Human carcinomas are comprised of complex mixtures of tumor cells that are known to compete indirectly for nutrients and growth factors. Whether tumor cells could also compete directly, for example by elimination of rivals, is not known. Here we show that human cells can directly compete by a mechanism of engulfment called entosis. By entosis, cells are engulfed, or cannibalized while alive, and subsequently undergo cell death. We find that the identity of engulfing ("winner") and engulfed ("loser") cells is dictated by mechanical deformability controlled by RhoA and actomyosin, where tumor cells with high deformability preferentially engulf and outcompete neighboring cells with low deformability in heterogeneous populations. We further find that activated Kras and Rac signaling impart winner status to cells by downregulating contractile myosin, allowing for the internalization of neighboring cells that eventually undergo cell death. Finally, we compute the energy landscape of cell-in-cell formation, demonstrating that a mechanical differential between winner and loser cells is required for entosis to proceed. These data define a mechanism of competition in mammalian cells that occurs in human tumors.

These observations suggest that myosin-II along with actin crosslinkers establish local cortical tension and elasticity, allowing for contractility independent of a circumferential cytoskeletal array. Furthermore, myosin-II and actin crosslinkers may influence each other as they modulate the dynamics and mechanics of cell-shape change.

Background Mechanosensing governs many processes from molecular to organismal levels, including during cytokinesis where it ensures successful and symmetrical cell division. While many proteins are now known to be force sensitive, myosin motors with their ATPase activity and force-sensitive mechanical steps are well poised to facilitate cellular mechanosensing. For a myosin motor to experience tension, the actin filament must also be anchored. Results Here, we find a cooperative relationship between myosin-II and the actin crosslinker cortexillin-I where both proteins are essential for cellular mechanosensory responses. While many functions of cortexillin-I and myosin-II are dispensable for cytokinesis, all are required for full mechanosensing. Our analysis demonstrates that this mechanosensor has three critical elements: the myosin motor where the lever arm acts as a force amplifier, a force-sensitive bipolar thick filament assembly, and a long lived actin crosslinker, which anchors the actin filament so that the motor may experience tension. We also demonstrate that a Rac small GTPase inhibits this mechanosensory module during interphase, allowing the module to be primarily active during cytokinesis. Conclusions Overall, myosin-II and cortexillin-I define a cellular-scale mechanosensor that controls cell shape during cytokinesis. This system is exquisitely tuned through the enzymatic properties of the myosin motor, its lever arm length and bipolar thick filament assembly dynamics. The system also requires cortexillin-I to stably anchor the actin filament so that the myosin motor can experience tension. Through this cross-talk, myosin-II and cortexillin-I define a cellular-scale mechanosensor that monitors and corrects shape defects, ensuring symmetrical cell division.