Developmental Biology of a Simple Organism: Bacillus subtilis
Transcript of Part 3: Stochasticity and Cell Fate
00:00:06.06 Hello, my name is Richard Losick 00:00:08.19 and this is the third part and last part of my presentation on 00:00:14.03 developmental biology of a simple organism. 00:00:17.10 This last part is devoted to the topic of stochasticity and cell fate. 00:00:23.04 I grew up in an era in which it was believed that development, biological development 00:00:30.20 is orchestrated by highly deterministic processes. 00:00:34.23 And indeed, that's largely the case. That's true in most cases. 00:00:39.23 But, increasingly, we're seeing that there are examples, especially in the microbial world 00:00:45.01 of cell fate decisions that are stochastic. 00:00:51.18 I'm going to tell you about a series of such examples from the spore forming 00:00:56.28 bacterium Bacillus subtilis. Let me begin with a famous quote 00:01:02.00 from Albert Einstein, the father of modern physics. 00:01:05.20 Einstein famously said that "I, at any rate, am convinced 00:01:10.10 that He does not play dice." 00:01:12.08 Thereby rejecting Heisenberg and his Uncertainty Principle. 00:01:17.21 Well, in biology, as I've said, most decisions are, indeed, highly deterministic. 00:01:26.01 But, itâ€™s also the case that some decisions, some cell fate decisions, 00:01:32.12 are, in fact, done by a role of the dice, as I'll explain. 00:01:37.21 So, I'm going to give you four examples from Bacillus subtilis 00:01:40.28 which, so far, appears to be the champion of stochasticity in the microbial world. 00:01:46.05 The four examples are: growth and competence, swimming versus chaining, 00:01:51.25 eating versus being eaten, and community versus individuality. 00:01:59.25 What do I mean by growth versus competence? 00:02:03.03 Well, B. subtilis can switch into an alternative state, 00:02:07.19 the state of competence, in which it stops growing and instead, acquires 00:02:12.29 the ability to take up DNA from its environment 00:02:15.20 which it can recombine into the chromosome. 00:02:21.13 Why would B. subtilis do such a thing if this means stopping to grow? 00:02:25.04 Well, what one imagines is that producing cells 00:02:30.13 that are able...that are on the prowl for 00:02:33.00 new genetic sequences improves the fitness of B. subtilis by allowing 00:02:38.21 it to uptake new genetic information that may help it cope 00:02:43.23 with changed circumstances in the future. So there must be a fitness benefit 00:02:48.15 to the bacterium for temporarily entering this non-growing state 00:02:52.21 so that it can be always on the look out for 00:02:55.28 potential new genetic sequences that can be useful to it. 00:03:00.16 Entry into this state of competence is controlled by a transcription factor 00:03:05.11 called ComK and as you'll see, the synthesis of ComK 00:03:10.22 is governed by a noise-driven stochastic switch. So in other words, 00:03:15.11 when cells are under conditions in which they're capable of entering the 00:03:20.08 competent state, only some of them do and the ones that decide 00:03:23.12 to do so, do so in a stochastic fashion. 00:03:26.17 Let me illustrate this to you with a beautiful experiment 00:03:31.02 from David Dubnau, who, along with others in the field are 00:03:34.06 responsible for our understanding of stochasticity in the competent state. 00:03:43.00 So what I'm going to show you is a field of cells in which all of the cells 00:03:46.10 harbor a fusion of the gene for the green fluorescence protein 00:03:50.09 to a promoter under the control of ComK. 00:03:54.09 And what you can immediately see is that only a subset of the cells 00:03:58.11 are brightly green. That is, most cells are off for ComK 00:04:03.25 and some cells are on for ComK. 00:04:07.23 These cells were grown and held in a homogeneous environment. 00:04:12.03 All of them, in principle, are capable of becoming competent, 00:04:16.25 of activating ComK but only some of them do so 00:04:20.15 and they do so in a random fashion, independently of 00:04:26.07 what the neighboring cells are doing. 00:04:28.25 How does this work? 00:04:30.07 So at the heart of this system is the following circuitry. 00:04:35.12 The comK gene, of course, encodes the ComK regulatory protein 00:04:40.04 which can bind to and activate the transcription of about 100 target genes 00:04:44.14 that define the state of competence. 00:04:48.03 But ComK also bind to the promoter for its own gene 00:04:52.18 in which case it sets up a positive feedback loop 00:04:56.18 which can stimulate transcription of its own genes. 00:04:59.26 So when ComK binds there, that leads to more transcription of comK 00:05:04.03 that results in yet more transcription which in turn leads to yet more ComK molecules 00:05:11.00 that lead to this on state in which large levels of ComK accumulate in the cell. 00:05:17.14 The key point is that this positive feedback loop has a threshold. 00:05:22.28 You can think of it as being poised on a knife edge. 00:05:26.10 And under the right conditions the cells have just less than a threshold 00:05:31.17 amount of ComK molecules in them and if, due to noise, 00:05:37.18 there are fluctuations in the amount of ComK from cell to cell then 00:05:43.21 some cells will have a bit more ComK than other cells. 00:05:47.18 Those cells that have a bit more have reached the threshold 00:05:51.13 and get the positive feedback loop going. Those that are below 00:05:55.20 the threshold can't get the positive feedback going. 00:05:59.08 And what makes this switch a bi-stable switch 00:06:03.10 is that multiple ComK molecules bind to the promoter in a cooperative fashion 00:06:11.05 by interacting with each other. This makes the switch highly sensitive 00:06:16.07 to small fluctuations in the level of ComK molecules. 00:06:20.24 So, when the amount of ComK in a cell is just below the threshold, 00:06:25.13 most cells will not activate ComK, but a few cells will have, 00:06:31.06 by noise driven processes, accumulated enough ComK molecules 00:06:36.19 to activate the positive feedback loop and get it going 00:06:39.27 and go into the competence on state. 00:06:42.12 Why does B. subtilis do this? 00:06:44.24 Well, we don't know for sure but, obviously, entering a state in which you're not 00:06:49.29 growing puts you at a disadvantage. But by deploying, stochastically, some cells 00:06:56.15 that are on the prowl for new genetic information, then B. subtilis is always 00:07:02.03 preparing itself for unexpected changes in its environment 00:07:06.07 when new kinds of genetic information may be important. 00:07:11.16 So, we can think of this as an example of bet hedging. 00:07:15.09 That is, B. subtilis is hedging its bets by deploying, stochastically, 00:07:20.02 a small proportion of cells that enter a non-growing state temporarily 00:07:25.05 so that should circumstances change 00:07:29.18 and should the right genetic sequences appear 00:07:32.14 then those cells will be at an advantage. 00:07:34.17 Remember, that evolution selects for the genome and not the individual. 00:07:39.00 So, deploying two kinds of cells in the population 00:07:42.02 can be advantageous to the genome 00:07:44.16 even if itâ€™s not advantageous to the individual. 00:07:47.19 Let me come to my second example, motility versus chaining. 00:07:52.26 This is a phase contrast micrograph that depicts B. subtilis cells as 00:07:58.04 we've traditionally seen them over many decades of research with this organism. 00:08:05.16 And if you look closely you can see there are two kinds of cells here. 00:08:09.03 There are long chains of cells that have completed cell division 00:08:12.01 but haven't separated from each other and there are also singlets and doublets. 00:08:17.17 The singlets and doublets, it turns out, are motile cells, 00:08:20.17 where as the long chains are non-motile cells. They're sessile cells, if you will. 00:08:26.03 Well, we saw this image for many, many years but didn't pay much attention to it. 00:08:33.01 But over time it emerged that a single transcription factor 00:08:36.07 called sigmaD is responsible for the production of enzymes that 00:08:42.29 degrade the cell wall material between newly divided cells, 00:08:46.25 enabling them to separate and also for the production of the machinery 00:08:51.18 that's responsible for motility. 00:08:55.16 So, with that in mind we revisited this field of exponential phase cells 00:09:02.08 but this time using cells that were tagged with green fluorescence protein reporter 00:09:07.21 gene fused to a promoter under the control sigmaD. 00:09:12.17 And at the same time we stained the cells with a red membrane dye 00:09:16.22 so that we could see the division septa. 00:09:18.25 And now, all of a sudden, we get a radically different view of 00:09:23.11 the field of cells that's very illuminating. 00:09:26.08 As you can see there are cells in two states. 00:09:30.04 There are sigmaD on cells that are doublets or singlets 00:09:34.28 and these are the motile cells. They undergo cell division 00:09:38.26 and then the products of cell division can separate from each other. 00:09:43.02 And then there are the sigmaD off cells which we see as long 00:09:47.06 red membrane staining cells. You can see the division septa in these cells. 00:09:51.19 The cells have divided but the daughter cells haven't separated from each other. 00:09:55.22 These cells are off for sigmaD and they're non-motile. 00:10:00.13 Why would B. subtilis do this? 00:10:02.03 Well, of course we don't know but itâ€™s attractive to imagine 00:10:05.26 itâ€™s another example of bet hedging. 00:10:09.04 Imagine that B. subtilis is in a particular niche where there are nutrients. 00:10:13.04 The sessile chains of cells can stay put and exploit the existing niche. 00:10:21.07 But the motile cells, the sigmaD on cells, we can think of those 00:10:25.11 as nomadic cells that wander off to look few niches. 00:10:29.28 B. subtilis doesn't know what the future holds in store and so its immediate niche 00:10:34.14 may run out of nutrients or exhibit other adverse environmental factors. 00:10:41.04 So, by deploying some cells to be motile and some cells to be sessile 00:10:46.06 the bacterium can hedge its bets. Some cells stay put 00:10:50.21 and exploit existing circumstances whereas other cells swim off 00:10:56.00 looking for new niches in anticipation of the possibility that the original 00:11:02.15 niche may become exhausted. 00:11:06.12 My third example is called eating versus being eaten. 00:11:11.13 So, as we saw in the first part of my presentation, when starved for nutrients 00:11:16.29 B. subtilis enters the pathway to sporulate. 00:11:20.26 So just as competence represents a distinct state, 00:11:24.00 sporulation represents a specialized developmental state. 00:11:28.13 Entry into sporulation is governed by a master regulatory protein called Spo0A. 00:11:35.01 Sporulation is a complex process that takes multiple hours. 00:11:40.17 It takes time and it takes energy, 00:11:43.16 and it culminates in the formation of a dormant cell type, 00:11:47.28 the spore that can remain dormant for many years. 00:11:50.22 So I think itâ€™s easy to imagine that the decision to sporulate 00:11:54.11 is not one that B. subtilis wants to take lightly. 00:11:58.01 Entry into this pathway is governed by Spo0A 00:12:01.18 which sits at the top of this regulatory sequence. 00:12:06.04 It becomes activated under conditions in which nutrients become limiting. 00:12:12.28 Well, there is a crucial window of time near the start of sporulation 00:12:17.25 when the cells can change their minds, so to speak. 00:12:22.17 Initially, when Spo0A become active but before the hallmark process 00:12:28.16 of asymmetric division takes place, if nutrients reappear, sporulation is arrested 00:12:34.16 and the cells can even start growing again. 00:12:37.21 But once the cells cross the Rubicon, so to speak, of asymmetric division 00:12:42.25 now, they're committed to making a spore, 00:12:44.18 even if lots of nutrients appear at later stages. 00:12:48.18 So there's this window of time up until asymmetric division takes place 00:12:54.05 when the process is reversible, 00:12:56.14 and then it becomes irreversible later on. 00:12:59.15 So hold that thought in mind when I tell you what I 00:13:02.29 think at this point will not come as a surprise, that Spo0A 00:13:06.21 is itself subject to a bi-stable switch. 00:13:11.01 That is, when cells are limited for nutrients, only some of the cells become on for Spo0A 00:13:18.02 whereas others remain off for Spo0A. 00:13:22.01 And once again, we can visualize this by using a green fluorescence protein gene 00:13:27.07 fusion to a promoter under the control Spo0A. 00:13:31.10 And as you can see there are two kinds of cells: 00:13:33.19 those in which Spo0A is off and those in which Spo0A is on. 00:13:38.20 Only some of the cells are on for Spo0A. 00:13:41.24 Well, why have a bi-stable switch? Why is Spo0A subject to a bi-stable switch? 00:13:49.00 Well we believe that the answer has to do with a phenomenon 00:13:52.18 that we refer to as cannibalism. 00:13:55.22 When the cells are deprived of nutrients 00:13:59.16 and Spo0A becomes activated, the Spo0A on cells produce and export 00:14:05.15 a toxin, a peptide toxin that kills their sibling cells, the cells that are off for Spo0A. 00:14:13.07 The producing cells, the on cells are immune to the toxin 00:14:16.20 but the non-sporulating cells are killed by it. They lyse and liberate their nutrients. 00:14:22.23 Now, remember, Spo0A activation is triggered by nutrient limitation in the first place. 00:14:28.19 So, if some of the cells are lysing and liberating nutrients 00:14:30.26 that will have the affect of impeding Spo0A activation 00:14:36.08 and that will arrest sporulation or perhaps even reverse it. 00:14:39.17 So, cannibalism is a process for slowing down sporulation. 00:14:44.26 Let me illustrate that for you with this single agar plate experiment. 00:14:52.08 So the left side of the slide shows a streak of wild type cells 00:14:58.08 and the right hand portion of the slide shows a cannibalism mutant. 00:15:05.00 When cells start to sporulate, the colonies become white and opaque 00:15:09.27 and so we can see that with the naked eye. 00:15:12.28 And as you can see at the early time when this photograph was taken 00:15:16.26 the wild type cells have only just begun to sporulate. 00:15:21.13 The colonies are not yet white and opaque. 00:15:23.11 But the mutant cells are filled with spores. 00:15:28.04 Now remember, cannibalism is a process for delaying sporulation. 00:15:32.12 So, therefore, in its absence sporulation is accelerated. 00:15:37.29 So we see rapid sporulation in a cannibalism mutant 00:15:42.20 and slow sporulation in the wild type. 00:15:45.03 Once again we can ask, "Why would B. subtilis do such a thing?" 00:15:50.17 And the appealing interpretation is, yet once again, itâ€™s hedging its bets. 00:15:55.27 Consider a population of B. subtilis cells that 00:15:59.18 experiences a drop in the availability of nutrients. 00:16:02.23 How does it know whether this decrease in nutrients is a simple 00:16:07.21 fluctuation, a temporary decrease in nutrients or the beginning of a famine? 00:16:13.07 If it willy-nilly committed itself to making a spore when nutrients were depleted 00:16:21.08 and went through this multi-hour, expensive process 00:16:24.12 and nutrients actually returned after the time of 00:16:28.21 commitment, well, it would put itself 00:16:30.25 at a disadvantage relative to other bacteria that would be simple waiting out 00:16:35.17 the period of low nutrients. 00:16:39.00 On the other hand, if itâ€™s in a period of prolonged starvation, 00:16:42.03 then going ahead to make spores makes good sense. 00:16:46.28 So, cannibalism is a way to stall for as long as possible 00:16:51.01 before crossing the Rubicon, before committing to spore formation 00:16:56.22 even at the expense of committing fratricide, killing and feeding on 00:17:01.07 genetically identical sibling cells. 00:17:05.19 Finally, I come to the example of individuality versus community. 00:17:11.19 This represents work done in collaboration with Roberto Kolter 00:17:16.21 and concerns the topic of multicellularity in biofilm formation 00:17:21.12 which was the subject of the second part of my presentation. 00:17:26.08 So, wild strains of B. subtilis can make architecturally complex communities. 00:17:31.24 In a standing culture, these communities form at the air liquid interface 00:17:37.22 in a structure known as the pellicle, that has an elaborate and distinctive architecture. 00:17:43.04 And on colonies, on solid medium, we also see an elaborate architecture 00:17:48.08 with thick veins and aerial structures. 00:17:51.25 The cells in these communities are held together 00:17:54.15 by an extracellular matrix, kind of a cement, that holds long chains 00:18:00.15 of cells together so that the architecture can be built. 00:18:05.08 This matrix consists of two components: a polysaccharide and a protein component 00:18:11.05 that are exported from the cells. 00:18:15.01 The matrix is subject to intricate regulation for its production. 00:18:20.02 And I've summarized, in a simplified form, the regulatory pathway 00:18:25.20 by which the matrix is produced. 00:18:29.12 But, most proximal to the genes for the matrix is 00:18:32.02 a repressor protein that holds them inactive. 00:18:35.20 The repressor protein is inactivated by another protein that we call an anti-repressor. 00:18:41.12 And finally, the anti-repressor is produced under the control 00:18:45.27 of our good friend Spo0A. And you will recall that Spo0A 00:18:50.14 is subject to a bi-stable switch. 00:18:54.18 Well, this predicts that if we look in cells that are about to form 00:18:59.15 a biofilm, we'll see the repressor being produced in all of the cells 00:19:04.27 but the anti-repressor, which is under the control Spo0A 00:19:08.15 will be produced in only a subset of the cells. 00:19:12.04 Let's look. First, I'll show you a field of cells that has 00:19:16.10 a green fluorescence protein gene fusion to the repressor gene. 00:19:20.08 And you can see, more or less, all of the cells are green. 00:19:25.06 All of them are producing repressor. 00:19:27.22 Now let's look at a comparable field of cells but this time 00:19:30.15 the green fluorescence protein gene is fused to the anti-repressor gene. 00:19:35.04 And now we get a radically different picture. 00:19:38.25 Only some of the cells, a minority of the cells are on for anti-repressor production. 00:19:45.18 That is, they're on for Spo0A and therefore, on for anti-repressor. 00:19:50.03 Hence, they're inactivating the repressor. Hence, these are the matrix producing cells. 00:19:54.18 And from this we conclude that some cells make matrix for the entire community. 00:20:00.02 This is a kind of altruism in which some cells are dedicated to making 00:20:05.12 matrix for the entire community of cells 00:20:10.13 and the other cells specializing in other directions. 00:20:17.05 OK, so, I've given you four examples from a bacterium, 00:20:20.21 a single bacterium in which cell choices are made in a stochastic manner. 00:20:26.15 But I don't want to leave you with the impression that stochasticity is unique to bacteria. 00:20:32.29 I'd like, in closing, to consider the case of the mouse olfactory neuron 00:20:37.25 and the eye of the fly which provide two examples of stochasticity 00:20:43.10 in complex metazoans. 00:20:46.16 So, the mouse devotes a great deal of its genetic material to the process of smell. 00:20:55.04 Fully 4% of its genes encode receptors, membrane receptors for odorants. 00:21:03.06 There are about 2000 such genes in the chromosomes of the mouse, 00:21:08.04 in the diploid mouse and...but itâ€™s the case that any given neuron 00:21:15.09 must express only a single receptor. 00:21:20.08 Otherwise, the mouse would be confused as to what odor it was sensing. 00:21:24.00 So, how does this work? 00:21:30.00 Here, in a cartoon form, is depicted a neuron that's expressing 00:21:36.06 a particular odorant receptor depicted in red. 00:21:40.18 Itâ€™s on and all of the other 999 odorant receptor genes on one haploid set 00:21:49.10 and the homolog on the homologous chromosome are off. 00:21:54.11 This cell expresses only one out of 2000 genes. 00:21:57.25 Now, how does this work? 00:21:59.07 Well, you could imagine, I suppose, a very complicated regulatory network 00:22:05.08 that was special for every neuron that ensured that only one 00:22:09.23 out of 2000 genes was turned on. 00:22:11.24 But, that would be so complicated itâ€™s hard even to imagine how it would work. 00:22:16.14 Instead, the mouse has evolved a very elegant strategy. 00:22:21.17 It turns on, in any given neuron a single receptor gene stochastically. 00:22:26.10 Each neuron throws a roll of the dice to decide which receptor gene to turn on 00:22:34.09 and then, by mechanisms that are not yet fully clear, all other genes in that neuron, 00:22:40.07 all other odorant receptor genes, are prevented from being expressed. 00:22:45.26 My last example, concerns the eye of the fly. 00:22:49.26 The eye of the fly is a compound eye. Flies don't have the simple eyes 00:22:55.02 that we have. They have many eyes as do other insects. 00:22:58.08 These compound eyes consist of many clusters of light sensitive cells 00:23:04.06 called ommatidia. Each of these ommatidia 00:23:07.12 can produce either of two color sensitive rhodopsins called rh5 or rh6. 00:23:15.13 So the eye is a field of many ommatidia and each of these ommatidia 00:23:21.07 switch on either the blue rh5 or the green rh6. 00:23:28.06 And they do so, they make this choice stochastically. 00:23:32.15 This is the work of Claude Desplan and I illustrate it to you 00:23:37.08 with this marvelous image from Desplan in which 00:23:40.24 you'll see a field of ommatidia in which some cells are producing the green rhodopsin 00:23:48.08 and others are producing the blue rhodopsin. 00:23:52.05 And if you stare at this image for a while you'll see that there's no consistent pattern. 00:23:56.25 It's stochastic. Itâ€™s not a simple flip of coin. 00:24:00.29 Itâ€™s not fifty-fifty. Itâ€™s a biased stochastic switch. 00:24:04.12 Itâ€™s biased in favor of the green choice in a ratio of about 70 to 30, 00:24:10.16 but any individual ommatidium is making its choice randomly, stochastically. 00:24:20.14 Such that, on average, the average decision is a ratio 70 to 30. 00:24:25.15 And if you look closely at this you'll see no consistent pattern 00:24:29.04 and if you look a the other eye of the fly you would see a different pattern. 00:24:32.21 If you looked at other flies you would see yet other patterns. 00:24:35.18 So the choice is stochastic. 00:24:38.21 So, in conclusion, we can say that nature does, indeed, know how to 00:24:44.20 make deterministic decisions, but, in contrast to Einstein's view of the universe, 00:24:50.07 she also knows how to leave certain decisions to a roll of the dice. 00:24:55.19 Thank you very much.