Small Talk - The Chemical Language of Bacteria

Bacteria communicate through the exchange of small signal molecules, allowing them to coordinate their behavior in amazing ways. At Garnet Homecoming and Family Weekend, Associate Professor of Chemistry and Biochemistry Stephen Miller discussed how bacteria use this communication to gain advantages, how his lab studies their language on a molecular level, and how we may someday exploit this dialogue to modify bacterial behavior.

Audio Transcript

Stephen Miller: I'm Stephen Miller. I'm a professor in Chemistry and Biochemistry Department, and I'm really delighted to have the chance to tell you about some of the work that's been going on in my lab that's been undertaken over the past five or six years by some really extraordinary Swarthmore undergraduates. So we study bacteria and actually how bacteria interact with each other. I think when most of us think of bacteria you think about individual, single-celled organisms that swim around, and they're looking for food, and reproducing, and making you sick if you get some bad hamburger or something. But it turns out that bacteria aren't entirely anti-social. They actually do communicate with each other and they do so in ways that are surprisingly complicated.

So, the thought is that if we can understand their communication all the way down to the molecular level that we might someday be able to exploit that communication to manipulate the bacteria behavior or at least change the way in which they interact with us and with their environment. So, this whole phenomenon of bacterial communication actually has a Swarthmore connection at its origin and that's through this man. Can we get the house light ... Yeah, thanks.

This is Woody Hastings, class of '47. I don't know how far alums go here today, but Woody is a professor of biology at Harvard. A number of years ago, he was studying a marine bioluminescent bacterium called Vibrio fischeri. What he discovered in his lab was that if you had this bacteria and you put it into a culture, just a couple of the bacteria into a big flask, they would sort of swim around but they wouldn't glow, it wouldn't give off any light. But over time, the bacteria would reproduce, there'd be more and more of them. The density of the bacteria would increase and when it reach a certain threshold all of a sudden the bacteria would all together begin to produce light, and you wind up with a nice glowing flask of Vibrio fischeri.

In trying to understand how this could happen, what they proposed in the Hastings lab was that the bacteria could be making small chemical signals that were being released into the environment. When there were just a couple of bacteria, those signals would diffuse away, but when there were a lot of the bacteria, the signal would accumulate and they would be able to detect it and turn on some behavior in response. This makes sense from the point of view of the bacteria. If you're a single bacteria floating around out in the ocean by yourself, and you start getting off light, a photon here or a photon there, you're wasting the energy involved in making that light because it's not really gonna have any impact on your environment. You're not getting any benefit from it. But if you wait, until there's a large population of bacteria and then start the behavior, then you give off enough light that it can have an actual impact and make a difference.

This, of course, begs the question why do bacteria want to make light in the first place, not to read by certainly. The answer, at least in the case of Vibrio fischeri has to do with another organization with which the fischeri have a symbiotic relationship, and that's the Hawaiian bobtail squid. So this squid lives in the shallows off of Hawaii and it has on its underside a light organ, a little compartment or sac in which it grows cultures of Vibrio fischeri. So it sleeps pretty much during the day so in the morning what it does, it ejects almost all the fischeri from the organ and it goes to sleep. Over the course of the day, the fischeri start to reproduce and they grow until to high cell density. By the evening, they're starting to bioluminesce again. That's when the squid comes out to hunt.

Now it looks big here. It's a very small little squid and it hunts at night swimming around the shallows and it actually has sensors on its back that allows it to detect the moonlight that's hitting the water. It control the light that comes out of this sac. What it does is it erases the shadow that it cast on the bottom of the seabed so that the predators can't see it. It's kind of a camouflage. So this is the symbiosis. The Vibrio fischeri get this lovely environment on which to grow plenty of nutrients and the squid gets to avoid being eaten. It's a very nice relationship. Anyway, it turns out that the method for communication that was proposed by the Hastings Lab is pretty much correct. This whole process is referred to as quorom sensing, the idea being that the bacteria are looking around and checking to see if there's enough of them to make it worthwhile to undertake some activity.

You have some protein inside of the bacteria that makes the signal, the signal gets released into the environment, at low cell density it diffuses away. At high cell density, it accumulates until it's able to be detected by another protein in the bacteria, which then turns on a series of genes that control some sort of behavior. It turns out that a lot of different bacteria use this kind of communication. These small molecules are referred to as autoinducers. They're used to control a lot of different behaviors, things from, we talked about bioluminescence and some symbiosis, but also virulence, antibiotic formation, biofilm formation.

Biofilms are these amazing structures that bacteria form, sort of layered. It's a film, in fact, that's what it sounds like, that can ... They're very difficult to treat and to remove. They're a real problem, actually on the hulls of ships and they foul ships. A lot of the early research on quorom sensing was sponsored by the Navy because they're interested in keeping the biofilms off the ship's hulls. It's also a problem with a lot of medical things that get implanted, these biofilms that form on them they're really hard to clean up. So these are all behaviors where you can imagine the bacteria would want to work together.

If a single bacterium gets inside of you and it becomes virulent, it attacks you, your immune system is not gonna have any problem dealing with it at that point. But if it waits until there's a large population of bacteria and then they attack, things could turn out very differently for the host at that point. So this whole process, of course, it's been studied for three or four decades now, I guess. Until fairly recently, it was thought that it was a very species-specific interaction. That is to say that any individual bacterial species would make a unique signal that would only be detected by that same species so you could have multiple different bacteria living in the same environment and they wouldn't know what the other one was doing at that time.

That was the paradigm for decades until about 10 years ago when Bonnie Bassler's Lab discovered another kind of autoinducer. Now there's another Swarthmore connection connection there. Bonnie Bassler was given honorary degree at commencement last year so that the Swarthmore connections continue. Excuse me. But what the Bassler Lab discovered, I think it's illustrated very nicely in this experiment. So what we have is just a petri dish, with some media on it, and you've painted different species of bacteria in different sections of this dish. So over here on this side and on this side there's a strain of E.coli called O157:H7, that's a strain that makes you sick. It's usually involved in food outbreaks. This top and bottom part is a different strain of E.coli, it's a lab strain called dh5 alpha. Then this arrow in the middle is a different bacterial species called Vibrio harveyi, it's another marine species but this one's also bioluminescent.

So what you can do is you can paint this on and wait a little while and come back and turn off the lights, and then this is what you see. The tips of the Vibrio harveyi that are near that wild type E.coli are starting to glow. The interpretation of this is that the Vibrio harveyi are responding to a signal that's being made by the E.coli. That's able to diffuse across the plate so that one species is responding to a signal that's being made by a different species. That's a whole different way of thinking about quorum sensing. Now we're thinking about inter-species communication, not just within a single species.

To support this idea, it turns out that there are a lot of different bacterial species that make this particular signal. Now this, I told you before that the signals are called autoinducers in general because this one was a new and novel autoinducer. The Bassler Lab named it Autoinducer-2 or AI-2. This is a incomplete list of the bacteria's that have the necessary genes to make this AI-2 signal. You can see there's a couple of really fun and exciting species on this list. We've got Anthrax up here, that's fun. We've got here at the back pylori, if you've had ulcers you know about that one. Salmonella, that's a good time. Cholera, cholera outbreak recently was in the news. Yersinia pestis, that's plague. So all kinds of good bacteria to be interested in.

This sort of gets at the motivation for setting AI-2 in particular because if you can really use this signal to influence in bacteria there's a whole host of biomedically important bacteria that are responding to this single signal molecule. So once this was established, as a chemist, you start to ask the question, "Well, what is the signal molecule?" If we're really going to make drugs and be able to play with this communication, you get to know what it is and specifically, what it is that's chemically. That turned out to be a very tricky question to answer and much more complex than people originally thought.

Again, this is work from the Bassler Lab and they discovered the biosynthetic pathway for AI-2 starts with this molecule, SrH. There's an enzyme called LuxS that cuts off this part and creates this molecule, which we call DPD. The thing about DPD, all the chemists who are sitting upfront here can tell you is that when that gets released into an environment where bacteria live, it's not gonna stay looking like that molecule very long. That's very reactive. You've got all those oxygens and double bonds, it's gonna interact with a lot of things in the environment to form all sorts of adducts, different molecules that start as this but can turn into a whole host of different things. This means that it's not clear what the bacteria are actually recognizing as the signal. It's not gonna be this, it's gonna be derived from that. But that's not what the bacteria are gonna recognize.

So what is, in fact, AI-2? Well, that's a really hard question to answer for us standing on the outside. But fortunately, the bacteria know what they're looking for. They know the answer to that question. So this is a schematic of the Vibrio harveyi. This is the cell and all we really care about is what's going on out here. This is a protein called LuxP and it acts as a sensor for AI-2. It binds to the AI-2 molecule and when it attacks AI-2, it turns on a long pathway that leads to different genes being turned on and light gets produced. That's all well and good, but we're just interested in this protein because it's that protein that knows what the small molecule is.

So here's a very rough schematic of what you can do. Imagine that you have some container that has a mixture of all of these things that are derived from DPD. You don't know which one is being recognized, but you have the receptor protein. So you can just take the protein, you dunk it into this mixture, but it binds to whatever it is that it's looking for you can pull it back out and it's going to bring with it whatever the particular signal is that you're looking for. Then all we have to do is go ahead and look to see what it is that's bound to the single protein. When I say look I literally mean that we go ahead and look. We actually visualize this at atomic resolution in three dimensions. We get to actually see that molecule bound to the protein. I want to show you what that looks like and tell you how we do that. Before I do it, I need to back up a little bit and tell you some things about protein structure and proteins in general.

So this is DNA, most of you are familiar with the molecule. That famous double helix was discovered by Watson and Crick. They won a Nobel Prize along with Maurice Wilkins for that work. It's a series of nucleotides that are strung together in this chain. It contains information about the secret of life, genetic information, yada, yada, yada, whatever. I think DNA is pretty boring. As a protein guy, I'm not that interested in DNA. Jack's Lab works on DNA, I'm sure he's gonna have words for me later. But DNA pretty much always looks like this. I mean that's not entirely true, these variations, but all it is a blueprint. It carries information that tells you how to make proteins. Proteins are where it's at. Proteins are what actually do things. Proteins catalyze the reactions inside of your cells. Protein makes up your hair and your nails. Proteins, actin and myosin, make up your muscles and they slide past each other allowing you to move and to run.

Amazing diversity and function and structure, what proteins can do. They're the business molecules, not DNA. So according to the central dogma of molecular biology, DNA contains the information, it gets transcribed into RNA, and then translated to make these proteins. So all that DNA is, is really a map for making proteins. So what a protein is, is a series of amino acids that are strung together. So here is one, the next one attached one after the other in a long chain. Now, small protein might be a hundred amino acids, large proteins can be tens of thousands of amino acids. There are actually only 20 different amino acids that are used. You can say they all have the same part up here that's ... Let's go back. Right. That's how they're linked together along this backbone and they just differ, if you look at this little R, this purple side chain as we call it, that's the part in red that is different here.

So all that DNA contains is the information that says, "Okay, grab this amino acid and the next grab this one, and then grab that one, and grab another one of these," it tells you how to link them together in this very long chain of amino acids. That's when things start to get interesting because unlike DNA, we go back to DNA, you know, it's just this long chain no matter what's going on with it. Do a first approximation, chain, quadruplet, yeah ... Fine. But a protein does not remain as a long extended chain once you put it in the solution. It turns out that if you take that protein chain and you throw it into some sort of solution, it will spontaneously fold up into an incredibly complicated three-dimensional structure. It will do so repeatedly. The three-dimensional structure is determined only by the sequence of amino acids that are strung together. There's no other machinery that's necessarily involved. It's just how those amino acids are strung together that's gonna determine that structure.

Our understanding of that process of how proteins fold repeatedly and what's involved in that actually has another Swarthmore connection going back. That's through Chris Anfinsen, class of '37, Nobel Laureate for his work. This is not his Nobel picture, this is his picture from the Halcyon. A quote from the Halcyon, "With nostrils that's [inaudible 00:17:10] denoting passion, Anfinsen strolls around campus under a mop of flaxen hair, looking soulfully at the coeds with big blue eyes." I don't know if we still write that kind of stuff in the Halcyon. I haven't looked. I don't know if we still use coeds, I tend to think not, but that's fine. On the other hand, this is what Anfinsen had to say at Swarthmore, which is a sentiment I think that many students have had at some point or another. A little bit of modesty there.

So Anfinsen's work allows us or helps us to understand what goes on when you start with this long extended sequence of amino acids strung together and you throw them into a solution and they begin to fold up to these shapes. This is an example of what one of those shapes looks like. This is a protein called myoglobin. It's actually myoglobin from sperm whale, but you have myoglobin in you, too in your muscle. It's involved in transporting oxygen. The chain, I'm just showing you the backbone here, the chain starts here sort of forms the spiral called the helix and it moves along those and then bond another helix. You can just keep following it through this structure. Amazingly complicated, and it folds up this way spontaneously every time.

Here's another view of it. This is showing you all the atoms. So we're talking about thousands of atoms. Now, you'll notice that it's very hard to get any sense of perspective from this two-dimensional picture because everything is just sort of a mess. It's a jumble. Thus, the 3D glasses, but we'll get there in a second. It's one of the great reasons [inaudible 00:18:49] proteins is because of the cool thing if you get to see it in three dimensions. But okay, so I'm telling you that this is the structure of myoglobin. How do we know that? How can we see how all of these atoms are positioned in space? It's still one of the grand problems in chemical sciences and biology. We can't predict what the proteins are gonna fold up to look like. The problem is still too hard. There's an active huge area of research for decades, and we're getting better at it.

There's actually a game you can download called Fold It. It's fun to play and it actually is helping science happen. There was a paper in Nature not long ago. One of the best folding labs in the country, David Baker's Lab, has put up this game and it lets humans try to fold these proteins and it uses that information. Anyway, check it out on the web if you like. It's fun. But you can't do that universally so instead, you have to do it experimentally. You can't predict the structure, you have to determine the structure. One of the ways to do it, the way that my lab does it, is to use a technique called X-ray crystallography. Now I love this figure, I think it's from, I don't know, the '50s or something like that. But it does a nice job of explaining the metaphor between light microscopy and crystallography.

I think we all have a sense of how light microscopy works. You've got an object, you shine light at it. The object scatters the light, then you have a lens that refocuses the image, and you can magnify the view of whatever it is you're looking at. It's either an amoeba or a pizza crust, I'm not sure which. Well, you can't use light microscopy to see atoms. The reason for that is that the wavelength of light is much bigger than the size of an atom. On the other hand, X-rays are about the same size as atoms. So you can think of crystallography as being sort of similar to doing X-ray microscopy. We have an object, our protein, in fact, the crystal of the protein. I'll show you what that looks like in a second. You shoot X-rays at it and the object scatters the X-rays.

Now the big problem is there's no such thing as an X-ray lens. People are sort of working on it, but you can't really refocus it to get an image yet. But what you can do is measure the way those X-rays are scattered and then using a computer, that notices about the same size of the crystallographer in this figure, and I guess this probably was the '50s [inaudible 00:21:17] and what-not, you can use a computer and [inaudible 00:21:21] punch cards to calculate what the image would look like if there was a lens. That's what we do in X-ray crystallography. That's what allows us to see the actual structure. Now it turns out that it is the electrons that scatter the X-rays so what you see is where the electrons are inside of your structure. I'll show you how we visualize that in just one second.

I told you we use crystals for this. There's a couple of reasons for that. Some of them are technical, but one is very simple. It's awfully hard to pick up a single protein. If instead, you can grow them into crystals, it becomes much simpler to manipulate in your experiment. These are some beautiful crystals of an AI-2 receptor that was grown by Anna [inaudible 00:22:09] who graduated a couple of years ago. Just to give you an idea of scale, we're actually looking at this through a microscope. These are pretty good sized crystals that's about, along the long axis, about that half a millimeter. So a very small, little, tiny crystals, but still much easier to work with than the proteins themselves.

Once you get these crystals, what you can do is you take them to an X-ray source. We went a couple of years back. This is a picture taken at the synchrotron at Brookhaven National Lab. It's one of those particle accelerators, those rings, and a byproduct of whatever the physicists are doing there, is that a lot of really bright X-rays are given off. So we can use those X-rays to get really good dating and a very good picture of our structure. So you go, you do your data collection. Then you come back, you do your calculations. What you wind up with is an electron density map. So now you can put on your 3D glasses. So this is what I'm about to show you is an electron density map for a protein called LsrF. This is a structure that was solved by Zamia Diaz in the lab. What you're seeing, that magenta-colored stuff, 3D, is where the electrons are. It's a little higher I'm gonna cut down on the slide just a little bit.

So you can start to see where the protein is and how that chain is going to move through space. You guys [inaudible 00:23:44] with 3D. So I think you can start to get a sense of why you need to visualize this in three-dimension. It's very hard to work with this in just 2D all the time. Let's take a slightly different view here. I think from this perspective you can see how the chain is, the backbone is running back and forth through the electron density, that string of amino acids that I was telling you about. The job of the crystallographer is to come along and interpret this. Since we know the sequence of the protein, we know the identity of all the atoms and how they're connected to one another. So the trick then is to come along and try to build that protein into the electron density.

So here I'm just showing you the way we would build the backbone going through there. You can actually put on all of the atoms eventually, build them into the density and this can either be very easy if you have very high quality density or sometimes it can be very difficult if the density is not very good if we try to zoom on over here. So you can see how it fits this ring that I've got that's situated over here. But it's not entirely obvious. This has a little bit of a low resolution map. But after you've gone ahead and done all this building, what you're left with is a model of where all of the atoms are in three dimensions for this protein, all thousands of them.

If we zoom out a little bit, this is the structure of that entire protein, LsrF. You can see this is 330 amino acids so it's thousands of atoms, quite complicated even in 3D. Difficult to understand exactly how everything fits together. So we often will represent this instead as a kind of a cartoon that allows you to follow the backbone, you can see the overall fold of the protein even though you don't necessarily have all the details. You can see that this one's got the sort of propeller-like organization where there's this central what's called the beta barrel running down the middle. Lots of helices around the outside, very complex, and again, remember, this only starts out as a long chain of amino acids. It spontaneously folds into this shape.

It's actually even more complicated than that because beyond just being a single protein chain, this protein actually organizes itself into a decamer. Ten of these protein chains spontaneously come together and they form this exquisite symmetric molecule, this sort of star-like shape. We've got one, two, three, four, five different proteins, and again, nothing helps, but they just assemble this way in solution. If we ... Excuse me, let's do this. Here's a surface representation. You can see there actually is a whole that runs through the middle of the disc, and if we take it and we rotate it, you can see there's actually two of these that are sitting on top of each other and they are actually sort of swapping back and forth, this sort of throwing an arm over each other to hang on.

So from an evolutionary standpoint it's just an amazing structure, this without any help the information performing this incredibly complicated structure is embedded just in the sequence of amino acids that comes together. Okay, so that's the structure. Sorry, it's the end of the 3D part of the presentation. But I think you'll see if we ... I had shown you this in, let's do this, in 3D before. Here's the 2D representation of the same protein. I think you can see the advantage of seeing it in 3D and be able to really get a sense of how things are put together and organized. So what do you get when you ... Why do all these work? Why see the structure? What can you get out of it? Well, in this case, this is LsrF. I told you we know what it does in a very general sense from genetic studies. We know it's involved in processing this quorom sensing signal, AI-2.

We don't really know how it does or what the product is yet. [Dan Li 00:28:09] who's in the lab is working on trying to figure that out. He has a poster and I should tell everybody that students from my lab and students from all of the labs in Chemistry departments have posters in the poster session will be going on. And the science that are [inaudible 00:28:24] again tomorrow from 10 to 11. 10 to 11? Yes? Okay, 10 to 11. I encourage you to stop by, the students are there, you can ask them questions about things. So it turns out that structures are very highly conserved even when sequences aren't. So this is LsrF, we saw this from this long tail coming over.

Here's another protein that's very well understood. It's involved in glycolysis. It's called an aldolase. If we overlay LsrF with this aldolase, you can see they have almost exactly the same structure even though they have almost no sequence similarity. Nature conserves structure because structure allows function much more than it conserves sequence. There's a lot of different sequences that can give you the same structure. So we don't know what it does, but because the structure is conserved, we have a pretty good idea, I'm not gonna go through all these, but this is what an aldolase does. And we've a pretty good idea the chemistry of LsrF is gonna be very similar to the aldolase because the structure is so very similar. So that's just some of the wonders of protein structure.

If you remember, 15, 20 minutes ago, I got on to the segue, [inaudible 00:29:44] this protein here, LuxP. Remember LuxP, quorom sensing, AI-2, all that stuff? Right. So here we are, LuxP, and now we can go in and you know of now how we can use X-ray crystallography to look at the protein and not just at the protein but also to the small molecule that's bound to it. So the structure of LuxP was determined by [Shin Chan 00:30:06] in Fred Hughson's Lab. This was back when I was doing my post doc for all the work there. Shin got the structure, and once he built the protein, there was extra electron density left over. That electron density has to belong to the signal molecule. If you zoom in on it, here's what the density of the signal molecule looks like when we know it's derived from DPD and they were able to figure out that they could build this into the density just by looking at it visually. And they could determine that Vibrio harveyi recognizes this molecule as the AI-2 signal, simply by looking at the three-dimensional structure.

There's something very funny about this to a biochemist because this atom here is boron. It may not be surprising to you, but as a biochemist I think about nitrogen, carbon, oxygen, hydrogen, a little bit of phosphate sometimes from sulfur, maybe a [inaudible 00:31:05] never boron. That's just a wacky result. But it turns out there's a lot of boron in seawater so it's maybe not entirely surprising that this borated form of the molecule is recognized by a marine bacterium. But it did raise the question, I mean might another species of bacteria recognize a different form of AI-2. So we got interested in studying what's going on in Salmonella. Salmonella also response to AI-2, but lives in, well, hopefully not, but sometimes in the gut in our environment. So there's not gonna be a lot of boron involved.

We could really just basically use the same tricks. We were able to figure out that the receptor protein is this protein called LsrB. You can go ahead and determine the structure of that protein. Once again, there's leftover electron density for the small molecule. If you zoom in on that you can see it's got a very different shape immediately. It's a single, excuse me, a single ring as opposed to that double ring we saw before. When you build into that electron density, you can see you got a very different kind of molecule. The boron is no longer present, it's a single molecule. A single ring, excuse me. And we're able to show this way that two different bacterial species recognize two different molecules as AI-2. So this complicates the language all of a sudden. There's a common root tongue, if you will, but then there's dialects the different species are speaking. There's a more complicated Lexicon than we would have originally guessed.

So by recognizing that there's this two different forms that allow us to draw a sort of schematic of how this could form, and basically we're starting from the same molecule here, DPD. It's possible that this could close in one of two ways because along this direction and boron gets added, here's the form that's recognized in the ocean by harveyi. Here's the form that's recognized by Salmonella. But the idea that follows from this is that by controlling what's in the environment, specifically by controlling the boron, we can control whether you make this form of the molecule or that form of the molecule. By adding or removing boron you should be able to push this equilibrium in one direction or the other. So this is a very rudimentary way to think about controlling how bacteria behave and respond by controlling the amount of boron in the environment.

We did a little proof of concept experiment, which I'll show you here. Now remember, Vibrio harveyi gives off light in response to the signal. So it's a very nice easy readout. You just look to see whether or not they glow. Well, if we don't give them any signal molecule, you get no light response. On the other hand, if we synthetically make DPD and give it to the real harveyi, if we do it in an environment where there's no boron, you get no light. But then if we add boron all of a sudden they start to give off light. Here, what we've done is we purified the protein from Salmonella, we heat it up to denature that makes it let go of the small molecule. Then we feed that back to the harveyi. The same thing, if there's boron around, we get a signal. If there's no boron, no signal.

We can do the opposite experiment in Salmonella. It's a little more complicated because there's no light response, but you can still tell if the Salmonella are responding or not. Sure enough, if there is no boron around, you get a strong response. If you flood this system with boron, the response is greatly reduced. So it's the most blunt force, rudimentary way of controlling bacterial behavior by manipulating their language. I am not recommending that if you have a Salmonella infection you start pounding down some boron. It's not a good treatment. It's gonna be a lot of work involved in developing therapeutics that are much more subtle and careful, but the principle that you can manipulate how bacteria behave by controlling their language, I think is illustrated here.

The utility of this, at least we can make argument about this, comes if you think about the real problem that we're having with bacteria nowadays. What's the big problem? Resistance, right? We're getting resistance strains to the antibiotics we have. One of the reasons that you get resistance is that when you use an antibiotic, you are hitting the bacteria with the strongest evolutionary pressure you can. You're killing them. So they either evolve, a way to become resistant, or they die. That's the end of it.

On the other hand, what we're talking about doing here is manipulating their language, changing their behavior. The theory is that the evolutionary pressure is weaker here. You can interfere with their language but it won't kill them so that means that in theory, at least, this kind of a drug will take longer for bacteria to evolve resistance to. They'll find a way. Bacteria are amazing creatures and they will beat us eventually. But we should be able to hold on a little bit longer if we could ever make this kind of technology work.

So that's the general principle of what we're talking about. Once we've established that there were a couple of different kinds of autoinducers or different forms of AI-2, one of the things my lab wanted to study was to see if there were still other bacterial species that recognize different forms of AI-2. So I'm not gonna go through all of these, but we basically went and looked in a lot of bacterial genomes. You guys probably all heard about the human genome project, everything associated with that. Well, hundreds and hundreds of bacterial genomes have been sequenced as well, and students in my lab went through and looked at those genomes and tried to look for bacteria that had the genes involved in responding to AI-2.

We were able to cope with a way to divide them into two groups, which we cleverly named group one and group two. Group one have all the components. We think they'll respond to AI-2. We're actually able to experimentally show that many of them do respond to AI-2. And group two don't have the components that are necessary. But what I really want to show you here is something that we noticed, we were compiling this table. This is a listing of the presence of this gene called LuxS. LuxS is responsible for making the AI-2 signal. So these are all the bacteria that we think can respond to AI-2. Look here, most of them have the ability to make AI-2. But there's two of them that don't. They seem to have what they need to listen for the signal but not what they need to make it. We thought this was really interesting.

So we focused on this species here, Sinorhizobium meliloti. It's a soil bacterium, it's a plant symbiont, nitrogen-fixing bacteria. So we decided it'd be kind of fun to play with this for a little while. We did that same light test I was telling you about. We were able to show that, in fact, the protein that we predicted would be the receptor of meliloti is able to bind AI-2. So it does seem to have that capability. A student in my lab a couple of years ago, Randall McAuley went and determined the structure of that receptor and sure enough, we can see the small molecule bound to it. Here's the electron density. It turns out it's exactly the same form of AI-2 that's recognized by Salmonella. So the soil protein and the gut protein are recognizing the same flavor of the language, if you will.

So beyond being able to show that it actually could bind the protein, my colleague, who's at the Instituto Gulbenkian de Ciencia in Portugal ... By the way, one of the things about being at Swarthmore is that we go on sabbatical every fourth year. It's very important that you have a European collaborator for when those sabbaticals come up. A lovely hotel room right on the coast, it was very nice. But Karina's lab is a microbiology lab, they do a lot more work in vivo, and they were able to show that when you add AI-2 to the solution the meliloti actually does change its gene expression. So it's not just detecting it, it's actually responding to it as well.

This is a whole different paradigm all of a sudden. So rather than having a conversation between different bacterial species, now you have a bacterial species that's eavesdropping on what's going on with the others. It's not participating, it's not giving away its quorom-ness, or how many there are, what it's doing, but it is, in some way, going to exploit that information coming from the other species. So another experiment that Karina's Lab did was they showed that if you add AI-2 ... Now remember, meliloti doesn't make its own AI-2 so we had to add outside AI-2, but here's the level of AI-2 in the solution and it goes down over time. That means that the meliloti is able to remove it from the solution even though it's not making it.

Why would it want to do this? Well, we had the idea there might be something going on and it's a sort of communications warfare between different species. So meliloti, it's a plant symbiont, so the idea was that maybe it's in some way trying to interfere with what's going on from a plant pathogen. So Karina did some experiments using another bacteria called Erwinia carotovora. Now carotovora is a plant pathogen, it makes AI-2, and there is some data that suggests that if pathogenicity is controlled by AI-2. What they did is that they grow up a [inaudible 00:41:45] culture of meliloti and carotovora and what they show was the level of AI-2, this one is what we care about is the one with the circles here, that it goes up over time then it comes back down. If you just grow the Erwinia by itself without the meliloti, that doesn't happen, it stays up.

So the idea is that the plant keeps the meliloti around because it gains an advantage from this. So here's a melon, you can see this is the rot, the kind of problem that's caused by the pathogen, Erwinia carotovora. And if that really is linked to AI-2 levels, then the idea is that by keeping meliloti around, meliloti can remove the AI-2 from the environment sort of fouling the communication the pathogen is using, and in that way, protecting the plant from this kind of a rot. So really it's a very subtle way of doing exactly what it is that we hope to be able to do someday. Interfering with the pathogenicity of a species by interfering with its communication. It seems to be happening in the biological world.

I can't give you a ... I wish I could give you better data to document this. We are still trying to do some experiments, growing [inaudible 00:43:04] cultures of bacteria is awful, and then you try ... We're trying it on potatoes and trying to measure the amount of damage in different circumstances. So I will say that this right now is a principle that we're investigating rather than so I can tell you it's factually going on. I mean I can definitely tell you the levels of AI-2 [inaudible 00:43:20] impact, whether or not this really confers a lot of defense of the plant.

I can't tell you that yet. We think that's what's happening. But certainly that's what we're aiming for as well, the idea that by interfering with this communication, that we can control how this pathogenicity happens that could be a real boon for us. We need, in order for that to work, of course, the organic chemists, synthetic organic chemists from the [inaudible 00:43:45] are here. So I figured out how to make really good drugs to get on top of this. It's not our problem. [inaudible 00:43:51] big ideas went with the organic chemists you know, sort out the details later on.

So with that, I think I will stop. I just want to thank the people who actually did this work. So what we have here is a picture of my lab as it stands right now. Changi and Dan and Cailin, all of whom have posters at the poster session, you should go ask them about their research. A lot of the structural work, the structure I showed you, were done by Zamia Diaz for LsrF, Randall McAuley did the work on the meloloti receptor, and everyone really contributed to all facets of the work. It's a really great working environment of students I've had. People collaborating and working well together in the lab. Of course, I also need to thank my collaborators from Portugal, that's Karina and her lab. It's a wonderful collaboration we've had over a number of years. Of course, the Dreyfus Foundation, Merck, AAAS, NIH, and of course, the college for supporting the research over these years. So thank you.