Aging and CNS Myeloid Cell Depletion Attenuate Breast Cancer Brain Metastasis

Cell lines

Stock vials of cells were tested to be free of infectious microbes (e.g., mycoplasma) by PCR (Animal Diagnostic Laboratory Services, NCI). Cells for injections were used within three passages after thawing or within 15 passages for in vitro experiments. Human and mouse cell lines were authenticated by short tandem repeat analysis, which was performed by Laragen Inc. (October 2016) and ATCC (September/October 2020), respectively.

MDA-MB-231BR-eGFP (231-BR) and 4T1-BR cells were cultured in DMEM supplemented with 10% FBS. MDA-MB-231T, a subline of MDA-MB-231 breast cancer cells were obtained from Dr. Danny Welch (University of Kansas Medical Center). 4T1-Luc2 were purchased from PerkinElmer and maintained in DMEM supplemented with 10% FBS. E0771, 6DT1, and MVT1 cells were cultured as described previously (11). The brain-selected E0771 subline was generated by first injecting 1 × 10⁵ parental E0771 into the left cardiac ventricle of 2-month-old female mice. Cells recovered from the brains of these mice were re-injected into a new cohort of mice. This was repeated five times to generate E0771-BR5 cells. 99LN-BrM cells that were derived from one round of in vivo selection to the brain was generously provided by Drs. Johanna Joyce and Florian Klemm (University of Lausanne). These cells were maintained in DMEM/F12 supplemented with 10% FBS and were selected in vivo three more times to generate 99LN-BrM4 cells. EOC2 cells were a gift from Dr. Jing Wu (Neuro-Oncology Branch, NCI) and cultured in DMEM supplemented with 10% FBS and 20% LADMAC conditioned media.

Animals

All animal experiments were approved by the Institutional Animal Care and Use Committees of the NCI (NCI-Bethesda and NCI-Frederick campuses), National Institute of Neurological Disorders and Stroke, Texas Tech University of Health Sciences Center, and Use Review Office of the United States Army Medical Research and Materiel Command. Unless specified otherwise, for aging studies in immunocompetent mice (BALB/C, C57BL/6, and FVB), young female mice were 4 to 6 months old and older female mice were 18 to 24 months old at the time of cancer cell injection. The older cohorts were obtained as retired female breeders and housed to the desired age. The young cohorts underwent one full cycle of pregnancy-lactation and were used >1 month after weaning of pups. Athymic nu/nu virgin mice were used at 2 to 3 months (young) and 12 to 15 months (older). Two-month-old female virgin mice were used for baseline immunophenotyping studies comparing naïve/healthy brains to metastatic brains and for immunodepletion experiments.

Brain and lung metastasis models

For brain metastasis models, cancer cells suspended in 100 μL sterile PBS were injected with aid of ultrasound into the left cardiac ventricle of mice under isoflurane anesthesia. 4T1-BR (5 × 10⁴) and 231-BR (1.75 × 10⁵) cells were injected into BALB/C and nu/nu mice, respectively. E0771-BR5 (1 × 10⁵ or 2 × 10⁵) and 99LN-BrM4 (5 × 10⁵) were injected into C57BL/6 mice. For lung metastasis models, MDA-MB-231T (7.5 × 10⁵) were suspended in 100 mL sterile PBS and injected into the tail-vein of nu/nu mice. 4T1-Luc2 (5 × 10⁵), E0771 (1 × 10⁵), MVT1 (2 × 10⁵), and 6DT1 (3.5 × 10⁴) cells were suspended in 50 mL sterile PBS and injected orthotopically into the #4 mammary fat pad (MFP) of syngeneic mice as described previously (11). For the 6DT1 model, tumors were resected at day 14. Primary tumor size was monitored twice a week and weighed after tissue collection.

Brain metastatic tumor burden was quantified from five (4T1-BR, MDA-MB-231BR, 99LN-BrM4) or ten (E0771-BR5) 8-μm-thick hematoxylin and eosin (H&E)-stained step-sections spaced 300 to 500 μm apart through the whole brain or one hemisphere. A metastatic cluster was defined as a group of four or more closely spaced metastatic deposits or one large lesion more than 300 μm in both dimensions. Total numbers of lung or liver metastases were quantified from one representative 5-μm-thick H&E-stained tissue section with all lung lobes or liver lobes visible.

Flow cytometry

Mice anaesthetized with isoflurane or heavily sedated with a terminal dose of chloral hydrate were transcardially perfused with saline. Single cell preparations were blocked with anti-CD16/32 and mouse IgG on ice for 10 minutes and then stained with a cocktail of fluorescently labelled antibodies for 30 minutes on ice. All samples were analyzed on a BD LSRFortessa or BD FACSymphony cell analyzer. Cell counts were determined using CountBright Absolute Counting Beads (Thermo Fischer Scientific, Catalog No. C36950). Data were analyzed using FlowJo version 10 (see Supplementary Materials and Methods).

Immune-depletion

To deplete CD4⁺ or CD8⁺ T cells, BALB/C were injected intraperitoneally with 250 μg of depleting antibodies on day -2, and then maintenance doses of 100 μg at day 2, 6, and 10. To deplete monocytic and granulocytic myeloid cells, BALB/C mice were injected intraperitoneally with 250 μg of anti-GR1 antibodies on day -1 and -2, followed by 100 μg doses at day 5 and 8. Control mice were injected with isotype rat IgG2b. See Supplementary Materials and Methods for antibody identifications. Mouse 4T1-BR breast cancer cells were injected into the left cardiac ventricle on day 0.

To deplete CNS myeloid cells in 2-month-old mice, mice were given PLX3397 (SelleckChem, Catalog No. S7818) formulated at 300 mg/kg in NIH-31 Open Diet (Envigo) ad libitum for 28 days prior to the injection of cancer cells. Mice remained on this diet until study termination. Control mice were given standard NIH-31 Open diet. To deplete resident CNS myeloid cells in a young versus older experiment, mice were given PLX3397 at 100 mg/kg body weight by daily oral gavage for 14 days prior to injection of cancer cells and then continuously thereafter (see Supplementary Materials and Methods). Control mice were given vehicle with an equivalent dose of DMSO.

Immunofluorescence staining of tissue sections

See Supplementary Materials and Methods for tissue acquisition, antibodies used, and image analysis. Frozen brain tissue sections (8- to 10-μm thick) were fixed or permeabilized with methanol for 5 minutes at −20°C, washed with PBS, blocked with 5% normal goat serum (Vector Labs) or 5% normal donkey serum (Jackson ImmunoResearch, Catalog No. 017–000–121) in PBS with 0.05% (v/v) Tween-20 (PBST), and then incubated overnight at 4°C with primary antibodies diluted in the appropriate blocking buffer. After washing slides with PBS, tissue sections were incubated with DAPI (2 mg/mL) and secondary antibodies diluted to 1:250 to 1:500 in 5% serum/PBST for 1 hour at room temperature. See Supplementary Materials and Methods for a list of primary and secondary antibodies used. Slides were coverslipped with Fluorescence Mounting Medium (Dako). Images of metastatic lesions and of uninvolved regions were taken using a Zeiss Axioskop with a 10×, 20×, or 40× objective. Metastatic lesions were identified on the basis of cytokeratin staining or dense clusters of DAPI. Uninvolved regions were defined as areas within the cerebral cortex without metastatic lesions and located at least 1 mm away from a lesion. Images of at least three distinct visual views of uninvolved brain regions were captured for each single tissue section.

In vitro assays

Boyden chamber motility and cell viability assays are described in Supplementary Materials and Methods.

Statistical analysis

Statistical tests were performed using GraphPad Prism 7, with P values < 0.05 considered significant. The Mann–Whitney U test and Wilcoxon rank test was used for comparing unpaired or pair-matched data, respectively. The Kruskal–Wallis test followed by Dunn test was used for multiple comparisons. All data presented, except for frequency data, are box-whisker plots or bar graphs depicting median ± interquartile range.

Refer to Supplementary Materials and Methods for blood–tumor barrier permeability assays, intracranial tumor model, microglia conditioned medium experiment, detailed methods of flow cytometry and immunofluorescence staining protocols including full list of antibodies used, tissue collection and preservation protocols, image analysis, RNA in situ hybridization (RNAscope), and methods used to analyze publicly available datasets.

Breast tumors from young and older patients have a similar mutational spectrum

In further support of previously published findings that primary breast cancers from young patients have few if any defining genetic alterations (12), we compared the mutational status of primary breast tumors from young (≤40 years) and older patients (>40 years) in two online datasets, TCGA (n = 1,083) and METABRIC (n = 1,980; Supplementary Figs. S1 and S2). Approximately 8% and 6% of patients in the TCGA and METABRIC databases were young, respectively. No consistent mutations were observed between the two datasets, which may be partly due to differences in patient characteristics, for instance a higher proportion of younger women have TNBC in the METABRIC dataset compared with TCGA (Supplementary Fig. S1C). Similar conclusions were obtained when varying the age cutoff (Supplementary Figs. S3–S6). Given the lack of somatic mutations consistently enriched in young tumors, we hypothesized instead that extrinsic microenvironmental factors contribute to age-related differences in breast cancer aggressiveness.

Young age promotes breast cancer metastasis specifically in the brain

Previous epidemiologic studies have demonstrated a strong association between young age at breast cancer diagnosis and risk of brain metastasis (13–15). To test whether the younger brain is more permissive for breast cancer metastasis, brain-tropic breast cancer cell lines (-BR) were injected into young and older mice side-by-side via the left cardiac ventricle and metastatic clusters were enumerated. The definition of young and older mice varied by strain as immunodeficient mice do not survive as long (see Materials and Methods). In general, ages of the young (2–6 months) and older (>12 months) cohorts were chosen to model human age less than or greater than 40 years old, respectively. Nulliparous versus parous status was held constant within each experiment.

For the triple-negative subtype 4T1 breast cancer model, young parous BALB/C mice developed 4.5-fold more 4T1-BR brain metastatic clusters (Fig. 1A). Young parous C57BL/6 mice similarly developed more brain metastatic clusters when injected with E0771-BR5 cells (Fig. 1B), a subline of the mouse triple-negative E0771 breast cancer cells that we selected in vivo for increased capacity to form metastases (Supplementary Fig. S7A–S7C). Young immunocompromised nulliparous nude mice injected with human MDA-MB-231-BR triple-negative breast cancer cells (231-BR) developed 2.3-fold more metastatic clusters (P < 0.05) compared with older mice (Fig. 1C). To test whether this phenomenon was restricted to the triple-negative subtype, a published 99LN-BrM luminal B model system (16) was subjected to three additional rounds of in vivo selection to produce 99LN-BrM4 cells with increased brain metastatic efficiency (Supplementary Fig. S7D and S7E). Young parous C57BL/6 mice injected with 99LN-BrM4 cells developed 3.5-fold more brain metastatic clusters compared with their older counterparts (Fig. 1D, top). Collectively, data from four models encompassing two molecular subtypes of breast cancer indicate that young age promotes breast cancer metastasis in the brain, or that older age protects against it.

We also determined the effect of age on metastasis to other sites. Many mice injected with 99LN-BrM4 cells simultaneously developed liver metastases but, unlike the brain, there was a trend towards reduced incidence and number of liver metastases in younger mice (Fig. 1D, bottom). Liver metastases were not reproducibly detected in the other brain-tropic model systems.

Breast cancer lung metastasis models were tested, including four spontaneous metastasis assays involving mammary fat pad injection with (Fig. 2A and B) or without (Fig. 2C and D) tumor resection, and one experimental metastasis assay where cells were injected into the tail vein (Fig. 2E). In most cases, lung metastasis did not vary by age, with the exception of 4T1-Luc2 where metastases were higher in older animals.

The young brain confers a survival advantage to breast cancer cells

We next ruled out several potential underlying causes of the increased brain metastasis in young mice. Breast cancer cells must traverse the blood–brain barrier (BBB) to establish a hematogenous metastasis. At the experimental endpoint of a 4T1-BR metastasis assay, the permeability of the blood–tumor barrier, the remnants of the BBB once a metastasis has established, was comparable between young and older mice (Supplementary Fig. S8A). To specifically test whether the effect of age is exerted on tumor cells at the extravasation step, 4T1-BR cells were directly injected into the brains of young and older BALB/C mice to bypass the BBB. At 8 days postinjection, young mice injected intracranially with cancer cells still developed larger tumors compared with older mice (Supplementary Fig. S8B).

High levels of estrogen can promote growth of triple-negative breast cancer cells in brain by inducing reactive glial fibrillary acidic protein (GFAP)-expressing astrocytes in the neuroinflammatory response to secrete metastasis-promoting factors (17). Serum samples from young and older mice with 4T1-BR and 231-BR brain metastases had comparable corticosterone and testosterone levels; no consistent differences in estradiol and progesterone concentrations were observed with age, possibly reflecting the fact that, unlike humans, mice do not undergo a true menopause (Supplementary Fig. S8C and S8D; ref. 18). Consistent with this finding, the astrocytic neuroinflammatory response of mice harboring 4T1-BR or 231-BR metastases did not differ with age (Supplementary Fig. S9A and S9B). In other experiments, tumor cell Ki67 and cleaved caspase 3 measurements of proliferation and apoptosis, respectively, were comparable between age groups (Supplementary Fig. S9C–S9F). These data suggest that initiation of metastatic colonization may vary with age, but once cancer cells have adapted to grow in either the young or older brain, they are equally able to thrive.

Brains with metastases have elevated numbers of lymphoid and myeloid cells

Aging is accompanied by prominent alterations to the immune system (19). We next examined whether immune cells contribute to this age effect on brain metastasis. In preliminary experiments, young mice injected with either 4T1-BR or E0771-BR5 brain colonizing cells were compared with sham injections (Supplementary Figs. S10–S13). The normal brain had few intravasated immune cells in agreement with the literature (20). Multiple alterations were observed in BALB/C brains with 4T1-BR metastases including significant increases in CD4⁺ and CD8⁺ T cells, Ly6G⁺ neutrophils, and Ly6C⁺F4/80⁻ monocytes. There was also an increase in regulatory T cells (Treg). No changes in numbers of Ly6C⁺F4/80⁺ macrophages and resident CNS myeloid cell numbers were observed as a result of metastasis. Similar trends were also observed in a second immunocompetent model, E0771-BR5, with the addition of a significant increase in macrophages and CNS resident myeloid cells. Immune cell invasion trends into the metastatic brain differed quantitatively from blood.

Young mice have more CNS myeloid cells compared with older mice

Aging in the CNS is accompanied by alterations to the brain immune landscape (21) and thus a changed immune contexture stood as a candidate mechanism to affect brain metastatic aggressiveness. We examined the effect of age on brain immune profile, both naïve and metastatic, in the 4T1-BR and 231-BR models (Fig. 3). CD4⁺ and CD8⁺ T cells did not significantly differ in metastatic mouse brains with age. CD3⁺ T cells that were FOXP3⁺ (Tregs) were scarce in metastatic lesions, with 17/27 and 21/26 4T1-BR metastatic clusters negative in young and older mice, respectively. For the Treg⁺ lesions, the number of Tregs was unchanged with age when normalized for lesion size (P = 0.2863; Supplementary Fig. S14).

Microglia, the resident myeloid immune cells of the brain, constituted the majority of brain immune cells in all groups. Young brains with 4T1-BR metastases had 1.5-fold more brain resident CNS myeloid cells compared with older brains with metastases (Fig. 3A). Microglia were accompanied by infiltrating blood-borne monocytes and macrophages, also higher in young brains.

To determine whether these age-related differences were inherent or induced by the brain metastases, brains from naïve young and older BALB/C mice were compared. Naïve brains from young BALB/C mice also contained 2-fold more CNS resident microglia compared with that in older BALB/C mice (Fig. 3B), indicating an inherent difference in myeloid numbers with age. No significant increase in infiltrating macrophage or monocyte populations were observed. Analysis of the 231-BR model system confirmed significantly increased microglia in young naïve and metastatic brains, with inconsistent trends in infiltrating myeloid cells (Fig. 3C and D). Increased microglia were also observed in naïve young nude mice when standard FACS gating or microglial-specific Tmem119 was used (Fig. 3D and E). The majority of immune cells in blood did not significantly differ with age in either model system (Supplementary Fig. S15). The data identify a prominent age-associated change in the resident CNS immune repertoire that could underlie differences in metastatic propensity.

CNS myeloid cells promote brain metastasis in young mice

Although the data nominated resident CNS myeloid cells as the most plausible age-associated determinants of brain metastatic ability, it remained possible that more subtle differences in other immune components actively recruited to the metastatic brain could also contribute. To eliminate this possibility, peripherally derived CD4⁺ T cells, CD8⁺ T cells, and GR1⁺ neutrophils and monocytes were depleted from young BALB/C mice by extended treatment with depleting antibodies. Depletion of these immune subsets at experimental termination was confirmed by flow cytometry (Supplementary Fig. S16). Loss of CD4⁺ T cells (Fig. 4A), CD8⁺ T cells (Fig. 4B), or GR1⁺ neutrophils and monocytes (Fig. 4C) did not significantly affect 4T1-BR metastasis burden in the young brain.

Resident microglia depend on colony-stimulating factor 1 receptor (CSF-1R) signaling for survival (22, 23). RNAscope analysis detected mRNA for CSF-1 or IL34, two ligands for CSF-1R, in some CNS myeloid cells, other cells of the brain microenvironment, and 4T1-BR cancer cells (Supplementary Figs. S17 and S18). To study whether disruption ofCSF-1R signaling, and thus survival of CNS myeloid cells, affected metastasis formation, BALB/C mice were fed chow containing the CSF-1R inhibitor PLX3397 to deplete myeloid cells. CNS myeloid cell depletion resulted in a highly significant 2.1-fold (P = 0.0001) reduction in brain metastatic burden compared with mice given control diet (Fig. 4D).

Additional markers were investigated to determine the effects of PLX3397 on brain myeloid subpopulations, in both the uninvolved brain away from metastases and the neuro-inflammatory response surrounding them (Fig. 4E–I). Iba1 is a dual microglia/macrophage marker that increases in staining intensity on cellular activation (24, 25). Iba1⁺ cells were elevated in the metastatic neuroinflammatory response as compared with the uninvolved brain (Fig. 4F). PLX3397 reduced the number of Iba1⁺ cells in the uninvolved brain and metastatic microenvironment by 33% and 31% over control, respectively, with no effect on Iba1 intensity, the latter indicative of activation (Fig. 4G). These Iba1⁺ metastasis-associated myeloid cells also expressed comparable intensities of the CD68 phagocytic activation marker, with or without PLX3397. The P2y12 marker binds a purinergic receptor that is specific for microglia and loses staining intensity upon chemotactic activation (26, 27). P2y12⁺ microglia were present in higher numbers in the metastatic neuroinflammatory response than in the uninvolved brain and were depleted in both compartments by PLX3397 (Fig. 4H). Both Iba1 and CD68 intensity in P2y12⁺ microglia, indicative of activation, were elevated in the metastatic microenvironment but unaffected by PLX3397 (Fig. 4I) confirming a depletion of microglia without altering activation status. Infiltrating border associated macrophages, identified by CD206 staining (28), were primarily present in the leptomeninges and covered a median of 1.4% of brain metastatic capillaries (Supplementary Fig. S19A–S19C). To the extent that they were present, CD206⁺ cells were reduced by PLX3397 (Supplementary Fig. S19B and S19C). In summary, treatment with the CSF-1R inhibitor PLX3397 demonstrated that resident CNS myeloid cells functionally support or promote brain metastasis in young mice, and that the drug works by reducing resident CNS myeloid cell numbers without affecting their activation status.

The CSF-1R inhibitor PLX3397 implicates CNS myeloid cells in metastasis promotion in young and older mice

Microglia undergo significant alterations with age (29). Two hypotheses could explain the contribution of CNS myeloid cells to age-related changes in brain metastasis: either young cells promoted, or older cells inhibited brain metastasis. Young and older BALB/C mice were given PLX3397 to deplete resident CNS myeloid cells (Fig. 5A). As previously, immunostaining for Iba1 (macrophage or microglia), P2y12 (microglia), or CD206 (border associated macrophage) markers showed substantial (>60%) loss of CNS myeloid cells following treatment with PLX3397 (Fig. 5B and C; Supplementary Fig. S19D–S19F). Young mice treated with PLX3397 showed a strong trend towards reduced numbers of brain metastatic burden (Fig. 5D), consistent with our earlier results (Fig. 4D). Few brain metastases were observed in older mice, but there was also a trend towards reduction in metastatic burden. Using the median number of metastatic clusters in a combined dataset of untreated BALB/C mice to establish an age-specific cut-off for high vs. low metastatic tumor burden status (Supplementary Fig. S20), 30.8% of young animals had a high brain metastatic burden in the vehicle arm. This percentage was reduced to zero in the PLX3397 arm (Fig. 5E; P = 0.035). No statistically significant effect was observed in older animals (P = 0.25). The data demonstrate that CNS myeloid cells promote brain metastasis at all ages, but that this effect is more pronounced in young animals.

CNS myeloid cells promote distinct age-related alterations in the uninvolved brain and metastatic microenvironment

The myeloid contribution to brain metastasis is likely multifactorial. Younger mouse brains contain greater numbers of resident myeloid/microglial cells which could contribute to the age-dependent difference in metastatic burden; myeloid cells could also vary in function with age. In addition, distinct age-related differences might be seen in the normal, uninvolved brain versus the metastatic neuroinflammatory response.

Known functions of CNS resident myeloid cells were investigated in both the uninvolved brain and the metastasis-associated neuroinflammatory response. Traditional markers of activation and resting state did not vary with CSF-1R inhibition in either location (Supplementary Fig. S21). Microglia are known to interact with and activate CSF-1R-negative astrocytes to either promote or inhibit a variety of brain pathologies (30, 31). Astrocytic activation has also been reported with normal brain aging (32), as was observed in our mouse cohorts (Fig. 5F). PLX3397 depletion increased astrocyte activation (GFAP positivity) in the uninvolved brain of young mice to levels observed in older mice (Fig. 5F; Supplementary Fig. S22A). PLX3397 did not affect the astrocytic neuroinflammatory response in either age group (Supplementary Fig. S22B), suggesting that any age-dependent differences in astrogliosis could only affect metastasis at the very earliest steps of colonization before development of a neuroinflammatory response.

We then asked whether CNS myeloid cells augmented metastatic colonization at later stages when a neuroinflammatory response was apparent. Conditioned medium from cultured EOC2 microglial cells was capable of stimulating proliferation of 4T1-BR tumors cells (Supplementary Fig. S23), suggesting that a component of the microglial secretome might promote metastasis. We therefore queried published RNA sequencing (RNA-seq) data on Ribotag-isolated microglia from young (3 months) and older (24 months) mice (29), with particular attention to secreted factors that could mediate myeloid–tumor crosstalk, and found that Semaphorin 3A (Sema3a; Log2FC 0.271, FDR = 0.0241) was significantly upregulated in younger mice. Sema3a is a member of semaphorin family that provides a repellant cue in axonal growth guidance during development. It binds the Plexin-A1/neuropilin1 co-receptors and is produced by microglia and other brain cells (33). Sema3a has been linked to diverse CNS diseasessuch astraumatic brain injury (34) and multiple sclerosis(33). It also contributes to aspects of cancer progression (35, 36). Similar to microglial conditioned medium, recombinant Sema3a stimulated tumor cell proliferation and motility in vitro (Fig. 5G and H). Immunofluorescence confirmed that Sema3a was produced by Iba1⁺ myeloid cells in vivo (Fig. 5I). Importantly, expression of Sema3a in the neuroinflammatory response of young metastatic brains showed a strong trend toward higher expression than in older brains (Fig. 5J and K). This effect of age was not observed in uninvolved brain (Supplementary Fig. S24). Not all myeloid products showed the same expression trend with age (Supplementary Fig. S25), demonstrating specificity in this pathway. Thus, factors secreted from young CNS myeloid cells in the developing tumor microenvironment change with age and can influence aspects of metastatic colonization.