Mouse glioma cell line CT-2A (Milipore Sigma) with characteristics of GBM phenotypes were used. CT-2A cells were cultured in Dulbecco's modified Eagle's medium (DMEM, Sigma-Aldrich) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Gibco), 2 ​mM glutamine (Gibco) and 1% penicillin/streptomycin (Gibco) in a 37 ​°C and 5 ​% CO₂ incubator. Fresh media was replaced every 2 days, and all cells were cultured for no more than 6 passages.

PA hydrogels were prepared according to a previously reported protocol [44]. Prior to gel being cast, glass coverslips (22 ​mm, Fisher Scientific) were attached to a Petri dish with a punctual hole of 13 ​mm diameter using Polydimethylsiloxane (PDMS, Dow Corning) of mass ratio 10:1 (Base Elastomer:curing agent) and baked in an 80°C oven. The coverslips attaching to petri dishes were then treated with 0.1 ​M sodium hydroxide (NaOH, Sigma-Aldrich) solution and were heated till the entire coverslips were covered with a uniform monolayer of NaOH. NaOH-coated coverslips were further functionalized with 3-aminopropyltriethoxysilane (Sigma-Aldrich) for 5 ​min and then rinsed three times with distilled water. The coverslips were then incubated with 0.5% glutaraldehyde (Sigma-Aldrich) for 30 ​min, washed three times with distilled water, and then dried at room temperature. An 18 ​μl mixture of 40% acrylamide, 2% bis-acrylamide (Sigma-Aldrich), 0.1% ammonium persulfate (APS, Sigma-Aldrich), and 0.1% tetramethylethylenediamine (TEMED, Sigma-Aldrich) were added to the center of the coverslips, and then covered by another 18 ​mm coverslip that was hydrophobically treated with Dichlorodimethylsilane (Sigma-Aldrich). To prepare PA gels for traction force microscopy, 0.2-μm, yellow-green fluorescent (505/515) microbeads (Thermo Fisher Scientific) were added to the 18 μl gel solution and mixed thoroughly before adding to the coverslips. After 60 ​min of polymerization, the top 18 ​mm coverslips were peeled off to have the PA gel attached on the 22 ​mm coverslips attaching to petri dishes.

PA gel substrates were first activated by treating with 200 ​μL of 1 ​mM Sulfo-SANPAH (sulfosuccinimidyl 6-(4′-azido-2′-nitrophenylamino)hexanoate, Pierce) in 50 ​mM HEPES (N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid, Gibco) buffer under ultraviolet (UV) light for 20 ​min followed by washing with phosphate-buffered saline (PBS) thrice, and were then dried at room temperature. PDMS stamps with different geometric patterns were first fabricated using standing photolithography and replica molding [45]. PDMS stamps were functionalized with a fibronectin solution (50 ​μg/ml; Sigma-Aldrich) for 60 ​min at room temperature, then were put into conformal contact with activated PA gels for 1 ​min and peeled off. PA gels functionalized with different patterns of fibronectin proteins were then washed with PBS twice and stored in PBS for experiments within one week. Prior to cell loading, functionalized PA gel substrates were sterilized under UV light for 15 ​min in PBS and for 15 ​min in cell culture media. 15,000 CT-2A cells were then seeded onto each sterilized PA gel substrate. These loaded cells would attach to the micropatterned adhesive areas and be cultured for 3 days in an incubator to form different geometric patterns of cell monolayers.

To measure traction forces generated from cells in the 2D cell monolayer, fluorescent microbeads embedded in PA gels underneath the patterned cells were imaged firstly to recode the position of each microbead using a Zeiss Observer microscope with a 40 ​× ​objective. Then cells cultured on PA gels were lysed by adding 100 ​μL of 1 ​N NaOH to the cell culture dish, and fluorescent images of the same locations were taken again to record the positions of these fluorescent microbeads as a control. The two sets of fluorescence microscopic images prior and post cell lysis and a phase contrast image of cells were used to measure bead displacement using particle image velocimetry. The resultant displacements were then used to calculate the traction force using Fast Fourier transform cytometry [46].

Cells were fixed with 4% Paraformaldehyde (Electron Microscopy Sciences) for 20 ​min and then permeabilized with 0.3% Triton X-100 (Roche Applied Science) for 10 ​min at room temperature. After permeabilization, cells were blocked with blocking buffer for 60 ​mins at room temperature. Primary antibodies (Table S1) were incubated with cells overnight at 4°C or 60 ​min at room temperature. Goat-anti mouse Alexa Fluor 488 and/or goat-anti rabbit Alexa Fluor 546 IgG secondary antibodies (Invitrogen) were used as secondary antibodies. The immunostained cells were imaged under an inverted fluorescent microscope (Zeiss Axio Observer.Z1).

To mimic the mechanical constraints experienced by cancer cells, we employed a 2D multicellular model of GBM by micropatterning of mouse GBM cells (CT-2A, Sigma Aldrich) into geometrically confined monolayers (Fig. 1a). Using a standard microcontact printing technique (Fig. S1) [45], mouse GBM cells were patterned into different shaped monolayers on PA hydrogel substrates with an elastic modulus of 3 ​kPa (Fig. 1b). PA gels were embedded with fluorescent microbeads to measure the traction force distribution within micropatterned multicellular monolayers [47]. We first examined the spatial distributions of CSCs in 500 ​μm-diameter circular GBM cell colonies by measuring expressions of key CSC makers [48], including canonical self-renewal transcription factors (Nanog, Oct4), CSC membrane marker (CD133) and intermediate filament marker (Nestin) after 3 days of cell culture (Fig. 1c). The fluorescent intensity of CSC makers on each single cell in a 2D colony after immunostaining was measured and presented as a heat map (Fig. 1d). The CSC surface maker analyses showed non-uniform spatial distributions of GBM CSCs in the micropatterned colonies, whereas GBM cells that reside in peripheral regions showed elevated expressions of all four CSC markers (Nanog, Nestin, Oct4, and CD133) compared to the those in the proximal regions (Fig. 1e-f).

To gain a deeper understanding of the mechanical factors that drive the transition of GBM cells into CSCs, we employed fluorescent bead-based traction force microscopy [47] to map the cellular traction force distributions within the micropatterned multicellular monolayers (Fig. 1g). The force analyses showed a strong correlation between the spatial distributions of CSCs and traction force in the micropatterned GBM colonies. In consistent with the measured trends of CSC marker expressions, traction forces generated by GBM cells increased from the central regions to the peripheral regions of the circular patterns (Fig. 1h). Such gradients of cellular force distributions in the patterned cell colonies can be explained by the unique mechanical constraints due to geometric confinement, as predicted by a finite-element model of multicellular sheet mechanics (Fig. 1g) [49]. We further investigated whether varying geometries of the multicellular patterns would alter the spatial distributions of CSCs and force maps. The same phenomenon was observed in all types of sizes (100, 300, and 500 ​μm-diameter circular) and geometries (circular, rectangular, T-shape) of cell patterns, whereas CSCs emerged more in the peripheral regions than in the central regions of micropatterned GBM cell monolayers (Fig. S2a-b). Moreover, traction forces generated by cells in the peripheral regions were all significantly higher than those in the inner regions (Fig. S2c-d). Meanwhile, such spatial CSC distribution was not observed under a nonconfined condition, indicating that geometric confinement is an inductive cue for emergence of CSCs (Fig. S2a-b).

We explored how the interplay between the cell-cell interaction and cell-ECM interaction under mechanical constraints mediates the force-driven emergence and spatial patterning of CSCs in the geometric confined GBM cell colonies. As cells in the GBM colonies interact with surrounding cells and ECM, mechanical signals are transmitted by cell adhesive molecules including integrins and cadherins to actomyosin cytoskeleton, thus regulate cellular forces and downstream mechano-responsive signaling [33] and mediate CSC phenotype in GBM [37,38]. We conducted immunostaining to assess the expression of cell adhesion molecules that are crucial for cell-cell adhesion (E-cadherin, N-cadherin, P-cadherin) and cell-ECM adhesion (integrin α₅β₁) (Fig. 2a). The results showed that the expressions of N-cadherin and P-cadherin increased significantly from the center to the peripheral regions of the patterned cell colonies, while E-cadherin expressed uniformly in the colonies (Fig. 2b). Such findings were expected, since E-cadherin expression has been found to be scarce in glioma cells, while N-cadherin upregulation has been widely exploited in glioma prognosis [50]. In addition, integrin adhesome takes a critical role in transmitting mechanical signals from the ECM to the cytoskeleton and regulating actomyosin-mediated traction force exert by the cells on the ECM components [51]. Previous studies have revealed that abnormal expression of integrin α₅β₁ in GBM cells results in a pathological alteration of cell-ECM interactions, and impact angiogenesis, migration, proliferation and therapy resistance of GBM [52,53,54]. Our results proved that integrin α₅β₁ expression in the patterned cell colonies increased significantly from the center to the peripheral regions, which implicated the corresponding force gradients in the patterned GBM colonies (Fig. 2a-b). The regression analysis between expression of adhesion molecules (E-cadherin, N-cadherin, P-cadherin, and integrin α₅β₁) and the CSC marker Nestin further showed that the expression level of Nestin in GBM cells in the colony is positively correlated with the expression levels of these adhesion molecules N-cadherin, P-cadherin, and integrin α₅β₁ but not E-cadherin (Fig. 2c).

Furthermore, we utilized pharmacological treatments to interrupt the cell-cell and cell-ECM interactions to delve into the mechanotransductive mechanisms underlying the emergence and spatial patterning of CSCs in the geometric confined GBM colonies. We used a low calcium medium, which effectively suppressed the adhesion activities of N-cadherins and P-cadherins, thereby interrupting the cell-cell interactions [55]. The traction force microscopy measurements revealed that disrupting cadherins greatly attenuated the traction force gradient in the GBM colonies, while the total traction force remained the same (Fig. 2d-e). We also applied an integrin α₅β₁ antibodies to block the cell-ECM adhesion and to disrupt the traction force that the cells exert on the PA substrate [31]. The force analyses suggested that while inhibition of integrin α₅β₁ reduced the overall magnitude of traction forces in most cells, the force gradient remains in the GBM colonies (Fig. 2d-e). Moreover, we directly inhibited cellular traction forces with Blebbistatin (20 μM), an inhibitor of myosin-II-specific ATPase [54]. Inhibition of actomyosin force also attenuated the force gradient in the GBM colonies (Fig. S3a-c). We then checked whether disrupting the mechanical constraints in the geometric confined GBM colonies would attenuate the emergence and spatial distribution of CSCs. Blocking cell-cell, cell-ECM interactions and traction force reduced the overall percentages and attenuated spatial distribution of CSCs in the GBM colonies of according to all four CSC markers (Nanog, Nestin, Oct4, and CD133) (Fig. 2e, Fig. S3d). Overall, our results suggest that the emergence and spatial patterning of CSCs require the development of gradient distribution of traction force mediated both by the cadherin-mediated cell-cell interaction and integrin-mediated cell-ECM interaction.

We next explored the Piezo1-mediated mechanotransductive mechanisms underlying the force-driven emergence and spatial patterning of CSCs in the geometric confined GBM cell colonies. Upregulation of Piezo1 mechanosensitive ion channel has been implicated in the growth of GBM stem cells and tumor development [41]. A previous study has also revealed a feedforward mechanism mediated by Piezo1 and integrin-focal adhesion kinase (FAK) signaling that promote glioma aggression by remodeling tumor mechanics [43]. We first examined the expression of Piezo1 in the 2D confined GBM cell colonies (Fig. 3a) and found that Piezo1 was prominently concentrated in the peripheral regions of the micropatterned colonies (Fig. 3b). As indicated in Fig. 3c, there is a positive correlation between Piezo1 expression and CSC marker Nanog, a transcription factor that promote GBM malignant phenotype [56]. Upon its activation by either mechanical or chemical stimulator, Piezo1 opens and allow influx of calcium ion (Ca²⁺) into the intracellular domain and regulate of cellular contraction [57]. We transfected the cell monolayer with Fluo4 AM, a calcium indicator, and found that calcium mostly concentrated in the peripheral regions of the micropatterned colonies (Fig. 3d-e). The transient activities of calcium upon Piezo1 activation were also tested using Yoda1, a Piezo1 chemical agonist [58]. Upon addition of Yoda1, we observed an immediate rise in intracellular calcium in all regions of the GBM colonies (Fig. 3f). However, the rise in intracellular calcium in cells in the peripheral regions was still significantly higher compared to cells in the central regions, which could be explained by the overexpression of Piezo1 in the peripheral regions. From these results, we confirmed that Piezo1 signaling plays a critical role in regulating GBM CSC phenotype.

Next, we tested whether focal adhesions and RhoA activities collaborate with Piezo1 to drive traction force elevation in GBM cells and CSC phenotype switch. This spatially segregated Piezo1 channel activity promotes efficient assembly and signaling of integrin-mediated focal adhesions, which serve as highly localized subcellular domains for Piezo1 to perceive mechanical stimuli at both the cell surface and the cell intracellular structures [43,59]. As an integrin-interacting molecule, Piezo1 also took part in assembly of focal adhesion [59]. From the co-staining of Piezo1 and FAK, we observed that Piezo1 colocalizes with FAK in individual CSCs (Fig. S4). Moreover, we found a strong correlation between Piezo1 expression and RhoA activities in the GBM cell colonies (Fig. 3g-i). RhoA is a downstream effector of intracellular calcium influx and initiate actin polymerization and thus generate actomyosin-mediated traction force [59]. Activation of Piezo1 and downstream RhoA activities mostly occur in the peripheral regions of the GBM cell colonies (Fig. 3i). These findings suggest that mechanosensation of the mechanically constrained microenvironment by Piezo1 triggers calcium influx in GBM cells, which in turn promotes RhoA activities to drive assembly of integrin-mediated focal adhesion and enhance traction force more in the peripheral regions of the cell colonies.

We next studied the interplay between integrin, cadherins, and Piezo1 in regulating mechanosensation of GBM cells to the mechanically constrained microenvironment (Fig. 4a). Interestingly, we first found that blocking of integrin α₅β₁ with antibodies significantly affected the overall Piezo1 expression and distribution in the micropatterned GBM cell colonies, blocking cadherins by using low calcium media did not alter Piezo1 expression in cells in the peripheral regions (Fig. 4b). To confirm whether Piezo1 recruitment precedes the change in traction force, we antagonized expression of Piezo1 using pharmacological treatment with 2.5 μM of GsMTx4 peptide for 1 h. Gsmtx-4 at μM concentration has been proven to inhibit Piezo1 mechanosensitive functions by 80% [59]. The efficacy of GsMTx4 was confirmed by a significant decrease in Piezo1 expression compared to the non-treated group, as demonstrated in Fig. 4c and d. We observed not only the overall intensity of Piezo1 decreased, but also the disappearance of the localization of Piezo1 at the edge of the circular GBM cell colonies. Furthermore, upon inhibition of Piezo1, we found the disruption in the distribution gradient of integrin α₅β_1, where the fluorescent intensity of integrin α₅β₁ in cells in the peripheral regions significantly decreased (Fig. 4d), implying the mutual interaction of Piezo1 and integrin adhesion during CSCs’ mechanosensing to mechanical constraints. Meanwhile, blocking Piezo1 did not affect N-cadherin activities among the GBM micropatterns and vice versa, as the intensity of N-cadherin in the cells located at the edge was still higher compared to those at the center region. Inhibition of Piezo1, along with the associated alleviation of integrin activities, resulted in the decrease in traction forces in GBM cell colonies and disappearance of the spatial patterning of CSC marker Nanog, confirming a Piezo1 force-driven mechanism in regulating the organization of GBM CSCs within the tumor (Fig. 4e-f).