Logic gates (AND, OR, NOT, etc.) are the basis of electronic computation. If we'd like to implement biological computation, one of our first steps has to be implementing similar logic gates using proteins and DNA. That is, we need to make devices that accept a few inputs, perform a logical operation on them, and then spit out the result. In the case of an AND gate, we want the output to be ON whenever both of the inputs are ON, and OFF when either input is OFF. It seems easy, but of course, this turns out to be a lot harder in biology (hence, people writing papers about it).
Figure 1: The result of this paper, if you abstract away all the interesting stuff. [Source]
What makes it hard to make an AND gate out of biochemical parts?
Lack of standard connectors. In electronics, every signal is carried by a current, and every connector is a wire. That isn't the case in biology. Biological signals are typically carried by the presence or absence of some protein that carries out some particular chemical reaction that affects other proteins. This is wildly nonstandardized, and it means if Protein A interacts with Receptor A, you can't just plug in Receptor B and expect things to work.
Fuzzy, non-discrete behavior. It's nearly impossible for a biological system to have a perfect ON or OFF state. Even if a signal is mostly off, there'll be a few molecules of it floating around somewhere. And when you go to turn it on, it'll take time. Basically, biological things tend to vary continuously and not discretely (in large jumps).
Crosstalk. If a biological device relies on some particular molecule, then that molecule is going to be everywhere in the cell. So, you can't put two copies of the same device into a cell and expect them to operate independently. They'll interfere or "crosstalk" with each other in ways you don't expect. In contrast, you can throw down dozens of electronic circuit elements onto a breadboard and they won't interfere with each other because they're separated by physical space. In biology, everything's in the same soup.
Anderson et al's AND gate goes a long way toward solving the first problem, standardizing connectors. Both its inputs and outputs consist of the connection between a promoter and the gene it controls. This is one of the most modular, most plug-and-playable connections known in biology today. You can stick basically any promoter in front of basically any gene, and it will work just fine.
Quick aside: A promoter is a short sequence of DNA that sits upstream of a gene and attracts RNA polymerase to read that gene. Some promoters are constitutive (always on), and some are inducible -- they turn on in response to some molecule or other condition. (Usually, the key molecule binds an activator protein and alters its shape, such that the activator can bind to the promoter and help recruit RNA polymerase.)
Figure 2: In the noble artistic tradition of molecular biologists (who draw all proteins as blobs), promoters are usually drawn as these bent-arrow things.
So that's how the authors of this paper gave their AND gate its modular connectors, which can be plugged in to a wide variety of inputs and outputs. You can choose any two promoters (which respond to any two molecules or environmental conditions you want), and hook them up to the input genes. Then once the signal travels from the input genes through the device and to the output promoter, you can hook up any output gene after that. The output gene drives the cell's response when both of its inputs are on.
Now, let's dive into the guts of the AND gate to see just how it works.
Recall that three bases of an mRNA (messenger RNA) code for a single amino acid. There are three stop codons (UAG, UGA, and UAA) that code for no amino acid at all, but tell the ribosome when to stop translating the protein. Stop codons work because there is no tRNA (transfer RNA) to match them.... usually. There are mutant tRNAs that will do the job. They feel a little dirty, running stop signs, and they can mess up the rest of the cell's workings, but they do exist. One of the AND gate's input promoters is hooked up to the gene that creates the mutant tRNA that reads through the UAG stop codon. UAG is called the "amber" stop codon, and its corresponding tRNA is called the amber suppressor. Hence the title of this post.
(The reason it feels dirty is, of course, that stop codons are there for a reason. Proteins have to end at the right place, after all. But it turns out that running through UAG is less harmful than running through either of the other two, as UAG stop codons are less than 10% of the cell's total stops. And the investigators did check; this system barely affects the cell's health at all.)
But a mutant tRNA won't do anything without an mRNA to match. Now here's the real logic of the AND gate! The other input promoter is hooked up to a gene with UAG stop codons sprinkled throughout. If this mRNA gets transcribed without the mutant tRNA around, it can't get translated into protein because a UAG stop codon will call the whole thing to a halt.
So you need both the gene with stop codons, and the mutant tRNA in order to get a functional protein.
This protein is an Ultra Special RNA polymerase called T7 (it comes from the T7 phage). T7 polymerase recognizes a different kind of promoter than regular polymerase, so it won't work on the rest of the promoters in the cell. It will only work on its partner Ultra Special T7 Promoter, which just so happens to be the output promoter of the AND gate! Man, that's so clever. I never would have thought of that. It's pretty elegant.
Figure 4: The whole shebang in all its glory. The mRNA is the gray line with red TAG (=UAG) stop codons in it, and the mutant tRNA is the yellow splat. Published as Figure 1 in the original paper.
To sum up: when we have both an mRNA with stop codons, and a mutant tRNA that can read through those stop codons, then we get a protein that's capable of transcribing the output gene. Cool!
Figure 5: It works, too. That's Inputs 1 and 2 on the X and Y axes, and fluorescent protein output on the Z axis. Published as Figure 2A in the original paper.
Now, at the very end of the paper, the authors did something that rather endeared them to me. Throughout the design, building, and testing stages, they used green fluorescent protein as an output, because it's easy to read and easy to quantify. This is exactly as it should be. But ultimately what we're trying to produce is behaviors. These are much harder to read and quantify. In most cases, it ultimately comes down to some poor student looking at cells (or worms, or whatever) under the microscope and saying "OK, these ones are behaving in X way, and these ones aren't". So, although behavior is not the most suitable output when you're trying to build a circuit, it's the acid test for when you've got something you're pretty confident in and you want to know, Does It Actually Work?
Accordingly, these investigators took their fancy AND gate and hooked it up to a completely different set of inputs and outputs (incidentally, demonstrating that that was even possible!). The new output was not fluorescence, but invasion -- invasion of eukaryotic cells (using a protein from a close relative of the Black Plague, creepily enough). Sure enough, the bacteria stayed quiet when they received no inputs, or only one of the two inputs; and they leapt into action when they detected both.
Even if we don't want to make a Turing-complete cell just yet, AND gates turn out to be particularly useful for all kinds of biology-specific applications. If you're a cell and you have to tell the difference between Environment A and Environment B (and respond accordingly), often you can't tell by only sensing one thing. You might have to sense the local pH AND the local salinity AND the presence of Random Molecule #32493 AND whether there are more of you in the vicinity... do you see where this is going? Cells in the wild have to do this kind of simple computation all the time, and we'd like our engineered bacteria to do the same. It'll enable them to make sophisticated decisions like "Hmm, this patch of cells is expressing Cancer Marker A, and Cancer Marker B, and Cancer Marker Q... I'll secrete poison and destroy it!" Et voila, cancer-killing bacteria! (In principle. You've still got to get around the immune system and so on and so forth...)