Constitutive activation of myosin-dependent contractility sensitizes glioma tumor-initiating cells to mechanical inputs and reduces tissue invasion

Abstract

Tumor-initiating cells (TIC) perpetuate tumor growth, enable therapeutic resistance, and drive initiation of successive tumors. Virtually nothing is known about the role of mechanotransductive signaling in controlling TIC tumorigenesis, despite the recognized importance of altered mechanics in tissue dysplasia and the common observation that extracellular matrix (ECM) stiffness strongly regulates cell behavior. To address this open question, we cultured primary human glioblastoma (GBM) TICs on laminin-functionalized ECMs spanning a range of stiffnesses. Surprisingly, we found that these cells were largely insensitive to ECM stiffness cues, evading the inhibition of spreading, migration, and proliferation typically imposed by compliant ECMs. We hypothesized that this insensitivity may result from insufficient generation of myosin-dependent contractile force. Indeed, we found that both pharmacologic and genetic activation of cell contractility through RhoA GTPase, Rho-associated kinase, or myosin light chain kinase restored stiffness-dependent spreading and motility, with TICs adopting the expected rounded and nonmotile phenotype on soft ECMs. Moreover, constitutive activation of RhoA restricted three-dimensional invasion in both spheroid implantation and Transwell paradigms. Orthotopic xenotransplantation studies revealed that control TICs formed tumors with classical GBM histopathology including diffuse infiltration and secondary foci, whereas TICs expressing a constitutively active mutant of RhoA produced circumscribed masses and yielded a 30% enhancement in mean survival time. This is the first direct evidence that manipulation of mechanotransductive signaling can alter the tumor-initiating capacity of GBM TICs, supporting further exploration of these signals as potential therapeutic targets and predictors of tumor-initiating capacity within heterogeneous tumor cell populations.

Figure 1. Human primary glioblastoma tumor-initiating cells (GBM TICs) retain stem-like properties when cultured on soft and stiff polyacrylamide gels coated with laminin

(A) ECM stiffness does not select against slow-cycling GBM TICs. GBM TICs were treated with carboxyfluorescein diacetate succinimidyl ester (CFSE) and analyzed by flow cytometry to determine distribution of cycling rate. Based on the persistence of the high-CFSE shoulder (arrow), the slow cycling population is retained on all gel stiffnesses. L2 curves are shown; L0 displayed similar results. (B) GBM TICs also express the stem cell markers,nestin (top row; red; 100% positive) and SOX2 (bottom row; red; 100% positive) on all ECM stiffnesses. (C) All ECM stiffnesses permit neuronal and glial differentiation. Addition of 5% serum to GBM TIC cultures results in differentiation into glial (GFAP, green) and neuronal lineages (βIII tubulin, red).

Figure 2. Tumor-initiating cells can spread, migrate, and proliferate on soft and stiff 2-D ECMs

(A) Effect of ECM stiffness on cell spreading. GBM TICs were cultured on laminin-coated polyacrylamide matrices with stiffnesses ranging from 0.08kPa to 119 kPa, and on laminin-coated glass. Phase contrast imaging reveals that GBM TICs can spread on ECMs of all stiffnesses. (B) Quantification of projected cell area shows that ECM rigidity does not regulate L2 TIC spreading area. n = 20 cells (pooled from at least 3 technical and 3 biological experiments) for all conditions. L0 TICs (not shown) exhibited qualitatively similar data. (C) GBM TICs do not form prominent stress fibers or focal adhesions on stiff ECMs. Cells were fixed and stained for the focal adhesion marker vinculin (red), F-actin (green), and nuclear DNA (blue). (D) Effect of ECM stiffness on cell motility. Time-lapse phase contrast imaging of random cell motility over 8–12 hours indicate that the migration speed depends very weakly on ECM stiffness, with a modest optimum at 0.8 kPa. *p<0.05 and **p<0.001 relative to glass; n > 80 cells for all conditions. (E) Effect of ECM stiffness on cell proliferation. Cells were incubated with BrdU for 45 minutes and stained with 7AAD before analysis by flow cytometry. Plots show sample means with standard error. n > 3 independent experiments for all conditions.

Figure 3. Increased Rho GTPase activation leads to stiffness dependent spreading and motility

(A) Genetic strategy for inducible activation of contractility. A constitutively active (CA) mutant of RhoA (or CA ROCK/CA MLCK as noted) was placed under the control of a doxycycline-inducible promoter, placed into a lentiviral vector, and used to transduce GBM TICs to yield stable cells lines. (B) RhoA induction increases myosin light chain phosphorylation. Western blot analysis confirms that induction of CA RhoA increases myosin light chain phosphorylation relative to control cells transduced with an empty vector. Data shown are for the L0 TIC line. n = 3 independent experiments. (C) CA RhoA preserves or enhances cell stiffness. AFM measurements reveal that overexpression of CA RhoA preserves mean cell stiffness in L0 cells (while pushing the range of observed stiffness values to higher values) and increases mean cell stiffness in L2 cells, consistent with an increase in cell contractility. n = 10–20 cells pooled from at least 3 biological and technical replicates. (D) CA RhoA restores ECM stiffness-dependent spreading. Phase contrast imaging reveals that overexpression of CA RhoA produces cell rounding on soft ECMs, in contrast to control-transduced cells, which spread on ECMs spanning the entire range of stiffness. (E) Quantification of projected cell area shows that CA RhoA rescues stiffness dependent cell spreading, with significantly decreased spread area for CA RhoA TIC on soft ECMs. *p<0.05 and ⁺p<0.001 relative to glass. n = 30 cells (pooled from at least three technical and three biological experiments) for all conditions. (F) CA RhoA rescues stiffness dependent cell motility. Time-lapse phase contrast imaging of random cell motility over 8–12 hours show significantly decreased motility of these cells on soft ECMs relative to stiff ECMs. *p<0.05 and **p<0.001 relative to glass. n=30 cells for all conditions. (G) CA RhoA does not alter cell proliferation. Cells were incubated with BrdU for 45 minutes and stained with 7AAD before analysis by flow cytometry. The data reveal that overexpression of CA RhoA does not produce statistically significant changes in proliferation for any given ECM stiffness nor does it render proliferation sensitive to ECM stiffness. n = 3 independent experiments. (H) CA RhoA TICs express the stem cell markers nestin (top row; red; 100% positive) and SOX2 (bottom row; red; 100% positive) on all ECM stiffnesses. (I) CA RhoA does not compromise the ability of TICs to form neurospheres *p<0.001 relative to control, n = 384 TICs. (J) Quantification of neurosphere size for L0 and L2 cells as a function of RhoA expression. *p<0.001 and ⁺p<0.05 relative to control, n = ~ 1500 spheres for all conditions.

Figure 4. Non-muscle myosin II regulates tumor-initiating cell mechanosensitivity

(A) CA RhoA-mediated suppression of cell spreading can be rescued with pharmacologic inhibition of myosin-II. Addition of 25 μM of blebbistatin results in spreading of CA RhoA GBM TICs on soft ECMs, similar to the control GBM TICs. (B) Quantification of projected cell area shows that addition of blebbistatin reverses CA RhoA effects, and caused CA RhoA TICs to spread on soft ECMs. *p<0.05 relative to glass, n=30 cells pooled from at least three technical and three biological experiments for all conditions. (C) Quantification of blebbistatin effects on cell migration speed. Time-lapse phase contrast imaging of random cell motility was conducted over 8–12 hours. *p<0.05 relative to glass, n=30 cells for all conditions. (D) CA RhoA TIC proliferation is not systemically altered by addition of blebbistatin. Cells were incubated with BrdU for 45 minutes and stained with 7AAD before analysis by flow cytometry. Plots show sample means with standard error. n= 3 independent experiments for all conditions. (E) Live/dead assay with propidium iodide staining shows that TICs remain viable in the presence of blebbistatin. (F) Western blot analysis of myosin II heavy chain isoforms A, B, and C shows that control and CA RhoA TICs express all three isoforms on all ECM stiffnesses (L2 shown; L0 had similar results). (G) Quantificatin of Western blot data, n>2 independent experiments. (H) Relative levels of myosin II heavy chain isoforms measured by liquid chromatography tandem mass spectroscopy. n=2 experiments per condition.

Figure 5. Tumor-initiating cells overexpressing other contractile activators also exhibit stiffness dependent motility and spreading

Lentiviral transduction of L0 TICs with (A) CA myosin light chain kinase (MLCK) and (B) CA Rho-associated kinase 1 (ROCK1) resulted in increased average cell speed on stiff ECMs compared to soft ECMs and restore mechanosensitive migration speed. *p<0.003 relative to glass, n = 10–20 cells pooled from at least 3 biological and technical replicates per condition.

Figure 6. Increased Rho GTPase activation limits invasion and motility in 3-D

(A) CA RhoA prevents GBM TICs from invading 3D collagen ECMs. Images depict spheroids formed from CA RhoA or control GBM TICs, implanted in 1.0 mg/ml collagen gels, and captured by phase contrast imaging 3 days later. n > 8 spheroids per condition for all conditions. (B) CA RhoA reduces transwell migration of GBM TICs. Transwell inserts were coated with laminin, and CA RhoA and control cells were allowed to migrate through pores for 24 hours before fixation. A 2x gradient of soluble EGF was used to drive chemotaxis through the pores, and the total number of cells migrated per transwell was counted. Data shown are for the L2 TIC line; L0 TICs exhibited qualitatively similar data.

Figure 7. Constitutive activation of RhoA in GBM TICs extends survival time and reduces tumor invasion in an orthotopic xenograft model

(A) Mice orthotopically implanted with L0 CA RhoA GBM TICs live significantly longer than mice implanted with L0 control TICs. Eight-week-old NSG mice were implanted intracranially with 200,000 cells. Mice were treated daily with doxycycline (625 mg/kg) 4 days post implant until endpoint. **p=0.0018, log rank test. n = 5 mice per condition. (B) CA RhoA results in a 30% increase in overall mouse survival. **p=0.0042, t-test. (C) Tumors formed from L0 CA RhoA GBM TICs do not infiltrate brain tissue as extensively as L0 control TICs. H&E stain shows (i) a large area of infiltration and secondary tumor formation (*) in mice implanted with L0 control TICs; and (ii) a defined area of infiltration with clear borders and no secondary tumor formation in mice implanted with L0 CA RhoA GBM TICs. (D) CA RhoA reduces diffuse invasion in GBM. H&E (i,v) and human nestin stains (iv,viii) reveal a clear border between CA RhoA tumors and the parenchyma whereas control GBM TICs infiltrate readily. Ipsilateral (ii) and contralateral (vi) hemisphere H&E stains show secondary tumor formation (*) for control cells and not for CA RhoA cells. (E) Human GBM cells were identified using anti-human Nestin antibody and visulaized using the ABC-Elite peroxidase method (Vector Laboratories). Counterstaining of the nuclei was performed using hematoxylin. Analysis of cros-sectional area shows significantly decreased tumor occupancy for mice brains implanted with L0 CA RhoA TICs compared to L0 control TICs. Graph shows average +/− SE over 5–8 sections per group. ***p<0.0001. (F) Ki67 staining and quantification shows no difference in cell proliferation across multiple regions in control and CA RhoA GBM TICs.