-L1 expression on human PMNs delays cellular apoptosis by triggering PI3K/AKT phosphorylation.³⁷ We confirmed Fn⁺ PMNs to be the PD-L1⁺ subgroup. PD-L1⁺ PMNs are considered to be a protumor because they can suppress cytotoxic T-cells and are propagated with disease progression, protecting both the host and colonizing bacteria from T-cell attacks.³⁸ PD-L1 expression was also observed in Fn-infected MΦs, and our previous study confirmed that Fn⁺ MΦs evade cell-intrinsic death to inhibit host cell apoptosis.¹¹ Interestingly, both our investigation and a recent investigation found that the presence of Fn-infected tumor cells weakened PD-L1 antibody therapy.¹⁵ These results show that Fn has formed mutualistic symbioses with phagocytes for immune escape and to facilitate tumor metastasis and further support Fn as a marker for poor treatment outcomes of immunotherapy.

Traditional antibiotic therapeutic strategies face a challenge against intracellular bacterial infection. Metronidazole is frequently used to treat anaerobic infections but does not effectively penetrate host cells to inhibit intracellular anaerobes,³⁹ which results in persistent and refractory infections. Both macrolides and quinolones are antibiotics frequently used against pathogens inside PMNs. Fortunately, Fn is macrolide-resistant due to its macrolide efflux protein.⁴⁰ Moreover, our studies show that moxifloxacin only partly eliminates intracellular Fn in human PMNs in vitro. It is known that H₂S acts by suppressing the oxidative component of antibiotic toxicity. H₂S-producing bacteria such as S. typhimurium confer resistance to fluoroquinolones by increasing H₂S levels.⁴¹ These results suggest that H₂S produced by intracellular Fn dampens moxifloxacin toxicity. Interestingly, mitochondrial targeting antibiotic doxycycline also has anti-tumor properties.⁴² Moreover, doxycycline can induce ROS production,⁴³ which may compensate for the Fn-induced ROS efflux and ensure that the infection is eradicated. Thus, in CRC treatment, it might be necessary to eliminate Fn using doxycycline which can effectively penetrate and/or accumulate inside phagocytes, strengthen intracellular ROS levels and avoid interference from H₂S.

In summary, we have demonstrated that Fn hijacks phagocytes, especially PMNs, as a “Trojan horse” to home to tumor tissues and we identified Fn-phagocyte mutualisms with the help of the novel virulence factor Fn1792. Phagocytes transfer Fn to tumor cells, and the infected tumor cells further recruit phagocytes regardless of whether they are Fn-infected or uninfected by various chemotaxis axes. We confirmed that Fn-infected PMNs are a special CX3CR1⁺PD-L1⁺ subgroup act as protumor PMNs in the TME. Our study showed that Fn, as an intracellular parasite, reprogrammed host cells, not only creating a haven for its survival but also maintaining the survival of the host cells themselves. Our study suggests that Fn can be eradicated by ROS-enhanced antibiotics doxycycline and Fn-loaded phagocytes may be explored as an important therapeutic target in CRC.

Limitations of the study

To investigate the pro-oncogenic mechanisms of Fn, we chose a commonly used strain, F. nucleatum ATCC 25,586.^44–47 While a very recent study found that the Fn subspecies, F. nucleatum subsp. animalis (Fna), predominates in the CRC niche.⁴⁸ Therefore, future research is essential to use more subspecies, especially Fna, for a fully comprehensive assessment.

Materials and methods

Clinical samples

A total of 10 CRC and paired adjacent normal tissue sections (Cohort 1), and fresh blood and fecal samples from 31 patients with CRC (Cohort 2) and 21 patients with CRC (Cohort 3) were obtained from the Sun Yat-sen University Cancer Center (SYUCC). Ethics approval was granted by the Ethics Committee of SYSUCC (No. G2023-089-01). Background information and grouping of the study cohorts are shown in Supplemental table S2–4.

Strains and growth conditions

Fusobacterium nucleatum (ATCC 25,586) was preserved in our laboratory and inoculated into brain heart infusion broth (BHI, Thermo Fisher, USA) under anaerobic conditions for 48–72 h at 37°C. Bacillus subtilis, Bacillus licheniformis, Streptococcus thermophilus, and Enterococcus faecalis were also cultured by the same methods as Fusobacterium nucleatum. Heat-killed Fn was heated at 100°C for 30 min. E. coli strain DH5α was enriched aerobically at 37°C overnight in Luria-Broth (LB) medium.

Cell preparation

Peripheral blood was collected from the author himself and was approved by the ethics committee. Human peripheral blood neutrophils (PMNs) and human peripheral blood mononuclear cells (PBMCs) were isolated by a human peripheral blood neutrophil isolation kit (Solarbio, Beijing, China), and then monocytes (MCs) were isolated from PBMCs by discarding the supernatant after adhering to the walls for 1 h. The procedure was performed according to the protocol provided by the manufacturer.

Human leukemia cell line HL60 cell differentiation was triggered by adding 1.35% DMSO for 4 days (Sigma, USA). The differentiation from undifferentiated (uHL60) to differentiated neutrophil-like (dHL60) cells was checked by flow cytometry with a CD11b-FITC antibody (5 μl/test, Elabscience, China).

Peritoneal macrophages (PMs) were isolated by peritoneal cavity lavage from mouse peritoneal cells, and erythrocytes were lysed. After 6 h, adherent cells were considered PMs. Mice were injected i.p. with 1.5 ml of 4% thioglycolate medium (Thermo Fisher, USA) three days in advance.

Cell culture

Human CRC cell lines SW480, RKO and HCT116; mouse CRC cell lines CT26 and MC38; mouse dendritic cell lines DC2.4, mouse monocyte cell lines RAW264.7; uHL60 and dHL60 cells; and isolated human PMNs, MCs and PMs were maintained in our laboratory and grown in RPMI-1640 or DMEM (Gibco, USA) containing 10% fetal bovine serum under a humidified 5% CO₂ atmosphere at 37°C.

Animal experiments

C57BL/6 mice and NOD. Cg-Prkdcscid IL2rgtm1Wjl/SzJ (NSG) mice (approximately six weeks old) were obtained from the Model Animal Center of Nanjing University (Nanjing, China) and were housed under pathogen-free barrier conditions in the SYSU Animal Center (Guangzhou, China). Ethical approval for the mice used in this project was obtained from the SYSU Animal Experimentation Ethics Committee [SYSU-IACUC-2022-001515].

To establish subcutaneous tumors, 1 × 10⁶ MC38 cells were implanted into C57BL/6 mice. To analyze Fn1792 activity, mice were randomly divided into three groups (5 mice/group): Con, Fn1792, and Fn1792+JMS-17-2. After the tumor size reached 100 mm³, Fn1792 (3 μg/mouse) or PBS were i.v. injected once every two days for three times. JMS-17-2 (10 mg/kg, MedChemExpress, USA) or solvent (10% DMSO + 90% corn oil) was i.p. injected every day for two weeks after MC38 cells inoculation.

To assess the effect of the PD-L1 antibody under Fn-infected conditions, C57BL/6 mice were randomly divided into four groups. After one week of MC38 cells inoculation, Fn (10⁷ CFU) were i.v. injected once every two days for three times. Then, mice were injected intraperitoneally once every four days for a total of three times with 100 μg anti-PD-L1 antibody (BioXcell, USA) or mouse IgG2a isotype control antibody (BioXcell, USA).

To evaluate the suppressive effect of Fn⁺ PMNs on tumor growth, NSG mice were subcutaneously implanted with 5 × 10⁶ HCT116 cells. After one week of HCT116 cells inoculation, T cells were co-cultured with autologous peripheral PMNs (T-cells+PMNs), or T-cells+Fn⁺ PMNs at a 1:1 ratio in the presence or absence of a neutralizing antibody against human PD-L1 (5 μg/mL, R&D Systems, MN) or a control IgG (5 μg/mL, R&D Systems, MN) for 24 h, and then mice were injected intraperitoneally every 3 days for a total of seven treatments.

To analyze the role of antibiotics in tumor therapy, C57BL/6 mice were randomly divided into four groups. One week after MC38 cells inoculation, Fn (10⁷ CFU) was i.v. injected once every two days three times. After a one-day interval, a group of Fn-infected mice was treated by oral administration with doxycycline (20 mg/kg, Sangon Biotech, China), metronidazole (100 mg/kg, Sangon Biotech, China) or gentamicin (20 mg/kg, Aladdin, USA). Then, the mice were treated with doxycycline (0.2 g/l, Sangon Biotech, China), metronidazole (1 g/l, Sangon Biotech, China), or gentamicin (0.5 g/l, Aladdin, USA) in sterile drinking water for two weeks.

Tumor growth was measured in three dimensions every three days with a caliper. Tumor volume was calculated using the following formula: (length×width²)/2. Mice were sacrificed after the end of administration and tissue samples were immediately collected.

In vivo tracking of DID-labeled Fn

C57BL/6 mice subcutaneously engrafted with 1 × 10⁶ MC38 cells. One week after inoculation, Fn and K-Fn were stained with DID dye (Beyotime, China) for 5 min and terminated by 2% FBS. Then, mice were infected by a single intragingival injection or tail vein injection of DID-Fn and DID-K-Fn (10⁷ CFU/mice). After 7 days of infection, mice and their viscera were fluorescently imaged by in vivo imaging systems (NightOWL II LB 983, Germany).

Protein purification

The Fn1792 gene was amplified by PCR (NdeI-F: CGGGATCCATGAGTTTATTCTTAGTAGCTTGTGGAGAAAAAAAAGAAGAAGA, XhoI-R: TGCTCGAGCTATTTAGCTTCAACAGTTACTGGTACTTCTTTAAGG), and the pET-28a (+) expression vector was constructed and transformed into E. coli BL21 (DE). E. coli carrying Fn1792 plasmids was grown in LB medium supplemented with 0.1 mm isopropyl-β-D-thiogalactopyranoside (IPTG) to express target proteins. After overnight incubation at 16°C, E. coli was collected and lysed by a high-pressure cell crusher (XM-Biotech, China). The supernatants were run through a Ni-NTA agarose resin (Sangon Biotech, China), washed with washing buffer (50 mm NaH₂PO₄ 300 mm NaCl, and 50 mm imidazole (pH 8.0)), and eluted with elution buffer (50 mm NaH₂PO₄ 300 mm NaCl, and 100 mm imidazole (pH 8.0)). The protein was concentrated by 10 kDa ultrafiltration tubes (Merck Millipore, GER). The endotoxin in Fn1792 was removed using the ToxinEraser Endotoxin Removal Kit (GenScript, China) and then quantified by a BCA protein assay kit (Beyotime, China), and an empty vector was used as a control protein.

Pull-down analysis

Mouse leukocytes were lysed with RIPA (Beyotime, China) buffer for 15 min, and the supernatants were retained. Fn1792 recombinant protein was run through a Ni-NTA agarose resin (Sangon Biotech, China) three times and washed with buffer (50 mm NaH₂PO₄ and 300 mm NaCl (pH 8.0)). Then, mouse leukocyte lysate was also run through the Ni-NTA agarose resin above three times, and washing buffer (50 mm NaH₂PO₄, 300 mm NaCl, and 50 mm imidazole (pH 8.0)) was used to wash away unbound proteins. Finally, bound proteins were eluted with buffer (50 mm NaH₂PO₄, 300 mm NaCl, and 100 mm imidazole (pH 8.0)) and examined using SDS‒PAGE.

Intracellular survival assays

Fn infection for intracellular survival and proliferation was assessed by the gradient dilution coating method. Briefly, cells were infected with live Fn at a multiplicity of infection (MOI) of 10:1 for 3, 6, and 12 h and washed with PBS, followed by incubation with gentamicin (50 μg/ml, Aladdin, USA) for 2 h to remove extracellular bacteria. Then, Fn-infected cells were lysed with 0.1% Triton X-100, and intracellular Fn was enumerated as colony-forming units (CFU) by plating on BHI plates.

Immunofluorescence (IF)

Tissue sections or cells were fixed with 4% paraformaldehyde for 30 min and permeabilized with 0.3% Triton X-100 for 15 min at room temperature (RT). Next, the samples were incubated with 10% goat blocking serum (Boster, China) for 30 min and incubated with primary antibodies, rabbit anti-Fn polyclonal antibody (homemade, 1:5,000), mouse anti-Fn polyclonal antibody (homemade, 1:3,000),^15,49 α-tubulin antibody (1:200, Rayantibody, China), CD68 (1:200, Servicebio, China), CD66b (1:100, Bioss, China), CD14 (1:500, Servicebio, China), Ly6G (1:200, Servicebio, China), PD-L1 (1:100, Proteintech, China) and CX3CR1 (1:100, Affinity, China) overnight at 4°C. Fn-IgG antibodies were prepared from mice or rabbits immunized with the whole bacteria protein and purified by Antibody Purification Kit. Then, the corresponding secondary antibodies were added for 45 min at 37°C. DAPI (Beyotime, China) was used to stain the cell nuclei for 10 min. Images were acquired with a laser scanning confocal microscope (LSCM, Olympus, FV3000).

Immunohistochemistry (IHC)

Tissues were fixed in 4% paraformaldehyde for 24 h, embedded in paraffin, and then cut into 5.0-μm thick sections. The sections were dewaxed and blocked with 10% goat serum (Boster, China). Then, the tissue sections were stained with hematoxylin and eosin (H&E) or incubated with primary antibodies against Fn (homemade, 1:5,000), CD66b (1:200, Servicebio, China), Ly6G (1:200, Servicebio, China), PD-L1 (1:5,000, Proteintech, China) and CX3CR1 (1:100, Affinity, China) overnight at 4°C. HRP-conjugated anti-mouse or anti-rabbit secondary antibodies (1:200, EarthOx Life Sciences, USA) were used as the secondary antibodies for incubation for 2 h at RT. The sections were developed with 3-diaminobenzidine tetrahydrochloride (CWBIO, China), followed by counterstaining with 10% hematoxylin (Beyotime, China) for 4 min.

Western blot analysis

Cells were collected, lysed with RIPA buffer (Beyotime, China), and boiled in loading buffer for 5 min. The protein samples were loaded onto SDS‒PAGE gels and transferred to PVDF membranes. The membranes were incubated with 5% skim milk blocking buffer for 1 h and then incubated at 4°C overnight with antibodies against CXCR2 (1:1,000, Affinity, China), CCR5 (1:1,000, Bioworld, USA), PD-L1 (1:1,000, Bioworld, USA), CX3CR1 (1:1,000, Bioworld, USA), p-P65 (1:1,000, CST, USA), P65 (1:1,000, CST, USA), p-IκBα (1:1,000, CST, USA), IκBα (1:1,000, HUABIO, China), p-P50 (1:1,000, HUABIO, China), P50 (1:1,000, HUABIO, China), p-STAT3 (1:1,000, CST, USA), STAT3 (1:1,000, Bioworld, USA), MMP9 (1:1,000, Proteintech, China) and VEGFA (1:1,000, Abcam, UK). Horseradish peroxidase-conjugated (HRP) secondary antibody (1:5,000, EarthOx Life Sciences, USA) was incubated with the membrane for 2 h at RT. GAPDH (1:5,000, Bioworld, USA) was used as a loading control. Protein bands were visualized with a chemiluminescence detection system (Tianneng, China).

Flow cytometry analysis

For animal experiments, tumor and spleen tissues were surgically excised and dissociated into single-cell suspensions. Then, the cells were stained using fluorescence-labeled antibodies against CD3, CD4, CD8a (5 μl/test, Elabscience, China), PD-L1 (5 μl/test, Thermo Fisher, USA), CX3CR1 (2.5 μl/test, Biolegend, CA), Ly6G (5 μl/test, Elabscience, China), F4/80 (5 μl/test, Elabscience, China) or isotype control. For PD-L1 expression in PMNs, dHL60 cells, and PMs, the cells were stained with PD-L1 (5 μl/test, Thermo Fisher, USA) using the manufacturer’s instructions and measured by CytoFLEX S (Beckman, USA). Data were analyzed by FlowJolnit software.

Leukocyte microbial 16S sequencing

The mouse leukocyte microbiome was analyzed using 16S sequencing high-throughput technology (PE250, NovaSeq 6000, Illumina Inc., USA). The V3–V4 region of 16S was amplified with primers 341F (5′-CCTAYGGGRBGCASCAG) and 806 R (5′-GGACTACNNGGGTATCTAAT). The data and statistical analyses of 16S sequencing were performed with Lianchuan (Lianchuan Biotech Co., Ltd.) online tools and were deposited in the SRA database (SRA: PRJNA1018212).

RT‒qPCR and RNA sequencing

Total RNA was extracted using TRIzol (Invitrogen, USA) with on-column DNase I (Sigma, USA) digestion following the manufacturer’s protocol. First-strand cDNA was synthesized using Evo M-MLV RT Master Mix (AGbio, China). RT‒qPCR was performed on a Light Cycler® 480 Instrument II (Roche, Germany) using SYBR Green Pro Taq HS Premix (AGbio, China). GAPDH was used as a reference gene for relative quantification. To quantify the relative abundance of Fn in the samples, we extracted bacterial genomic DNA using the TIANamp bacteria DNA kit (TIANGEN, Germany), and performed with qPCR value of Fn was normalized to the host 18S gene or per gram of tissues. The primers used in the real-time PCR are presented in the Supplementary Table S6. The 2^−ΔΔCt method was utilized to quantify gene expression.

RNA sequencing was completed by IGE Biotechnology Ltd. (Guangzhou, China). Briefly, total RNA was isolated using an RNeasy Mini Kit (Qiagen, USA) following the manufacturer’s protocol. Paired-end sequencing was performed with an Illumina NovaSeq 6000 system (Illumina, USA). The data were deposited in the GEO database (GEO: GSE243383).

Nano LC-MS/MS

The purified recombinant Fn1792 was freeze-dried and trypsin-digested (Sigma, USA). The products of the pull-down assay were extracted from SDS‒PAGE gels, the target band was cut off, treated with DL-dithiothreitol and indole-3-acetic acid, and then digested with trypsin. Subsequently, the tryptic protein hydrolyzates were analyzed using Nano LC-Q Exactive Plus MS (Thermo Fisher, USA). The protein peptide information was obtained by Mascot 2.2 software to search the Swiss-Prot and NCBI protein databases.

Transfection with small interfering RNA (siRNA)

Cells were cultured in 6-well plates at a density of 5.0 × 10⁵/ml. The cells were then transfected with siCX3CR1 and siNC (SR301082, Origene, USA) using a siRNA transfection reagent (Invitrogen, USA) for 24 h. The expression of the target gene CX3CR1 was detected by Western blotting.

Transwell migration assay

Transwell inserts (Corning, USA) in 24-well plates were used. 1.0 × 10⁵ PMNs or dHL60 cells in 100 µl serum-free RPMI-1640 were added to the upper chambers treated with or without CXCR2 neutralizing antibody (CXCR2 nAb, 5 μg/ml, R&D Systems, MN), and 600 μl medium with 1.0 × 10⁵ HCT116 and RKO cells treated with or without Fn (MOI 100: 1) and CXCL2/CXCL8 (CXCL2/8, 20 ng/ml, Peprotech, USA) was added to the lower chamber to serve as a chemotactic agent. Transwell chemotaxis assay of MCs: MCs were added to the upper chambers treated with or without CCR5 neutralizing antibody (CCR5 nAb, 5 μg/ml, R&D Systems, MN), and HCT116 or RKO cells were seeded in the lower chamber with or without Fn (MOI 100:1) and CCL5 (20 ng/ml, Peprotech, USA).

To examine the cellular migration ability of PMNs toward vascular endothelial cells, PMNs were seeded into the upper chambers with or without Fn (MOI 10: 1) or Fn1792 (0.01 μg/ml) for 12 h or pretreated with JMS-17-2 (10 nM, MedChemExpress, USA) for 1 h. HUVECs were added to the lower chamber.

After 3 h, cells across the pores were fixed with 4% paraformaldehyde (Beyotime, China) and stained with 1% crystal violet solution (Beyotime, China). For each chamber, five fields were randomly chosen, and the cells were counted. The migration was analyzed by ImageJ. Each Transwell assay was repeated in three independent experiments.

Cell-to-cell transmission assays

Cell-to-cell transmission assays were performed using Transwell inserts. Briefly, PMNs and MCs were infected with Fn (MOI 10:1) for 12 h, and extracellular bacteria were removed by gentamicin (50 μg/ml) for 2 h. Then, the cells were seeded into 0.4 μm pore size Transwell inserts, and the PMNs and MCs inserts were placed on top of uninfected HCT116 or RKO cells per well in 6-well plates. PMNs and tumor cells were both cultured in their corresponding media. After 12 h, 24 h, 48 h or 72 h, extracellular bacteria of tumor cells were removed by gentamicin (50 μg/ml) for 2 h. Then, intracellular Fn was detected by intracellular survival assays and immunofluorescence.

T cell proliferation and IFN-γ levels analysis

CD3⁺ T-cells were isolated from human PBMCs using CD3-immunomagnetic beads (Miltenyi Biotech, USA) and stimulated with plate-bound anti-CD3 antibody (Miltenyi Biotech, USA). The purified CD3⁺ T-cells were stained with CFSE (5 μM, Thermo Fisher, USA) at 37°C for 8 min under gentle agitation. The labeling reaction was terminated for 5 min by adding an equal volume of FBS. The CFSE-labeled cells were then washed twice with PBS, resuspended to a final concentration of 1 × 10⁵ cells/mL in RPMI-1640 medium, and treated with PMNs, Fn⁺ PMNs or Fn⁺ PMNs plus PD-L1 neutralizing antibody (20 μg/ml, R&D Systems, MN) at a 1:1 ratio for 5 days. Cells were further stained with the APC-anti-IFN-γ antibody (5 μl/test, Thermo Fisher, USA) and analyzed by flow cytometry.

T-cell/CRC cell coculture systems

HCT116 and RKO cells were cultured in 6-well plates to 60% confluency and stained with CFSE (5 μM, Thermo Fisher, USA). Purified CD8⁺ T-cells treated with PMNs, Fn⁺ PMNs or Fn⁺ PMNs plus PD-L1 neutralizing antibody (20 μg/ml, R&D Systems, MN) effectors were cocultured with tumor cell targets (E:T of 50:1) in a 6-well plate for 24 h. Following coculture, the cells were washed three times with PBS and analyzed by flow cytometry. Additionally, tumor cells were cocultured with the above-treated CD8⁺ T-cells and analyzed by Cell Counting Kit-8 (New Cell & Molecular, China).

ROS detection

To determine intracellular ROS production, PMNs were seeded in 6-well plates (1 × 10⁵ cells/well) treated with Fn (MOI 10: 1) or E. coli strain DH5a (Ec) (MOI 10: 1). After 12 h, the cells were incubated with reagents from the Reactive Oxygen Species Assay Kit (Beyotime, China), ROS Brite™ HPF (AAT Bioquest, USC), or dihydroethidium (KeyGEN, China), and fluorescence was measured by flow cytometry.

To determine extracellular ROS production, the fluorescent probe 2,7-dichlorodihydrofluorescein (DCFH) was converted from DCFH-DA (0.5 mL, 1 mm in ethanol) by reacting with an aqueous solution of NaOH (2 mL, 10 mm) for 30 min at RT. The hydrolyzate was then neutralized with PBS buffer solution to obtain the stock solution and was mixed separately with cell culture supernatants of Fn- or Ec-treated PMNs, and the final concentrations were 10 μM. Then, the fluorescence was measured by a fluorescence microplate reader.

H₂S detection

PMNs were treated with Fn (MOI 10:1) and Sodium Hydrogen Sulfide (NaSH, 50 μM) for 6 h. Then, the cells were incubated with the WSP-5 h₂S probe (Maokangbio, China) at 37°C for 30 min, and fluorescence was measured by flow cytometry.

Cytotoxicity assay

Cell survival was quantified using an Annexin V-FITC/PI Apoptosis Detection Kit (Meilunbio, China), LDH Cytotoxicity Assay Kit (Beyotime, China), or Cell Counting Kit-8 New (Cell & Molecular, China) according to the instruction manual.

ELISA

The levels of CXCL2, CXCL8, CCL5, VEGF and MMP9 were detected by ELISA kits (Multi Sciences, Hangzhou, China) according to the manufacturer’s instructions.

Database analysis

The DIANA database (https://dianalab.e-ce.uth.gr/peryton/#/.) and gutMEGA online database (http://gutmega.omicsbio.info/) were used to analyze microbe-disease associations. The colorectal carcinoma dataset of The Cancer Genome Atlas (TCGA) database (https://docs.gdc.cancer.gov/Data/Bioinformatics_Pipelines/Expression_.

mRNA_Pipeline/, accessed on 5 January 2024) were downloaded and visualized using the R language (version: 4.1.3).

Quantification and statistical analysis

The data are represented as mean ± SEM. An unpaired two-tailed Student’s t-test, one-way ANOVA, two-way ANOVA or Mann‒Whitney test was used to analyze the statistically significant differences as defined in the figure captions, and data were considered statistically significant when the values of p < 0.05. Statistical analyses were performed using GraphPad 8.0.

Funding Statement

This work was supported by the National Natural Science Foundation of China [No. 82172890; 32300761]. The Guangdong Basic and Applied Basic Research Foundation [No. 2023A1515012256].

Disclosure statement

No potential conflict of interest was reported by the author(s).

Author contributions

F.-F.C. and S.-H.G. performed most of the experiments and data analysis. Y.-Q.L. and Y.-F.L. assisted in the in vivo or in vitro experiments. S.-H.G. and L.L led the collection of fecal, blood, tissue sections and related clinical data. F.-F.C. and S.-H.G. wrote the original draft. F.-F.C., S.-H.G., Y.-Q.L. and S.-X.C contributed in experiments guidance. G.Z. and J.A. designed the project and revised and edited the manuscript.

Data availability statement

16S sequencing and RNA-seq data have been deposited at SRA (PRJNA1018212) and GEO database (GSE243383), and are publicly available as of the date of publication. Any additional information required to reanalyze the data reported in this paper is available from the corresponding author, Ge Zhang (zhangge@mail.sysu.edu.cn).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19490976.2024.2442037