Supplementary Fig. S4: Cell existence analysis on Visium and cell-to-cell communication analysis

Supplementary Video 2: Diversity of microglial responses to DTCs visualized in vivo

Supplementary Fig. S5: Extended imaging analysis of the DTC imaging model

Heterogeneity of microglia in human brain metastasis captured by single-cell transcriptome. A, Single-cell analysis workflow. B, Dimensional reduction of expression profiles from 23,934 cells using single-cell variational inference (scVI) plotted on a UMAP plane and color-coded according to patient origin. LUAd, lung cancer adenocarcinoma; LUSq, lung cancer squamous cell carcinoma; SCLC, small cell lung cancer. C, Characterization of clustered cell types, colored by cell type. D, Top left, second round clustering to identify microglia. Bottom left, specific microglia and MDMs marker expression for evaluation. E, Volcano plot of DEGs in the defined microglia and MDMs. F, UMAP cluster map showing microglia from five patients in the same plane. G, Application of TAM-MG education markers (see main text; top) and tumor-adjacent score (bottom) on human samples. Red, more tumor-associated tumors. MG, microglia. H, Phenotype annotation based on clustering. A strongly estimated phenotype annotation is indicated by a large red circle. I, Estimated phenotypic heterogeneity of human microglia. GO-BP, Gene Ontology Biological Process.

Supplementary Table 2: A list of genes shown to inhibit macrophage phagocytosis activity

Heterogeneous microglial transcriptomic features in the DTC microenvironment. A, Schematic of scRNA-seq of the DTC model. The cortex was collected 7 and 10 days after injecting CMT167mCh cells into the internal carotid artery (ICA). The mCherry-positive microglia were also isolated. Vehicle-injected mice served as controls, and all data were combined. Each sample was pooled from five mice. B, UMAP cluster map of all cell types across the samples. C, Summary of signature genes defining each cell type. Full list in Supplementary Fig. S3C. D, Classification of microglia and MDMs. E, Volcano plot of DEGs in microglia and MDMs. F, UMAP cluster map showing microglia (MG) from the DTC model, vehicle-injected mice, and mCherry+ microglia on the same plane. G, Velocity analysis using the ExDyn model. Black arrows, estimated cell differentiation paths. H, Phenotypic annotation based on DEGs in each cluster. Fifteen clusters were identified in the microglia; 10 clusters had no DEGs (light blue), whereas five clusters showed DEGs (top). Bottom, phenotypes were assigned to clusters based on the DEGs. The full DEG list is provided in the Supplementary Data. CL, cluster; GO-BP, Gene Ontology Biological Process. I, Left, schematic of scRNA-seq on the DTC model with TGFβ receptor inhibitor (LY2109761) administered via intracerebroventricular catheter. Middle, successful administration was confirmed by toluidine blue injection. Right, the experimental timeline. TGFβRI, TGFβ1 receptor inhibitor. J, Qualitative comparison of CL2 (top) and post–TGFβ inhibitor treatment (bottom) in the LY2109761-treated DTC model, showing opposite positions on the UMAP plane. K, Minimal overlap between the TGFβ–responsive population CL3 and LY2109761 treatment on the UMAP. L, Summary diagram. Microglia were hypothesized to shift from a healthy state to a DTC-specific state. TGFβ inhibition may redirect microglia toward an IFN-responsive, antigen-presenting phenotype.

Supplementary Fig. S9: CD24 and/or CD47 positive brain metastasis in human

Supplementary Fig. S2: Antigen-Processing molecule enrichment in mCherry+ microglia

Heterogeneous microglial responses and DTC fate. A, Overview of the imaging model for DTCs. mCherry-labeled tumor cells were injected into the internal carotid artery (ICA) of Cx3cr1GFP/+ mice to create a brain metastasis model. Craniotomy was performed, and two-photon microscopy (2PM) was used for in vivo imaging. B, Representative images from the imaging model of DTCs, visualizing the insides of a 2 mm cranial window. Microglia and tumor cells can be visualized at a single-cell resolution. C, Protocol for imaging of DTCs. Internal carotid artery injection on day 0, craniotomy on day 1, and consecutive daily in vivo imaging starting on day 2. D, Representative in vivo images of morphologic and density changes in Cx3cr1+ cells during tumor formation. E and F, The time course of cell density (E) and cell body size (F) of Cx3cr1+ cells within <100 μm from the tumor. G, Phagocytic response of Cx3cr1+ cells to tumors. See Supplementary Video S1. H, A variety of microglial responses and tumor cell fates were observed in one host. Displacement (top right row), elimination (middle right row), and metastasis (bottom right row) were tracked using the imaging model of DTCs. See Supplementary Video S2. I, The frequency of tumor fate. Summary of eight imaging models generated by CMT167mCh (top row) and the fate of each mouse (bottom row). J, Response of Cx3cr1+ cells to tumor cells. Cx3cr1+ cells around “Elimination” or “Metastasis” tumor cells have significantly larger cell bodies compared with physiologic microglia. K, Response of Ccr2+ cells to tumor cells. The time course of cell density of Ccr2+ cells within <100 μm from the tumor. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, nonsignificant.

Supplementary Fig. S8: Supplemental analysis to evaluate Cd24 and Cd47 function to prevent metastasis

Supplementary Fig. S3: Extended single-cell RNA sequencing analysis on microglia in the DTC Model

Microglia dominate the initial response to DTCs in the brain cortex. A, Scheme of the brain DTC model. Tumor cells were injected into the internal carotid artery (ICA), and DTCs/micrometastases were identified 14 days after injection. B, Representative immunofluorescence images of the DTC model on day 10 showing microglia (Iba1+P2ry12+) surrounding the tumor. C, Composition of microglia (MG) and MDMs surrounding DTCs/micrometastases, as determined by immunofluorescence. Microglia were defined as Iba1+P2ry12+, and MDMs as Iba1+Cd206+. D, Left, the density of microglia and MDMs in the cerebral cortex was measured using FCM. Brief gating protocol for microglia and MDMs (middle, top row) and DTCs/micrometastases (middle, bottom row). For the detailed gating protocol, see Supplementary Fig. S1D. Densities of microglia and MDMs are shown in the right column. E, Definition of “Microglia-dominant DTC microenvironment.” We defined this phase as 0–12 days after tumor inoculation. F, mCherry+ microglia was determined using immunofluorescence. Iba1+P2ry12+ cells incorporated cell-sized (right, top row, white arrowheads) and fragment-sized mCherry (right, bottom row, blue arrowheads). G, FCM analysis of mCherry+ microglia and mCherry+ MDMs. A combined plot from three mice (left) and quantified data (right). H, Ratio of mCherry+ microglia to mCherry+ MDMs as measured by FCM. I and J, Increased expression of “antigen processing and presentation” molecules in mCherry+ microglia at the transcript level (I) and protein levels (FCM; J). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, nonsignificant.

Enhanced tumor elimination via Cd24a and Cd47 deletion in DTCs. A, Experimental scheme showing the time-course transcriptome of DTCs/micrometastases. Tumor cells were isolated from the cortex of DTC models on the indicated days. Five mice were included for each of the 11 samples (represented by orange triangles). ICA, internal carotid artery. B, Strategy for identifying surface proteins that promote metastasis. FPKM, fragments per kilobase of exon per million reads mapped. C, Volcano plot of 238 selected genes with eight factors identified as potential therapeutic targets. D, FCM analysis showing Cd47 and Cd24 protein expression levels. E, Both Cd24 and Cd47 protein expression increased on day 14 in DTCs. F, Schematic of pipeline for the functional analysis of Cd24 and Cd47 in metastasis. Cd24 and Cd47 were deleted from the WT cell line to create “ΔCd24/47” cells, and a “Rescue” cell line was generated by restoring the expression of these two factors. Three cell lines were evaluated using imaging models to confirm the roles of Cd24 and Cd47. 2PM, two-photon microscopy. G, Frequency of metastasis formation. H, Elimination of DTCs in brain vasculature using imaging model mice. The indicated cell lines were then evaluated. I, Cell proliferation curves for three cell lines. Tumor volumes were assessed using intravital imaging and normalized to the volume on day 2. J, Survival time of the imaging model mice. MST, median survival time. K,In vivo time-lapse imaging of the indicated cell lines in an imaging mouse model. A time-lapse video is available in Supplementary Video S3. Increased Cx3cr1+ cell body size in the ΔCd24/47 transplantation model. Error bars, SDs. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; ns, nonsignificant.

Supplementary Video 1: Phagocytosis of DTCs by microglia

Transcriptional feature of tumor-adjacent microglia in intravital imaging. A, Experimental design. Left, the PSmOrange2 protein fluoresces upon exposure to a 930 nm two-photon laser. In the Cx3cr1-PSmOrange2 mice, CMT167-EGFP cells were injected into the internal carotid artery (ICA) to create a DTC model. Microglial clustering around the tumor was specifically targeted for fluorescence changes. Pinpoint stimulation was achieved using holographic two-photon microscopy (2PM), and labeled cells were isolated using FCM and collected in 96-well plates. B, Intravital imaging of labeled DTCs/micrometastasis-adjacent microglia. Only the Cx3cr1+ cells near the tumor exhibited near IR + fluorescence. C, Volcano plot identifying marker genes characterizing tumor-adjacent microglia with an adjusted P < 0.05. The most representative markers (P < 0.01) are listed. Unlabeled microglia, N = 7; labeled microglia, N = 7. D, Tumor-adjacent score intensity projected onto the UMAP to estimate the location (see Fig. 2F). The tumor-adjacent score was calculated as follows: (tumor-adjacent positive score) − (negative score). E, Two-dimensional plot based on tumor-adjacent positive and negative scores. The mean scores for each cluster were plotted. F, Right, gene set enrichment analysis showing enrichment of inflammation-related genes in tumor-adjacent microglia. GO-BP, Gene Ontology Biological Process; NES, normalized enrichment score. G, Enrichment of gene sets “acute inflammatory response to antigen stimulus” in tumor-adjacent microglia. The enriched gene set included Mhc-I genes (indicated in red). H, Immunofluorescence images of Tmem119 and H2-D1. DTC-adjacent microglia (Tmem119+) showed high H2-D1 expression.

Supplementary Fig. S11: Single-cell analysis on human brain metastasis from lung cancer

Supplementary Fig. S7: Evaluation of activated microglial function in DTC models

Supplementary Fig. S10: Supportive experiments to confirm microglial tumor elimination