Research Article

Seeing around corners: Cells solve mazes and respond at a distance using attractant breakdown

See allHide authors and affiliations

Science  28 Aug 2020:
Vol. 369, Issue 6507, eaay9792
DOI: 10.1126/science.aay9792

You are currently viewing the abstract.

View Full Text

Log in to view the full text

Log in through your institution

Log in through your institution

You can find your own way

Migration of cells through tissues and embryos is often steered by gradients of attractive chemicals in a process called chemotaxis. Cells are best at navigating complex routes, for which they use “self-generated chemotaxis” and create their own attractant gradients. An example of this is when neutrophils migrate into tissues to attack infection. Using modeling and live-cell data, Tweedy et al. found that self-generated chemotaxis allows cells to obtain surprising amounts of information about their environment. Cells of the slime mold Dictyostelium discoideum and mouse pancreatic cancer–derived cells were able to use the diffusion of attractants to identify the best route through complex mazes, even when the correct path was long and twisted, without ever entering incorrect paths.

Science, this issue p. eaay9792

Structured Abstract


Cells migrate over long distances during normal embryonic development and in disease, and they navigate through complex, branched paths. Chemotaxis, cell migration steered by gradients of attractive chemicals, is a central regulator of this behavior. Its sensitivity is limited by the responsiveness of attractant receptors. Simple chemotaxis is unable to guide cells over long distances and cannot resolve complexities and branches.

Self-generated chemotaxis, a process in which cells create their own local, dynamic gradients by breaking down an attractant in their environment, is different. Gradients can be steeper because they are only made in the vicinity of the cells. They also work over long distances because they are remade locally as the cells migrate. Furthermore, self-generated gradients can follow complex paths because they are constantly generated by interactions between the cells and their environment.


The migration of cells through complex structures such as developing embryos or vascularized tumors involves both long distances and numerous decisions about which path to follow. Therefore, we examined chemotaxis in artificial complex environments, which are represented here by mazes. We predicted the pathfinding of cells in mazes using computational and mathematical modeling of self-generated gradients and tested the outcomes with real cells in microfluidic mazes.


We modeled cells as they navigated through various junctions between a connection to an attractant well and a dead end. We then tested the model’s predictions with cells that typically use long-range navigation: Dictyostelium discoideum cells (which find each other over large distances in the environment) and metastatic cancer cells (which spread around the human body). Both successfully solved a range of mazes, even remarkably complex ones (see figure), and could identify optimum paths. The results confirm that a model of self-generated chemotaxis captures all essential decision-making in these mazes: Cells broke down attractants locally, which were replenished at different, predictable rates by each possible path. The resulting local attractant gradients allowed cells to sense upcoming maze junctions before reaching them, even around corners, and to make correct decisions about paths they had not yet encountered. Attractant breakdown thus allowed cells to create a detailed picture of their surroundings.

Real cell behavior closely agreed with the models for both rapidly moving D. discoideum and slower-moving pancreatic cancer cells, implying that the underlying conceptual model is an effective explanation of general cell behavior. In particular, the diffusivity of attractant, which is ignored by simple models, is crucial to cells’ responses. We simulated and then constructed “easy” and “hard” mazes, which appear similar but caused radically different pathfinding by cells. Finally, we designed maze geometries that passively created a temporary peak of attractant flux, deliberately misdirecting cells into a dead end and creating a chemotactic “mirage.”


Self-generated chemoattractant gradients allow cells to navigate complex paths with great efficiency. Diffusion and attractant breakdown allow cells to obtain detailed information about their surroundings that could not be provided by simple attractant gradients. Cell migration in vivo cannot be understood without considering the interplay among cells, attractants, and the structure of the local environment. Microfluidic mazes are an excellent experimental system for studying these interactions.

Cells solving mazes by using self-generated chemotaxis and identifying a shortcut.

(Top) D. discoideum solving a representation of the maze from Hampton Court Palace. (Bottom) Cells sensing a shortcut using the self-generated gradient and selecting a new minimal route. Cells enter from the left and exit at the right. Time is represented by cell color; blue shows early times and red later times.


During development and metastasis, cells migrate large distances through complex environments. Migration is often guided by chemotaxis, but simple chemoattractant gradients between a source and sink cannot direct cells over such ranges. We describe how self-generated gradients, created by cells locally degrading attractant, allow single cells to navigate long, tortuous paths and make accurate choices between live channels and dead ends. This allows cells to solve complex mazes efficiently. Cells’ accuracy at finding live channels was determined by attractant diffusivity, cell speed, and path complexity. Manipulating these parameters directed cells in mathematically predictable ways; specific combinations can even actively misdirect them. We propose that the length and complexity of many long-range migratory processes, including inflammation and germ cell migration, means that self-generated gradients are needed for successful navigation.

View Full Text

Stay Connected to Science