< Back to Articles

Cells Communicate Via the Telephone Game


View Web Version


In Andre Levchenko’s lab at Yale University, an organoid floats in a gel in a well of water. The well contains epidermal growth factor (EGF), and there’s a smidgen more of it on one side than the other. The gradient is subtle, delicate. A cell shouldn’t be able to tell the difference between the sides. Yet after a few days, the organoid begins branching, reaching an arm of cells into the EGF-concentrated side of the well.

Cells use chemical gradients like these as maps to guide their movements and make life decisions. In the 1970s and 1980s, Berg and Purcell calculated the limit of a cell’s ability to sense a shallow gradient, but researchers know much less about how larger groups of cells such as organoids make decisions to move or build organs.

Levchenko and Andrew Ewald at Johns Hopkins University noticed the ability of a cell collective to sense slight chemical differences, but discovered one additional caveat: the bigger organoids were no better at it than the smaller ones. So they turned to biophysicist Ilya Nemenman at Emory University to figure out what was going on and a new theory was born.

3D Printing a New Type of Petri Dish

As a cancer cell biologist, Ewald studies mammary epithelial cells. “The mammary gland carries a network of branched tubes; they’re the ones that will make and carry the milk during lactation. And we knew that in the postnatal animal, a very small fraction of it contains tubes,” he said. “Weeks later in mouse, or months to years later in humans, you end up having this large increase in size and complexity in this ductal network, and we wanted to understand how that worked.”

Levchenko, a systems biologist and biomedical engineer, and Ewald, a cancer cell biologist, teamed up to see how mammary cells responded to gradients of EGF, the signal that helps form the ducts. But exposing these large cell collectives to a shallow gradient would be difficult. They needed a centimeter by centimeter by millimeter chip since the organoids are 100-200 microns across, plus room for extracellular matrix and protein scaffolding.

“Andre and I were sitting there asking, can we build a class of devices that would be big enough for these several hundred micron structures where we could expose them to really precise gradients?” said Ewald.

Levchenko decided to use stereolithography , a type of 3D printing, to build the device. He used a computer to design a mold and printed the mold using Prototherm 12120. They poured a flexible polymer into the mold and baked it overnight.

Conversations Between Cells

With their chambers in hand, the researchers seeded differently sized groups of mammary epithelial cells in an EGF gradient. The cells branched according to the concentration of EGF. “They orient very strongly in even a shallow gradient,” said Ewald.

The big surprise was the gradient was even shallower than they had thought, a change of just 0.2% per 10 microns: “It was in fact below what we would expect single cells to be able to detect,” said Ewald. “Andre Levchenko said, 'Wait a minute, this is way too efficient, much more efficient than single cells.'” When they put solitary cells in the chamber, the cells could not detect this shallow gradient.

“What happens is the cells talk to each other. The reason why they are able to detect gradients that are smaller than they should be is they essentially compare notes,” said Nemenman. “Cells are trying to get information from different far corners of the cell collective and by expanding the range, they are able to sense gradient concentration changes that are smaller than any individual small cell would be able to detect.”

If each cell talked to all the other cells, there would be too much communication going on, so Nemenman and Levchenko devised a theory: when an external signal molecule such as EGF binds to the cell, it triggers activation of two molecules in the cell: one locally active molecule that stays in the cell, and one globally active molecule that can travel from one cell to another through gap junctions. All of the cells produce the same global molecule, so the average concentration of the exterior signal molecule is represented by the interior global molecules that all of the cells contribute. A cell compares the signals from its local molecule to the global molecule to see if it’s at the lower or the higher end of the gradient, and, therefore, determine which direction it should move. It’s almost as if all the cells become one big cell. “This is the first time it has been shown that the Berg and Purcell theory needs to be modified to understand how ensembles of communicating cells behave,” said Levchenko.

The team blocked the gap junctions with four different drugs, and all gave the same answer: the cells stayed healthy, but the directional branched pattern disappeared. In addition, depleting calcium stores disrupted gradient sensing. “It was as though we had very directly taken away their ability to communicate with each other and very directly taken away their ability to sense this shallow gradient,” said Ewald. “We can say for sure that the collective response to the gradient requires gap junctions and calcium release.”

The Telephone Game

Theoretically, this gradient sensitivity should increase indefinitely as the size of the sensing tissue increases, but the team saw that the sensitivity improved only up to a certain size of organoid and then stalled.

They reasoned that noise in the signals within and between cells of the organoid limits sensitivity. “The key additional insight is that chemical communication within a cell or a group of cells is a noisy process, something that was neglected in various embodiments of the Berg and Purcell theory previously,” said Levchenko. Transport of the global molecule is noisy, so the information cannot be communicated over a distance that is greater than a certain limit.

In a separate paper, Levchenko, Nemenman, and Andrew Mugler, now at Purdue University, determined the limit of gradient sensitivity accounting for noise in the internal communication between cells, and also how many cells can compose the organoid before the sensitivity peaks. Surprisingly, they found that sensitivity improves if the local reporter molecule also diffuses, though much more slowly, which decreases measurement noise. They call this new theory regional excitation-global inhibition (REGI). “We argue that this theory allows for a better gradient performance, and can indeed be more consistent with the known biology,” said Levchenko.


1. Ellison D, Mugler A, Brennan MD, Lee SH, Huebner RJ, Shamir ER, Woo LA, Kim J, Amar P, Nemenman I, Ewald AJ, and Levchenko A. “Cell-cell communication enhances the capacity of cell ensembles to sense shallow gradients during morphogenesis,” PNAS, January 20, 2016, E679-688.

2. Mugler A, Levchenko A, and Nemenman I. “Limits to the precision of gradient sensing with spatial communication and temporal integration,” PNAS, January 20, 2016, E689-695.

< Back to Articles