Spatial Regulation of Mitochondrial Heterogeneity by Stromal Confinement in Micropatterned Tumor Models

Abstract

Heterogeneity of mitochondrial activities in cancer cells exists across different disease stages and even in the same patient, with increased mitochondrial activities associated with invasive cancer phenotypes and circulating tumor cells. Here, we use a micropatterned tumor-stromal assay (μTSA) comprised of MCF-7 breast cancer cells and bone marrow stromal cells (BMSCs) as a model to investigate the role of stromal constraints in altering the mitochondrial activities of cancer cells within the tumor microenvironment (TME). Using microdissection and RNA sequencing, we revealed a differentially regulated pattern of gene expression related to mitochondrial activities and metastatic potential at the tumor-stromal interface. Gene expression was confirmed by immunostaining of mitochondrial mass, and live microscopic imaging of mitochondrial membrane potential (ΔΨ_m) and optical redox ratio. We demonstrated that physical constraints by the stromal cells play a major role in ΔΨ_m heterogeneity, which was positively associated with nuclear translocation of the YAP/TAZ transcriptional co-activators. Importantly, inhibiting actin polymerization and Rho-associated protein kinase disrupted the differential ΔΨ_m pattern. In addition, we showed a positive correlation between ΔΨ_m level and metastatic burden in vivo in mice injected with MDA-MB-231 breast cancer cells. This study supports a new regulatory role for the TME in mitochondrial heterogeneity and metastatic potential.

Figure 1

Spatial regulation of mitochondrial, metabolic, and metastatic pathways in micropatterned tumor-stromal assays (µTSA) assessed by RNA sequencing. (A) Schematics depicting the steps to create a MCF-7/BMSC µTSA; (B) Representative micropattern of an MCF-7 island (green: Pan-Cytokeratin) surrounded by BMSCs (red: Vimentin). Blue: nuclei. Scale bar: 100 μm. Cancer cells from the center and edge of the µTSA (opaque annular ring with dotted outlines) were isolated by laser capture microdissection, and their RNA was extracted and sequenced; (C) Gene sets found to be enriched at the interface vs. center by RNA sequencing and gene set enrichment analysis (glycolysis has negative NES indicating enrichment in the center). (See Methods for more details. FDR: False Discovery Rate; FDR < 0.25 was considered significant.) MCF-7 cells from >6 micropatterns/experiment were microdissected and combined before RNA isolation and sequencing; N = 2 independent experiments.

Figure 2

Differential regulation of mitochondrial membrane potential (ΔΨ_m) and mass at the interface vs. center in the µTSA. (A) ΔΨ_m levels assessed by TMRM fluorescence on Day 4 MCF-7-BMSC µTSA before and after addition of 20 μM FCCP uncoupler. Scale bar: 500 μm; (B) Representative radial distribution of TMRM fluorescence in a MCF-7-BMSC µTSA on day 4 before and after FCCP treatment. The red dot (r = 0.52) on the no-treatment (NTX) curve indicates that the TMRM fluorescence to the right of the dot is significantly higher than that of all FCCP-treated samples; from N = 6 independent experiments (p < 0.05, Welch’s t-test); (C) µTSA stained for mitochondrial mass with anti-TOM20. Right panels: confocal scans of the edge and the center. Green: TOM20; purple: vimentin; blue: DAPI. Scale bars: 500μm in widefield (left) and 25μm in confocal (right). (D) Fold difference in TMRM and TOM20 fluorescence between the edge and the center. N = 3 independent experiments. P-values: ordinary one-way ANOVA.

Figure 3

Redox imaging of cancer cells in a µTSA. Representative images of (A) NAD(P)H fluorescence; (C) FAD fluorescence; and (E) optical redox ratio (defined as FAD/(FAD + NAD(P)H)) at the center and edge of the µTSA on day 4. Scale bar: 25 μm. Quantification of (B) NAD(P)H and (D) FAD fluorescence intensities, and (F) the optical redox ratio at the single-cell level from the center (blue dots) or the edge (red dots) areas within the µTSA. Mitochondrial regions are segmented from FAD images and applied to the optical redox ratio images (green regions in C,E). Color scale is to the right of each image. (Representative dataset from N = 4 independent experiments; p-values: Welch’s t-test.)

Figure 4

Correlation of ΔΨ_m and YAP/TAZ nuclear translocation in micropatterns. (A) Schematics of the three micropattern cultures used in this experiment; (B) TMRM staining of ΔΨ_m (scale bar: 500 μm) and YAP/TAZ immunostaining (scale bar: 25 μm) in the three micropatterns on day 4; (C) Quantification of cancer cells with nuclear YAP/TAZ localization at the center and edge of the µTSA (n.s.: not significant; ****p < 0.0001 by ordinary one-way ANOVA); Representative dataset from N = 3 independent experiments. (D) Linear regression of YAP/TAZ nuclear localization and TMRM fluorescence in cancer cells in the monoculture and co-culture µTSA. Three locations (center, edge and approximately 700 μm away from the edge) were taken from the monoculture µTSA and two locations (center and edge) from the co-culture. (Representative dataset, N = 2 independent experiments; p-value: zero-slope hypothesis in linear regression).

Figure 5

Regulation of cancer cell ΔΨ_m by stromal density. (A) TMRM fluorescence in a µTSA with BMSC seeding densities varying from 25,000, 50,000, to 100,000 cells per micropattern. Monocultures without or with PDMS constraint were used as controls. Scale bars: 500 μm; (B) Normalized radial distribution of TMRM fluorescence in micropatterns; (C) Percentage of YAP/TAZ nuclear localization in cancer cells at the edge and center of the micropatterns (n.s.: not significant, *p < 0.05, ****p < 0.0001 by ordinary one-way ANOVA); (D) Areas of cancer islands as a function of the initial stromal density on Day 4. Non-liner regression: single-exponential decay; (E) TMRM peak area at the interface normalized to total cancer area of the respective micropattern as a function of initial stromal density. Non-linear regression: single-exponential decay; (F) Cancer cell densities at the center (red curve) and edge (black curve) of micropatterns. (*p < 0.05, ***p < 0.001, by Kruskal-Wallis test); (G) Normalized TMRM peak area as a linear function of cancer cell density at the interface. Linear regression R² = 0.9957. Representative dataset shown from N = 3 independent experiments for MCF-7, MCF-7 + 50k BMSC, and MCF-7 + PDMS micro-patterns, N = 1 for MCF-7 + 25k BMSC and MCF-7 + 100k BMSC micropatterns.

Figure 6

Loss of ΔΨ_m heterogeneity by inhibition of mechanotransduction. (A,B) TMRM fluorescence in monoculture µTSA on day 4 treated with inhibitors of mechanotransduction: Y-27632 (50 μM, ROCK inhibitor) and Latrunculin A (LatA, 0.5 μM, actin polymerization inhibitor) from 30 to 240 minutes. Scale bar: 500 μm; (C,D) Changes of TMRM fluorescence in cancer cells at the center and edge as a function of treatment time (Representative dataset from N = 3 independent experiments; *p < 0.05, ***p < 0.001, ****p < 0.0001, by ordinary one-way ANOVA).

Figure 7

Correlation of ΔΨ_m with metastatic potential in vivo. (A) MCF-7 cells and (B) MDA-MB-231 cells, both transduced with GFP/luciferase, were sorted into a ΔΨ_m-high and -low subpopulations for tail-vein injection into NSG mice; Quantification of the metastatic burden in the lungs of mice injected with (C) MCF-7 cells at week 5 and (D) MDA-MB-231 cells at week 4, via ex vivo quantification of luciferase activity in tissue lysate (p-values: Mann-Whitney test).