Metastatic site-specific polarization of macrophages in intracranial breast cancer metastases

Abstract

In contrast to primary tumors, the understanding of macrophages within metastases is very limited. In order to compare macrophage phenotypes between different metastatic sites, we established a pre-clinical mouse model of intracranial breast cancer metastasis in which cancer lesions develop simultaneously within the brain parenchyma and the dura. This mimics a situation that is commonly occurring in the clinic. Flow cytometry analysis revealed significant differences in the activation state of metastasis-associated macrophages (MAMs) at the two locations. Concurrently, gene expression analysis identified significant differences in molecular profiles of cancer cells that have metastasized to the brain parenchyma as compared to the dura. This included differences in inflammation-related pathways, NF-kB1 activity and cytokine profiles. The most significantly upregulated cytokine in brain parenchyma- versus dura-derived cancer cells was Lymphotoxin β and a gain-of-function approach demonstrated a direct involvement of this factor in the M2 polarization of parenchymal MAMs. This established a link between metastatic site-specific properties of cancer cells and the MAM activation state.

Keywords: breast cancer brain metastases; dural metastases; lymphotoxin β; metastasis-associated macrophages; tumor-associated macrophages.

Figure 1. 4T1 breast cancer model with simultaneous metastasis to the brain parenchyma and the dura

(A) The distribution of metastatic lesions 10 days after administration of Fluc-tagged 4T1 cancer cells into the external or internal carotid artery was analyzed by ex-vivo bioluminescence imaging of the brain parenchyma (brain) and the skull/dura (skull). (B) Quantification of bioluminescence signal shown in A. Int: internal; Ext: external; (C) 3D in vivo bioluminescence imaging of cancer cells 10 days after their administration into the internal carotid artery. Dorsal (left) and side view (right) are shown. The majority of cancer lesions are localized at the top of the head, suggesting predominant tumor burden at the skull/dura. (D) H&E staining of coronal head sections containing dural metastases (red arrows). (E) Verhoeff-Van Gieson staining of dural metastases. Dura mater is marked with black arrows (left image). Invasion of cancer cells into the skull is marked with red arrows (left and right image). (F and G) Distribution of cancer lesions between the skull/dura and the brain parenchyma was analyzed by ex vivo bioluminescence imaging at 16 and 45 days post-cancer cell injection into the internal carotid artery using PyMT (F) and MDA-MB-231 cancer cell lines (G), respectively. Statistical significance in B, F and G was determined using two-tailed Student's T-test with unequal variance (p ≤ 0.05); n = 4.

Figure 2. Inflammatory tumor microenvironment in dural and parenchymal brain metastases

(A) Infiltration of immune cells into dural (dura) and parenchymal 4T1 cancer lesions (parenchyma) in intracarotid artery model; n = 5. (B) Representative flow cytometry analysis of microglia and macrophages within parenchymal (top) and dural lesions (bottom) in 4T1 breast cancer model. Microglia were identified as CD45^lowCD11b^low cells (red) and macrophages as CD45^highCD11b^high cells (blue) within the CD11b⁺F4/80⁺ gate. (C) Quantification of microglia and macrophages in 4T1, PyMT and MDA-MB-231 (231) models based on the flow cytometry analysis shown in Figure 2B, Supplementary Figure S5A–S5B; n = 4. (D) Representative flow cytometry analysis of MHCII and CD11c expression in the microglia (red) and macrophages (blue) within parenchymal (top) and dural (bottom) 4T1 metastases. The contour plots were gated on the CD11b⁺F4/80⁺ population shown in B. (E) Quantification of MHCII⁺ and CD11c⁺ cells within the microglia and macrophage populations in 4T1, PyMT and MDA-MB-231 models based on the flow cytometry analysis shown in Figure 2D, Supplementary Figure S5A–S5B; n = 4. (F) Mean fluorescent intensity (MFI) for MHCII expression in dural and parenchymal MAMs; n = 4. Statistical significance in A, C, E and F was determined using one-tailed Student's T-test with unequal variance (p ≤ 0.05). Error bars represent standard deviations (SD).

Figure 3. 4T1 cancer cells evolve distinct molecular profiles after they have metastasized to the dura versus brain parenchyma

(A) Experimental outline for the in vivo selection of 4T1 cancer cell variants with site-specific characteristics. Dural cancer lesions are illustrated in black and parenchymal lesions in red. (B) Heat map and hierarchical clustering of genes differentially expressed between the 4T1-Par3 and 4T1-Dura 3 cancer cell variants. (C) Upregulation of the NF-kB1 and TCF7L2 (TCF4) transcription factor activity in parenchymal versus dural 4T1 cell variants. E-value scores and intersection percentage for significantly inhibited transcription factors (e-value < = 0.05) in dural versus parenchymal cell variants were determined using TFactS software. (D) Summary of the signaling pathways that were differentially regulated between the 4T1-Par3 and 4T1-Dura 3 cancer cell variants. (E) Graphic summary of the cytokines that were significantly upregulated in parenchymal versus dural 4T1 cancer cell variants. (F) Quantification of Ltβ, Ltα, and Ltβr mRNA by qRT-PCR. Statistical significance was determined using two-tailed Student's T-test with unequal variance (p ≤ 0.05). Error bars represent SD. (G) Analysis of total LTβR protein expression in whole cell lysates by Western blot. One out of three independent experiments is shown. (H) LTβR surface expression (MFI) is reduced in 4T1-Par3 versus 4T1-Dura3 cell lines as quantified by flow cytometry.

Figure 4. Distinct polarization state of dural and parenchymal MAMs

(A) Gating strategy for CD11b⁺F4/80⁺Gr1⁻CD45^high macrophages. An example of dural lesions is shown. (B) Quantification of CD11b⁺F4/80⁺Gr1⁻CD45^high MAMs within dural 4T1-Dura3 and parenchymal 4T1-Par3 cancer lesions. (C) Representative histograms showing MFI for the expression of M1 (iNOS, MHCII) and M2 (Arginase-1 (Arg-1), CD206) macrophage markers in dural 4T1-Dura3 (black) and parenchymal 4T1-Par3 lesions (red) in intracarotid artery model. (D) Quantification of MFI for iNOS, MHCII, CD11c, Arg-1, CD206, TNFα and IFNγ; n = 9/11 for dural/parenchymal lesions, respectively. Statistical significance in B and D was determined using one-tailed Student's T-test with unequal variance (p ≤ 0.05). Error bars represent SD.

Figure 5. Cancer cell-derived factors contribute to site-specific polarization of parenchymal MAMs

(A) Expression of macrophage polarization markers in MAMs isolated from the 4T1-Par3 and 4T1 parent-derived cancer lesions established within brain parenchyma after direct intracranial implantation of cancer cells; n = 5. (B and C) Expression of macrophage polarization markers in MAMs isolated from parenchymal (B) and dural (C) brain metastases established from the 4T1-Dura3-LTβ and 4T1-Dura3 control cell lines; n = 7. Statistical significance in A–C was determined using one-tailed Student's T-test with unequal variance (p ≤ 0.05). Error bars represent SD.

Figure 6. Proposed model for site-specific regulation of MAM phenotypes in intracranial metastases

Metastatic site-specific stroma is likely implicated in the establishment of site-specific molecular profiles in cancer cells, which then co-determine MAM phenotypes. The top 3 up-regulated cytokines/chemokines in cancer cells growing at the dura and within brain parenchyma, as well as differences in macrophage polarization markers between the two sites are indicated. Notably, we demonstrated a functional link between LTβ and MAM polarization within the brain parenchyma, while the functions of other cytokines remain to be determined. Based on our findings and evidence from the literature, we propose that MAM activation state is co-determined by the metastatic site-specific cancer cell characteristics (e.g. cytokines), the organ-specific stroma, and by the origin of macrophages (e.g. Gr1⁺Ly6C^high inflammatory monocytes versus Ly6C^low resident monocytes [20]).