Making Cellular Memories

Cells permanently change their behavior in response to temporary changes to the environment, a kind of biological memory that controls processes as important and complex as how stem cells differentiate into specific tissues or how the immune system "remembers" dangerous pathogens. At its simplest, cellular memory is achieved with a positive feedback loop--once activated by some external signal, the feedback loop will continually activate itself, even as the cell divides and the signal is taken away. In synthetic biology we can recreate such simple feedback loops, genetic circuits built of parts that activate in response to signals and keep turning themselves on, remembering past chemical events. A few years ago in my lab, David Drubin and Caroline Ajo-Franklin made such a synthetic genetic device in yeast that remembers if the cells have been grown in media containing the sugar galactose.

They connected genetic elements that turn on when the cells taste galactose to a protein that fluoresces red and to a protein that activates the synthetic positive feedback loop. The positive feedback loop in turn is made of a protein that fluoresces yellow and the protein that activates the loop, starting a permanent cycle of feedback. In the video below you can see the red protein turn on and then off again as the galactose is removed, but the yellow protein (which looks green in the video) stays on in the population, even as the cells divide and grow:

We don't usually need to know whether or not cells have tasted galactose, but the parts of the synthetic circuit are modular, they can be swapped out and activated with different triggers to create larger networks of interconnected feedback loops to create more complicated behaviors, or to ask questions about cell biology.

In galactose almost all of the cells responded and turned on the feedback loop, but in more natural conditions, for example in tissues in the body or in mixed populations of microorganisms in the environment, cellular responses to signals are rarely uniform. When radiation or carcinogens damage cells' DNA in a tissue, some cells may have more mutations and more strongly activate the cellular stress response to fix their DNA. These different responses to DNA damage between different cells show up even in populations of single-celled organisms and can have implications for how we understand cancer progression, where a cell's response to DNA mutation can have an impact on whether or not that cell starts dividing out of control.

My awesome labmate Devin wanted to use synthetic memory to be able to track yeast cells that "remembered" having experienced significant DNA damage, to study how they are different from their neighbors that escaped with minimal mutations. She swapped out the genetic part that tastes galactose in the old yeast memory circuit to one that turns on when the cell's DNA is mutated by radiation or chemical carcinogens, cutely and somewhat strangely named HUG1. When she poisoned the yeast cells that had the synthetic memory in place with EMS, a chemical that causes DNA mutations, she saw something very similar to the previous memory circuit. The red fluorescent protein (RFP) stayed on for a short time after the carcinogen was washed off of the cells, but the yellow protein (YFP) stayed on for several days after that, identifying cells that remembered the DNA damage and their offspring.


This is already pretty cool, but the really interesting part of the story started when she started to study how the cells that remembered the damage were different from the ones that didn't. Because the memory cells were fluorescent, she could use fluorescent activated cell sorting to sort the two populations, the dim cells that didn't experience as many mutations from the bright cells that were harder hit.

While the dim cells were indistinguishable from yeast cells that had never been mutagenized, the fluorescent cells grew much slower and had a very different mutation rate. As cells divide and grow there is always a low percentage of mutations, even without carcinogens, errors that arise in the copying of the DNA. We can get an estimate of how many random mutations there are by seeing how many cells out of a billion will mutate to a different behavior, like being able to grow in conditions that are typically bad for the cell. Devin found that even many generations after the mutagen was removed, the fluorescent cells that remembered the chemical had a much lower rate of mutation. Remembering the past DNA damage left them hyper-vigilant against future mutations, keeping the stress response that can fix mutations active and the mutation rate low.


This hormesis--a beneficial response to a low dose of a toxin--would be impossible to detect without the synthetic memory, especially since the slower growth rate of the memory cells would dilute them out of the population as the cells grew. Synthetic biology often uses metaphors from computers to help understand and promote the field, including when it comes to memory. But biological memory can do a lot more than store bits, it can help us to understand something fundamental about how cells work.

You can check out Devin's amazing paper here: "Synthetic circuit identifies subpopulations with sustained memory of DNA damage," Genes and Development 25: 434-439, 2011.

More like this

One day, your iPod will be made out of biological flesh. Just kidding. In general, I'm a pretty staunch skeptic of The Singularity, but I've got to admit that experiments like this are pretty rad: A team in Silver's HMS lab led by Caroline Ajo-Franklin, now at Lawrence Berkeley National Laboratory…
We do a pretty good job at appreciating the visible intricacies of nature: the antennae and legs and claws of a lobster, the geometrical order of the spots on a butterfly's wings. But a lot of nature's intricacies are hidden away inside single-celled creatures, such as the baker's yeast that makes…
Last month I wrote about my friend Devin Burrill's paper about synthetic memory in yeast cells. There were a lot of really interesting questions left in the comments, and I asked Devin if she would write a guest post to answer them. She agreed and here it is, answers to your questions straight from…
My friend Devin's essay, "Making Cellular Memories" just came out today in Cell (and it's open access!) It's a great look at how natural cells and synthetic biological systems use feedback loops in order to maintain memory of past events. Like whole animals, cells need to remember what happened to…

Super interesting!

I feel like there must be a way to build an organic computer out of this.

The cited paper mentions "heritable damage responses." Is damage restricted to random acts of nature, or can there be such a thing as self damage for some unknown purpose that will nevertheless be heritable?

Do the cells that retain this memory of an experience then pass on that memory to those they have been divided into. Inheritance of acquired memories of experience perhaps? And how many generations does this affect? And can the effects of this experience ever be completely erased from the genome if the experience itself is replicated or repeated in a particular environment? And is this perhaps one of the ways that cells evolve to anticipate and deal strategically with a multitude of problems?

I see two people have left interesting questions here, but there have been no interesting answers. Are such questions then to be treated as rhetorical?

hello commenters! this is devin burrill, the first author on the paper, "synthetic circuit identifies subpopulations with sustained memory of DNA damage." christina has notified me that she's been getting a lot of questions, and so we have decided to write a short blog post that answers all of these questions. keep a look out for it later this week here on christina's' blog! and thank you for all of the interest!

By devin burrill (not verified) on 03 May 2011 #permalink

A very interesting article, I haven't finished reading but I can't help speculate how this work can be generalized. There are some specific carcinogenic models where p53 signaling is not activated however carcinogenesis still continues, in those models DNA-damage does however accumulate. So I wonder if the DNA-damage sensing part of this can be used to label those cells and then their continuing generations will still have the color expressed. Does that sound plausible?

By dhillonv10 (not verified) on 20 Jan 2012 #permalink