T19 - Programmable Transitions from Orbiting to Matrix Invasion in Epithelial Spheroids via Dynamic Physical Confinement

Introduction: : Epithelial tissues dynamically self-organize from spherical morphologies towards budding or branched architectures during embryonic development, wound healing and tumor invasion. Coordinated orbiting or matrix invasion is shaped by extracellular matrix architecture, but these activities can subsequently remodel the extracellular matrix. How this dynamic reciprocity, whereby the matrix interface and cell migration mutually influence each other, regulates 3D collective cell migration remains poorly understood.
Here, we show that multicellular spheroids orbit and invade based on an asymmetric morphology coupled to localized matrix deformations. Remarkably, orbiting motion drives more uniform rounding of a spheroid, leading to localized matrix deformation at regions of initially high curvature. The concentrated matrix deformations disrupted the collective orbiting and promoted the emergence of radial strands, restoring high curvature at those sites and resulting in morphological similarity between early and late spheroids. We further explored the effects of mechanical cues on this transition. Strikingly, osmotic pressure could reversely program these coordinated dynamics, driving invading strands to retract back into the spheroids or to arrest orbiting motion. Next, we generated “mosaic” spheroids by mixing in cells induced to mesenchymal states via Snail1. In mosaic spheroids, collective orbiting was progressively disrupted while invasive behaviors persisted with the increasing fraction of the EMT-induced cells. Finally, we developed an agent-based model that recapitulates our observations based on geometric perturbations of local curvature and cellular heterogeneity. Overall, this work reveals the dynamic interplay at a spheroid-matrix interface, with implications for epithelial tissue morphogenesis in development and disease.

Materials and

Methods: : MCF-10A stably transfected with H2B-mCherry was a gift from M.R. Ng and J.S. Brugge (Harvard Medical School). MCF-10A variant with tamoxifen-inducible Snail expression expressing H2B-mCherry and cytoplasm-GFP was a gift from D. A. Haber (Massachusetts General Hospital). Snail expression was induced by treatment with 500 nM 4-hydroxy-tamoxifen (4-OHT). Multicellular spheroids were prepared using a hanging drop method where each drop contains 500 cells. Harvested spheroids were embedded in bovine collagen I solution prepared at a final collagen concentration of 6 mg/mL and tracer bead concentration of 0.05 mg/mL. Spheroids embedded in the collagen matrix were imaged using spinning disk confocal microscope. To apply osmotic pressure on the embedded spheroids, growth media containing 4% wt/vol (100 mOsml−1) polyethylene Glycol 400 used.
The velocity profiles of cellular migration and bead displacement were obtained using Farneback optical flow algorithm. Displacement Arrays of Rendered Tractions (DART) analysis was performed on the 2D displacement data obtained from optical flow analysis of tracer beads. A computational agent-based model was constructed by modeling cells as spherically symmetric self-propulsive (active) particles. Cells also experience drag-like forces which is proportional in magnitude to the cube of its speed. Cells interact with other cells and with the ECM via soft repulsion at short distances and adhesion-based attraction at longer scales.

Results, Conclusions, and Discussions:: We first investigated transitions from orbiting to invasion transition based on cellular migration. The cells exhibited minimal migration during the first 6 hours (Pre-orbiting). From 6–12 h, coordinated angular motion was observed (Orbiting), followed by the emergence of invading strands (Post-orbiting) (Fig. A). Coordinated circumferential velocity profile at the orbiting phase and radial invasion ‘hotspots’ at the post-orbiting phase are visualized by optical flow analysis (Fig. B).
Interestingly, invading strands consistently emerged from regions corresponding to initial curvature peaks. Initially, typical spheroids exhibited roundness around 0.9, which later increased to nearly 1.0 as local curvature variations were canceled out during the orbiting phase. Upon transition to invasion, new curvature peaks reappeared at the same site as the initial ones (Fig. C). Matrix deformation analysis revealed localized inward tractions near the initial curvature peaks during orbiting due to the retractions of the spheroid-matrix boundary. The anisotropic matrix deformations built up monotonically, concluding the orbiting phase (Fig. D).
We then sought to manipulate collective migration. Spheroids in isotonic media exhibited the usual orbiting and invasion. However, those in hypertonic conditions remained the same in size and shape over 5 days, arrested their motilities. Remarkably, the distinct behaviors under isotonic and hypertonic conditions were reversible by replacing the media. Even long strands (> 100μm) formed under isotonic condition retracted back into the spheroids upon switching to hypertonic condition (Fig. E). Incorporating more than 30% of EMT-induced cells into the spheroids also disrupted the collective orbiting motions (Fig. F).
Finally, we developed a computational model with tunable parameters for cell-cell and cell-matrix interactions. The model revealed regimes where collective orbiting was sustained or suppressed. Further, this model recapitulated migratory behaviors of mosaic spheroids, aligning with experimental observations.
In conclusion, we show that transitions from orbiting to matrix invasion are mediated by the local curvature of the spheroid-matrix interface, mediated by cell-generated matrix deformations. We manipulate these transitions via volumetric compression via osmotic pressure, showing dynamic arrest or reversal of collective invasion. Overall, this work provides new insights into the role of local asymmetry in driving budding and branching for epithelial tissue morphogenesis.

Acknowledgements and/or References (Optional):: We acknowledge funding from Brown University’s Hibbitt Engineering Postdoctoral Fellowship (JK), NIH R01GM140108 (JK, HJ, AMH, AM, HY, MG, IYW), “la Caixa” Foundation Fellowship 100010434 with code LCF/BQ/EU21/11890128 (C. Falc´o), EPSRC grant P/R014604/1 (C. Falc´o, WDM, JAC), ERC Horizon 2020 Research and Innovation Program Advanced Grant Non-local-CPD 883363 (WDM, JAC), ONR Panther Award N000142212828 (MA, C. Franck), MIT School of Engineering Takeda Fellowship (HY), and ARO W911NF2310385 (IYW).