Synaptic proximity enables NMDAR signalling to promote brain metastasis

Abstract

Metastasis-the disseminated growth of tumours in distant organs-underlies cancer mortality. Breast-to-brain metastasis (B2BM) is a common and disruptive form of cancer and is prevalent in the aggressive basal-like subtype, but is also found at varying frequencies in all cancer subtypes. Previous studies revealed parameters of breast cancer metastasis to the brain, but its preference for this site remains an enigma. Here we show that B2BM cells co-opt a neuronal signalling pathway that was recently implicated in invasive tumour growth, involving activation by glutamate ligands of N-methyl-D-aspartate receptors (NMDARs), which is key in model systems for metastatic colonization of the brain and is associated with poor prognosis. Whereas NMDAR activation is autocrine in some primary tumour types, human and mouse B2BM cells express receptors but secrete insufficient glutamate to induce signalling, which is instead achieved by the formation of pseudo-tripartite synapses between cancer cells and glutamatergic neurons, presenting a rationale for brain metastasis.

Extended Data Figure 1. Association of the NMDAR signaling pathway with human breast cancer.

a, mRNA expression of 6 NMDAR-associated genes (GRIN2A, GRIN2B, GRIN2C, GRIN2D, GRIN1 and DLGAP1) in 1,100 breast cancer (BC) patients from The Cancer Genome Atlas (TCGA). Samples were sorted by NMDAR score (lower panel), which combines, for each sample, the Z-scores of normalized expression values of all 5 genes. b, mRNA expression of 4 AMPAR-associated genes (GRIA1, GRIA2, GRIA4 and GRIA3) in 1,100 breast cancer (BC) patients from TCGA. Samples were sorted by AMPAR score (lower panel), which combines, for each sample, the Z-scores of normalized expression values of all 4 genes. c, mRNA expression of 5 Kainate-associated genes (GRIK2, GRIK4, GRIK3, GRIK1 and GRIK5) in 1,100 breast cancer (BC) patients from TCGA. Samples were sorted by Kainate score (lower panel), which combines, for each sample, the Z-scores of normalized expression values of all 5 genes. d, mRNA expression of 8 Metabotropic glutamate-associated genes (GRM8, GRM3, GRM4, GRM7, GRM2, GRM6 and GRM5) in 1,100 breast cancer (BC) patients from TCGA. Samples were sorted by Metabotropic score (lower panel), which combines, for each sample, the Z-scores of normalized expression values of all 5 genes. e, Boxplot comparison of mRNA expression of GRIN2A, GRIN2B, GRIN2C, GRIN2D, DLGAP1 and GRIN1 (from left to right) across the five breast cancer PAM50 subtypes. P-values were computed using a two-sided ANOVA test. f, Comparison between NMDAR score (X-axis) and a previously derived NMDAR expression signature (Y-axis) across all breast cancer samples. Points are color coded based on points concentration (warm color = high concentration, cold colors = low concentration), n = 1100. g, Boxplot comparison of scores from the previously derived NMDAR expression signature across the five PAM50 breast cancer subtypes. h, Kaplan-Meier plots comparing distant-relapse free survival between samples with high GRIN2B expression (top 25%, red line) versus remaining cases (blue line) in Luminal B, Her2-enriched, Luminal B, Normal-like, and Basal-like breast cancer subtypes (from left to right). Log-rank tests are two-sided. In all comparisons, the number of samples for each breast cancer subtype (panels a-e and g) are: Luminal A (n = 467), Luminal B (n = 198), Her2-enriched (n = 72), Normal-like (n = 28), Basal-like (n = 155). The thick central line of each box plot represents the median number of significant motifs, the bounding box corresponds to the 25th–75thpercentiles, and the whiskers extend up to 1.5 times the interquartile range.

Extended Data Figure 2. GluN2B-NMDAR signaling is activated in brain metastasis.

a, Representative H&E stained images of two paired human primary breast cancers (primary) and brain metastases (brain mets) are shown. Scale bar, 100 μm. Two independent experiments. b, Quantification of pGluN2B Y1472 staining in 16 primary breast tumors and 18 brain metastases, including 9 matched primary cancers and brain mets, along with 7 unmatched primary tumors and 9 unmatched brain mets. Mean ± s.e.m. Mann-Whitney test (two-tailed) was used. Subtype categorization for those patients are: 4 TNBC, 6 Her2+, 2 luminal and 13 unknown. c-e, Representative images (c) and quantification (d) of IHC staining of pGluN2B Y1252 in 12 paired human primary breast cancers (primary) and brain metastases (brain mets). Mean ± s.e.m. Wilcoxon test analysis (two-tailed). Subtype categorization for these twelve patients: 1 TNBC, 4 HER2+, 2 luminal and 5 unknown. (e), Quantification of pGluN2B Y1252 staining in 16 primary breast tumors and 22 brain metastases, including 12 matched primary cancers and brain mets, along with 4 unmatched primary tumors and 8 unmatched brain mets. Mean ± s.e.m. The Mann-Whitney (two-tailed) statistical test was used. Subtype categorization for those patients: 4 TNBC, 6 Her2+, 2 luminal and 13 unknown. Scale bar, 100 μm. f-h, IHC staining of GluN2B in 16 human primary breast cancers and 23 brain metastases. Representative images (f), Wilcoxon test analysis (two-tailed) of the 15 matched primary cancers and brain mets (g), and Mann-Whitney (two-tailed) test analysis (h) of all samples. Mean ± s.e.m. (It remains to be ascertained whether the few brain mets lacking GluN2B expression have activated NMDAR signaling via another GluN2 subunit.) Scale bar, 100 μm. Subtype categorization for those patients: 4 TNBC, 6 Her2+, 2 luminal and 13 unknown. i-k, IHC staining of GluN1 in 16 human primary breast cancers and 16 brain metastases. Representative images (i), Wilcoxon test (two-tailed) analysis of the 11 matched primary cancers and brain mets, (j) and Mann-Whitney test (two-tailed) analysis (k) of all samples. Mean ± s.e.m. Scale bar, 100 μm. Note that due to variations in the presence of cancer in the sections taken from the TMA block for analysis, not all 25 samples were represented in each of the immuno-stainings performed in Figure 2 and in this figure. Subtype categorization for those patients: 4 TNBC, 6 Her2⁺, 2 luminal and 13 unknown. l, Representative IHC staining of pGluN2B Y1472, total GluN2B, and luciferase in primary tumours, brain mets, and lung mets formed by MDA231-B2BM cells (day 28~35 after injection). Scale bar, 100 μm. The areas indicated by black boxes are shown in Figure 2e. m, n, IHC staining of pGluN2B Y1252, total GluN2B, and luciferase (m) and quantification of pGluN2B Y1252 staining (n) in primary tumors, brain mets, and lung mets formed by MDA231-BrM cells (day 28~35 after injection). Scale bar, 100 μm. Insets are magnified 3.2-times relative to the main field. Two-tailed Student t-test, mean ± s.e.m., n = 3 mice per group.

Extended Data Figure 3. NMDAR-mediated calcium signaling and single channel currents in BrM cells.

To test for functional NMDARs, fluorescence imaging and patch-clamp recordings were carried out in 2D cultures of TS1-BrM cells, 14 view-fields (301 X 301 μm) from 10 different culture dishes. Total of 151 responding cells out of 1,070 imaged cells = 14.11%. Data from bath addition of 50-100 uL of 100 mM KGlut, 50-200 uL of 10 mM NMDA, or local perfusion-pipette puffing of 100 uM K Glutamate, or of 500 uM NMDA, all of which produced calcium responses. a-c, Representative calcium imaging experiments. a, TS1BrM cells were loaded with Oregon Green BAPTA-1 AM fluorescent calcium indicator, and imaged with a sCMOS camera at 5 frames per second. A subset of cells in the field of view (13/75) showed calcium elevation in response to addition of glutamate to the bath (50 μL of 10 mM glutamate, final concentration 125 μM). b, Calcium signals from 4 cells as indicated in (a), with glutamate application at the arrow. c, Average signal over the 13 agonist-responsive cells during application of glutamate (blue trace) or preceding control application of the same volume of Ringer solution (red trace). d, Response to application of 50 uL of 10 mM NMDA (final concentration 125 μM), averaged over 6 responding cells within a field of view containing 30 cells (blue trace), compared to an example response averaged over 52 cells to the same amount of NMDA, but in the presence of 100 μM APV and 2 mM Mg²⁺ (red trace). Overall, calcium transients stimulated either by NMDA or glutamate were detectable in about 14% of cells (151 of 1,070 cells, from 10 culture dishes). e-g, Puff application (pipette at left) of NMDA (200 μm, gray bar) activates inward single-channel currents recorded through the pipette at right (e, f), with amplitudes of ≈4.5 pA at a holding potential of -90 mV (g), equivalent to a chord conductance of 50 pS, characteristic of GluN2A/GluN2B-containing NMDARs. 4.5 pA single-channel currents were detectable in low-noise whole-cell recordings on application of glutamate (100 μM) or NMDA (200 μM) in 45% (9/20) of recorded cells. h, Resting membrane potentials were measured with whole-cell current-clamp using pipettes filled with cytoplasm-like high-potassium solutions (box plot, -52 ± 10.6 mV (mean ± SEM, n = 9 cells). For comparison, the voltage-dependence of GluN2A/2B inward current in physiological (1 mM) magnesium is superimposed^,, indicating that membrane potentials are sufficiently depolarized to overcome a substantial fraction of voltage-dependent magnesium block of NMDARs.

Extended Data Figure 4. Brain mets are not inducing the neuronal apoptosis anticipated for autocrine secretion of glutamate by cancer cells.

a, WB analysis of Western blotting of the xCT and vGlut2 glutamate transporters in pairs of human and mouse parental breast cell lines and their corresponding brain metastatic derivatives. Protein levels in BrM cells are quantitated relative to cognate parental cell, after normalization to GAPDH. Three independent experiments. GAPDH was run on separate gels from xCT and vGlut2 as sample processing controls. b, Immunohistochemical-(DAB)-based staining to detect apoptotic cells in mouse brains with metastases formed by MDA231-BrM cells (day 28~35 after injection). DNAase I treatment on brain tissue sections was used as a TUNEL⁺ control. Images shown are representative of an analysis of >20 brain mets from 4 mouse brains, 2 sections/mouse brain. Scale bar, 10 μm. c, Immunofluorescent staining for cleaved-caspase 3 (green), luciferase (cancer cells; magenta) and NeuN (neurons; red) in mouse brain with metastases formed by MDA231-BrM cells (day 28~35 after injection). Images shown are representative of an analysis of >20 brain mets from 4 mouse brains, 2 sections/mouse brain. Scale bar, 10 μm. d, Immunofluorescence staining for TUNEL (green), NeuN (red), GFAP (glioma cells; magenta) in a brain harboring a glioma that arose in a genetically engineered mouse model. Images shown are representative of an analysis of > 6 gliomas from 6 mouse brains, 1 section/mouse brain. As previously reported, glioma cells secrete glutamate that causes neuronal apoptosis, in contrast to BrM cells, which do not. Scale bar, 50 μm.

Extended Data Figure 5. Interactions between BrM cells and neurons.

a-c, Immunofluorescence staining of luciferase (B2BM; blue), vGlut2 (red) and pGluN2B 1252 (green) in a mouse brain metastasis formed by MDA231-BrM cells (day 28~35 after injection), imaged by STED super-resolution microscopy. (a, b) Tilted 3D images of the area circumscribed by the white solid border box in Fig. 4a. Arrow, vGlut2⁺ in presynaptic neuron; arrowhead, pGluN2B Y1252⁺ in a blue-labelled B2BM cell, potentially in close apposition to a postsynaptic neuron. (c) Tilted 3D image of the area circumscribed by the white dotted border box in Fig. 4a depicting a normal synapse revealed by close association between vGlut2⁺ puncta in a presynaptic neuron (arrow) with pGluN2B⁺ puncta in a postsynaptic neuron (arrowhead). Side length for each square in 3D view, 400 nm (a-b) and 300 nm (c). Images shown are representative of an analysis of 9 brain mets from 3 mouse brains, 2 sections/mouse brain. See Supplemental Videos further describing the images shown in a-c. d, e, Immunofluorescence staining of luciferase (green) and synaptobrevin 1 (red) in a mouse brain metastasis formed by MDA231-BrM cells (day 28~35 after injection), imaged by STED super-resolution microscopy. Shown in part d are a merged image (left panel) and individual fluorescent images (middle and left panels), scale bar, 10 μm. Part e shows a tilted 3D image at lower and higher magnifications, highlighting a synaptobrevin 1⁺ bouton chain (red) localized proximal to luciferase⁺ tumor cells (green). Images shown are representative of an analysis of nine brain mets from three mouse brains, 2 sections/mouse brain. Side length for each square in 3D view, 10 μm. f,g, Immunofluorescence staining for post-synaptic proteins PSD-95 (red, f), and Neuroligin 2 (red, g) along with luciferase (green) in mouse brains with metastases formed by MDA231-BrM cells (day 28~35 after injection). Scale bar, 10 μm. The area circumscribed by the white box is shown in the leftmost column. Scale bar, 1 μm. Images shown are representative of an analysis of >20 brain mets from 4 mouse brains, 2 sections/mouse brain.

Extended Data Figure 6. NMDAR signaling in BrM cells is activated in co-culture with neurons.

a, Schematic of cancer cell - primary neuron co-culture system and experimental design. b, Immunofluorescence staining for MAP2 (neuronal marker, green) and GFAP (astrocyte marker, red) in primary cultures of cortical neurons, revealing the preponderance of neurons. Scale bar, 10 μm, except for the leftmost panel, in which the area circumscribed by the white box is shown at higher magnification; scale bar, 1 μm. Three independents experiments. c, Representative brightfield (BF) and fluorescent (mKate⁺) images (left panel), and quantification (right panel) of MDA231 parental and MDA231-BrM cells co-cultured with primary cortical neurons for 14 days. Two-tailed Student t-test, mean ± s.e.m., n = 3 biological replicates over three independent experiments. Scale bar, 100 μm. d, Representative bright-field and fluorescent images (left panel) and quantification (right) of TS1 parental and TS1-BrM cells co-cultured with primary cortical neurons for 14 days. Data are represented as mean ± s.e.m. Student t-test (two-tailed) was used. n = 3 biological replicates over three independent experiments. e-g, Quantification of MDA231 parental and MDA231-BrM cell numbers after 14 days of culture in poly-D-lysine coated plates in three different conditions: in complete neuronal culture medium (e); in conditioned medium from primary cortical neuron cultures (f); and in a Boyden chamber with neurons in the top chamber and cancer cells in the bottom chamber (g). Mean ± s.e.m. Student t-test (two-tailed) was used. n = 3 biological replicates over three independent experiments. The panels below each bar graph show representative images of the cancer cells at the end of the assay, revealed by mKate fluorescence. h, Schematic inducible miR-E-based shRNA knock-down system carrying 3 distinct miR-E-based shRNA sequences that bind to different regions of a targeted mRNA (top vector), and tet-on inducible vector for GRIN2B open reading frame (ORF) used for rescue experiments (bottom vector). TRE; tet-on inducible promoter, rtTA; reverse tetracycline transactivator. i, Knockdown of GRIN2B in cultured MDA231-BrM cells with tet-on inducible shRNAs, as assessed by western blotting after DOX treatment (1 μg/ml) for 2 days. Three independent experiments. The numbers above indicate levels of pGluN2B Y1252 and total GluN2B protein normalized to GAPDH. GAPDH was run in the same gel as pGluN2B Y1252, and run on separate gels from GluN2B as sample processing controls. j, Representative brightfield and fluorescent images (bottom panel) and quantification (top panel) of MDA231-BrM cells transfected with inducible shCTRL and shGRIN2B (DOX; 1 μg/ml), co-cultured with primary cortical neurons for 14 days. Two-tailed Student t-test, mean ± s.e.m., n = 3 biological replicates over three independent experiments. Scale bar, 100 μm. k, Rescue expression of GluN2B in cultured MDA231-BrM cells with tet-on inducible shRNAs along with a GRIN2B cDNA, as assessed by western blotting after DOX treatment (1 μg/ml) for 2 days. Three independent experiments. GAPDH was run in the same gel as GluN2B. l, Representative fluorescent (mKate⁺) images (left panel), and quantification (right panel) of MDA231 shCTRL, shGRIN2B and GluN2B rescue cells co-cultured with primary cortical neurons for 14 days. Two-tailed Student t-test, mean ± s.e.m., n = 3 biological replicates over three independent experiments. Scale bar, 100 μm. m, Cell proliferation in shCTRL and shGRIN2B BrM cells as determined by MTT assays, starting with 5,000 or 10,000 cells per well, after 72 hrs in culture. Two way ANOVA, mean ± s.e.m., n = 3 independent experiments.

Extended Data Figure 7. In vivo assessment of the functional importance of GluN2B-NMDAR signaling in brain-metastatic human breast cancer cells.

a-b, Representative immunofluorescent images of FMRP (green), luciferase (red) and DAPI (blue) in brain metastases formed by shCTRL and shGRIN2B MDA231 cells (a), and quantification of mean FMRP fluorescence in Luciferase⁺ tumour cell clusters (b). Two-tailed Student t-test, mean ± s.e.m.; n = 25 for shCTRL and n = 32 shGRIN2B group, all from 3 mice per group. c, Weight of orthotopic breast tumors formed by MDA231-BrM cells transfected with inducible shCTRL or shGRIN2B inoculated into the 4th mammary fat pads of female mice. DOX food was added to induce the shRNAs concomitantly with MFP injection. Two-tailed Student t-test, mean ± s.e.m.; n = 9 mice for shCTRL group, and 10 mice for shGRIN2B group. d, Bio-luminescent imaging (BLI) and quantification of lung metastatic lesions formed by MDA231-BrM cells transfected with inducible shCTRL or shGRIN2B inoculated IV. DOX food was added to induce the shRNAs concomitantly with IV injection. Two-tailed Student t-test, mean ± s.e.m.; n = 5 mice per group, two independent experiments. e, Bio-luminescent imaging (BLI) and quantification of brain metastatic lesions formed by MDA231-BrM cells transfected with inducible shCTRL or shGRIN2B, or additionally with rescue expression of a GRIN2B cDNA. DOX food was supplied concomitantly with ICD injection to induce shRNA expression. Tumor burden was assessed 4 weeks later by BLI. Two independent experiments. Two-tailed Student t-test, mean ± s.e.m.; n = 10 mice for shCTRL group, 9 mice for shGRIN2B group and 9 mice for rescue group.

Extended Data Figure 8. Functional analysis of GluN2B expression in brain-metastatic mouse breast cancer cells.

a, Knockdown of GluN2B expression in cultured mouse TS1-BrM cells with tet-on inducible shRNAs, as assessed by western blotting 2 days after DOX treatment in vitro (1 μg/ml). Three independent experiments. The numbers above indicate levels of GluN2B normalized to GAPDH. GAPDH was run in the same gel as GluN2B. b, Weight of primary orthotopic breast tumors formed by TS1-BrM cells transfected with inducible shCTRL or shGRIN2B, and inoculated into the 4th mammary fate pad of female mice. DOX food was added to induce the shRNAs concomitantly with MFP injection. Two-tailed Student t-test, mean ± s.e.m.: n = 6 mice for shCTRL group, and 8 mice for shGRIN2B group. c-e, In vivo bioluminescent imaging (c) and quantification of brain (d) and thoracic cavity (e) photon flux in mice bearing TS-BrM cells transfected with inducible shCTRL or shGRIN2B four weeks after intracardiac injection. DOX food was supplied concomitantly with ICD injection to induce shRNA expression. Two-tailed Student t-test, mean ± s.e.m.; n = 9 mice per group, two independent experiments. f, Ex vivo bioluminescent imaging (BLI) and quantification of excised brains with metastatic lesions formed by TS1-BrM cells transfected with inducible shCTRL or shGRIN2B. DOX food was supplied concomitantly with ICD injection to induce shRNA expression. Two-tailed Student t-test, mean ± s.e.m.; n = 9 mice per group, two independent experiments.

Fig. 1. Expression of GluN2B-NMDAR signaling components is associated with poor prognosis across all human breast cancer types.

a, mRNA expression of 5 NMDAR-associated genes (GRIN2A, GRIN2B, GRIN2C, GRIN2D, and DLGAP1) in 1,100 breast cancer (BC) patients from The Cancer Genome Atlas. Samples were sorted by NMDAR score (b), which combines, for each sample, Z-scores of normalized expression values of all 5 genes. b, Boxplot comparison of NMDAR scores among PAM50 BC subtypes, ranked by median score within each subtype. NMDAR scores in Basal-like BC were tested against scores obtained for all other subtypes. Wilcoxon two tail test. c, NMDAR mRNA expression score for the TCGA BC cohort (n = 1,100). Scores for all BC samples are ranked in increasing order and color-coded based on the PAM50 subtype. d, Boxplot comparison of GRIN2B mRNA expression (z-scores of log2-normalized RNA-seq RSEM-normalized values) between samples classified as triple negative breast cancer (TNBC; red dots) vs. remaining samples (gray dots). Wilcoxon two-tail test. e, Boxplot comparison of GRIN2B mRNA expression (z-scores of log2-normalized RNA-seq RSEM-normalized values) between samples with high risk of recurrence (RoR; red dots) vs. medium/low RoR (blue dots). Wilcoxon two-tail test. f,g Percentage of tumors with detectable GRIN2B (f) and GRIN1 (g) mRNA (RSEM-normalized counts > 0) classified by high RoR, medium/low RoR, and the most frequent PAM50 subtypes. h, Distant relapse-free survival (DRFS) analysis of an independent cohort of BC patients (GSE25065) with over-expressed GRIN2B (top 25% - red line) vs. remaining patients (blue line). Two-sided log-rank test. i,j, DRFS analysis of basal-like (i) or TNBC (j) patients (GSE25065) with over-expressed GRIN2B (top 25% - red line) vs. remaining patients (blue line). Two-sided log-rank test. The thick central line of each box plot represents the median number of significant motifs, the bounding box corresponds to the 25^th–75^th percentiles, and the whiskers extend up to 1.5 times the interquartile range.

Fig. 2. GluN2B-NMDAR signaling is highly activated in human brain metastasis.

a-b, Representative images (a) and quantification (b) of IHC staining of pGluN2B Y1472 in paired human primary BC (primary) and brain metastases (brain mets). Mean ± s.e.m. Two sided Wilcoxon two-tail test. Scale bar, 100 μm. BC subtype: 3 Her2⁺, 1 luminal and 5 unknown. c, Western blotting of pGluN2B Y1252, pGluN2B Y1472, GluN2B, and GluN1 in pairs of human (left column, MDA231) and mouse (right column, TS1) parental BC cell lines and their corresponding B2BM derivatives. Three independent experiments. Protein levels in B2BM cells were quantified relative to cognate parental cells, after normalization to ß-actin or GAPDH controls, for which blotted membranes were cut into two parts, of <50 K for beta-actin or GAPDH staining, and 50-250 Kd for pGluN2B Y1252/Y1472 staining. Additional, equally loaded gels were blotted and stained for GluN2B or GluN1, using the loading control from the pGluN2B Y1252 gel. d, Schematic of assays performed with luciferase-expressing B2BM cells either inoculated orthotopically into the mammary fat pad (MFP) to develop primary tumours, or intracardiac (ICD) to seed brain mets, or intravenous (IV) into the tail vein to seed lung mets. e,f, IHC staining of pGluN2B Y1472, total GluN2B, and luciferase (e) and quantification of pGluN2B Y1472 staining (f) in primary tumours, brain mets, and lung mets formed by MDA231-B2BM cells (day 28~35 after injection). Scale bar, 10 μm. These images correspond to the inserts shown in Extended Data Figure 2l. Two-tailed Student t-test, n = 3 mice per group, mean ± s.e.m.

Fig. 3. Autocrine secretion of L-glutamate is not sufficient to explain the brain metastasis-specific induction of GluN2B-NMDAR signaling.

a, Western blotting of pGluN2B Y1252 and Y1472 in MDA231-BrM cells treated with increasing concentrations of L-glutamate and 10 μM glycine for 1 hr in artificial cerebrospinal fluid. The numbers below indicate levels of pGluN2B Y1252, with Y1472 expression normalized to GAPDH. Three independent experiments. GAPDH control was run in the same gel as pGluN2B Y1252, and on separate gels from pGluN2B Y1472. b, Secreted L-glutamate concentrations in conditioned medium from MDA231 and TS1 parental and derivative B2BM cells after 48 hrs of culture. Two-tailed Student t-test, n = 3 independent experiments. c, Real-time RT-PCR analysis of glutamate transporter gene expression in MDA231 parental and B2BM cells. Two-tailed Student t-test, mean ± s.e.m., n = 3 independent experiments. d Immunofluorescence staining to detect apoptotic cells (TUNEL; green) in mouse brains with metastases formed by MDA231-BrM cells (day 28~35 after injection). Luciferase (magenta) reveals the cancer cells, and NeuN (red) reveals neurons; Bottom row: DNAase I treatment on brain tissue sections was used as a TUNEL-positive control. Images shown are representative of an analysis of >20 brain mets from 4 mouse brains, 2 sections/mouse brain. Scale bar, 10 μm. Three independent experiments.

Fig. 4. B2BM cancer cells form pseudo-tripartite synapses with neurons

a,b, Immunofluorescence staining of luciferase (B2BM; blue), pGluN2B Y1252 (green) and vGlut2 (red) in a mouse brain metastasis formed by MDA231-BrM cells (day 25 after injection), imaged by STED super-resolution microscopy. (a) Image with 3D volume. The white dotted border box indicates a normal synapse shown in Extended Data Fig. 5c. (b) Tilted 3D image of the area circumscribed by the white solid border box in (a) shows a pseudo-tripartite synapse between B2BM cell(s) and neurons, revealed by vGlut2⁺ neuronal puncta (arrow) and pGluN2B⁺ puncta in B2BM cells and presumptive neuronal processes in close proximity (arrowhead). Images are representative of 9 brain mets from 3 mice, 2 sections/mouse brain. Side length for each square in 3D view = 5 μm (a) and 400 nm (b). (Two Supplemental videos further describe the boxed images.) c, Electron micrograph of a brain metastasis formed by MDA231-BrM cells in the cerebral cortex. The cancer cell is purple-shaded. Scale bar, 5 μm. Two independent experiments. d, Three serial EM images from the region indicated by the black box in (c) showing protrusions from a B2BM cell in close apposition to an excitatory synapse.(Representative of 6 regions within 2 ultrathin sections from one mouse brain.) B2BM cells, purple-shaded; D = dendrite, yellow-shaded; Ax = axon, green-shaded; P = postsynaptic density. Scale bar, 1 μm. e, 3D reconstruction of the region in (d) illustrating a metastatic cell’s protrusion proximal to pre- and postsynaptic elements and the synaptic density (shown in red). f, Brain-metastatic cancer cells supplant astrocytes at neuronal synapses to enable NMDAR-dependent colonization. The tripartite synapse composed of pre- and post-synaptic neurons and astrocytes (left) is subverted by metastasizing breast cancer cells, which mimic astrocytes to form pseudo-tripartite synapses (right) that provide a source of glutamate ligand, activating NMDAR signaling to stimulate tumor growth in the brain.

Fig. 5. NMDAR activity mediates colonization but not seeding of brain metastases.

a, Bio-luminescent imaging (BLI) and quantification of brain metastatic lesions formed by MDA231-BrM cells transfected with doxycycline (DOX) inducible shCTRL or shGluN2B. DOX food was supplied concomitantly with ICD injections and metastasis burden was assessed 4 weeks later by BLI. Two-tailed Student t-test, mean ± s.e.m., n = 10 mice per group, two independent experiments. b, Brain metastasis-free survival of mice seeded by intracardiac inoculation with MDA231-BrM cells transfected with shCTRL or shGluN2B. DOX food was supplied concomitantly with ICD injection. Two-sided log-rank test. c, Evaluation of brain seeding capability of luciferase⁺ shGluN2B vs. shCTRL MDA231-BrM. DOX food was supplied concomitantly with ICD injection; mice were sacrificed at day 7. Every 10^th brain section was stained with luciferase antibody, and all luciferase⁺ cells were counted. Representative images show single luciferase⁺ B2BM cells (brown staining; black arrowhead). Two-tailed Student t-test, mean ± s.e.m., n = 3 mouse brains per group. Scale bar, 100 μm. d, Evaluation of Luciferase⁺/Ki67⁺ shGluN2B vs. shCTRL MDA231-BrM cells after initial colonization of brain-metastatic lesions. DOX food was supplied 7 days after ICD injection; mice were sacrificed at day 14. Every 10^th brain section was stained with luciferase and Ki67 antibody, and all luciferase⁺ and Ki67⁺ cells were counted. Representative images show luciferase (green) and Ki67 (red) staining. DAPI; nuclear staining. Two-tailed Student t-test, mean ± s.e.m., n = 3 mouse brains per group. Scale bar, 100 μm. e, Bio-luminescent imaging (BLI) and quantification of brain metastatic growth by MDA231-BrM cells transfected with inducible shGluN2B or shCTRL. DOX food was supplied at day 7 after ICD injection, and continued for 3 weeks, until day 28. Two-tailed Student t-test, mean ± s.e.m., n = 9 mice for shCTRL, and 8 mice for shGluN2B.