Prophylactic TLR9 stimulation reduces brain metastasis through microglia activation

Abstract

Brain metastases are prevalent in various types of cancer and are often terminal, given the low efficacy of available therapies. Therefore, preventing them is of utmost clinical relevance, and prophylactic treatments are perhaps the most efficient strategy. Here, we show that systemic prophylactic administration of a toll-like receptor (TLR) 9 agonist, CpG-C, is effective against brain metastases. Acute and chronic systemic administration of CpG-C reduced tumor cell seeding and growth in the brain in three tumor models in mice, including metastasis of human and mouse lung cancer, and spontaneous melanoma-derived brain metastasis. Studying mechanisms underlying the therapeutic effects of CpG-C, we found that in the brain, unlike in the periphery, natural killer (NK) cells and monocytes are not involved in controlling metastasis. Next, we demonstrated that the systemically administered CpG-C is taken up by endothelial cells, astrocytes, and microglia, without affecting blood-brain barrier (BBB) integrity and tumor brain extravasation. In vitro assays pointed to microglia, but not astrocytes, as mediators of CpG- C effects through increased tumor killing and phagocytosis, mediated by direct microglia-tumor contact. In vivo, CpG-C-activated microglia displayed elevated mRNA expression levels of apoptosis-inducing and phagocytosis-related genes. Intravital imaging showed that CpG-C-activated microglia cells contact, kill, and phagocytize tumor cells in the early stages of tumor brain invasion more than nonactivated microglia. Blocking in vivo activation of microglia with minocycline, and depletion of microglia with a colony-stimulating factor 1 inhibitor, indicated that microglia mediate the antitumor effects of CpG-C. Overall, the results suggest prophylactic CpG-C treatment as a new intervention against brain metastasis, through an essential activation of microglia.

Fig 1. A single systemic prophylactic treatment with CpG-C results in long-term reduction of metastatic burden in the brain.

(a) In the experimental metastasis models, we used the aECAi approach [32] for injection of tumor cells (see Methods): a method that improves brain targeting and preserves cerebral hemodynamics. (b) Histological images of D122 brain metastases from C57BL/6J mice on day 21 post–tumor inoculation show well-demarcated metastases, as well as vessel co-option growth. PC14-PE6 brain metastases from nude animals on day 25 post–tumor inoculation show large, well-demarcated metastases. Scale bar is 500 μm for the images on the left and middle, and 50 μm for the images on the right. (c-f) A single prophylactic systemic (i.p.) injection of 4 mg/kg CpG-C resulted in reduced growth of experimental brain metastases, as indicated by bioluminescence and fluorescence imaging. (c) C57BL/J6 mice injected with syngeneic D122 tumor cells and pretreated with CpG-C had reduced tumor burden compared with control animals, becoming significant on day 14 (ci; n = 6–7; F_(1,11) = 19.02, p = 0.0011) and reaching a 77-fold difference in total flux on day 21 (two-tailed Mann–Whitney U = 3, p = 0.0082; cii). Interestingly, in two CpG-C–treated animals, the bioluminescent signal gradually decreased and disappeared on day 21. (d) Ex vivo bioluminescence imaging of the brains from the syngeneic model indicated a 48-fold reduced tumor burden (total flux) in CpG-C–treated animals (n = 6; two-tailed unpaired Student t test, t₍₁₀₎ = 3.722, p = 0.0040). (e) Athymic nude mice injected with human (xenograft) PC14-PE6 tumor cells and pretreated with CpG-C had reduced tumor burden compared with control animals, becoming significant on day 4 (e.i; n = 7; F_(1,12) = 77.45, p < 0.0001) and reaching a 82-fold difference in total flux on day 25 (two-tailed unpaired Student t test, t₍₁₂₎ = 7.090, p < 0.0001; e.ii). (f) Using Maestro fluorescence imaging, a reduction in brain tumor burden (i.e., tumor area) was evident in the human xenograft model in CpG-C–treated animals (n = 7; two-tailed Mann–Whitney U = 8, p = 0.0373). (g) (left) Timeline for spontaneous melanoma brain metastasis model [33] (see Methods). (right) CpG-C treatment during seven perioperative days resulted in reduced micrometastases in the brain (measured by mCherry mRNA expression; n = 9 and n = 12 for control and CpG-C, respectively; two-tailed unpaired Student t test, t₍₁₉₎ = 2.278, p = 0.0345). The background of images was manually removed. Box plot whiskers represent minimum–maximum range. See S1 Fig for comparison of bioluminescent signal at day 1. The underlying data for this figure can be found in S1 Data. aECAi, assisted external carotid artery inoculation; CCA, common carotid artery; ECA, external carotid artery; ICA, internal carotid artery; i.p., intraperitoneal; RT-qPCR, real-time quantitative polymerase chain reaction.

Fig 2. The effects of CpG-C on brain metastases are not mediated by NK cells or monocytes.

(a) Depletion of NK cells using NK1.1 antibody resulted in a 5-fold elevation in D122 LLC tumor retention in the lungs (n = 8; t₍₁₄₎ = 4.4781, p = 0.0001), and partially blocked the beneficial effects of CpG-C (n = 8; t₍₁₄₎ = 1.1517, p = 0.0038), evident in naïve animals (n = 8; t₍₁₄₎ = 0.7002, p = 0.0019). (b) In brains of the same animals, NK depletion had no effect on tumor retention (n = 8; t₍₁₄₎ = 0.1894, p = 0.3935), nor mediated the beneficial effects of CpG-C (n = 8; t₍₁₄₎ = 0.1099, p = 0.0811), evident in both naïve (n = 8; t₍₁₄₎ = 0.7973, p < 0.0001) and NK-depleted animals (n = 8; t₍₁₄₎ = 0.4979, p = 0.0056). (c) Depletion of monocytes using clodronate liposomes resulted in increased lung tumor retention of D122 LLC cells (n = 7–8; t₍₁₃₎ = 0.9072, p = 0.0292), an effect rescued by CpG-C (n = 8–9; t₍₁₅₎ = 1.2700, p = 0.0003), indicating that lung tumor retention is mediated by monocytes, while they do not mediate the effects of CpG-C. (d) In brains of the same animals, monocyte depletion did not affect tumor retention (n = 7–8; t₍₁₃₎ = 0.3028, p = 0.3081), and CpG-C reduced tumor retention in naïve (n = 7; t₍₁₂₎ = 0.7910, p = 0.0006) and monocyte-depleted animals (n = 8–10; t₍₁₆₎ = 1.0377, p = 0.0001). Two-way permutations were used for the above analyses. Box plot whiskers represent minimum–maximum range. The underlying data for this figure can be found in S1 Data. LLC, Lewis lung carcinoma; NK, natural killer; n.s., nonsignificant.

Fig 3. CpG-C infiltrates the brain and is internalized by endothelial cells, astrocytes, and microglia.

(a-b) TAMRA-labeled CpG-C was injected intraperitoneally; 24 hours later, brains were perfused, and CpG-C internalization in endothelial cells (CD31), astrocytes (GFAP and GLAST), and microglia (CX3CR1 and CD11b) was visualized in histological sections using confocal microscopy (a; top panels are 15-μm z-max projections in the primary somatosensory cortex, and lower panels are partial reconstructions) and quantified using ImageStream FACS analysis (b). The majority of each of the three cell populations internalized CpG-C, indicating that CpG-C crosses the BBB into the parenchyma (n = 4). Scale bar is 5 μm. Data presented as mean (±SEM). The underlying data for this figure can be found in S1 Data, and our gating strategies are provided in S9 Fig. BBB, blood-brain barrier; CD, cluster of differentiation; CX3CR1, CX3C chemokine receptor 1; FACS, fluorescence-activated cell sorting; FITC, fluorescein isothiocyanate; GFAP, glial fibrillary acidic protein; GLAST, glutamate aspartate transporter; TAMRA, tetramethylrhodamine.

Fig 4. CpG-C does not affect BBB leakage or cellular permeability.

(a-c) Biocytin-TMR was intravenously injected into animals expressing GFP under the claudin-5 promotor 24 hours following CpG-C or control treatment, and 90 minutes later brains were perfused and removed. Biocytin-TMR intensity (normalized to intensity levels in the liver, not shown) and IgG staining intensity were similar in CpG-C–treated and control animals (two-tailed unpaired Student t test, t₍₂₂₎ = 0.3758, p = 0.7106; a-b). Moreover, no difference in number of gaps in claudin-5 strands was found between control and CpG-C–treated animals (two-tailed unpaired Student t test, t₍₂₂₎ = 0.4283, p = 0.6726; c). (d) Brain sections of WT animals treated with CpG-C or PBS were stained for CD4 or CD68 24 hours following treatment. No infiltration of immune cells was detected (spinal cords of EAE mice served as positive controls; right panels). (e-f) Using two-photon imaging, we followed leakage of a low (NaF; 376 Da) and a high (Texas-Red; 70kDa) molecular weight dextrans. Intensities of representative images (e) were auto-adjusted in Fiji for display purposes only. No differences were found between control and CpG-C–treated animals at baseline (p = 0.8567 and p = 0.8421 for NaF and Texas-Red, respectively), following a single CpG-C treatment (p = 0.9243 and p = 0.2419 for NaF and Texas-Red, respectively), and following two CpG-C treatments (p = 0.4656 and p = 0.3918 for NaF and Texas-Red, respectively. See Methods for an explanation of the quantification; f). For (a-c), each sample consisted of an average of at least five images that were analyzed. Samples were taken from four different anatomical brain regions (cortex, midbrain, cerebellum, and hippocampus) in three mice/group (See S5 Fig for regional presentations). Scale bar is 50 μm. Box plot whiskers represent minimum–maximum range (a-c), and data in (f) are presented as mean (±SEM). The underlying data for this figure can be found in S1 Data. A.U., arbitrary units; BBB, blood-brain barrier; CD, cluster of differentiation; EAE, experimental autoimmune encephalomyelitis; GFP, green fluorescent protein; IgG, immunoglobulin G; TMR, tetramethylrhodamine; WT, wild type.

Fig 5. Microglia, but not astrocytes, mediate the effects of CpG-C in vitro.

(a-d) Primary cultures of microglia and of astrocytes were subjected to 100 nM/L CpG-C or non-CpG ODN (control) for 24 hours. ¹²⁵IUDR-labeled D122 cells were plated with the treated primary cultures or subjected to their conditioned media alone, and cytotoxicity (percent of D122 lysis) was assessed by measuring radioactivity in the media 8 hours later. Primary cultured astrocytes, subjected to CpG-C or non-CpG ODN, did not cause tumor cell lysis when in contact (F_(3,12) = 0.7755, p = 0.5298; a), nor did their conditioned media (F_(3,9) = 0.6923, p = 0.5794; b). In contrast, primary cultured microglia cells induced lysis in tumor cells, and treatment with CpG-C significantly increased it when in contact with tumor cells (F_(3,28) = 64.1, p < 0.0001; c), while their conditioned media had no effect (F_(3,9) = 0.6923, p = 0.5794; d). (e-i) The microglia cell line, N9, was subjected to CpG-C (see above). Luc2-mCherry–labeled D122 cells were plated with the N9 cultures with contact (e) or without contact (coculture; f), or with their conditioned media alone (g), and bioluminescent signal was measured, indicating viability of tumor cells. There was a reduced signal in tumor cells cultures that were in direct contact with N9 cells (F_(3,12) = 14.6, p = 0.0003; e), while no difference was evident in cocultures (no contact; F_(3,18) = 0.3535, p = 0.7872; f) or in cultures subjected to conditioned media (F_(3,12) = 0.1425, p = 0.9325; g). Two-tailed one-way ANOVA with Bonferroni multiple comparison correction was used for a-g. (h) Annexin V binding (a marker for early-stage apoptosis) was quantified in D122 cells cultured with pretreated N9 cells using FACS. Tumor cells cultured with N9 cells pretreated with CpG-C exhibited increased annexin V staining (compared with scrambled CpG-C; two-tailed Student t test for unpaired samples, t₍₉₎ = 2.3060, p = 0.0465). (i) N9 cultures treated with CpG-C or non-CpG ODN (control) for 24 hours were washed and plated with pH-sensitive bio-particles to assess phagocytosis capacity. CpG-C–treated N9 cells exhibited a 3-fold increased phagocytic capacity (two-tailed Student t test for unpaired samples, t₍₁₄₎ = 6.6960, p < 0.0001). Box plot whiskers represent minimum–maximum range. Refer to S7 Fig for comparison between PBS, non-CpG ODN, and CpG-C. The underlying data for this figure can be found in S1 Data, and our gating strategies are provided in S9 Fig. FACS, fluorescence-activated cell sorting; IUDR, iododeoxyuridine; ODN, oligodeoxynucleotide.

Fig 6. Microglia mediate the in vivo effects of CpG-C.

(a-e) Chronic in vivo two-photon imaging in CX3CR1^GFP/+ mice indicated that microglia cells (green) have dynamic relations with tdTomato-labeled D122 tumor cells (red; 15-μm stacks, with 1-μm z-steps) and that CpG-C treatment increases tumor internalization by microglia. (a) A microglia cell (arrow) interacting with a tumor cell (arrowhead; two hours post–tumor cells inoculation), phagocytizing it (one day later), and dismantling it (day 2 post–tumor cell injection, inset shows an engulfed tdTomato-positive cell or part of it). (b) Different levels of interaction between microglia and tumor cells—(i) no interaction; (ii) contact; and (iii) microglia phagocytized a tumor cell. (c) Partial reconstitution of a 15-μm stack with 1-μm z-steps demonstrating the microglia-tumor cells’ “battlefield” four hours after tumor cell injection. (d) Representative images and quantification (e) of microglia-tumor interactions in control and CpG-C–treated mice four hours and one day following tumor cell inoculation (arrows for contact and arrowheads for internalization). CpG-C treatment resulted in increased contacts four hours following tumor inoculation (n = 3; F_(1,4) = 2.875, p = 0.0218) and in microglia internalization of tumor cells/debris 4 (p = 0.0372) and 24 hours (n = 3; F_(1,4) = 3.400, p = 0.0041) following tumor inoculation. Scale bar for (a-d) is 20 μm. (f-h) CX3CR1^GFP/+ mice were treated with a single systemic prophylactic CpG-C treatment, injected mCherry-labeled D122 tumor cells using the aECAi approach, and brains were analyzed using ImageStream FACS. While CpG-C treatment did not affect the number of microglia cells (n = 5; two-tailed Mann–Whitney U = 10, p = 0.6905; f) or capacity of tumor cell infiltration (indicated by total mCherry area in perfused brains; two-tailed Mann–Whitney U = 9, p = 0.5476; g), it resulted in increased phagocytosis of tumor cells by microglia (two-tailed Student t test for unpaired samples, t₍₄₎ = 3.8850, p = 0.0178; h). Scale bar for (e-g) is 5 μm. Data in (e) are presented as mean (±SEM) and box plot whiskers represent minimum–maximum range (f-h). The underlying data for this figure can be found in S1 Data. aECAi, assisted external carotid artery inoculation; FACS, fluorescence-activated cell sorting; GFP, green fluorescent protein; n.s., nonsignificant; ROI, region of interest.

Fig 7. Blocking microglia activation or complete depletion hinders the effects of CpG-C on brain metastasis.

(a-b) Microglia activation was blocked in vivo using systemic treatment with minocycline. (a) Minocycline treatment resulted in increased brain tumor retention of D122 cells (n = 15–16, t₍₂₇₎ = 66.3229, p = 0.0118), while CpG-C treatment reduced tumor retention in naïve mice (n = 16 for control and n = 14 for minocycline-treated mice; t₍₂₈₎ = 63.0149, p < 0.0001), but not in minocycline-treated animals (n = 14–15; t₍₂₇₎ = 42.7850, p = 0.1863). The effects of CpG-C were completely blocked by minocycline treatment (n = 14; t₍₂₆₎ = 86.5528, p < 0.0001), indicating that microglia activation mediates the beneficial effects of CpG-C. (b) ImageStream FACS analysis indicated that minocycline blocked (p = 0.0493) the beneficial effects of prophylactic CpG-C treatment (p = 0.01) on the ability of microglia to phagocytize tumor cells (n = 2 for control and n = 3 for CpG-C and CpG-C + minocycline animals; two-tailed one-way ANOVA with Tukey multiple comparisons test; F_(2,5) = 12.85, p = 0.0107). (c) Without stimulation with CpG-C, microglia cells do not affect brain tumor seeding, as indicated by depletion of microglia cells using the colony stimulating factor 1 receptor inhibitor, PLX5622. Microglia depletion did not affect D122 tumor retention in the brain (n = 14–15; t₍₂₇₎ = 0.0851, p = 0.7490), while it blocked the beneficial effects of CpG-C (n = 14 for depleted animals and n = 16 for depleted animals treated with CpG-C; t₍₂₈₎ = 0.0460, p = 0.8637), evident in naïve animals (n = 14–15; t₍₂₇₎ = 0.0460, p = 0.0087). Accordingly, microglia-depleted animals treated with CpG-C had increased brain tumor retention compared with naïve animals treated with CpG-C (t₍₂₈₎ = 0.5417, p = 0.0068). Two-way permutations were used for analyses of (a) and (c). Box plot whiskers represent minimum–maximum range (a,c), and data in (b) are presented as mean (±SEM). The underlying data for this figure can be found in S1 Data. FACS, fluorescence-activated cell sorting; n.s., nonsignificant.

Fig 8. CpG-C treatment results in elevated in vivo expression of apoptosis-inducing, phagocytosis-related, and inflammatory factors.

(a) CX3CR1^GFP/+ mice were injected with CpG-C or vehicle, and 24 hours later mRNA expression levels in sorted microglia cells were quantified using RT-qPCR. In one experiment, six animals of each group were pooled into a single sample, and in the second experiment, two CpG-C–treated animals and three controls were analyzed separately (n = 3–4 from 8–9 animals). As expected, Tmem119, a general microglia marker, was unaffected by the treatment (t₍₅₎ = 0.371, p = 0.7258). (b) The death ligands, Tnfsf10 and Fasl, were elevated by 3–4-fold by a single CpG-C injection (t₍₅₎ = 2.564, p = 0.0437; and t₍₅₎ = 2.36, p = 0.0324, respectively). (c) Expression levels of receptors related to phagocytosis were significantly higher in microglia of CpG-C–treated animals. While no change was apparent in Cd36 (t₍₅₎ = 0.3966, p = 0.7080) and Cd68 (t₍₅₎ = 0.01655, p = 0.9874), a significant increase was evident in Cd47 (t₍₅₎ = 2.819, p = 0.0186), Trem2 (t₍₅₎ = 2.762, p = 0.0199), and Marco (which was not detected in control animals) (t₍₄₎ = 4.499, p = 0.0108). (d) While RNA of the inflammatory cytokines Il-6 and Il1-β was not affected by CpG-C treatment (t₍₅₎ = 0.04089, p = 0.9690; t₍₅₎ = 0.4417, p = 0.6772, respectively), Tnf (t₍₄₎ = 3.207, p = 0.0163) and Inf-γ (t₍₄₎ = 2.394, p = 0.0374), which synergistically induce apoptosis in tumor cells [54], and Nos2 (t₍₅₎ = 2.744, p = 0.0203), which is tumoricidal at high concentrations [60], were increased following CpG-C injection. Data are presented as mean (±SEM). The underlying data for this figure can be found in S1 Data, and our gating strategies are provided in S9 Fig. Cd, cluster of differentiation; Fasl, Fas ligand; GFP, green fluorescent protein; Il, interleukin; Inf-γ, interferon gamma; Marco, macrophage receptor with collagenous structure; Nos2, nitric oxide synthase 2; RT-qPCR, real-time quantitative polymerase chain reaction; Tmem119, transmembrane protein 119; Tnf, tumor necrosis factor; Tnfsf10, tumor necrosis factor superfamily member 10; Trem2, triggering receptor expressed on myeloid cells 2.

Fig 9. Proposed mechanism.

Systemic prophylactic treatment with CpG-C during the perioperative period activates microglia to induce apoptosis in tumor cells and phagocytize them, resulting in reduced brain metastases colonization. A few weeks to months may pass from the time of cancer diagnosis to the time of primary tumor excision [95]. During this period, and a few weeks after surgical excision (known as the perioperative period), there is a high risk for developing brain metastasis with terminal consequences. CpG-C, a TLR9 agonist, given as a systemic prophylactic treatment during this crucial period, infiltrates the brain and activates microglia (1), increasing their expression of Tnfsf10 and Fasl, resulting in contact-dependent induced apoptosis of tumor cells (2). Furthermore, Cd47, Trem2, and Marco expression is increased, triggering enhanced microglial phagocytosis and dismantling of tumor cells (3), thereby reducing brain metastasis colonization. CD, cluster of differentiation; FasL, Fas ligand; Marco, macrophage receptor with collagenous structure; TLR9, toll-like receptor 9; Tnfsf10, tumor necrosis factor superfamily member 10; TREM2, triggering receptor expressed on myeloid cells 2.