doi: 10.1038/s41467-019-14134-w.
Cancer-associated fibroblast heterogeneity in axillary lymph nodes drives metastases in breast cancer through complementary mechanisms
Affiliations
- PMID: 31964880
- PMCID: PMC6972713
- DOI: 10.1038/s41467-019-14134-w
Item in Clipboard
Cancer-associated fibroblast heterogeneity in axillary lymph nodes drives metastases in breast cancer through complementary mechanisms
Nat Commun.
.
Abstract
Although fibroblast heterogeneity is recognized in primary tumors, both its characterization in and its impact on metastases remain unknown. Here, combining flow cytometry, immunohistochemistry and RNA-sequencing on breast cancer samples, we identify four Cancer-Associated Fibroblast (CAF) subpopulations in metastatic lymph nodes (LN). Two myofibroblastic subsets, CAF-S1 and CAF-S4, accumulate in LN and correlate with cancer cell invasion. By developing functional assays on primary cultures, we demonstrate that these subsets promote metastasis through distinct functions. While CAF-S1 stimulate cancer cell migration and initiate an epithelial-to-mesenchymal transition through CXCL12 and TGFβ pathways, highly contractile CAF-S4 induce cancer cell invasion in 3-dimensions via NOTCH signaling. Patients with high levels of CAFs, particularly CAF-S4, in LN at diagnosis are prone to develop late distant metastases. Our findings suggest that CAF subset accumulation in LN is a prognostic marker, suggesting that CAF subsets could be examined in axillary LN at diagnosis.
Conflict of interest statement
The authors declare no competing interests.
Figures
a Representative FACS plots showing FAP, CD29, PDPN, αSMA and PDGFRβ protein levels in DAPI− EPCAM− CD45− CD31− CD235a− cells from a primary breast tumor (PT, left) and its corresponding metastatic axillary lymph node (LN, right). b, c FlowSom trees built on CAF from LNs (b, n = 20) and PTs (c, n = 16) and annotated for each CAF marker expression. Colors show CAF marker intensities. Node size depends on the number of phenotypically close cells. d Specific mean fluorescence intensity (speMFI) for each marker per CAF subset in PT and LN. Values are in log2 adjusted with offsets per marker. Each dot represents one sample (n ≥ 13 PT/LN pairs). Boxplots are median ± 25%–75% quantiles, whisker values range 1.5 × IQR above 75th or below 25th percentiles. p Values from Wilcoxon signed rank test. e Correlation plots between each marker speMFI in PT and LN, matched by patient and CAF subset (n ≥ 13 PT/LN pairs). p Values from Spearman’s test. f Same as in a for an invaded axillary LN (left) and its corresponding non-invaded LN (right). g Correlation plots between the percentage (%) of each CAF subset among total CAF and EPCAM+ cells among live cells, in invaded axillary LN (n = 19). p Values from Spearman’s test. Source data provided in Source Data file, with R scripts used.
a Percentage (%) of stroma in invaded LN according to BC subtypes (n = 124), LumA (violet), LumB (blue), HER2 (light gray), Lum B/HER2 (dark gray), TN (black). b Histological scores (H-Scores) for each CAF marker in invaded LN according to BC subtypes (n = 124). c Representative images of CAF marker staining on serial LN sections (LN#1 HER2; LN#2 TN). Scale bar, 50 μm. d Decision tree defining CAF subset identity based on quartile (Q) and median (Mdn) distribution of CAF markers. Thresholds and decision rules were first established on FACS data and next applied to IHC data. e Repartition of CAF subset enrichments (CAF-S1 (red), CAF-S2 (orange), CAF-S3 (green) and CAF-S4 (blue)) in LN according to BC subtypes (n = 124). f Representative views of CAF marker immunostaining on serial LN sections used for building maps of CAF subsets at cellular scale using the decision tree algorithm, shown in d. CAF-S1 are in red, CAF-S4 in blue and epithelial cells in black. CAF-S1- and CAF-S4-enriched LNs are shown. Scale bar, 200 μm. g Repartition of CAF subset enrichments in PTs and matched LNs (n = 41 pairs). h Same as in g with unmatched samples (N+ cases, n = 75 PT and 84 LN). i Same as in h according to BC subtypes (Lum A: n = 38 PTs, 30 LNs; HER2: n = 16 PTs, 26 LNs; TN: n = 21 PTs, 28 LNs). j Contingency table showing repartition of CAF subset enrichments in PTs and corresponding invaded LNs (n = 41 pairs). In all panels, boxplots are median ± 25%–75% quantiles, whisker values range 1.5 × IQR above 75th or below 25th percentiles. b
p Values from Mann–Whitney test. e, g, h
p Values from Fisher’s exact test. Non-significant p values are not mentioned. Source data provided in Source Data file, with R scripts used.
a PCA based on the 500 most variant genes from RNAseq data of CAF-S1, CAF-S4 and EPCAM+ cells sorted from paired PT and LN (n = 5 pairs for CAF-S1 and CAF-S4; n = 4 pairs for EPCAM+ cells). b Hierarchical clustering on the same samples and same genes as in a using Ward’s method with Pearson distances. Rows represent samples and columns genes. Color saturation shows gene expression deviation from the mean (above in red, below in blue). c PCA based on the 500 most variant genes from RNAseq data of EPCAM+ cells (n = 4 PT/LN pairs, top left), CAF-S1 (n = 5 PT/LN pairs, top right) and CAF-S4 (bottom left: n = 5 PT/LN pairs; bottom right: PCA restricted to n = 8 samples, as indicated). d Venn diagram showing overlaps between the transcriptomic signatures of CAF-S1 from PT, CAF-S1 from LN, CAF-S4 from PT and CAF-S4 from LN. Circled numbers show common genes between tissue localization for each cell type. p Values from hypergeometric test indicate significance of overlaps between subgroups. Source data provided in Source Data file, with R scripts used.
a CAF-S1 (red) and CAF-S4 (blue) doubling time (n = 6). b Left, Transwell membrane underside images. Scale bar, 200 μm. Right, CAF-S1/-S4 migration capacity (cells/mm2) (n = 6). c CAF-S1/-S4 velocity, persistence and direction (|sin(α)|, assessed by cell exclusion assay. d Left, Images of CAF-S1/-S4 spheroids embedded into collagen. Scale bar, 200 μm. Middle, CAF-S1/-S4 invaded areas/core spheroid areas. Each dot is one spheroid (n ≥ 6 per CAF subset). Right, Same as Middle for median values per CAF subset. e Left, Images of collagen-gel contraction by CAF-S1/-S4. Scale bar, 2 mm. Right, Percentage of collagen contraction by CAF-S1/-S4 (n = 6). f–h Contractility of CAF-S1/-S4 by traction force microscopy. f Images of traction stress applied by CAF-S1/-S4 on substrate. Traction forces (arrows) and cellular outlines (dashed lines) shown. Scale bar, 20 μm. Traction stress magnitudes in Pascal (Pa). g Left, CAF-S1/-S4 strain energy density (Joules (J)/m2). Each dot is one cell (n ≥ 21 cells/CAF subset). Right, Same as left for median strain energy densities per CAF subset (n = 3). h Left, CAF-S1/-S4 traction stress. Right, Same as left for median traction stress per CAF subset (n = 3). i Left, Images of collagen (blue) by CAF-S1/-S4 (red) assessed by second harmonic generation. Scale bar, 20 μm. Right, Collagen density in each cell stack. Each dot is the average value of collagen density around one cell (n ≥ 10 cells per CAF subset) (n = 2). In all panels, boxplots are median ± 25%–75% quantiles, whisker values range 1.5 × IQR above 75th or below 25th percentiles. a, b, e right, g right, h right: p Value from paired t-test. c, g left: p Values from Mann–Whitney test. d middle, h left, i right for two first CAF pairs: p Values from Mann–Whitney test (1st pair) and Student’s t-test (2nd pair). d right: p Value from Student’s t-test. At least three CAF-S1/-S4 pairs tested, except in c/i, two pairs. Source data provided in Source Data file, with R scripts used.
a Total number of viable BC cells (Dapi− cells by FACS) in co-culture with CAF-S1 (red) or CAF-S4 (blue) relative to control (− gray, without CAF) (n = 6 per BC cell type). b Total number of viable BC cells (Resazurin staining) with CAF-S1- or CAF-S4-conditioned medium (CM) relative to control (−, without CM) (n = 9 per BC cell type). c Total number of viable CAF-S1 (red) and CAF-S4 (blue) (Dapi− cells by FACS) in co-culture with BC cells relative to control (−, without BC cells) (n ≥ 6 per BC cell type). d CAF-S1 and CAF-S4 capacities to attract BC cells (n ≥ 8 per BC cells). e Left, Images of co-staining of Vinculin (red), F-actin (green) and DAPI (blue) in MCF7 (left) or T47D (right) cultured alone, or in presence of CAF-S1 or CAF-S4. Scale bars, 20 μm; inset, 10 μm. Arrows show reduced BC cell cohesion in presence of CAF-S1/-S4; asterisks show F-actin interconnections between BC cells in presence of CAF-S4. Vinculin and F-actin individual staining in Supplementary Fig. 4c. Right, Number of BC cells per tumor area (at least 13 images per condition, n = 3 per BC cell type). f Left, Images of E-cadherin (red), F-actin (green) and DAPI (blue) co-staining (top) or of E-cadherin (red) and DAPI (blue) staining (bottom) in BC cells alone or in presence of CAF-S1/-S4. Scale bars, 20 μm; inset, 10 μm. Right, Quantification of E-cadherin staining per BC cell area (at least eight images per condition) (n = 2 per BC cell type). In all panels, barplots are mean ± SEM and n number of independent experiments; a. u., arbitrary units. a, d: p Values from Wilcoxon signed rank test. b, c: p Values from paired t-test. e right, f right: p Values from Student’s t-test (MCF7) and Mann–Whitney test (T47D). At least two CAF-S1 and CAF-S4 pairs tested, except in e, f for T47D, where one CAF-S1/-S4 pair is used. Source data provided in Source Data file, with R scripts used.
a Representative 3D views of MDA-MB-231 invasion assessed by inverted Transwell assays, in CAF-free (left), CAF-S1- (middle) or CAF-S4- (right) embedded collagen matrix. In left images in each condition, colors show the different cell types: BC cells in green; CAF-S1 in red; CAF-S4 in blue. In right images in each condition, colors indicate the distance browsed by BC cells on the vertical (z) axis in 3D (maximal distance 165 μm in red). Scale bar, 50 μm. b Maximal vertical distance browsed by BC cells in CAF-free, CAF-S1- or CAF-S4-embedded collagen (n = 3, ~500 BC cells per analyzed z-stack). c Mean frequency (%) of BC cells along vertical axis (z, μm) in CAF-free, CAF-S1- or CAF-S4-embedded collagen. d Proportion of BC cells that invaded above 50 μm in CAF-S1- or CAF-S4-embedded collagen, relative to CAF-free condition (n = 3, ~500 BC cells per analyzed z-stack). e Left, Representative tumor-on-chip experiment showing velocity of MCF7 in CAF-free, CAF-S1- or CAF-S4-embedded collagen. Each dot represents one cell (n ≥ 90 cells per condition). Right, Median velocity of MCF7 in CAF-S1- or CAF-S4-embedded collagen matrix relative to CAF-free condition (n = 3). f Same as in e for T47D. In all panels, barplots are mean ± SEM. Boxplots are median ± 25%–75% quantiles, whisker values range 1.5 × IQR above 75th or below 25th percentiles; n indicates number of independent experiments. b, d, e right: p Values from paired t-test. e left: p Values from Mann–Whitney test. At least two CAF-S1 and CAF-S4 pairs have been tested. Source data provided in Source Data file, with R scripts used.
a Up, Representative images of E-cadherin (red), F-actin (green) and DAPI (blue) in BC cells alone (−) or with CAF-S1 transfected with non-targeting (siCTL) or CXCL12-targeting (siCXCL12) siRNA. CAF-S4 images in Supplementary Fig. 5c. Scale bars, 20 μm; inset 10 μm. Down, BC cell density and E-cadherin staining alone (gray) or with CAF (CAF-S1 red; CAF-S4 blue) (≥12 images/condition; n = 8). b Same as a with siCTL- or siCXCR4-transfected BC cells, with/without CAF-S1 (red/gray) (≥8 images/condition; n = 4). c Same as a with/without CAF-S1 with/without TGFβ-R inhibitor (LY2109761). CAF-S4 images in Supplementary Fig. 5e. Quantifications without (gray) or with CAF (CAF-S1 red; CAF-S4 blue) (≥7 images/condition; MCF7: n = 7; T47D: n = 3). d Percentage (%) of collagen contraction by CAF-S4 (blue) or CAF-S1 (red) without (DMSO) or with DAPT (n ≥ 3). e Strain energy density of CAF-S4 without (DMSO, blue) or with DAPT (gray). Each dot is one cell, n ≥ 46 cells/condition. f Representative 3D views of MDA-MB-231 (green) invasion by inverted Transwell assays in CAF-S4 (blue)-embedded collagen matrix without (DMSO) or with DAPT. Colors indicate distance browsed by BC cells on z axis (dmax180 μm, red). Scale bar, 50 μm. g Maximal distance of BC cells in CAF-S4-embedded collagen with DAPT relative to control (DMSO) (n = 6, ~700 BC cells/z-stack). h Same as in g for proportion of BC cells that invaded above 50 μm. i Left, Velocity of MCF7 in tumor-on-chip in CAF-free (−, gray) or CAF-S4 (blue)-embedded collagen treated or not with DAPT. Each dot is one cell, n ≥ 95/condition. Right, Same as left for median velocity of MCF7 (n = 3). j Same as i for T47D. In all panels, boxplots are median ± 25%–75% quantiles, whisker values range 1.5 × IQR above 75th or below 25th percentiles. Barplots mean ± SEM. a–e
p Values from Mann–Whitney test. g–i
p Values from paired t-test. At least four CAF-S1 and CAF-S4 tested. Source data provided in Source Data file, with R scripts used.
a Kaplan–Meier curves showing patient disease-free survival according to percentage (%) of stroma relative to epithelium in LN sections (n = 119). Patient subgroups defined by median. b Same as in a for overall survival. c Overall survival according to both LN stromal quantity (as in a) and CAF-subset enrichment, as defined in Fig. 2 (n = 119). a–c
p Values from log rank test. d Left, Distribution of patients with low- or high-LN stromal quantity (as in a) without (M0, n = 88) or with (M1, n = 31) metastases. Right, Same as in left per BC subtypes (65 Lum; 26 HER2; 28 TN). e Same as in d according to LN stromal quantity and CAF-subset enrichment (same groups as in c). f, g Left, Number of patients with metastases in any distant site except liver (gray) or with at least one metastasis in liver (black), according to CAF content in LNs at diagnosis (n = 31) either considering both CAF quantity and CAF-subset enrichments (same groups as in c) (f) or considering CAF-S1 and CAF-S4 enrichments (g). d–g
p Values from Fisher’s exact test. h Model: Left: four CAF subsets (CAF-S1 to -S4) are detected in metastatic axillary LN, as in PT. Two myofibroblast subsets (CAF-S1, red and CAF-S4, blue) are highly abundant in invaded LNs and correlated with tumor cell invasion. Middle: While patients with low-CAF content in LNs at diagnosis have less risk to develop late distant metastasis, those presenting high CAF-S1/S-4 quantity are prone to develop distant metastases, in particular in liver if LNs are CAF-S4-enriched. Right: Both CAF-S1 and CAF-S4 display pro-invasive properties through distinct mechanisms. CAF-S1 promote cancer cell proliferation, attraction and EMT initiation through CXCL12-CXCR4 and TGFβ axes. Highly contractile and matrix remodeler CAF-S4 induce cancer cell invasion in 3D via NOTCH signaling pathway. Thus, our work reveals the clinical interest of defining CAF subsets content in LNs at diagnosis. Source data provided in Source Data file, with R scripts used.
References
Publication types
MeSH terms
Substances
LinkOut - more resources
Full Text Sources
Other Literature Sources
Medical
