Membrane to cortex attachment determines different mechanical phenotypes in LGR5+ and LGR5- colorectal cancer cells

Abstract

Colorectal cancer (CRC) tumors are composed of heterogeneous and plastic cell populations, including a pool of cancer stem cells that express LGR5. Whether these distinct cell populations display different mechanical properties, and how these properties might contribute to metastasis is poorly understood. Using CRC patient derived organoids (PDOs), we find that compared to LGR5- cells, LGR5+ cancer stem cells are stiffer, adhere better to the extracellular matrix (ECM), move slower both as single cells and clusters, display higher nuclear YAP, show a higher survival rate in response to mechanical confinement, and form larger transendothelial gaps. These differences are largely explained by the downregulation of the membrane to cortex attachment proteins Ezrin/Radixin/Moesin (ERMs) in the LGR5+ cells. By analyzing single cell RNA-sequencing (scRNA-seq) expression patterns from a patient cohort, we show that this downregulation is a robust signature of colorectal tumors. Our results show that LGR5- cells display a mechanically dynamic phenotype suitable for dissemination from the primary tumor whereas LGR5+ cells display a mechanically stable and resilient phenotype suitable for extravasation and metastatic growth.

Fig. 1. LGR5+ and LGR5- single cells display differences in adhesion, polarity and YAP localization.

a CRC PDOs cultured for 1 week in culture matrix gel stained for CK20 and nuclei. LGR5+ cells are labeled with Tdtomato. Scale bar, 20 μm. Representative images from two independent experiments. b Scheme illustrating the preparation of PDOs for single cell analysis on 2D soft substrates. c Cell roundness measured for sorted cells seeded on collagen-I coated gel substrates of 0.5, 3, 5, 11 kPa in stiffness. d LGR5-, LGR5^med and LGR5+ cells on 0.5 and 11 kPa gel substrates. Representative images of four independent experiments. Scale bar, 50 μm. e Single PDO cells seeded on 0.5 kPa gel substrates stained for Actin, nuclei and YAP. LGR5+ cells are labeled with Tdtomato. Representative images of four independent experiments. Scale bar, 10 μm. f Quantification of YAP nuclear/cytoplasmic ratio of PDO cells seeded on 0.5, 3, 30 kPa gels. In (c) and (f) n ≥ 60 cells/condition from four independent experiments. Statistical significance was determined using two-way analysis of variance, followed by a Šidák multiple-comparison test. Quantification of cell velocity (g) and mean traction (h) exerted by sorted cells on gels of 0.5, 1.5, 5 and 11 kPa in stiffness. In (g) and (h) data are represented as the mean ± s.d. of n ≥ 60 cells/condition from four independent experiments. Statistical significance was determined using Shapiro-Wilk normality test, followed by a Kruskal-Wallis multiple-comparison test. i LGR5-, LGR5^med and LGR5+ cells seeded on 3 kPa gels and their corresponding traction stress field and force dipole. The traction stress vectors and their amplitude are represented by the yellow arrows and colormap, respectively. The two eigenvectors of the dipole matrix are represented by red arrows. Representative images of four independent experiments. Scale bar, 10 μm. j Quantification of the polarization Mδ of LGR5-, LGR5^med and LGR5+ cells. n ≥ 95 cells/condition from four independent experiments. Statistical significance was determined using Shapiro-Wilk normality test, followed by a Kruskal−Wallis multiple-comparison test. All data are represented as the mean ± s.d. Source data are provided as a Source Data file.

Fig. 2. LGR5+ are stiffer and respond to confinement with a higher survival rate and slower ameboid migration.

a Scheme of the real time-deformability cytometry (RT-DC) setup. Cells suspended in a highly viscous solution are flushed through a microfluidic channel of 20 μm height and width, where they are deformed by shear stress and pressure gradients. Cells are imaged in the region of interest (dashed rectangle) to determine their degree of deformation. b Representative images of LGR5-, LGR5^med and LGR5+ analyzed with RT-DC. Scale bar, 5 μm. c, d Cell deformation and derived elastic modulus of LGR5-, LGR5^med and LGR5+ cells. Data are represented as the mean ± s.d. from one experiment representative of four independent ones. n ≥ 300 cells/subgroup. Statistical significance was determined using Shapiro-Wilk normality test, followed by a Kruskal−Wallis multiple-comparison test. e Scheme of dynamic cell confiner. The central PDMS piston holds a glass slide to which PDMS pillars of fixed height (4.5 μm) are attached. The device acts as a suction cup. When negative pressure is applied, the central piston is pressed down, thereby confining the cells. f Representative examples of the four response categories identified in confined LGR5- and LGR5+ cells. Scale bar, 10 μm. g Percentage of dying/dead, polarized blebs, rounded blebs and no blebs in LGR5- and LGR5+ cells under confinement on a non-adhesive surface. Data are represented as the mean ± s.d. of percentages of n ≥ 112 cells/ subgroup from four independent experiments. Statistical significance was determined using two-way analysis of variance, followed by a Šidák multiple-comparison test. h Representative time lapse images of LGR5- and LGR5+ no blebs, LGR5+ and LGR5- polarized blebs. Scale bar, 10 μm. In (f, h) images are representative images of four independent experiments, with a total of 303 cells. i Migration velocity of tracked nuclei of LGR5- and LGR5+ cells, divided into categories according to the confinement response. Data are represented as the mean ± s.d. of n ≥ 112 cells/ subgroup from four independent experiments. Statistical significance was determined using two-way analysis of variance, followed by a Šidák multiple-comparison test. Source data are provided as a Source Data file.

Fig. 3. LGR5- clusters are rounder, migrate faster and exert less traction.

a Scheme of the preparation of PDOs for cluster analysis on 3kPa 2D substrates. b Time lapse of LGR5^low and LGR5^high clusters (first and second rows, respectively). Bottom rows show the time lapse of the traction stress field of the clusters shown in the top rows. Representative images of n ≥ 24 clusters/ subgroup from four independent experiments. Scale bars, 50 μm. Tdtomato fluorescence intensity (c), cluster roundness (d), migration speed (e), and mean traction (f). Data are represented as the mean ± s.d. of n = 128 clusters from four independent experiments. Statistical significance was determined using Shapiro-Wilk normality test, followed by a Kruskal−Wallis multiple-comparison test. g LGR5^low, LGR5^med and LGR5^high clusters labeled with Sir-Actin and their corresponding XZ and YZ lateral planes. Scale bars, 50 μm. Representative images of two independent experiments. n = 49 clusters. h Mean contact angle θ between the cell cluster and the 3 kPa substrate in LGR5^low, LGR5^med and LGR5^high clusters. Data are represented as the mean ± s.d. of n ≥ 6 clusters/ subgroup from two independent experiments. Statistical significance was determined using one-way analysis of variance, followed by a Šidák multiple-comparison test. i 3D rendering of the same clusters in (g). Source data are provided as a Source Data file.

Fig. 4. LGR5+ adhere better to endothelial monolayers and form transendothelial gaps.

a Schematic of the preparation of PDOs for adhesion analysis on an endothelial monolayer. b PDO clusters seeded on HUVEC monolayers formed on collagen I coated 3 kPa gel. Representative time lapse of LGR5^low, LGR5^med and LGR5^high clusters on a HUVEC monolayer. The last two columns show the fluorescence image of the HUVEC monolayers at 8 h including a zoom at the contact point between clusters and monolayers. Scale bars, 50 μm. Representative images from four independent experiments. c Mean Tdtomato fluorescence labeling LGR5+ cells in CRC clusters that either remained attached to the monolayer during the whole time-lapse or detached. Data are represented as the mean ± s.d. of n ≥ 31 clusters/ subgroup clusters from four independent experiments. Statistical significance was determined using Shapiro-Wilk normality test, followed by a Kruskal−Wallis multiple-comparison test. d Percentage of clusters that formed a gap in the endothelial monolayer for LGR5^low, LGR5^med and LGR5^high cell clusters. Data are represented as the mean ± s.d. of percentages from four independent experiments. Statistical significance was determined using two-way analysis of variance, followed by a Šidák multiple-comparison test. e Time for gap formation for LGR5^low, LGR5^med and LGR5^high cell clusters. f Gap area measured 8 h after beginning of time lapse acquisition. In (e) and (f) data are represented as the mean ± s.d. of n = 77 clusters that remained attached during time lapse acquisition from four independent experiments. Statistical significance was determined using Shapiro-Wilk normality test, followed by a Kruskal−Wallis multiple-comparison test. Source data are provided as a Source Data file.

Fig. 5. ERM proteins determine the mechanical differences between LGR5+ and LGR5- cells.

a Normalized Enrichment Scores (NES) from enrichment analysis ROAST-GSA for selected GO cellular component (GOCC) gene sets in LGR5-, LGR5^med and LGR5+ cells. p values are listed in Supplamentary Table 1. Data obtained from two independent replicates (at least 20,000 cells per condition). b Relative expression of Lgr5, Krt20 and ERM proteins for sorted cells. Data are represented as the mean ± s.d. from nine independent experiments. Statistical significance was determined using Shapiro-Wilk normality test, followed by a Kruskal−Wallis multiple-comparison test. c Levels of ERM and phospho-ERM proteins in sorted PDO cells with quantification of band intensity normalized to GAPDH from two independent experiments. d Scheme of the iMC-linker. e mRNA levels of ERM proteins in wild type and shRNA ERM silenced organoids. Data are represented as the mean ± s.d. from two independent experiments. f LGR5+ cells treated with doxycycline to induce the iMC-linker expression, untreated (no doxy) and LGR5- ERM silenced cells on 3 kPa gel substrates. Representative images of two independent experiments. Scale bar, 50 μm. g Cell roundness for sorted cells seeded on 3 kPa gels. Data are represented as the mean ± s.d. of n ≥ 58 cells/condition from two independent experiments. Statistical significance was determined using Shapiro-Wilk normality test, followed by a Kruskal−Wallis multiple-comparison test. h Representative images of LGR5-, LGR5+ and iMC-linker expressing LGR5+ cells under confinement. Scale bar, 10 μm. i Percentage of dying/dead, polarized blebs, rounded blebs and no blebs in LGR5-, LGR5+ and iMC-linker expressing LGR5+ cells. Data are represented as the mean ± s.d. of percentages from four independent experiments (n ≥ 119 cells/subgroup). Statistical significance was determined using two-way analysis of variance, followed by a Šidák multiple-comparison test. j Migration velocity of LGR5^high, LGR5^low, iMC-linker expressing LGR5^high clusters and LGR5^low clusters with ERM silencing. Data are represented as the mean ± s.d. of n ≥ 46 clusters/condition from two independent experiments. Statistical significance was determined using Shapiro-Wilk normality test, followed by a Kruskal−Wallis multiple-comparison test. k Time lapse of LGR5^high, LGR5^low, iMC-linker expressing LGR5^high and ERM silenced LGR5^low clusters seeded on Collagen-I coated 3 kPa gels. Representative images of two independent experiments. Scale bars, 50 μm. Source data are provided as a Source Data file.

Fig. 6. Expression of LGR5 and Ezrin are anticorrelated in transcriptomic data from CRC patients.

a, b Gene expression of Lgr5 and Ezr in epithelial tumor cells from CRC patients in the SMC cohort. Each dot corresponds to the average expression levels of one patient. The box center line represents the median, and the limits represent the first and third quartiles. Whiskers indicate maximum and minimum values. n = 20. A linear model was fitted to the data to assess significance. c Violin plots showing Ezr expression levels in single epithelial tumor cells from patients in the SMC cohort. d−g Expression of Lgr5, Ezr, Rdx and Msn in colon epithelial cells from healthy donors. h Summary scheme showing how membrane to cortex attachment determines different mechanical phenotypes in LGR5+ cancer stem cells and LGR5- cancer cells.