The temporal progression of lung immune remodeling during breast cancer metastasis

Abstract

Tumor metastasis requires systemic remodeling of distant organ microenvironments that impacts immune cell phenotypes, population structure, and intercellular communication. However, our understanding of immune phenotypic dynamics in the metastatic niche remains incomplete. Here, we longitudinally assayed lung immune transcriptional profiles in the polyomavirus middle T antigen (PyMT) and 4T1 metastatic breast cancer models from primary tumorigenesis, through pre-metastatic niche formation, to the final stages of metastatic outgrowth at single-cell resolution. Computational analyses of these data revealed a TLR-NFκB inflammatory program enacted by both peripherally derived and tissue-resident myeloid cells that correlated with pre-metastatic niche formation and mirrored CD14⁺ "activated" myeloid cells in the primary tumor. Moreover, we observed that primary tumor and metastatic niche natural killer (NK) cells are differentially regulated in mice and human patient samples, with the metastatic niche featuring elevated cytotoxic NK cell proportions. Finally, we identified cell-type-specific dynamic regulation of IGF1 and CCL6 signaling during metastatic progression that represents anti-metastatic immunotherapy candidate pathways.

Keywords: NF-κB; TLR; cancer immunology; immunosuppression; inflammation; metastasis; myeloid; natural killer cells; single-cell genomics.

Figure 1:. Longitudinal scRNA-seq cell analysis of PyMT mouse lung immune cells captures dynamics of the metastatic microenvironment.

(A) Schematic of experimental approach. Lungs from PyMT⁺ mice were dissected, dissociated, and cryopreserved before thawing, MULTI-seq barcoding, CD45⁺ immune cell enrichment, and scRNA-seq. (B) UMAP visualization of immune gene expression space colored by cell type with dot plot showing the top 3 differentially-expressed genes for each cell type. Dot color indicates expression level and size indicates the proportion of cells expressing each gene. (C) Jensen-Shannon divergence (JSD) heatmap demonstrating (dis-)similarity between PyMT lung samples. Hierarchical clustering-defined clades representing WT/early, mid, and late metastatic stages highlighted. (D) Bar charts describing shifts in cell type proportions during metastatic progression. Enriched, depleted, and transiently-enriched cell types highlighted with red, blue, and green boxes, respectively. Data represented as the cross-sample mean ± s.e.m proportion relative to total immune cells. Significant shifts highlighted in inset p-value heatmap (n=29 total samples; propeller test). See also Figure S1 and Table S1.

Figure 2.. AM inflammation and IM wound healing transcriptional signatures are linked to pre-metastatic niche formation.

(A) UMAP visualization of IM gene expression space colored by subtype or annotation marker genes with bar charts describing shifts in subtype proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all IMs (n=29 total samples). * p<0.05 propeller test. (B) Z-score heatmap of the average expression of Crip1^high Cav1⁺ signature genes in IM subtypes. (C) UMAP visualization of AM gene expression space colored by subtype or Cd14⁺ inflammatory AM cell marker genes. (D) UMAP visualization of metastatic stage densities in AM gene expression space with bar charts describing shifts in subtype proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all AMs (n=29 total samples). * p<0.05 propeller test. See also Figure S2 and Table S2.

Figure 3.. NMF and GSEA links Cd14⁺ inflammatory AM signature to TLR-NFκB inflammation and CD14⁺ ‘activated’ MDSCs.

(A) UMAP visualization of AM gene expression space before and after sub-clustering of major AM subtypes (e.g., antigen, allergy, and lipid AMs) colored by NMF8 module score or subtype. Inflammatory subpopulations (red) identified as clusters with elevated TLR-NFκB signature marker expression and NMF8 module scores. (B) UMAP visualization of sub-clustered AM gene expression space colored by TLR-NFκB inflammatory signature marker genes. (C) Bar charts describing shifts in Cd14⁺ inflammatory AM proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all AMs (n=29 total samples). * p<0.05 propeller test. (D) Lollipop plot depicting Hallmark gene sets that are enriched amongst NMF8 transcriptional signature genes. Dotted line denotes statistical significance threshold (p<0.01 hypergeometric test). (E) UMAP visualization of AM gene expression space colored by CD14⁺ ‘activated’ MDSC module score. (F) Scatter plot illustrating correlation between CD14⁺ ‘activated’ MDSC and NMF8 module scores in AMs. Cells colored by inflammation state. See also Table S3.

Figure 4.. BM-derived myeloid subtype characterization uncovers metastasis-associated DC subtype frequency shifts and TLR-NFκB inflammation in neutrophils.

(A) UMAP visualization of monocyte gene expression space colored by subtype with bar charts describing shifts in subtype proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all monocytes (n=29 total samples). * p<0.05 propeller test. (B) UMAP visualization of neutrophil gene expression space colored by subtype, TLR-NFκB inflammation signature genes, or CD14⁺ ‘activated’ MDSC module score with bar charts describing shifts in subtype proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all neutrophils (n=29 total samples). * p<0.05 propeller test. (C) UMAP visualization of mature neutrophil gene expression space colored by metastatic stage with z-score heatmap of stage-specific differentially-expressed genes. Z-scores for each gene binned by metastatic stage clustered using hierarchical clustering. (D) UMAP visualization of DC gene expression space colored by subtype with bar charts describing shifts in subtype proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all DCs (n=29 total samples). * p<0.05 propeller test. See also Figure S3 and Table S4.

Figure 5.. Myeloid TLR-NFκB inflammation signature is detected in PyMT pre-metastatic niche myeloid cells, human metastasis-associated macrophages, and multiple mouse models of metastatic breast cancer.

(A) UMAP visualization of monocyte and IM gene expression space colored by NMF19 module score, subtype, or CD14⁺ ‘activated’ MDSC signature score. Inflammatory subpopulations (red) identified as clusters with elevated TLR-NFκB signature marker expression and NMF19 module scores. (B) Bar charts describing shifts in inflammatory monocyte and IM proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all monocytes or IMs (n=29 total samples). * p<0.05 propeller test. (C) Scatter plot illustrating correlation between CD14⁺ ‘activated’ MDSC and NMF19 module scores in monocytes and IMs. Cells colored by inflammation state. (D) Lollipop plot depicting Hallmark gene sets that are enriched amongst genes comprising the NMF19 transcriptional signature. Dotted line denotes statistical significance threshold (p<0.01 hypergeometric test). (E) Density plots and box plots describing changes in CD14 membrane protein abundance on neutrophils, AMs, and monocytes during metastatic progression measured using flow cytometry. Boxes represent the interquartile range and median of the CD14 abundance geometric mean, and the whiskers extend ±1.5-fold the interquartile range (dots correspond to individual lungs, n=5 lungs per stage; * p<0.05 in Wilcoxon rank-sum test). Samples merged by metastatic stage for density plot visualizations. (F) UMAP visualization of PyMT validation cohort monocyte and neutrophil gene expression space colored by subtype or NMF19 module score with bar charts describing shifts in inflammatory subpopulation proportions. Inflammatory subpopulations (red) identified as clusters with elevated TLR-NFκB signature marker expression and NMF19 module scores. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all monocytes or neutrophils (n=18 total samples). * p<0.05 propeller test. (G) UMAP visualization of 4T1 monocyte and basophil gene expression space colored by subtype or NMF19 module score with bar charts describing shifts in inflammatory subpopulation proportions. Inflammatory subpopulations (red) identified as clusters with elevated TLR-NFκB signature marker expression and NMF19 module scores. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all monocytes or basophils (n=29 total samples). * p<0.05 propeller test. (H) UMAP visualization of brain MAM gene expression space colored by donor, MAM subtype, or NMF19 module score with scatter plots illustrating relationships between NMF19, S100A8⁺ MAM, and APOE⁺ MAM module scores. Cells in scatter plot colored by subtype. See also Figure S4, Figure S5, and Table S5.

Figure 6.. Lymphocyte subtype characterization reveals details of the inflammatory and immunosuppressive lung metastatic microenvironment.

(A) UMAP visualization of NK cell gene expression space colored by subtype with bar charts describing shifts in subtype proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all NK cells (n=29 total samples). * p<0.05 propeller test. (B) Box plots describing the changes in lung NK subtype proportions during metastatic progression measured using flow cytometry (n=5 lungs per stage; dots correspond to individual lungs). * p<0.05 t-test. Boxes represent the interquartile range and median of NK subtype proportions, and the whiskers extend ±1.5-fold the interquartile range. (C) UMAP visualization of human lymph node and lung NK cell gene expression space colored by subtype, tissue, or NK subtype markers with bar charts describing differences in subtype proportions between sample types. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all NK cells (n=39 total samples). * p<0.05 propeller test. mLN: metastasized lymph node; nLN: normal lymph node; tLung: tumor-bearing lung; nLung: normal lung. (D) UMAP visualization of T cell gene expression space colored by subtype with bar charts describing shifts in subtype proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all T cells (n=29 total samples). * p<0.05 propeller test. (E) UMAP visualization of B cell gene expression space colored by subtype with bar charts describing shifts in subtype proportions during metastatic progression. Proportional data represented as the cross-sample mean ± s.e.m proportion relative to all B cells (n=29 total samples). * p<0.05 propeller test. See also Figure S6 and Table S6.

Figure 7:. Intercellular communication modeling reveals metastasis-associated changes in immune cell signaling network.

(A) Weighted network graph of CXLC2-CXCR2 signaling binned by metastatic stage. Nodes colored by cell type, edges weighted by signaling probability and colored by sender cell type. (B) Violin plots showing Igf1 and Igf1r expression in all cell types binned by metastatic stage. Black dots denote mean expression. (C) Violin plots showing Igf1 expression in each IM subset and in Mrc1⁺ IMs binned by metastatic stage. Black dots denote mean expression. * p<0.05 Wilcoxon rank-sum test. (D) Weighted network graphs of CCL6-CCR1/2 signaling. (E) Violin plot showing Ccl6 expression across all cell types binned by metastatic stage. Black dots denote mean expression. Neutrophils and AMs highlighted with box. (F) Violin plots showing Ccl6 expression in neutrophil and AM subtypes binned by metastatic stage. Black dots denote mean expression. Significant shifts in Ccl6 expression highlighted with p-value heatmaps (Wilcoxon rank-sum test). See also Figure S7.